Acquired Substrate Preference for GAB1 Protein Bestows Transforming Activity to ERBB2 Kinase Lung Cancer Mutants

Background: Activating ERBB2 mutants drive tumor formation. Results: Oncogenic ERBB2 has a striking substrate preference for GAB1 in vitro, and GAB1 hyper-phosphorylation is required for mutant ERBB2-induced cell signaling and transformation. Conclusion: Acquired substrate preference for GAB1 is critical to the ERBB2 mutant-mediated oncogenesis. Significance: Understanding the activation mechanism of mutant ERBB2s may lead to the development of therapies targeted against these oncogenic kinases. Activating mutations in the αC-β4 loop of the ERBB2 kinase domain, such as ERBB2YVMA and ERBB2G776VC, have been identified in human lung cancers and found to drive tumor formation. Here we observe that the docking protein GAB1 is hyper-phosphorylated in carcinomas from transgenic mice and in cell lines expressing these ERBB2 cancer mutants. Using dominant negative GAB1 mutants lacking canonical tyrosine residues for SHP2 and PI3K interactions or lentiviral shRNA that targets GAB1, we demonstrate that GAB1 phosphorylation is required for ERBB2 mutant-induced cell signaling, cell transformation, and tumorigenesis. An enzyme kinetic analysis comparing ERBB2YVMA to wild type using physiologically relevant peptide substrates reveals that ERBB2YVMA kinase adopts a striking preference for GAB1 phosphorylation sites as evidenced by ∼150-fold increases in the specificity constants (kcat/Km) for several GAB1 peptides, and this change in substrate selectivity was predominantly attributed to the peptide binding affinities as reflected by the apparent Km values. Furthermore, we demonstrate that ERBB2YVMA phosphorylates GAB1 protein ∼70-fold faster than wild type ERBB2 in vitro. Notably, the mutation does not significantly alter the Km for ATP or sensitivity to lapatinib, suggesting that, unlike EGFR lung cancer mutants, the ATP binding cleft of the kinase is not significantly changed. Taken together, our results indicate that the acquired substrate preference for GAB1 is critical for the ERBB2 mutant-induced oncogenesis.

Activating mutations in the ␣C-␤4 loop of the ERBB2 kinase domain, such as ERBB2 YVMA and ERBB2 G776VC , have been identified in human lung cancers and found to drive tumor formation. Here we observe that the docking protein GAB1 is hyper-phosphorylated in carcinomas from transgenic mice and in cell lines expressing these ERBB2 cancer mutants. Using dominant negative GAB1 mutants lacking canonical tyrosine residues for SHP2 and PI3K interactions or lentiviral shRNA that targets GAB1, we demonstrate that GAB1 phosphorylation is required for ERBB2 mutant-induced cell signaling, cell transformation, and tumorigenesis. An enzyme kinetic analysis comparing ERBB2 YVMA to wild type using physiologically relevant peptide substrates reveals that ERBB2 YVMA kinase adopts a striking preference for GAB1 phosphorylation sites as evidenced by ϳ150-fold increases in the specificity constants (k cat /K m ) for several GAB1 peptides, and this change in substrate selectivity was predominantly attributed to the peptide binding affinities as reflected by the apparent K m values. Furthermore, we demonstrate that ERBB2 YVMA phosphorylates GAB1 protein ϳ70-fold faster than wild type ERBB2 in vitro. Notably, the mutation does not significantly alter the K m for ATP or sensitivity to lapatinib, suggesting that, unlike EGFR lung cancer mutants, the ATP binding cleft of the kinase is not significantly changed. Taken together, our results indicate that the acquired substrate preference for GAB1 is critical for the ERBB2 mutantinduced oncogenesis.
Protein kinases are key players in cellular signaling, and their activity is strictly controlled by multiple layers of autoinhibition under normal physiological conditions (1)(2)(3). Mutations can cause relief of autoinhibitory constraints and constitutive activation of kinases. Many human cancers arise from and are addicted to constitutively activated protein kinases (1,4). The development of cancer therapies targeting mutated kinases requires detailed characterization of their specific roles in the process of tumorigenesis and investigation of the underlying activation mechanisms.
