ErbB Receptor-induced Activation of Stat Transcription Factors Is Mediated by Src Tyrosine Kinases*

Epidermal growth factor (EGF) binding to its receptor, ErbB1, triggers various signal transduction pathways, one of which leads to the activation of signal transducer and activator of transcription (Stat) factors. The mechanism underlying ErbB1-induced Stat activation and whether Stats are downstream targets of other ErbB receptors have not been explored. In this report we show that ErbB2, ErbB3, and ErbB4 do not potentiate Stat5 phosphorylation by EGF. However, neu differentiation factor-induced heterodimers of ErbB2 and ErbB4 activated Stat5. In A431 cells, Stat1, Stat3, and Stat5, were constitutively complexed with ErbB1 and rapidly phosphorylated on tyrosine in response to EGF. Neither mutation of the conserved tyrosine residue (Tyr694) nor inactivation of the Stat5a SH2 domain disrupted this association. However, an intact SH2 domain was necessary for EGF-induced Stat5a phosphorylation. In contrast to prolactin, which induced only Tyr694 phosphorylation of Stat5a, EGF promoted phosphorylation on Tyr694 and additional tyrosine residue(s). Janus kinases (Jaks) were also constitutively associated with ErbB receptors and were phosphorylated in response to EGF-related ligands. However, we provide evidence that EGF- and neu differentiation factor-induced Stat activation are dependent on Src but not Jak kinases. Upon EGF stimulation, c-Src was rapidly recruited to Stat/ErbB receptor complexes. Pharmacological Src kinase inhibitors and a dominant negative c-Src ablated both Stat and Jak tyrosine phosphorylation. However, dominant negative Jaks did not affect EGF-induced Stat phosphorylation. Taken together, the experiments establish two independent roles for Src kinases: (i) key molecules in ErbB receptor-mediated Stat signaling and (ii) potential upstream regulators of Jak kinases.

for phosphotyrosine binding, cytoplasmic signaling molecules that activate numerous intracellular signaling pathways. In some cases, for example the mitogen-activated protein (MAP) kinase pathway, the important intermediates in the signaling cascade have been well described (3). Receptor tyrosine kinasemediated activation of signal transducer and activators of transcription (Stat) proteins, however, is less well understood.
Stats were initially identified as a family of transcription factors activated by engagement of interferon (IFN) and cytokine receptors, which lack intrinsic kinase activity and depend upon the receptor-associated Janus kinases (Jaks) to transduce signals. Cytokine receptor oligomerization brings the Jaks into juxtaposition, leading to their cross-phosphorylation and activation. Jaks in turn phosphorylate receptors on tyrosine residues, thereby providing docking sites for downstream signaling proteins. The Stats that are recruited to the Jak-receptor complex via their Src homology 2 (SH2) domains are phosphorylated on a conserved tyrosine residue in the C terminus. This phosphorylation is sufficient for Stat dimerization, nuclear translocation, and binding to specific DNA sequences, affecting the expression of multiple target genes (reviewed in Refs. 4 and 5).
Ligand-activated platelet-derived growth factor and EGF receptors stimulate Stat phosphorylation and DNA binding activity (reviewed in Ref. 6). The intrinsic kinase activity of these receptors, but not that of Jaks, appears to be required for Stat activation (7,8). Immune complex-purified ErbB1 was shown to phosphorylate Stat1 on the conserved tyrosine residue in the C terminus (9), and the baculovirus-expressed kinase domain of ErbB1 phosphorylated Stat3 in vitro (10). Whether or not additional proteins are involved in receptor tyrosine kinasemediated Stat activation in vivo has not been resolved. In addition, the role that other ErbB family members play in Stat activation has not been analyzed.
ErbB2, which has no direct ligand but heterodimerizes with all other family members upon ligand binding (11,12), as well as ErbB3 and ErbB4, which both bind neu differentiation factor (NDF), were tested for their ability to activate Stats. We used NIH3T3 sublines expressing pairs of ErbB receptors (13) to examine dimerization-dependent differences in Stat activation. ErbB2, which is known to potentiate EGF-and NDFinduced MAP and S6 kinase activation (14), did not affect EGF-induced Stat activation, suggesting that ErbB1 expression is sufficient for this activity. Interestingly, NDF treatment of cells coexpressing ErbB2 and ErbB4, but not ErbB2 and ErbB3, led to Stat activation. This is a novel signaling property specific for the heterodimer because homodimers of either receptor were inefficient at activating Stat5.
In addition to receptor tyrosine kinases and Jaks, v-Src and v-Abl kinases have been implicated in Stat signaling (15)(16)(17). Transformation of cells by these oncoproteins induces constitutive activation of Stat transcription factors. Because the c-Src tyrosine kinase lies downstream of ErbB receptors (18 -22), we examined the role of Src kinases in ErbB receptor-mediated Stat activation. Using pharmacological Src kinase inhibitors and dominant negative (DN) c-Src and Jak expression vectors, we provide evidence that Src but not Jak kinase activity is required for ErbB receptor-mediated Stat activation. Thus, Stat activation by ErbB receptors proceeds via a mechanism distinct from the classical Jak-dependent pathway described for IFN and cytokine receptor signaling.
