FER Kinase Activation of Stat3 Is Determined by the N-terminal Sequence*

p94 fer and p51 ferT are two tyrosine kinases that share identical SH2 and kinase domains but differ in their N-terminal regions. To further explore the cellular func-tions of these two highly related tyrosine kinases, their subcellular distribution profiles and in vivo phosphorylation activity were followed using double immunofluorescence assay. When combined with immunoprecipitation analysis, this assay showed that p94 fer can lead to the tyrosine phosphorylation and activation of Stat3 but not of Stat1 or Stat2. Native p94 fer exerted this activity when residing in the cytoplasm. However, modified forms of p94 fer , which are constitutively nuclear, could also lead to the phosphorylation of Stat3. Endogenous Stat3 and p94 fer co-immunoprecipitated with each other, thus proving the interaction of these two proteins in vivo . Unlike p94 fer , p51 ferT did not induce the phosphorylation of Stat3 but led to the phosphorylation of other nuclear proteins. Replacing the unique 43-amino acid-long N-terminal tail of p51 ferT with a parallel seg-ment from the N-terminal tail of p94 fer did not change the subcellular localization of p51 ferT but enabled it to activate Stat3. Thus the different N-terminal sequences of p94 fer and p51 ferT can affect their ability to induce phosphorylation

From the Faculty of Life Sciences, Bar-Ilan University, Ramat Gan 52900, Israel p94 fer and p51 ferT are two tyrosine kinases that share identical SH2 and kinase domains but differ in their N-terminal regions. To further explore the cellular functions of these two highly related tyrosine kinases, their subcellular distribution profiles and in vivo phosphorylation activity were followed using double immunofluorescence assay. When combined with immunoprecipitation analysis, this assay showed that p94 fer can lead to the tyrosine phosphorylation and activation of Stat3 but not of Stat1 or Stat2. Native p94 fer exerted this activity when residing in the cytoplasm. However, modified forms of p94 fer , which are constitutively nuclear, could also lead to the phosphorylation of Stat3. Endogenous Stat3 and p94 fer co-immunoprecipitated with each other, thus proving the interaction of these two proteins in vivo. Unlike p94 fer , p51 ferT did not induce the phosphorylation of Stat3 but led to the phosphorylation of other nuclear proteins. Replacing the unique 43-amino acid-long N-terminal tail of p51 ferT with a parallel segment from the N-terminal tail of p94 fer did not change the subcellular localization of p51 ferT but enabled it to activate Stat3. Thus the different N-terminal sequences of p94 fer and p51 ferT can affect their ability to induce phosphorylation of Stat3 and most probably direct their different cellular functions.
Non-receptor tyrosine kinases are localized in various subcellular compartments, where they exert specific functions. These include receptor-associated kinases that mediate signals for cell growth and differentiation (1,2), membrane-associated kinases that regulate cytoskeletal-mediated signal transduction pathways (3,4), and other tyrosine kinases that can be detected in both the cytoplasm and the nucleus of cells. The last group contains c-Src-and c-Abl-related tyrosine kinases (5)(6)(7) and the Fes/FER family of non-receptor tyrosine kinases (8 -10). The Fes/FER family includes c-fes and the FER tyrosine kinases whose kinase domains are 70% homologous to the kinase domain of c-fes (11)(12)(13)(14).
Two tyrosine kinases are encoded by the FER locus in mice, the evolutionary conserved p94 fer (15,16) and its meiotic truncated variant p51 ferT , which is expressed in meiotic cells (17,18). These two kinases share common SH2 and kinase domains but differ in their NH 2 regions (see Fig. 1 and Refs. 12 and 17). p94 fer carries a 412-aa 1 N-terminal tail in which three potential coiled-coil-forming domains were identified (10,19). These have been implicated in directing the subcellular localization (10) and oligomerization (20) of this molecule. In p51 ferT , the N-terminal 412 aa of p94 fer are replaced via differential splicing, with a novel 43-aa-long N-terminal tail (17).