ERBB2 (also known as Her2/Neu) is a member of the ERBB receptor-tyrosine kinase family that also includes EGF receptor (EGFR, 3 ERBB1/Her1), ERBB3 (Her3), and ERBB4 (Her4). A unique characteristic of ERBB2 is that it does not have a specific ligand and has been considered a preferred dimerization partner for other ERBBs. Overexpression or dysregulation of ERBB2 kinase activity has been found in various human cancers, and accordingly, the receptor has been extensively studied as a therapeutic target (5,6). An ERBB2 monoclonal antibody, Herceptin, and the tyrosine kinase inhibitor, lapatinib, have been approved for the clinical treatment of ERBB2-overexpressing breast cancer, and many inhibitors are currently in development (6). ERBB2 kinase mutations in the ␣C-␤4 region have been identified in non-small cell lung cancers (7,8). Our previous work has demonstrated that the unique glycine-rich ␣C-␤4 loop plays a critical autoinhibitory role in control of the ERBB2 kinase activity (9), and this observation has been verified by a recent crystallographic structure (10). An earlier work revealed that the most common oncogenic ERBB2 mutation with an in-frame duplication of four residues YVMA before Gly-776 (ERBB2 YVMA ) has increased autokinase activity and is more potent than wild type (WT) kinase in cell signaling and tumorigenicity (11). Using a transgenic animal model, ERBB2 YVMA has been demonstrated to drive rapid development of lung tumors, which have striking histological and radiologic phenotypic similarities to human bronchogenic adenocarcinomas (12).
Adaptor/docking proteins are required for transmitting key downstream cellular signals from receptor kinases and contribute to the specificity and amplification of the signaling pathways by selectively recruiting and activating signaling proteins (13). GAB1 is the prototype of a docking protein family that also includes GAB2 and GAB3 (for reviews, see Refs. 14 -16) and contains a pleckstrin homology domain, several proline-rich motifs, and conserved tyrosine residues that can be phosphorylated by activated receptor-tyrosine kinases. These GAB1 phosphotyrosines selectively bind Src homology 2 domain containing signaling proteins such as SHP2 and PI3-kinase (PI3K), and GAB1 is a critical mediator of EGFR-and ERBB2-mediated signaling (17)(18)(19)(20). EGFR-or ERBB2-mediated AKT and MAPK signaling pathways are impaired in GAB1-deficient cells, and GAB1 has been identified as a direct substrate for EGFR and other receptor kinases (21,22). Our previous work demonstrated that ligand-activated EGFR has an increased preference for GAB1 versus receptor autophosphorylation sites by selectively decreasing the K m for the major GAB1 phosphorylation site (23).
The aim of our work is to understand the molecular mechanism for oncogenic signaling by ERBB2 lung cancer mutants and the role of GAB1 in this pathological process. We found that GAB1 phosphorylation is required for mutant ERBB2-induced activation of oncogenic signaling pathways, cell transformation, and tumorigenesis. To directly assess the activation mechanism, we performed a systematic enzymatic study using physiologically relevant peptide substrates corresponding to major phosphorylation sites derived from GAB1 and ERBB receptors. The results demonstrate that the oncogenic ERBB2 not only possesses dramatically increased catalytic activity relative to the WT kinase but has adopted a striking preference for GAB1 via increased binding affinity for phosphorylation sites in GAB1.
Retroviruses and Lentiviruses-To generate retroviruses encoding Myc-tagged WT or mutant ERBB2s, cDNAs were subcloned into RevTet-off vectors, and retroviruses were produced in PT67 cells. The pSLIK-Neo/TRE Pitt lentiviral vector plasmids encoding Myc-tagged ERBB2 or FLAG-tagged GAB1 and the pNL(CMV)/CMV/WPRE⌬U3 lentiviral vector plasmid encoding FLAG-tagged GAB1 were constructed as described previously (24,25). Mutations were introduced using the QuikChange kit (Stratagene). Identities of all expression constructs were confirmed by DNA sequencing. GAB1 and ERBB3 lentiviral shRNAs were purchased from Santa Cruz.
Transduction and Generation of Stable Cell Lines-BEAS2B cells were grown in DMEM containing 10% FBS. H1781 and T47D Tet-off cells were cultured as instructed by the providers. To generate T47D cells stably expressing ERBB2s, cells were infected with WT or mutant ERBB2 retroviruses, selected with 200 g/ml hygromycin, and expression was induced by removal of doxycycline for 18 h. To generate BEAS2B cells stably expressing ERBB2s, cells were transduced by WT or mutant ERBB2 lentiviruses and selected with 800 g/ml G418 for 2 weeks, and expression was induced with 0.1 g/ml doxycycline. Expression of WT or mutant GAB1s in BEAS2B or T47D cells was achieved by transduction with CMV-driven lentiviruses, and expression of GAB1s in H1781 was achieved by transduction with doxycycline-inducible lentiviruses followed by 800 g/ml G418 selection and 0.1 g/ml doxycycline induction. Stable GAB1 knockdown cell lines were generated by shRNA lentivirus transduction followed by selection in 4 g/ml puromycin.
Immunohistochemical Staining-Analyses were performed on formalin-fixed paraffin sections of lung tissues of agematched normal or transgenic mice that express ERBB2 YVMA (17). Slides were deparaffinized in xylene, rehydrated sequentially in ethanol, treated with target retrieval solution (Dako), quenched in hydrogen peroxide (3%), blocked in 10% normal murine serum, and incubated overnight at 4°C with primary antibody against pGAB1 Tyr-627 or normal rabbit IgG. Slides were developed using avidin-biotin peroxidase complex (Vector) and AEC substrate (Invitrogen).