Cell Lines and Cell Culture-NIH3T3 derivative cell lines ectopically expressing human ErbB receptors (13) and the vulval carcinoma cell line A431 were maintained in Dulbecco's modified Eagle's medium containing 10% fetal calf serum. NE1/2/3 and NE1/2/4 cells were generated by stable introduction of pBabe-based retroviruses (29) encoding human ErbB2 into NE1/3 and NE1/4 cells, respectively. Prior to growth factor stimulation, cells were starved for 18 h in serum-free medium (Dulbecco's modified Eagle's medium containing 1 mg/ml fetuin (Sigma) and 10 g/ml transferrin (Sigma)). HC11 mammary epithelial cells were grown to confluency and maintained in RPMI 1640 supplemented with 10% fetal calf serum, 10 ng/ml EGF, and 5 g/ml insulin. The cells were starved for 12 h in serum-free medium prior to induction for 30 min with the lactogenic hormones (10 Ϫ6 M dexamethasone, 5 g/ml insulin, and 5 g/ml ovine prolactin) (30).
Immunoprecipitations and Western Blotting-Whole cell extracts (WCE) were obtained by solubilizing cells in Nonidet P-40 extraction buffer (50 mM Tris, pH 7.5, 5 mM EGTA, 120 mM NaCl, 1% Nonidet P-40, 2 mM sodium orthovanadate, 50 mM sodium fluoride, 20 mM ␤-glycerophosphate, 15 mM sodium pyrophosphate, 10 g/ml leupeptin, 10 g/ml aprotinin, 0.5 mM phenylmethylsulfonyl fluoride) for 10 min on ice. The lysates were clarified by centrifugation at 16,000 ϫ g for 10 min. For immunoprecipitations, equal amounts of protein were incubated with specific antibodies for 2 h on ice. Immune complexes were collected with protein A-or protein G-Sepharose (Sigma), washed three times with lysis buffer, and washed once with TNE buffer (50 mM Tris, pH 7.5, 140 mM NaCl, 5 mM EDTA). Precipitated proteins were released by boiling in sample buffer and were subjected to sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE). The proteins were blotted onto polyvinylidene difluoride membranes (Roche Molecular Biochemicals). After blocking with 20% horse serum (Life Technologies, Inc.) in TTBS (50 mM Tris, pH 7.5, 150 mM NaCl, 0.05% Tween 20), filters were probed with specific antibodies. Proteins were visualized with peroxidase-coupled secondary antibody using the ECL detection system (Amersham Pharmacia Biotech). Stripping of membranes was performed in SDS buffer (62.5 mM Tris, pH 6.8, 2% SDS, 100 mM ␤-mercaptoethanol) for 30 min at 60°C; membranes were then washed in TTBS and reprobed with the indicated antibodies.
Src Kinase Assays-Cells were lysed in Nonidet P-40 extraction buffer as described under "Immunoprecipitations and Western Blotting." After preclearing of lysates with protein G-Sepharose (Sigma), c-Src was immunoprecipitated. Immunoprecipitates were washed twice with extraction buffer and twice with kinase buffer (50 mM Tris, pH 7.5, 5 mM MgCl 2 ). The immune complexes were incubated with 5 g of acid-denatured enolase (Sigma) for 10 min at 30°C in 30 l of kinase buffer containing 5 M ATP and 5 Ci [␥-32 P]ATP. The reaction was stopped with sample buffer; proteins were subjected to SDS-PAGE (10% gel) and blotted onto polyvinylidene difluoride membranes. The membrane was analyzed using a PhosphorImager (Molecular Dynamics) and ImageQuant software.
Plasmids and Transient Transfections-An HA-tagged wild type (WT) Stat5a cDNA was cloned into the mammalian expression vector pcDNA3.1. Mutations (Y694F and R618L) were introduced by sitedirected mutagenesis. The expression plasmid encoding WT c-Src was provided by G. Superti-Furga (EMBL, Heidelberg); the plasmid encoding DN c-Src was provided by S. Courtneidge (Sugen, Inc.). DN Src has a K298M substitution in the ATP binding site and a Y530F substitution in the regulatory tyrosine, thereby rendering the protein in an open conformation that enables binding of substrates (32). DN Jak1 and Jak2 expression vectors were provided by O. Silvennoinen (University of Tampere, Finland). The expression vector encoding the prolactin receptor was provided by B. Groner (Georg Speyer Haus, Frankfurt). Plasmids were introduced into cells using Fugene transfection reagent (Roche Molecular Biochemicals) according to the manufacturer's protocol. Eight hours posttransfection, cells were starved for 16 h in serumfree medium prior to stimulation with growth factors.

ErbB Receptor-mediated Activation of the Stat Signaling
Pathway-It has previously been shown that signaling through ErbB1 leads to phosphorylation and activation of Stat1, Stat3, and Stat5 (33)(34)(35)(36). In this study we investigated whether Stat proteins are downstream targets of other ErbB receptors. To examine the contribution of ErbB2, ErbB3, and ErbB4 to EGFinduced Stat activation and the ability of the ErbB3/ErbB4 ligand, NDF, to activate these proteins, we used NIH3T3 sublines expressing defined combinations of ErbB receptors as a model cellular system (13). NIH3T3 cells lack ErbB1, ErbB3, and ErbB4 and have very low levels of endogenous ErbB2. Ectopic expression of human ErbB1 and ErbB2 in NIH3T3 cells gave rise to the NE1 and NE2 clones, respectively. NE1-derived cell lines coexpressing ErbB2, ErbB3, or ErbB4 were designated NE1/2, NE1/3, and NE1/4, respectively. Likewise, NE2derived cell lines coexpressing ErbB3 or ErbB4 were designated NE2/3 and NE2/4, respectively (13). In NE1 cells, EGF at 1 nM was efficient at activating the Stat5 isoforms a and b as measured by their tyrosine phosphorylation (not shown). In addition to homodimers of ErbB1, EGF can induce heterodimers of ErbB1 with ErbB2, ErbB3, or ErbB4 (11). However, these heterodimers appeared not to contribute to Stat5b activation because phosphorylation of Stat5b in the NE1 derivatives was unaffected by coexpression of ErbB2, ErbB3, or ErbB4 (Fig. 1A). Because ErbB2 enhances EGF-mediated activation of ErbB3 and ErbB4 (11) we introduced ErbB2 into NE1/3 and NE1/4 cells, thereby generating the NE1/2/3 and NE1/2/4 cell lines, respectively. Expression of ErbB2, however, failed to potentiate EGF-induced activation of Stat5b (Fig. 1A). These results suggest that ErbB1 expression is sufficient for maximal EGF-induced Stat5 phosphorylation.