The FER kinases also differ in their tissue distribution profiles. Although the presence of p94 fer was documented in most mammalian cell lines analyzed (11,12), except for pre-B, pre-T, and T cells (21), p51 ferT was shown to accumulate solely in meiotic pachytene spermatocytes (18,22). The two FER enzymes not only accumulate in different tissues, but they also exhibit different subcellular distribution patterns. p94 fer is mainly cytoplasmic though it enters the nucleus upon transition of cells from G 1 to the S phase (10). In the cytoplasm, p94 fer associates with cell-cell adhesion molecules (23,24), and its activity is induced in growth factor-stimulated cells (19). Moreover p94 fer was shown to associate with activated epidermal growth factor and platelet-derived growth factor receptors in fibroblasts (19) and with the Fc⑀RI receptor in mast cells (25). Thus p94 fer is linked to growth promoting processes. Unlike p94 fer , the meiotic FER tyrosine kinase p51 ferT has not been detected in the cytoplasm of cells that express it, and it accumulates constitutively in the cell nucleus (10,22). The functional implications of these differences between p94 fer and p51 ferT have not been explored. Downstream effectors of these enzymes have not been identified, and their cellular roles in somatic or meiotic cells are not well understood. To further understand the cellular role of these tyrosine kinases, and to learn whether they can phosphorylate similar repertoire of cellular substrates, a panel of FER variants was subjected to double immunofluorescence assay, in addition to immunoprecipitation and Western blot analysis. These assays allowed the characterization of the subcellular distribution and in vivo phosphorylation activity of the FER kinases. This approach revealed that the N-terminal regions of the two FER kinases direct their different substrate specificity, and it implied that p94 fer could serve as a novel activator of Stat3.

EXPERIMENTAL PROCEDURES
Expression Vectors-The construction of the expression plasmids that were used in this study has been previously described (10). Native p94 fer , fused to a single influenza HA epitope at its NH 2 -terminal tail, was expressed from the pCDNA3 fer vector under the control of the cytomegalovirus promoter. HA-p94 fer devoid of a functional NLS (see Fig. 1, fer KR652/3NQ and Ref. 10) was also expressed from the pCDNA3 plasmid under the control of the cytomegalovirus promoter. The HA-tagged p94 fer variants, fer⌬1-299, fer1-315, fer⌬1-376, and fer⌬1-427, were expressed from the PECE vector under the control of the SV40 early promoter. HA-p51 ferT was also expressed from the PECE vector (see Fig. 1 and Ref. 10).
Cells and Transfections-COS1 and CHO cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum. COS1 cells (5.5 ϫ 10 5 ) were transfected with 7 g of DNA mixed with 30 l of LipofectAMINE (Life Technologies, Inc.) in 100-mm dishes. CHO cells (7.5 ϫ 10 5 ) were transfected with 5 g of DNA mixed with 20 l of LipofectAMINE PLUS reagent (Life Technologies, Inc.) and 30 l of LipofectAMINE in 100-mm dishes. During transfection the cells were grown in Opti-MEM medium (Life Technologies, Inc.). Two or three dishes were taken for each sample. The murine myogenic C2C12 cells (26) and NIH3T3 cells were grown in Dulbecco's modified Eagle's medium supplemented with 15 (C2C12) or 10% (NIH3T3) fetal calf serum.
Immunoprecipitation-Extracted proteins (750 -1000 g) were incubated overnight at 4°C with 1:350 diluted 4G10 monoclonal anti-phosphotyrosine antibody (Upstate Biotechnology, Inc.), 1:25 dilution of ␣Stat3 monoclonal antibody, or 1:125 polyclonal ␣FER antibody. Antigen-antibody complexes were precipitated with protein A/G-Sepharose for 1 h at 4°C and were washed four times with HNTG buffer, which contained 10% glycerol, 0.1% Triton X-100, 20 mM Hepes, pH 7.5. The first two washes were carried out with HNTG buffer containing 150 mM NaCl, and the third wash was done with HNTG containing 80 mM NaCl. The last wash was done with HNTG buffer lacking NaCl. Precipitated proteins were then resolved by SDS-PAGE, blotted onto nitrocellulose membranes, and were then reacted with monoclonal ␣HA (Babco), ␣FER/C2, ␣PT-66, or monoclonal ␣Stat1, ␣Stat2, and ␣Stat3 antibodies. Co-immunoprecipitation of p94 fer and Stat3 was carried out as follows: cell lysates (0.8 -1 mg of protein) were incubated with 1:125 diluted pre-serum or ␣FER antibody (directed against the SH2 domain of p94 fer ) for 4 h at 4°C. The reactions were then transferred for overnight incubation at 15°C to avoid nonspecific association of cellular proteins. Antigen-antibody complexes were precipitated as described above.