Cell Proliferation and Viability Assay-H1781 cells were seeded in triplicate at a density of 2 ϫ 10 4 /well in 6-well plates in complete medium, and the number of cells was subsequently counted with Bio-Rad CT10 Automated Cell Counter. For T47D viability study, cells were plated in five replicates at a density of 8 ϫ 10 3 /well in 96-well plates in complete growth medium and changed to serum-free medium the next day, and survival cells were monitored with Cell Titer AQ kit (Promega).
Soft Agar Colony Formation Assay-0.5 ml of 0.5% agarose (Invitrogen) was solidified in the bottom of each well of six well plates. BEAS2B (5 ϫ 10 4 ) cells in growth medium were mixed with 0.3% agarose in four replicates and laid on the bottom agar. Colonies measuring Ն50 m were counted under a microscope with a grid in the eyepiece. Eight randomly selected fields were counted from each well after 14 days.
Mouse Tumorigenicity Assay-Cells at ϳ80% confluence were harvested by trypsinization and resuspended in serumfree medium, and 5 ϫ 10 6 cells in 0.1 ml were injected subcutaneously into 4-week-old female athymic nude mice (NCI, National Institutes of Health). Tumor formation was monitored twice a week, and the size of each tumor was measured with a digital caliper. The volume of the tumors was calculated by the formula: volume ϭ width 2 ϫ length/2.
Kinase Assays-Reactions and steady-state kinetic parameter measurements were performed as described previously (9,26). The kinetic parameters and standard errors were obtained using Enzyme Kinetics of Sigma Plot.
Statistical Analyses-Means, standard deviations, and p values from t test were obtained using Sigma Plot software.

Oncogenic ERBB2 Mutant-induced GAB1 Phosphorylation
Correlates with Activation of AKT and MAPK-Previous work has demonstrated that expression of ERBB2 YVMA in human BEAS2B bronchial epithelial cells potently activates AKT and MAPK signaling pathways and induces cellular transformation (11). We stably transduced BEAS2B cells with lentiviral vectors that express green fluorescent protein (GFP), myc-tagged WT, or mutant ERBB2s from a doxycyclineinducible promoter. In addition to ERBB2 YVMA and ERBB2 G776VC identified in cancer, we also included the activated ERBB2 G776S/G778D described in our previous work (9). The cell lysates were analyzed using specific antibodies against major phosphorylation sites in ERBB2 (Tyr-1221/ 1222) and GAB1 (Tyr-627). As shown in Fig. 1, A, left, and B, the expression levels of all myc-tagged ERBB2s were comparable, whereas the phosphorylation levels of ERBB2 YVMA , ERBB2 G776VC , and ERBB2 G776S/G778D were significantly higher than WT receptor. GAB1 phosphorylation was elevated in cells expressing ERBB2 YVMA or ERBB2 G776VC , accompanied by markedly reduced mobility of the GAB1 bands in SDS-PAGE. Interestingly, GAB1 phosphorylation in ERBB2 G776S/G778D expressing cells was significantly lower than in cells expressing the two cancer mutants. Correlated to the hyper-phosphorylation of GAB1, expression of ERBB2 YVMA and ERBB2 G776VC resulted in increased levels of activated AKT and MAPK. These results indicate that the oncogenic ERBB2 mutants are much more potent at inducing constitutive GAB1 phosphorylation and subsequent signaling than WT receptor and the activated ERBB2 G776S/G778D . FIGURE 1. Oncogenic ERBB2 mutants induce phosphorylation of GAB1 and activation of AKT and MAPK signaling pathways. A, BEAS2B and T47D cells stably expressing indicated proteins were generated as described under "Experimental Procedures." These cells and H1781 lung cancer cells were serumstarved overnight, and cell lysates were subjected to SDS-PAGE or immunoprecipitation with GAB1 antibody. Lysates and anti-GAB1 immunoprecipitates (GAB1 IP) were analyzed by Western blotting with indicated antibodies. B and C, relative phosphorylation levels of ERBB2, ERBB3, GAB1, AKT, and MAPK in BEAS2B and T47D cells were quantified using Kodak Image Station 4000R and normalized to the YVMA mutant. The data points represent the means Ϯ S.D. of triplicate experiments. D, relative SHP2 and PI3K protein levels co-immunoprecipitated with GAB1 in BEAS2B and T47D cells were quantified, and data analysis was performed as in B and C.