NDF, which binds and activates ErbB2/ErbB3 heterodimers in NE2/3 cells, did not promote Stat5b phosphorylation (Fig.  1B). In NE2/4 cells, however, in which NDF activates ErbB4 homodimers and ErbB2/4 heterodimers, a strong increase of Stat5b phosphotyrosine was observed (Fig. 1B). In NDFtreated NE1/4 cells, in which mainly ErbB4 homodimers are activated (13), there was basically no increase in Stat5b phosphorylation (Fig. 1B). The mAb FRP5, which induces homodimers of ErbB2 and activates MAP and S6 kinases (37), also did not promote phosphorylation of Stat5b (not shown). Taken together, the results suggest that coexpression of ErbB2 and ErbB4 is required for Stat5b activation in response to NDF. The very weak Stat5b phosphorylation seen in NE1/4 cells is most likely because of heterodimers of ErbB4 with the low levels of endogenous ErbB2.
The ErbB1-overexpressing A431 carcinoma cell line also expresses ErbB2 and ErbB3, but lacks ErbB4. Treatment of these cells with the ErbB1 agonists EGF, heparin-binding EGF, and betacellulin led to Stat5b phosphorylation (Fig. 1C). NDF treatment, however, failed to induce Stat5b tyrosine phosphorylation, which is in accordance with the results obtained with the NIH derivatives.
EGF-induced Activation of Stat1, Stat3, and Stat5 and Their Constitutive Association with ErbB1-A431 cells express Stat1, Stat3, and Stat5b and low levels of Stat5a. In these cells, EGF treatment leads to activation of Stat1 and Stat3 (33,36) and of both Stat5 proteins (Fig. 2). We analyzed the kinetics of EGFinduced Stat activation in these cells by treating them with growth factor for various times ranging from 5 min to 2 h. EGF promoted a rapid increase in the tyrosine phosphorylation of all Stat proteins (Fig. 2). Activation of Stat3 was the most transient (Fig. 2B), and Stat5b phosphorylation was the strongest and most sustained (Fig. 2D).
In cytokine and IFN signaling, Stat transcription factors are recruited to ligand-activated receptors. We, therefore, asked whether there was a physical interaction of Stat proteins with ErbB1. Probing of ErbB1 immunoprecipitates from A431 cells with Stat-specific antibodies demonstrated that there was stable EGF-independent association of Stat1 (mainly the p91 isoform) and Stat3 with the receptor (Fig. 3, A and B). Detection of endogenous Stat5 in immunoprecipitates of ErbB1 proved to be difficult. However, following transient expression of HAtagged Stat5a in A431 cells, ErbB1 was readily detected in FIG. 3. Stat1, Stat3, and Stat5 are associated with ErbB1. A431 cells were serum-starved for 18 h and, prior to lysis, were stimulated with 1 nM EGF for the indicated times. C, HA-tagged Stat5a was transiently introduced into A431 cells 8 h prior to serum starvation and subsequent growth factor treatment. A and B, ErbB1 was immunoprecipitated with mAbs EGFR1 and 528. C, Stat5a was immunoprecipitated with HA-specific mAb. Immune complexes were resolved by SDS-PAGE and blotted. A and B, the lower parts of the membranes were probed with Stat1-and Stat3-specific antibodies, respectively, and the top parts was probed with ErbB1-specific antibody 1005. C, The top part of the membrane was probed with ErbB1-specific antibody 1005, and the lower part was probed with HA-specific mAb. Lane on the far left, 50 g of WCE were loaded as a control. IP, immunoprecipitation; WB, Western blot. HA-Stat5a immunoprecipitates (Fig. 3C). Reprobing of the membrane using HA-specific mAb showed that equal amounts of HA-Stat5a were expressed and immunoprecipitated (Fig.  3C). We confirmed by probing with Stat5a-specific serum that the top band comigrates with Stat5a (not shown); the lower band is nonspecific. In summary, all Stat proteins were found complexed with ErbB1 even in the absence of growth factor. Although ErbB1 from nonstimulated A431 cells contains high basal phosphotyrosine (not shown), suggesting some degree of activation, the phosphorylation of associated Stats was strictly ligand-dependent (Fig. 2).
EGF Promotes Phosphorylation of Stat5a on Additional Tyrosine(s)-To test which regions of Stat proteins are essential for their interaction with and activation by ErbB receptors, we constructed HA-tagged mutants of Stat5a with (i) a Y694F substitution in the C-terminal consensus tyrosine residue (Tyr mutant) and (ii) a R618L substitution in the SH2 domain, which prevents Stat interactions with phosphorylated tyrosine residues (SH2 mutant). HA-tagged WT Stat5a and the two Stat5a mutant proteins were transiently expressed in NE1 cells and then immunoprecipitated from control and EGF-stimulated cultures using HA-specific mAb. Probing of Western blots with ErbB1-specific antibody showed that neither mutation disrupted Stat5a association with ErbB1 (Fig. 4A, top  panel). However, inactivation of the SH2 domain prevented EGF-induced tyrosine phosphorylation of Stat5a (Fig. 4A, middle panel). Surprisingly, the Y694F mutant still displayed tyrosine phosphorylation in response to EGF (Fig. 4A, middle  panel). To test whether this was specific to EGF, we transiently expressed the Stat5a variants together with the prolactin receptor in NE1 cells and analyzed prolactin-induced Stat5a tyrosine phosphorylation. Prolactin is known to promote Stat5a activation via Tyr 694 phosphorylation (38). As seen with EGF, prolactin failed to induce phosphorylation of the SH2 domain mutant (Fig. 4B). However, in striking contrast to EGF, the mutant lacking Tyr 694 was not phosphorylated in response to prolactin (Fig. 4B). Apparently, EGF induces phosphorylation of Tyr 694 and additional tyrosine residues(s). The signal is not because of dimerization and coimmunoprecipitation of endogenous Stat5 molecules because EGF also induced tyrosine phosphorylation of the Y694F mutant in COS7 cells, which do not express Stat5 proteins (not shown). The lower panels in Fig. 4, A and B show that equal amounts of Stat5a were expressed and immunoprecipitated from all the cell lysates.