Immunohistochemical Analysis-COS1 (3 ϫ 10 4 ) cells or CHO (3.5 ϫ 10 3 ) cells were seeded in eight-well chamber slides with well areas of 1 cm 2 and were then transfected with 125 ng of DNA mixed with 0.5 l of LipofectAMINE PLUS reagent and 1.5 l of LipofectAMINE, in a total volume of 125 l. Cells were fixed 40 h post-transfection using 4% paraformaldehyde and were subsequently treated with 0.5% Triton X-100 for 30 min. Blocking was carried out with 6% skim milk, 3% bovine serum albumin, and 0.2% Tween 20 in 100% fetal calf serum. Cells were then exposed overnight at 4°C to the following antibodies: 1:500 diluted monoclonal ␣HA and 1:500 diluted polyclonal ␣PT (Transduction Laboratories) and 1:500 diluted ␣pStat3 or ␣pStat1 (New England BioLabs) antibodies. Reacting antibodies were visualized with fluorescein isothiocyanate-conjugated donkey anti-mouse and Lyssamine Rhodamine-conjugated donkey anti-rabbit antibodies (Jackson Laboratories) using a Bio-Rad MRC 1024 upright confocal microscope with a krypton-argon ion laser. Confocal microscope image analysis was performed using Bio-Rad software, and figures were complied using the Laser Sharp 3.0 software package.
Preparation of Nuclear Extracts-Nuclear extracts were prepared essentially as described before (28). Cells were washed with phosphatebuffered saline, spun down, and resuspended in buffer containing the following: 10 mM Hepes, pH 7.9, 10 mM KCl, 1.5 mM MgCl 2 , 0.5 mM dithiothreitol, 0.2 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 1 g/l Pefabloc, 0.5 g/l aprotinin, 0.5 g/l leupeptin, and 1 g/l benzamidine. Cells were stored for 10 min on ice to prevent swelling and were then lysed in the presence of 0.1% Nonidet P-40. Nuclei were spun down by centrifugation at 10,000 ϫ g for 15 s and suspended in high salt extraction buffer containing the following: 10 mM Hepes, pH 7.9, 420 mM NaCl, 1.5 mM MgCl 2 , 0.2 mM dithiothreitol, 0.2 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 1 g/l Pefabloc, 0.5 g/l aprotinin, 0.5 g/l leupeptin, and 1 g/l benzamidine. Nuclear proteins were extracted by rotating the nuclei for 30 min in the extraction buffer at 4°C, followed by centrifugation at 20,000 ϫ g for 15 min. Supernatants were collected, divided into aliquots, and stored at Ϫ70°C.
Electromobility Gel Shift Assay-A double-stranded synthetic fragment that carries the Stat3 binding site (29) was labeled with [␣-32 P]␣ATP by fill in reaction using Klenow fragment and was then purified on 12% acrylamide gel (30). 10-g nuclear protein extracts were incubated at 25°C for 15 min with 400 ng of poly(dI⅐dC) and 0.1 ng of labeled probe in the following binding buffer: 4 mM Tris-HCl, pH 8.0, 40 mM NaCl, 4 mM MgCl 2 , and 5% glycerol in total volume of 25 l. The assay mixture was then separated on 4% acrylamide Tris-borate EDTA gel at 100 V for 2-3 h. The gel was dried and exposed overnight on Kodak film (X-OMAT AR).