Interactions of phosphorylated GAB1 with PI3K and SHP2 are required for receptor kinase-mediated activation of AKT and MAPK, respectively (17,18,27). Phosphorylation of Tyr-627 and 659 in GAB1 is essential for SHP2 association and MAPK activation (27,28), whereas phosphorylation of Tyr-447, -472, and -589 is required for binding to the p85 subunit of PI3K and AKT signaling (17). Accordingly, we immunoprecipitated GAB1 from cell lysates and analyzed for the presence of PI3K and SHP2. As shown in Fig. 1, A, left, bottom, and D, expression of ERBB2 YVMA or ERBB2 G776VC in BEAS2B cells resulted in significantly increased phosphorylation of GAB1 and association of SHP2 and PI3K with GAB1, relative to WT ERBB2-expressing cells. In the ERBB2 G776S/G778D cells, the increase in association of GAB1 with either SHP2 or PI3K is modest. Immunoprecipitation of SHP2 in the reciprocal experiment confirmed these findings (data not shown).
Next, we characterized signaling in T47D breast carcinoma cells that stably express WT or ERBB2 YVMA (Fig. 1, A, middle, and C). Similar to that observed in BEAS-2B cells, phosphorylation levels of GAB1, AKT, and MAPK were much higher in T47D/ERBB2 YVMA cells relative to WT. Correlated with its phosphorylation, GAB1 association with either SHP2 or PI3K also dramatically increased in T47D/ERBB2 YVMA cells (Fig. 1, A, middle, bottom, and D). Furthermore, we detected high levels of GAB1, AKT, and MAPK phosphorylation as well as GAB1 interaction with SHP2 and PI3K in H1781 lung cancer cells, which harbor the homozygous oncogenic ERBB2 G776VC mutation (Fig. 1A, right). These results demonstrate that signaling by the oncogenic ERBB2 mutants results in constitutive hyperphosphorylation of GAB1, association of SHP2 and PI3K with GAB1, and concomitant activation of AKT and MAPK. It is noteworthy that the levels of ERBB3 protein and phosphorylation in these three cell lines were very different. In BEAS2B cells, ERBB3 protein level was very low, and phosphorylation was undetectable. T47D cells had a moderate level of ERBB3, and ERBB2 YVMA induced higher ERBB3 phosphorylation than WT kinase. Relative to BEAS-2B and T47D cells, H1781 cells had the highest level of ERBB3 protein and phosphorylation.
To extend these observations, we performed immunohistochemical staining for phosphorylated GAB1 Tyr-627 on serial sections of lung tissues from transgenic mice expressing ERBB2 YVMA . In contrast to histology of the normal lung tissues, intrabronchial carcinomas were clearly identified in the transgenic mice (Fig. 2, top). Positive staining for phospho-GAB1 Tyr-627 was observed in the carcinoma cells of the tumors, whereas the control lung tissues were negative (Fig. 2, middle). The specificity of the immunostaining was confirmed using a normal rabbit IgG (Fig. 2, bottom panels). These results demonstrate that GAB1 is highly phosphorylated in ERBB2 YVMAdriven lung carcinomas, coincident with the previously documented activation of AKT and MAPK in carcinomas of the transgenic mice (12).
To further examine the role of GAB1 in mutant ERBB2-mediated signaling, we generated stable GAB1 shRNA cell lines in the BEAS2B/YVMA, T47D/YVMA, and H1781 cell backgrounds. As shown in Fig. 4A, GAB1 shRNA specifically reduced endogenous GAB1 protein, whereas control shRNA had no effect, and ERBB2, ERBB3, AKT, and MAPK protein levels were not changed. Consistent with the results obtained from the GAB1 dominant negative mutants, GAB1 knockdown by shRNA significantly inhibited both AKT and MAPK signaling in BEAS2B and T47D cells expressing ERBB2 YVMA (Fig. 4, A  and B). However, in H1781 cells, GAB1 knockdown only suppressed MAPK activation without an effect on AKT signaling. Nevertheless, these data demonstrate that tyrosine phosphorylation of GAB1 plays an important role in mutant ERBB2-mediated AKT and/or MAPK signaling pathways.
ERBB3 Knockdown Reduces AKT Activation in H1781 Cells-The inability of either dominant negative GAB1 mutants lacking PI3K docking sites (GAB1-3F or -5F) or GAB1 knockdown to affect AKT activation in H1781 cells may be related to the high expression and phosphorylation level of ERBB3. The catalytically inactive ERBB3 receptor contains at least six canonical PI3K binding sites in its C terminus and thus may provide a redundant pathway for ERBB2 G6776VC -induced AKT stimulation (17,29). To test this hypothesis, we knocked down endogenous ERBB3 in H1781 using lentiviral shRNA in the absence or presence of GAB-5F expression. As shown in Fig. 5, ERBB3 protein was specifically reduced, and AKT activation was suppressed independent of the expression of GAB1-5F. The results demonstrate that rather than GAB1, ERBB3 plays a predominant role in AKT activation in H1781 cells.