Next we tested the ability of the different Stat mutants to bind DNA by performing an EMSA with the 32 P-labeled Stat5 binding site from the ␤-casein promoter (31). Nuclear extracts from EGF-treated NE1 cells transiently expressing WT Stat5a demonstrated increased DNA binding activity compared with extracts from cells transfected with empty vector (Fig. 4C, lane  2 versus 11). This additional binding activity could be supershifted by HA-specific mAb (Fig. 4C, lane 3). The remaining endogenous Stat binding activity could be supershifted with Stat5-specific antiserum (not shown). The SH2 domain mutant had no detectable DNA binding activity (Fig. 4C, lanes 7-9). Although still phosphorylated by EGF, the Tyr 694 mutant also failed to bind DNA (Fig. 4C, lanes 4 -6). Analysis of Stat5a present in cytoplasmic and nuclear fractions showed that equal FIG. 4. Phosphorylation and activation of Stat5a wild type and mutants by EGF and prolactin. A and B, plasmids encoding HA-tagged WT Stat5a, Y694F Stat5a (Y mut.), and R618L Stat5a (SH2 mut.) were transfected into NE1 cells. B, a plasmid encoding the prolactin receptor was cotransfected. 8 h posttransfection cells were serum-starved for 18 h and, prior to lysis, were stimulated with 1 nM EGF (A) for 10 min or 5 g/ml prolactin (PRL) (B) for 15 min. Stat5a was immunoprecipitated with HA-specific mAb; immune complexes were resolved by SDS-PAGE and blotted. The membrane was probed with phosphotyrosine-specific mAb and, after stripping, reprobed with Stat5a-specific antibody. A, the top part of the membrane was probed with ErbB1-specific antibody 1005. C, plasmids encoding HA-tagged WT, Y694F (Y mut.), and R618L (SH2 mut.) or an empty vector were transfected into NE1 cells. 8 h posttransfection cells were serum-starved for 18 h and were stimulated with 5 nM EGF for 15 min. Nuclear and cytoplasmic extracts were prepared, and 10 g of nuclear protein were subjected to an EMSA using a radiolabeled Stat5 binding site as a probe. Where indicated, samples were preincubated with 2 g of HA-specific mAb. D, 50 g of cytoplasmic (C) and nuclear (N) protein from the experiment described in C were subjected to SDS-PAGE, blotted, and probed with Stat5a-specific antiserum. PY, phosphotyrosine; WB, Western blot. amounts of Stat5a wild type and mutant proteins were expressed (Fig. 4D). Both mutants, however, failed to translocate to the nucleus upon EGF stimulation (Fig. 4D, lanes 8 and 12  versus lane 4). The low amount of Stat5a detected in these lanes is most likely because of translocation of endogenous Stat5a in response to EGF.
The results indicate that the stable association of Stat5 with ErbB1 is not mediated via interaction of its SH2 domain with a basally phosphorylated receptor. However, the SH2 domain is apparently necessary for ensuing EGF-induced tyrosine phosphorylation. The fact that EGF, as opposed to prolactin, promotes Stat5a phosphorylation on additional tyrosine(s) makes it tempting to speculate that the ErbB receptor-mediated activation of Stat transcription factors occurs by an alternate mode of signaling.
c-Src Is Activated by EGF and Recruited to the Stat-ErbB1 Complex-The mechanism underlying ErbB receptor-mediated activation of Stat transcription factors has not been resolved. In vitro, ErbB1 is able to phosphorylate Stat1 and Stat3 (9, 10). However, it is not known whether additional kinases are involved in vivo. In v-Src transformed cells, Stat3 is constitutively activated and complexed with V-Src (15,17), suggesting that this kinase might be involved in Stat3 phosphorylation. Because c-Src lies on a pathway downstream of ErbB1 and ErbB2 (18 -20), we examined its role in ErbB receptor-induced phosphorylation of Stats. c-Src was immunoprecipitated from EGF-treated A431 cells, and its activity was measured in an in vitro kinase assay using acid-denatured enolase as a substrate. After 30 min of EGF treatment, c-Src activity increased, reaching its maximum (3.2-fold) at 60 min (Fig. 5A).
Next we examined whether c-Src was associated with ErbB1 and Stat proteins. In A431 cells, ErbB1 and both p84 and p91 Stat1 could be detected in immunoprecipitates of c-Src (Fig.  5B). A low degree of association was observed in nonstimulated cells. However, increased amounts of c-Src were recruited to ErbB1 and Stat1 within 5 min of EGF treatment, and the highest level of complex formation was seen after 60 min of EGF treatment (Fig. 5B). A ligand-induced increase in the complex of c-Src with ErbB1 and ErbB2 was also observed in EGF-treated NE1 cells and NDF-treated NE2/4 cells, respectively (not shown). Following transient coexpression of HAtagged Stat5a and c-Src in A431 cells, association of these two proteins was readily detected (not shown). The fact that the c-Src-ErbB1 interaction peaked after 60 min of EGF stimulation, the time at which c-Src kinase activity reached its maximum, supports the hypothesis that c-Src creates its own binding site on ErbB1 (39).