RESULTS
In Situ Staining of the Tyrosine Phosphorylation Induced by p94 fer -To follow the subcellular distribution of the p94 fer kinase and the tyrosine phosphorylation that it induces, we utilized double immunofluorescence staining. This assay allows one to follow the subcellular distribution and phosphorylation activity of a kinase at the single cell level, thus enabling the dissection of a mixed population of cells. A native HA-tagged p94 fer protein (see Ref. 10 and Fig. 1) was transiently expressed in COS1 cells. Immunohistochemical analysis using monoclonal ␣HA antibody revealed the presence of the ectopically expressed p94 fer in the cytoplasm of most transfected cells (Fig.  2, A and B). To test whether the subcellular localization of the enzyme parallels the distribution of its induced tyrosine phosphorylation, the transfected cells were also co-stained with ␣PT antibodies. A prominent ␣PT signal high above the background was obtained in HA-p94 fer -expressing cells (Fig. 2). The induced phosphotyrosine staining was dependent on the kinase activity of p94 fer , because an inactive HA-p94 fer mutant ( Fig. 1, fer⌬685-756) did not induce tyrosine phosphorylation in transfected cells (Fig. 2C). The subcellular distribution profile of the phosphotyrosine staining varied, however, among different transfected cells and could be seen in two typical patterns. In the majority of the cells, both HA-p94fer and the induced phosphorylation signal were located in the cytoplasm ( Fig. 2A). However, in around 30% of more than 200 ␣PT-positive cells, while p94 fer was localized in the cytoplasm, the induced phos-FIG. 1. Schematic description of the FER enzymes that were studied in this work (10). Wild type fer and wild type ferT represent the native p94 fer and p51 ferT , respectively. fer⌬685-756 is an inactive kinase with a 71-aa deletion in the large lobe of its kinase domain. FerKR652/3NQ is a cytoplasmic form of p94 fer that contains a mutated and inactive NLS. fer⌬1-299 is a truncated, nuclear form of p94 fer that carries two of the three N-terminal coiled-coil domains of p94 fer . fer⌬1-315 is a truncated form of p94 fer that lacks the first, and part of the second, N-terminal coiled-coil domains of p94 fer . This mutant accumulates mainly in the cytoplasm of transfected cells. fer⌬1-376 bears the last 14 aa of coiled-coil domain III in p94 fer . This modified p94 fer accumulates mainly in the nuclei of transfected cells. fer⌬1-427 lacks all three N-terminal coiled-coil domains and accumulates mainly in the nuclei of transfected cells (10). Box C in p94 fer , coiled-coil-forming region; SL, kinase domain small lobe; LL, kinase domain large lobe; gray box, represents the SH2 domains of p94 fer and p51 ferT ; dotted box in p51 ferT represents the unique 43 N-terminal aa of the enzyme; kinase act, kinase activity in vivo; WT, wild type. phorylation signal was mainly in the nucleus (Fig. 2B). We could never see clear cytoplasmic and nuclear phosphorylation signals combined in the same cell. Moreover, one could clearly see that native p94 fer led to the accumulation of cytoplasmic or nuclear phosphorylation signals when the kinase was expressed in the cytoplasm (Fig. 3, A and B). About 30% of the transfected cells did not show an obvious induced tyrosine phosphorylation signal. A similar heterogeneous subcellular distribution profile of phosphotyrosine staining was also seen in CHO cells that transiently expressed the HA-p94 fer protein (data not shown).