GAB1 Is Critical for Mutant ERBB2-induced Cell Transformation, Proliferation, and Survival-Two hallmarks of the oncogenic transformation of cells are morphology change and acquisition of anchorage-independent growth. As shown in Fig.  6A, top, BEAS2B cells expressing ERBB2 YVMA or ERBB2 G776VC underwent striking morphological transformation. The cells expressing WT ERBB2 or GFP displayed a normal epithelial cell morphology, which is flat and non-refractile. In contrast, cells expressing either ERBB2 YVMA or ERBB2 G776VC had a typical sharply elongated, spindle-shaped, and highly refractile appearance of transformed cells. To determine whether GAB1 is involved in the mutant ERBB2-induced morphological change, we assessed the effects of expression of WT or mutant GAB1s on the cells. As shown in Fig. 6A, bottom, expression of WT GAB1 had no obvious effects on the transformed morphology of BEAS2B/YVMA cells. Conversely, expression of either GAB1-2F, -3F, or -5F blocked the ERBB2 YVMA -mediated morphology change.
Next, we determined the role of GAB1 in ERBB2 YVMA -induced BEAS2B cell transformation using an anchorage-inde-FIGURE 3. Expression of GAB1 mutants lacking SHP2 or PI3K docking sites suppresses mutant ERBB2-induced MAPK and AKT signaling. A-C, stable expression of FLAG-tagged WT or mutant GAB1s in BEAS2B/YVMA, T47D/ YVMA, or H1781 cells was achieved by transduction using lentiviruses as described under "Experimental Procedures." Cells were serum-starved overnight, and cell lysates were subjected to Western blotting using indicated antibodies. D, relative phosphorylation levels of MAPK and AKT are shown. Quantification and analysis were performed as Fig. 1. The data points were normalized to GFP control, and the p values from t test were calculated by comparison of WT or mutant GAB1s to the control (GFP). *, p Ͻ 0.01; **, p Ͻ 0.001, and ***, p Ͻ 0.0001.

FIGURE 4. GAB1 knockdown inhibits mutant ERBB2-mediated MAPK and/or AKT signaling.
A, BEAS2B/YVMA, T47D/YVMA, or H1781 cells were transduced with lentivirus encoding either control-or GAB1-targeting shRNA. Lysates were prepared from serum-starved cells and subjected to Western blotting using indicated antibodies. B, relative phosphorylation levels of MAPK and AKT are shown. Quantification analysis was performed as in Fig. 1. The data points were normalized to control shRNA. *, p Ͻ 0.01; **, p Ͻ 0.001.

FIGURE 5. ERBB3 mediates AKT activation in H1781 cells.
Stable cells expressing GAB1-5F from a doxycycline-inducible promoter were transduced with a lentivirus encoding either control or ERBB3 targeting shRNA. GAB1-5F expression was induced with 100 ng/ml doxycycline. Lysates were prepared from serum-starved cells and subjected to Western blotting using indicated antibodies. pendent colony formation assay. Two weeks after cells were seeded in soft agar, colonies larger than ϳ50 m were counted (Fig. 6B). Consistent with a previous report (11), cells expressing ERBB2 YVMA and ERBB2 G776VC formed many more colonies than WT ERBB2-or GFP-expressing cells. However, the colony-forming activity of BEAS2B/YVMA cells was suppressed dramatically by expression of either GAB1-2F, -3F, or -5F. Finally, we evaluated the role of GAB1 in ERBB2 YVMA -driven tumorigenicity in vivo. BEAS2B cells expressing WT or mutant ERBB2s or BEAS2B/YVMA cells also expressing either WT or mutant GAB1s were implanted in athymic nude mice. As depicted in Fig. 6C, tumors were observed in mice injected with BEAS2B/YVMA 12 days after inoculation and grew to large sizes in 3 weeks, but no tumor was observed in mice injected with cells expressing WT ERBB2 or GFP during the same time frame (data not shown). Tumor formation and growth was not significantly changed by WT GAB1 expression, whereas expression of GAB1-2F, -3F, or -5F markedly delayed tumor formation and significantly inhibited tumor growth (Fig. 6C).
We also investigated the role of GAB1 in proliferation of H1781 cells and survival of T47D/YVMA cells in serum-free medium. As shown in Fig. 6D, H1781 cells stably expressing GAB1-2F or -5F grew at considerably slower rates, in contrast to cells expressing GAB1-3F, which proliferated at a rate similar to WT GAB1 or GFP cells. Serum starvation resulted in a rapid decline in the viability of T47D cells expressing WT ERBB2 or empty vector, whereas T47D/YVMA cells survived in serumfree medium up to 7 days (Fig. 6E). Expression of GAB1-2F, -3F, or -5F all inhibited ERBB2 YVMA -induced T47D cell survival in the absence of serum (Fig. 6E), indicating that GAB1dependent signals are important for ERBB2 YVMA -promoted survival of T47D cells. Together, these results demonstrate that GAB1 phosphorylation is critical for oncogenic mutant ERBB2induced transformation of BEAS2B cells, proliferation of H1781 cells, and survival of T47D in serum-free medium.