Src Kinase Activity Is Required for ErbB Receptor-mediated Stat Activation-To explore the role of Src kinases in ErbB receptor-mediated activation of Stat proteins, we examined the effect of two nonrelated pharmacological Src kinase inhibitors, PP1 (27) and CGP77675 (28), on EGF-and NDF-induced Stat phosphorylation. Prior to growth factor treatment, NE1 cells and NE2/4 cells were treated for 90 min with 10 M PP1 or 1 M CGP77675. Both PP1 and CGP77675 completely blocked Stat5b tyrosine phosphorylation in EGF-treated NE1 cells and NDF-treated NE2/4 cells (Fig. 6A). To determine whether Src kinases were also involved in Stat1 and Stat3 activation, we investigated the effects of the inhibitors in A431 cells. Both inhibitors abolished EGF-induced tyrosine phosphorylation of Stat1, Stat3, and Stat5b (Fig. 6B). Despite this block in Stat activation, the inhibitors had little or no effect on ligand-induced tyrosine phosphorylation of ErbB1, ErbB2, and ErbB4 (data not shown). Furthermore, in NE1 cells, EGF-induced activation of extracellular signal-regulated kinase 1/2 and p38 MAP kinases was not altered in the presence of PP1 or CGP77675, indicating that receptor signaling was not impaired (data not shown).
In a second approach, we examined the effect of a DN c-Src protein on Stat5a phosphorylation. HA-tagged Stat5a was coexpressed with either the wild type or the DN mutant of c-Src in NE1 cells. Whereas expression of wild type c-Src resulted in constitutive phosphorylation of Stat5a, the level of which could not be increased any further by EGF (Fig. 6C, lanes 3 and 4), expression of DN c-Src inhibited both basal and EGF-induced tyrosine phosphorylation of Stat5a (Fig. 6C, lanes 5 and 6).
Phosphorylation on the conserved C-terminal tyrosine residue in Stats is essential for dimerization and specific DNA binding (4,5). Because Src inhibitors prevented ligand-induced Stat phosphorylation, we also expected them to affect Stat DNA binding. To test this, total cellular extracts from A431 cells treated with 5 nM EGF, with or without preincubation with Src inhibitors, were prepared, and an EMSA analysis was performed. Two bands were observed after EGF treatment (Fig. 6D). Supershift analyses with specific antibodies revealed that the more rapidly migrating band contained Stat1 and the slower band contained Stat5 (not shown). Extracts from inhibitor-treated cells displayed a marked reduction in DNA binding activity (Fig. 6D). Tyrosine phosphorylation of Stat1 and Stat5 was completely abolished by both Src inhibitors (Fig. 6B), yet DNA binding activity was not fully inhibited by the PP1 inhibitor. This is most likely because of the higher amounts of EGF used to visualize DNA binding (5 nM versus 1 nM for the experiment in Fig. 6B), whereas the inhibitor concentration was not increased. Thus, inhibition of Src kinase activity blocks both tyrosine phosphorylation and DNA binding of Stat proteins. Taken together, these results support the idea that functional Src kinases are required for ErbB receptor-mediated Stat signaling.

FIG. 5. EGF-induced c-Src kinase activity and c-Src association with ErbB1 and Stat1
. A431 cells were serum-starved for 18 h and, prior to lysis, were stimulated with 1 nM EGF for the indicated times. A, c-Src was immunoprecipitated from 200 g of WCE with mAb Ab-1, and immune complexes were subjected to an in vitro kinase assay using acid-denatured enolase as a substrate. Following SDS-PAGE and blotting, c-Src activity was quantified using a PhosphorImager. The values represent the mean of duplicate samples. B, c-Src was immunoprecipitated with mAb Ab-1; immune complexes were resolved by SDS-PAGE and blotted. The top part of the membrane was probed with ErbB1-specific antibody 1005, and the lower part was probed with Stat1-specific antibody and, after stripping, reprobed with Src-specific mAb. B, 50 g of WCE were loaded as a control. WB, Western blot; IP, immunoprecipitation. (4,5). However, the role of Jak kinases in EGF-mediated Stat activation is not well understood. In an attempt to establish Jak function in ErbB receptor signaling we analyzed ligandinduced phosphorylation of these kinases in NIH derivatives and in A431 cells.

Jak1 and Jak2 Are Phosphorylated in Response to EGF and NDF and Are Complexed with ErbB Receptors-IFNs and cytokines activate Stat transcription factors via Jak kinases
Following EGF and NDF treatment of NE1 and NE2/4 cells, respectively, an increase in Jak2 phosphorylation was observed (Fig. 7A, top panel). In A431 cells, the ErbB1 agonists EGF, heparin-binding EGF, and betacellulin but not NDF promoted Jak2 phosphorylation (Fig. 7B, top panel). This pattern of Jak2 phosphorylation correlates with Stat5 tyrosine phosphorylation in response to these ligands (see Fig. 1C). Interestingly, a ligand-induced tyrosine-phosphorylated band of approximately 180 kDa, the size of ErbB receptors, coimmunoprecipitated with Jak2 (Fig. 7, A and B, top panels, see arrowheads). Reprobing of the membranes revealed that in extracts from A431 cells, ErbB1 was present in Jak2 immunoprecipitates (Fig. 7B). The sensitivity, however, was too low to allow detection of coimmunoprecipitating ErbB1 in extracts from NE1 cells. Similarly, ErbB2 could be detected in Jak2 immunoprecipitates from A431 and NE2/4 cells (Fig. 7, A and B, bottom panels). Both receptors were associated with Jak2 even in nonstimulated cells.