To test whether the p94 fer kinase activity requires cytoplasmic localization, the cellular phosphorylation activity of three HA-tagged modified forms of p94 fer were compared. These included fer KR652/3NQ, which bears a mutated NLS (see Ref. 10 and Fig. 1). Mutating the monopartite NLS of p94 fer excludes the protein from the nucleus without compromising its kinase activity (Fig. 3, A and B). Another cytoplasmic variant of p94 fer that was tested was fer⌬1-315, which lacks the first, and part of the second, N-terminal coiled-coil domains of the enzyme (see Ref. 10 and Fig. 1). The in situ phosphorylation profiles of these cytoplasmic forms of fer were compared with the phosphorylation profile of fer⌬1-299. Deleting the first N-terminal coiled-coil domain of p94 fer in fer⌬1-299 (see Ref. 10 and Fig. 1, fer⌬1-299) leads to its nuclear accumulation without affecting its kinase activity (Fig. 3D). The three kinases were transiently expressed in COS1 cells. Unlike the native p94 fer , the accumulation of the NLS negative FER variant (KR652/3NQ) was mainly cytoplasmic, and no preferential accumulation of this protein in the nucleus of any transfected cell could be found. However, as seen with the native p94 fer , transient expression of this variant led to the accumulation of cytoplasmic phosphotyrosine in some transfected cells and to nuclear accumulation of phosphotyrosine in others (Fig. 3, A  and B). Transiently expressed fer⌬1-315 also accumulated in the cytoplasm. This resulted in a perinuclear accumulation of a phosphotyrosine signal in some cells (Fig. 3C). The constitutively nuclear FER variant fer⌬1-299 exhibited exclusive nuclear accumulation and induced nuclear phosphorylation signal in the transfected cells (Fig. 3D). It therefore appears that both modified forms of p94 fer expressed in the cytoplasm or in the nucleus can elicit the accumulation of tyrosine phosphorylation in the cell nucleus. This profile differs from the activity exhibited by the native p94 fer , which induces phosphorylation only when residing in the cytoplasm.
Both Cytoplasmic and Nuclear Forms of FER Induce Tyrosine Phosphorylation of Stat3-To further identify the cellular substrates of p94 fer , whole cell proteins from HA-p94 fer -transfected CHO cells were immunoprecipitated with ␣PT antibodies. Immunoprecipitates were resolved by SDS-PAGE and then probed with ␣PT antibodies by Western blot analysis. Several proteins were found to be highly tyrosine-phosphorylated in transfected but not in untransfected cells. The sizes of these proteins clustered between 75 and 95 kDa and between 97 and 110 kDa, respectively (Fig. 4, upper panel). Similar results were obtained when HA-p94 fer was transiently expressed in COS1 cells (data not shown). Based on the migration distances of the phosphorylated proteins seen in SDS-PAGE, the overlapping accumulation of pHA-94 fer , and the in situ phosphotyrosine signals in the cytoplasm, one would predict the presence of phosphorylated HA-p94 fer in the 97-110-kDa cluster of tyrosine-phosphorylated bands (Fig. 4, upper panel). Indeed reacting the ␣PT precipitates with ␣HA antibody in a Western blot revealed the immunoprecipitation of a phosphorylated HA-p94 fer form that migrated slightly above the 97-kDa marker (Fig. 4, bottom panel, lane 1).
In situ immunofluorescence of induced tyrosine phosphorylation in COS1 cells transiently overexpressing HA-p94 fer (A and B) is shown. Cells were double stained with monoclonal ␣HA antibody, which detects the ectopic HA-p94 fer protein (green) and polyclonal ␣PT antibodies (red). A, HA-p94 fer localization (a); induced cytoplasmic tyrosine phosphorylation (b); the merged images (c). B, HA-p94 fer localization (a); induced tyrosine phosphorylation accumulates mainly in the nucleus (lower cell) or cytoplasm (upper cell) (b); the merged image (c). C, transient overexpression of HA-tagged non-active p94 fer mutant (fer⌬685-756) in COS1 cells (a); HA-fer⌬685-756 localization (b); lack of induced tyrosine phosphorylation merged images (c). These photographs represent stacked confocal laser sections taken 1 m apart. Experiments were repeated more then ten times, and more than 100 cells were scored in each experiment. Representative photographs are shown.Scale bar, 50 m.
The accumulation of nuclear phosphotyrosine signals in cells overexpressing cytoplasmatic forms of FER suggests the transduction of the phosphorylation activity of FER from the cytoplasm to the nucleus. Possible downstream effectors in this process could be the signal transducers and activators of tran-scription, which were shown to be phosphorylated by several non-receptor tyrosine kinases (31,32). These proteins are tyrosine-phosphorylated in the cytoplasm and are then translocated to the cell nucleus (33). The molecular mass of these proteins varies between 84 and 113 kDa (33), a size range that coincides with the size of some of the phosphorylated proteins seen in Fig. 4 (upper panel).