ERBB2 YVMA Acquires a Substrate Preference for GAB1 Phosphorylation Sites-To elucidate the mechanism of mutant ERBB2-mediated GAB1 hyperphosphorylation, we performed steady-state kinetic studies, which provide measurements of Michaelis (K m ) and catalytic rate (k cat ) constants. The specificity constant, k cat /K m , serves as the best parameter for comparisons of catalytic efficiencies between enzymes. We determined the kinetic parameters for the WT and mutant kinases toward peptides containing sequences for the major phosphorylation sites of GAB1, ERBB2, and ERBB3. Fig. 7A contains representative Lineweaver-Burk plots of GAB1 Tyr-627 peptide phosphorylation catalyzed by WT and ERBB2 YVMA kinases, and the obtained kinetic parameters for all tested peptide substrates are summarized in Table 1. As illustrated in Fig. 7C, relative to WT kinase, the k cat values of ERBB2 YVMA for all tested peptides were increased ϳ12-fold. However, the mutation differentially decreased the K m values for the peptides (Fig. 7C, Table 1). The K m values of ERBB2 YVMA for ERBB peptides were modestly decreased by ϳ2-3-fold relative to WT, but strikingly, the K m for GAB1 Tyr-447, Tyr-627, and Tyr-659 peptides were more dramatically decreased (Ͼ10-fold). The decreased K m and concomitantly increased k cat yielded striking increases in catalytic efficiency of ERBB2 YVMA for GAB1 peptides as indicated by the specificity constants (Fig. 7C, bottom; Table 1). The k cat /K m values for all 3 GAB1 peptides were increased Ͼ150-fold relative to the WT kinase. ERBB2 YVMA had an extremely high k cat /K m value of 2438 min Ϫ1 mM Ϫ1 for GAB1 Tyr-627 peptide, which was ϳ29-fold higher than the best ERBB peptide substrate. In contrast to the findings with the peptides, K m values of ERBB2 YVMA for ATP were only modestly altered relative to WT (Ͻ2-fold) (Fig. 7B and Table 2). Taken together, these results demonstrate that the YVMA insertion not only increases the catalytic activity of the kinase but, more importantly, results in an alteration in substrate specificity and a preference for GAB1 phosphorylation sites.
To validate that ERBB2 YVMA has a remarkable preference for GAB1 phosphorylation sites within the GAB1 protein, we expressed and purified GST-tagged full-length and truncated GAB1 566 -694 . These fusion proteins were subjected to kinase  Table S1. D, H1781 cells expressing WT or mutant GAB1s were seeded in triplicate in 6-well plates in complete medium, and cells were counted at day 7. Results represent the mean Ϯ S.D. for triplicates. E, T47D cells expressing the indicated proteins were plated in five replicates in 96-well plates in complete growth medium and serum-starved starting the next day. Cell survival was monitored with Cell Titer AQ kit (Promega) at day 7, and results represent the mean Ϯ S.D. for five replicates. All p values were calculated with comparison to control (GFP). *p Ͻ 0.01; **, p Ͻ 0.001; ***, p Ͻ 0.0001. reactions catalyzed by WT, ERBB2 G776S/G778D , or ERBB2 YVMA kinases. We previously found that ERBB2 G776S/G778D has an ϳ8-fold increased k cat relative to WT kinase but an unaltered K m for GAB1 Tyr-627 peptide substrate (9). The concentrations of full-length and truncated GAB1 in the reactions were 1 and 2 M, respectively, which is much lower than the K m values for all tested GAB1 peptides. Under these conditions, the catalytic efficiency is a function of both turnover number and substrate binding affinity. As shown in Fig. 7, D and E, both GST-GAB1 and GST-GAB1 566 -694 were phosphorylated ϳ8and ϳ70-fold faster by ERBB2 G776S/G778D and ERBB2 YVMA , respectively, than by the WT kinase. These results are consistent with the fact that the measured K m values for the GAB1 peptides were much lower for ERBB2 YVMA (Fig. 7C and Table 1) than for the activated G776S/G778D mutant or for WT ERBB2 kinase (9). The differential impact of the YVMA and G776S/G778D mutations on the abilities of the kinases to phosphorylate GAB1 peptides and protein highlights the unique catalytic nature of the ERBB2 YVMA mutant found in human lung cancer.