Next we examined the kinetics of Jak2 phosphorylation and its association with ErbB1. Treatment of A431 cells with 1 nM EGF led to a strong increase in Jak2 phosphorylation, which was rapid and sustained for 30 min (Fig. 8A). After EGF treatment, a slight increase in the complex of the two proteins was noted, which reached its maximum after 30 min. This was also observed when probing for associated Jak2 in ErbB1 immunoprecipitates (Fig. 8B).
Higher concentrations of EGF were necessary to visualize phosphorylation of Jak1 in A431 cells. At a concentration of 16 nM (100 ng/ml), Jak1 phosphorylation was rapid and was still elevated even after 2 h of growth factor treatment (Fig. 8C). The tyrosine-phosphorylated band of approximately 180 kDa at the top of the membrane is most likely ErbB1. Several additional EGF-responsive, tyrosine-phosphorylated proteins coimmunoprecipitated with Jak1 (see arrowheads), demonstrating that many proteins physically interact with Jak1.
Jaks Are Downstream of Src but Are Not Required for ErbB Receptor-induced Stat Activation-Our results provide evidence that Src kinases are required in ErbB receptor-mediated Stat signaling. We have also shown that Jak kinases are constitutively associated with ErbB receptors and become phosphorylated in response to ligand binding. To examine the possibility that Src kinases mediate Stat activation via Jaks, the effect of Src inhibitors on the phosphorylation of Jak2 was examined. The EGF-induced increase in Jak2 tyrosine phosphorylation was completely abolished by pretreatment of A431 cells with either PP1 or CGP77675 (Fig. 9A), suggesting that Src kinases are upstream of Jaks. To control for specificity of the inhibitors, we analyzed their effects on prolactininduced activation of Stat5, which is known to be mediated by Jak2 (38). Treatment of HC11 mammary epithelial cells with the lactogenic hormones dexamethasone, insulin, and prolactin leads to Stat5 activation (30). Pretreatment of HC11 cells with PP1 or with CGP77675 had no effect on dexamethasone, insulin, prolactin-induced Stat5 tyrosine phosphorylation (Fig. 9B) or DNA binding activity (not shown). Thus, the Src inhibitors do not have nonspecific effects on Jak kinases and FIG. 6. Src inhibitors and dominant negative c-Src block ErbB receptormediated phosphorylation and DNA binding of Stat proteins. A and B, NE1, NE2/4, and A431 cells were serumstarved for 18 h and then treated with 10 M PP1 or 1 M CGP77675 for 90 min prior to a 10 min stimulation with 1 nM EGF and NDF. Cells were lysed, and Stat5b (A) and Stat1, Stat3, and Stat5b (B) were immunoprecipitated with specific antibodies. Immune complexes were resolved by SDS-PAGE and blotted. Membranes were probed with phosphotyrosine-specific mAb and, after stripping, reprobed with the respective Stat-specific antibodies. C, plasmids encoding HAtagged Stat5a and WT c-Src or DN c-Src were cotransfected into NE1 cells. 8 h posttransfection, cells were serumstarved for 18 h and then stimulated with 1 nM EGF for 10 min prior to lysis. Stat5a was immunoprecipitated with HA-specific mAb; immune complexes were resolved by SDS-PAGE and blotted. The membrane was probed with phosphotyrosine-specific mAb and, after stripping, reprobed with HA-specific mAb. D, A431 cells were serum-starved for 18 h and then treated with 10 M PP1 or 1 M CGP77675 for 90 min prior to a 15 min stimulation with 5 nM EGF. Proteins were extracted by hypertonic lysis of cells and repeated cycles of freezing and thawing. 10 g of protein were incubated with radiolabeled Stat5 binding site. Complexes were electrophoresed on a 4% native polyacrylamide gel. PY, phosphotyrosine; WB, Western blot; IP, immunoprecipitation.
do not block Stat activation by lactogenic hormones.
Finally, we tested whether Jak activation was necessary for Stat phosphorylation by performing transient transfection studies with DN Jak constructs. HA-tagged Stat5a and DN Jak1 or DN Jak2 were transiently expressed in NE1 cells. No inhibition of EGF-induced Stat5a phosphorylation was observed ( Fig. 10, lanes 1-6). Coexpression of HA-tagged Stat5a, the prolactin receptor, and DN Jak2 completely blocked prolactin-induced Stat5a phosphorylation (Fig. 10, lanes 7-12), demonstrating that DN Jak2 was capable of interfering with Jak2 function in our assay. Finally, we conclude that, in ErbB receptor signaling, Src kinases promote Stat1, Stat3, and Stat5a/b phosphorylation by a direct, Jak-independent mechanism. DISCUSSION Stat transcription factors are activated by numerous cytokines and peptide growth factors (4 -6). EGF activates Stats in vitro in cultured cell lines (33,35,36,40) and in vivo in the liver (34). Our results show that in A431 cells, ErbB1 is constitutively associated with latent Stat1, Stat3, Stat5a/b, and Jak kinases even in the absence of EGF. Ligand-induced recruitment of active Src kinases to the receptor complex appears to be a prerequisite for both Stat and Jak phosphorylation. Src kinases phosphorylate Stats, in a Jak-independent fashion, on the conserved C-terminal tyrosine residue, thereby promoting Stat dimerization, nuclear translocation, and DNA binding. As opposed to prolactin, EGF induces Stat phosphorylation on additional tyrosine(s), the function of which remain to be elucidated. A model of the proposed mechanism by which ErbB receptors mediate Stat activation is presented in Fig. 11.