To test the possible link between p94 fer and signal transducers and activators of transcription, phosphotyrosine precipitates from COS1 cells transfected with various FER variants were exposed to ␣Stat1, ␣Stat2, ␣Stat3 (Fig. 5, A and C), and ␣HA antibodies (Fig. 5B, top panel). The phosphorylation of Stat3 (Fig. 5A, lane 1), but not that of Stat1 and Stat2 (Fig. 5C), was specifically induced by the native HA-p94 fer enzyme. Both cytoplasmic (fer⌬1-315 and ferKR653/3NQ) and nuclear (fer⌬1-299) modified forms of p94 fer also induced tyrosine phosphorylation of Stat3 (Fig. 5A, upper panel). The phosphorylation was dependent on the kinase activity of the fer enzymes, because an inactive mutated form of p94 fer (fer⌬685-756, Fig. 5A, upper panel, lane 5) could not induce the tyrosine phosphorylation of Stat3. Similarly, native or nuclear forms of FER overexpressed in CHO cells were able to induce the tyrosine phosphorylation of Stat3 (data not shown). To verify the specificity of tyrosine phosphorylation of Stat3 in p94 fer -overexpressing cells, Stat3 was directly immunoprecipitated utilizing ␣Stat3 antibodies, and precipitates were reacted with ␣PT   3 and 4,  respectively). The membranes were exposed to ␣PT (upper panel) or ␣HA (lower panel) antibody. IP, immunoprecipitations; IB, immunoblotting; WCL, whole cell lysates. Numbers on the right indicate migration distances of known molecular mass markers.
antibody. Stat3 was efficiently immunoprecipitated from transfected, as well as from untransfected, cells. Again both native and nuclear forms of FER (fer⌬1-299) efficiently induced the accumulation of a tyrosine-phosphorylated Stat3 (Fig. 6). This experiment confirmed the increased tyrosine phosphorylation of Stat3 in cells overexpressing a FER kinase.
Stat3 Is Tyrosine-phosphorylated in Most of the Cells Ex-pressing HA-p94 fer -After identifying Stat3 as a downstream effector of p94 fer , we turned to check to what extent does the tyrosine phosphorylation signal induced by p94 fer in transfected COS1 cells reflect the tyrosine phosphorylation of Stat3. COS1 cells were transiently transfected with the HA-p94 fer expression vector and were then co-stained with ␣HA and ␣pStat3 antibodies. The ␣pStat3 specifically detects Stat3, which became phosphorylated on Tyr-705. Tyrosine-phosphorylated Stat3 was detected in 70% of more than 200 HA-p94 ferexpressing cells, which were scored (Fig. 7). This signal was prominent in the cytoplasm or in the nucleus of the HA-p94 ferexpressing cells (Fig. 7, A and B). ␣pStat1 antibodies, on the other hand, gave only a faint signal, which was barely detected under the experimental conditions used in this work (Fig. 7, C  and D). Thus, both cytoplasmic and nuclear phosphotyrosine signals elicited in HA-p94 fer -expressing cells (Fig. 2, A and B) could reflect the accumulation of tyrosine-phosphorylated Stat3, and Stat3 is therefore activated in most cells expressing the exogenic p94 fer . Tyrosine-phosphorylated Stat3 was not detected in cells expressing the inactive p94 fer mutant, fer⌬685-756 (data not shown). These experiments were repeated in HeLa cells and in the murine myogenic cell line, C2C12 (26), and gave similar results (data not shown). Thus, p94 fer can activate Stat3 in various cell types. Induction of Stat3 DNA Binding Activity in FER Overexpressing Cells-Tyrosine phosphorylation of Stat3 lead to their dimerization via reciprocal SH2-phosphotyrosine interaction. The signal transducers and activators of transcription dimers then enter the nucleus and bind specific DNA elements involved in activation of specific gene transcription (34). To test whether ectopic expression of an FER kinase leads to the activation of Stat3, the DNA binding activity of Stat3 was analyzed in COS1 cells expressing or non-expressing an ectopic FER kinase. Overexpression of fer⌬1-299 induced the DNA binding activity of Stat3 to a synthetic Stat3 binding site (Fig. 8, lane  4). This binding, which was not induced by an inactive p94 fer mutant (Fig. 8, lanes 6 and 7), was specific and could be competed by an unlabeled probe (Fig. 8, lane 5). Thus, a FER kinase can induce the activation of Stat3 (33,34).