DISCUSSION
The role of ERBB2 in cancer involves a complicated biological network in which adaptor/docking proteins, such as GAB1, occupy pivotal points. Previous studies have revealed that GAB1 orchestrates signaling from a variety of cell surface receptor kinases such as EGFR (21,30), ERBB2 (18), and c-MET (31,32). GAB1 is phosphorylated by the activated receptor kinases at multiple tyrosine sites and recruits downstream signaling proteins such as SHP2 and PI3K, which in turn lead to activation of MAPK and AKT, respectively. MAPK and AKT are considered as two major cancer-related signaling pathways. In the present work we found that, correlated to activation of MAPK and AKT signaling pathways, GAB1 is hyperphosphorylated in mutant ERBB2-expressing cells and in ERBB2 YVMAinduced lung tumors from an animal model (Figs. 1 and 2). Expression of GAB1 mutants lacking canonical tyrosine residues for SHP2 interaction (GAB1-2F and -5F) or endogenous GAB1 knockdown suppressed oncogenic ERBB2 mutant-induced MAPK activation in BEAS2B/YVMA, T47D/YVMA, FIGURE 7. ERBB2 YVMA has increased catalytic activity and substrate preference toward GAB1 tyrosine phosphorylation sites. A, shown are Lineweaver-Burk plots for GAB1 Tyr-627 peptide using 0.1 M purified WT or ERBB2 YVMA . Initial velocities were measured and analyzed using 100 M ATP, and data points represent the mean Ϯ S.E. (n ϭ 4) from a representative experiment. B, shown are Lineweaver-Burk plots for ATP using 1.2 mM GAB1 Tyr-627 peptide. Data points represent mean Ϯ S.E. (n ϭ 4). C, shown is a comparison of kinetic parameters of WT and ERBB2 YVMA for various peptides derived from major phosphorylation sites in GAB1, ERBB2, and ERBB3. D, purified GSTtagged GAB1 (1 M) or truncated GAB1 556 -694 (2 M) were phosphorylated by 150 nM WT, 150 nM ERBB2 G776S/G778D , or 50 nM ERBB2 YVMA in the presence of 100 M [ 32 P]ATP, and reactions were terminated at indicated times by the addition of 0.1 M EDTA and subjected to SDS-PAGE, Coomassie staining (GAB1 and GAB1 556 -694 ), and autoradiography (pGAB1 and pGAB1 556 -694 ). E, individual GAB1 protein bands were excised, and incorporated 32 P was quantified using a scintillation counter. Activity was calculated using the slope of time course and is presented as relative activity to WT kinase.  and H1781 cells (Figs. 3 and 4). Dominant negative GAB1 mutants lacking PI3K binding sites or GAB1 knockdown inhibited ERBB2 YVMA -mediated AKT activation in BEAS2B and T47D cells but had no obvious effect in H1781 cells (Figs. 3 and  4). However, knockdown of ERBB3 significantly reduces AKT activation in H1781 cells, indicating that ERBB3, rather than GAB1, plays a predominant role in AKT activation in H1781 cells (Fig. 5). Furthermore, expression of dominant negative GAB1s also inhibited ERBB2 YVMA -driven BEAS2B cell morphology change, colony formation in soft agar, and tumor formation in nude mice and ERBB2 YVMA -mediated survival of T47D cells in serum-free medium (Fig. 6). Taken together, our results provide compelling evidence that oncogenic ERBB2mediated hyperphosphorylation of GAB1 plays a major role in the transforming activity of these mutants. Despite the dramatic benefits from ERBB2-targeted therapies, the outcomes of these therapies are uniformly limited by the development of drug resistance (33,34). Two major possible mechanisms are acquired secondary mutation of the target kinase and induced amplification of other receptor kinases, such as c-MET (35) and IGF-1 receptor (36). GAB1 has been demonstrated to play essential roles in transformation elicited by these receptors and, more interestingly, to serve as a bridge mediating the cross-talk between ERBB2 and other receptorsignaling pathways. Therefore, GAB1-targeted anti-cancer therapies could be more efficacious by blocking oncogenic signaling mediated by individual receptor kinases and their crosstalk. Strategies to target GAB1 and other docking proteins have been proposed (15,37).
The catalytic activity of all protein kinases is regulated by two key coupled structural elements, namely the activation loop and ␣C helix (2). Compelling evidence has revealed that, in contrast with most receptor-tyrosine kinases, phosphorylation of the activation loop is not required for activation of ERBB kinases (9,38). Structural studies have provided evidence that EGFR (39) and ERBB4 (40) kinases are activated through an allosteric mechanism via formation of an asymmetric dimer. Upon dimer formation, the helix ␣C of one kinase domain is restrained to the catalytically competent conformation by the C-lobe of the other. The proper orientation of ␣C helix for kinase activation in EGFR and ERBB4 is highly controlled through hydrogen bonds and hydrophobic interactions of residues in the ␣C-␤4 loop with residues in the activation loop and the C-lobe (10). Conversely, the unique Gly-rich ␣C-␤4 region in ERBB2 fails to form these interactions and imparts greater flexibility to the enzyme active site. Compiled evidence has demonstrated that changes in flexibility or slight conformation adjustment within the enzyme active site can result in dramatic functional changes (41,42). Indeed, our previous work revealed that ERBB2 has much lower intrinsic catalytic activity relative to EGFR and ERBB4 kinases (9). The ␣C-␤4 loop mutations in human lung cancer might introduce interactions, which stabilize the ␣C-helix in an active orientation.