Using NIH3T3 sublines expressing defined combinations of ErbB proteins (13), we examined the ability of distinct receptor dimers to induce Stat activation. In a previous study we have shown that ErbB2 can potentiate EGF-induced signaling through MAP and S6 kinases (14). However, ErbB2 expression did not enhance EGF-induced Stat activation, even when ErbB3 or ErbB4 were coexpressed. This suggests that ErbB1 homodimers are sufficient for maximal Stat activation. We provide the first report showing NDF induction of Stat activity, specifically through heterodimers of ErbB2 and ErbB4. We have recently shown that ErbB receptor phosphorylation is modulated by the dimerization partner (13). This explains how heterodimers can acquire novel signaling properties that are not the sum of the activity of the individual receptors. Intriguingly, tryptic phosphopeptide mapping of ErbB2 from NDFtreated NE2/4 cells revealed a phosphorylated peptide that was not present in ErbB2 from NDF-treated NE2/3 cells (13). It is tempting to speculate that this peptide (h) contains a phosphorylated residue involved in Stat activation by the ErbB2/ErbB4 heterodimer. The significance of heterodimer-specific signaling is seen in cardiac development where trabeculae formation absolutely depends on the presence of NDF and the coexpression of ErbB2 and ErbB4 (reviewed in Ref. 41). Therefore, in specific tissues NDF induction of the Stat pathway may be of biological importance.
NDF-activated ErbB2/ErbB3 dimers did not induce Stat phosphorylation, although we observed c-Src activation downstream of the ErbB2/ErbB3 dimer in SKBR3 cells (42). One striking difference between ErbB2/ErbB3 and ErbB2/ErbB4 dimers is their differential ability to activate the phosphatidylinositol 3-kinase (PI3K) pathway because of the multiple bind- ing sites for the p85 subunit of PI3K on ErbB3 (43,44). Treatment of cells with an inhibitor of PI3K, LY294002, led to increased phosphorylation of Stat5, 2 suggesting that high levels of PI3K activity could negatively regulate Stat signaling. We are continuing to investigate the role of this pathway in Stat activation.
Our results show that there is constitutive association of Stat proteins with ErbB1 in nonstimulated A431 cells. This is in contrast to a previous report demonstrating ligand-induced recruitment of Stat1 to ErbB1 (33). However, in these earlier experiments A431 cells were serum-starved for 48 h, which may have disrupted basal Stat-ErbB1 complexes. There are other examples of Stat association with nonstimulated receptors, e.g. Stat1 and Stat2 interact with a subunit of the IFN␣ receptor in the absence of ligand (45). Fu and Zhang (33) further reported that the SH2 domain of Stat1 was necessary for association with ErbB1. Because we observed that the Stat5a SH2 domain mutant still complexed with ErbB1, Stat1 and Stat5a may interact with the receptor via different mechanisms. The SH2 domain, however, was necessary for ensuing EGF-induced tyrosine phosphorylation of both Stat1 (33) and Stat5a (Fig. 4A).
Src family members have important roles in transducing mitogenic signals from receptor tyrosine kinases (reviewed in Refs. 21 and 22). In ErbB receptor signaling, we have identified a novel function of Src kinases as activators of Stat transcription factors. The pharmacological Src kinase inhibitors PP1 (27) and CGP77675 (28) used in this study blocked Stat activation in response to EGF and NDF. These inhibitors inactivate not only c-Src but also other Src family members. c-Src, c-Yes, and Fyn are ubiquitously expressed and thus, may act in concert to activate Stats. In addition to c-Src, c-Yes has been implicated in ErbB receptor signaling. Both kinases were found associated with ErbB2/Neu and demonstrated elevated kinase activity in Neu-induced tumors (20).
A431 cells have high c-Src kinase activity in the absence of growth factors (46), yet we detect no basal phosphorylation of Stat proteins. c-Src kinase activity peaked after 60 min of EGF treatment, whereas recruitment of c-Src to latent ErbB-Stat complexes was more rapid, occurring within 5 min of EGF treatment. Because Stat activation was also rapid, we consider ligand-induced recruitment of c-Src, rather than its level of activation, crucial for Stat and Jak phosphorylation. Interactions between Src and ErbB1 appear to be bidirectional. Src itself can phosphorylate ErbB1, leading to hyperactivation of the receptor and potentiation of downstream signals (47,48 9. Src inhibitors block EGF-induced phosphorylation of Jak2 but do not inhibit lactogenic hormone-induced activation of Stat5. A, A431 cells were serum-starved for 18 h and then treated with 10 M PP1 and 1 M CGP77675 for 90 min prior to a 10 min stimulation with 1 nM EGF. Cells were lysed, and Jak2 was immunoprecipitated with polyclonal antibody C20. Immune complexes were resolved by SDS-PAGE and blotted. The membrane was probed with phosphotyrosine-specific mAb and, after stripping, reprobed with Jak2specific antiserum (UBI). B, HC11 cells were serum-starved for 12 h and then treated with 10 M PP1 and 2 M CGP77675 for 90 min prior to a 15 min stimulation with lactogenic hormones. Cells were lysed, and Stat5a was immunoprecipitated with specific antiserum. Immune complexes were resolved by SDS-PAGE and blotted. The membrane was probed with phosphotyrosine-specific mAb and, after stripping, reprobed with Stat5a-specific antiserum. DIP, dexamethasone, insulin, prolactin; PY, phosphotyrosine; WB, Western blot; IP, immunoprecipitation.