The catalytic specificity of protein kinases toward their physiological targets is believed to be critical for fidelity in signaling (43). A systematic enzyme kinetic analysis comparing ERBB2 YVMA to WT kinases with peptide substrates containing major physiological phosphorylation sites is a powerful method to determine whether the gain-of-function oncogenic mutations alter substrate specificity of the kinase and, therefore, provides important insights into the molecule basis of mutant kinase-induced signaling. An important finding in the work presented here is that the YVMA insertion resulted in significantly greater relative increases in specificity constants for GAB1 peptides compared with ERBB2 and ERBB3 phosphorylation site peptides. Therefore, ERBB2 YVMA adopts a strikingly increased preference for GAB1 phosphorylation sites. The alteration of the specificity is predominantly attributed to the increased substrate binding affinity as reflected by the decreased K m values. It should be noted that ERBB2 YVMA may also have an acquired preference for other physiological substrates that we did not test in our current steady-state kinetic studies using peptides. The concept of altered substrate specificity is illustrated in Fig. 7C in which the kinetic parameters, K m , k cat , and k cat /K m of WT and mutant ERBB2 kinases for a number of physiologically relevant peptides were compared. The concomitant effects of the decreased K m and the increased k cat yield ϳ150-fold increases in the specificity constants (k cat / K m ) for several GAB1 peptides. These kinetic findings with peptide substrates were verified by our observation that ERBB2 YVMA phosphorylated purified GAB1 proteins at an ϳ70-fold faster rate relative to WT ERBB2 (Fig. 7, D and E). A decrease in the K m value for specific phosphorylation sites may play a more important role in pathological signaling in cancer cells in which the concentrations of these substrate targets are much lower than the K m . In our previous work we found that G776S/G778D mutations in the ␣C-␤4 loop also dramatically activate the ERBB2 kinase. However, even though the G776S/ G778D mutation is localized to the same region, the activation mechanism is very different from that of ERBB2 YVMA . In contrast to the YVMA insertion, G776S/G778D does not significantly alter the K m values for the peptide substrates. Even though ERBB2 G776S/G778D has a k cat value similar to ERBB2 YVMA , expression of this mutant led to only slight increases in GAB1 phosphorylation and MAPK and AKT activation in BEAS2B cells (Fig. 1). Thus, mutational activation per se, as in the case of ERBB2 G776S/G778D , need not result in a change in substrate specificity of the kinase. Collectively, our results demonstrate that (i) the ␣C-␤4 loop plays a vital role of in the regulation of ERBB2 kinase activity and alterations in this region can result in different activated states of the kinase, and (ii) the acquisition of the substrate preference for GAB1 via decreased K m values is one of the primary mechanisms for transforming activity of ERBB2 mutations in lung cancer.
Another noteworthy finding from our enzyme kinetic analyses is that the K m value of ERBB2 YVMA for ATP is not significantly altered compared with WT ERBB2 (Fig. 7B and Table 2). Furthermore, GAB1 phosphorylation by ERBB2 YVMA or by ligand-activated WT ERBB2 in cells was equally responsive to lapatinib, an ATP competitive inhibitor for the ERBB2 kinase (supplemental Fig. S1). These results are consistent with earlier findings (11) demonstrating that cellular signaling by WT or ERBB2 YVMA are similarly sensitive to lapatinib. A number of activating EGFR mutations have also been identified in nonsmall cell lung cancers (44 -46), and the most frequent mutants identified in lung cancer are an ATP binding loop deletion (Del 746 -750) and an activation loop point mutation (L858R). A comparison of oncogenic mutation in ERBB2 to EGFR and other protein kinases is shown in supplemental Fig. S2. The EGFR mutations lead to constitutive kinase activation and, importantly, confer exquisite sensitivity to the small molecule kinase inhibitors, gefitinib and erlotinib (44 -47). These activated EGFR mutants have been found to have compromised affinity for ATP relative to the WT EGFR kinase (48,49). Of interest, EGFR ␣C-␤4 loop insertions, analogous to ERBB2 YVMA , were also identified in non-small cell lung cancers and do not confer sensitivity to either gefitinib or erlotinib (50). These results suggest that, in contrast to what was postulated previously, the insertions in the ␣C-␤4 loop region in both EGFR and ERBB2 kinases do not alter the ATP binding cleft dramatically (51). The "Achilles heel" for the ␣C-␤4 loop insertion mutants of ERBB receptor kinases may not exist in the ATP binding region. Therefore, treatment of lung cancers harboring ERBB2 mutations may benefit from alternative inhibition strategies aimed at targeting other important structural elements such as the peptide binding site.