FIG. 10. Jak kinases are not required for EGF-induced phosphorylation of Stat5. Plasmids encoding DN Jak1 or Jak2 were cotransfected with a plasmid encoding HA-tagged Stat5a and, where indicated, prolactin receptor (Prl-R) into NE1 cells. 8 h posttransfection cells were serum-starved for 18 h and, prior to lysis, were stimulated with either 1 nM EGF for 10 min or 5 g/ml prolactin for 15 min. Stat5a was immunoprecipitated with HA-specific mAb; immune complexes were resolved by SDS-PAGE and blotted. The membrane was probed with phosphotyrosine-specific mAb and, after stripping, reprobed with HA-specific mAb. PY, phosphotyrosine; WB, Western blot; IP, immunoprecipitation.

FIG. 11. Model for ErbB receptor-mediated Stat activation.
EGF binds to its receptor, ErbB1, which is preassociated with latent Jak and Stat proteins and induces receptor dimerization, autophosphorylation, and recruitment of Src tyrosine kinase. Src phosphorylates Jaks, and it phosphorylates Stats on the consensus C-terminal tyrosine residue, which enables dimerization and nuclear translocation, and on additional tyrosine(s). P, phosphotyrosine. vitro (9, 10), we cannot rule out the possibility that ErbB receptors hyperactivated by Src kinases directly phosphorylate Stats. c-Src is also involved in Stat3 activation in response to colony-stimulating factor 1 and interleukin-3 (15,49). A more complex mechanism involving both Fyn and Jak2 activity has been proposed for Stat1 phosphorylation downstream of the angiotensin II receptor (50).
Tyrosine phosphorylation of Jak kinases in response to EGF (51) and platelet-derived growth factor (8) has been observed. However, the role of Jaks in receptor tyrosine kinase-mediated signaling is still unclear. Our results using DN Jak1 and Jak2 suggest that these kinases are not required for ErbB-induced Stat activation. This conclusion is supported by the fact that EGF-and platelet-derived growth factor-induced Stat activation were not affected in Jak1-, Jak2-and Tyk2-deficient cells (7,8). This is the first report showing that ErbB receptorinduced Jak phosphorylation requires Src kinase activity. Elevated phosphotyrosine levels and increased activity of Jaks have been observed in v-Src transformed cells (52), suggesting that Jaks lie downstream of Src. However, this may be influenced by the cellular context because Jak phosphorylation was Src-independent in other studies (49,50,53).
We also present the first evidence of a direct interaction between ErbB receptors and Jaks. Despite the fact that Jaks are apparently not necessary for Stat activation, they may fulfill other important roles in ErbB receptor signaling. Jaks can be phosphorylated on multiple tyrosine residues, thereby providing docking sites for phosphotyrosine-binding signaling proteins which, in turn, may serve as substrates for Jak kinases. In addition to Stat activation, Jaks have been implicated in the phosphorylation of Shc and the insulin-regulated substrate 1/2 signaling molecules (54,55) and in the activation of Raf1 and the MAP kinase pathway in response to IFN␤, IFN␥, and oncostatin M (56,57). Immunoprecipitates of Jak1 contained multiple tyrosine-phosphorylated proteins (Fig. 8C), lending credence to the hypothesis that Jaks have a function as scaffold molecules. Jaks may in fact serve as adaptors for Stats, enabling interaction with ErbB receptors. We show that Jaks are associated with ErbB1 and ErbB2. Furthermore, we detected Jak2 in immunoprecipitates of transiently expressed Stat5a (not shown). Indirect coupling of Stats to ErbB receptors via constitutively associated Jaks is consistent with the observation that autophosphorylation sites on ErbB1 are not required for Stat activation (40).
In striking contrast to prolactin, EGF induced Stat5a phosphorylation not only on the consensus Tyr 694 but also on additional tyrosine(s). We currently do not know whether this is specific for Stat5a or applies to EGF-activated Stat proteins in general. In tryptic phosphopeptide maps of EGF-activated Stat5a, we observed that Tyr 694 is the major site of phosphorylation. 3 Additional phosphorylated tyrosine(s) may have a function in modulating Stat activity by altering DNA binding or interaction with other transcription factors. Alternatively, these tyrosine(s) may be involved in signaling properties unrelated to transcriptional activation. Because the Stat5a Tyr 694 mutant failed to translocate to the nucleus, it will be interesting to see if this mutant can participate in other responses. For example, Stat3 has been proposed to act as an adaptor protein, coupling PI3K to the IFN receptor via phosphorylation on an additional tyrosine residue (58). Furthermore, mutants of Stat1 proteins, which are deficient in dimer formation, restored sensitivity to tumor necrosis factor-induced apoptosis in Stat 1 Ϫ/Ϫ cells (59).
Coexpression of ErbB1 and c-Src in murine fibroblasts re-sults in a synergistic increase in growth and tumorigenicity (47). Furthermore, c-Src is activated/overexpressed in many human carcinomas that possess elevated levels of ErbB receptors (60,61). In this study we show that Stats are downstream of both ErbB receptors and Src kinases. In NE1 cells, overexpression of Stat5a and wild type c-Src led to constitutive activation of Stat5a (Fig. 6C), and in A431 cells, it led to constitutive association of c-Src with Stat5a and ErbB1 (data not shown). In addition, Src kinase inhibitors blocked the activation of Stats by various ErbB ligands. Two recent reports demonstrated that v-Src-induced malignant transformation is dependent on Stat3 (62,63). It is possible that Stat proteins are also involved in ErbB-mediated tumorigenesis. However, the precise role of Stats in ErbB receptor signaling awaits the elucidation of Stat target genes. It will be interesting to determine whether Stats activated by Src kinases as opposed to Jak kinases elicit different biologic responses.