The Endogenous p94 fer Kinase Associates with Stat3 in NIH3T3 and C2Cl2 Cells-To substantiate the interaction of p94 fer and Stat3 in vivo, the association of these two proteins was analyzed in two unrelated cell types. These were the fibroblastic NIH3T3 cell line and the murine myogenic C2C12 cell line (26). The endogenous p94 fer protein was immunoprecipitated from whole cell extracts, and the presence of Stat3 was tested in the obtained precipitates. Stat3 co-immunoprecipitated with p94 fer in extracts prepared from the two analyzed cell lines. Polyclonal ␣FER antibodies directed against the SH2  1, 3, 5,  7, and 9) and from non-transfected cells (lanes 2, 4, 6, 8, and 10) were immunoprecipitated with ␣PT antibody (lanes 1, 2, 5, 6, 9, and 10) or were left untreated (lanes 3, 4, 7, and 8). domain of p94 fer specifically immunoprecipitated the FER enzyme and Stat3, which co-immunoprecipitated with it but failed to co-immunoprecipitate Stat1 (Fig. 9). These results clearly demonstrate the specific interaction of p94 fer and Stat3 in vivo and strongly implies that p94 fer is an activator of Stat3.
In Situ Staining of the p51 ferT Phosphorylation Activity-Our observation that both cytoplasmic and nuclear forms of p94 fer can lead to the tyrosine phosphorylation of Stat3 prompted us to study the ability of the nuclear meiotic form of FER, the p51 ferT kinase, to induce a similar phosphorylation profile. Transiently transfected COS1 cells expressing an HA-tagged p51 ferT were immunostained with ␣PT and ␣HA antibodies. As expected, HA-p51 ferT accumulated in the nucleus of all expressing cells (Fig. 10A). Staining the transfected cells with the ␣PT antibody revealed the induction of tyrosine phosphorylation signal by p51 ferT , which was confined to the cell nucleus (Fig. 10A). This profile was similar to the one seen in cells expressing the nuclear truncated form of the somatic p94 fer (Fig. 3D, fer⌬1-299). Western blot analysis showed that p51 ferT was able to induce phosphorylation of proteins in two main bands, around 80 and around 105 kDa (data not shown). p51 ferT Does Not Induce the Phosphorylation of Stat3-To check whether the in situ nuclear phosphorylation signal seen in HA-p51 ferT -expressing cells (Fig. 10A) reflects, at least in part, increased phosphorylation of Stat3, immunoprecipitation and Western blotting were performed. Whole cell proteins were immunoprecipitated from HA-p51 ferT -expressing cells using ␣PT antibodies. The precipitates were resolved by SDS-PAGE and then exposed to ␣Stat3 and ␣HA antibodies. Although HA-p51 ferT was efficiently expressed in the transfected cells (Fig. 11B, lower panel, lane 2), it did not lead to the precipitation of a tyrosine pStat3 (Fig. 11A, upper panel, lane 2), Stat1,   6 and 7) were incubated with a 32 P-labeled probe in the absence (lanes 2, 4, and 6) or presence (lanes 3, 5, and 7) of 175 ng of unlabeled competitive fragment, which was used as a probe in that assay. Lane 1, free probe. or Stat2 (data not shown). In addition, HA-p51 ferT was not autophosphorylated in the transfected cells, because it was not precipitated by the ␣PT antibodies (Fig. 11B, upper panel, lane  2). Thus, the in situ nuclear tyrosine phosphorylation signal seen in p51 ferT -expressing cells (Fig. 10A) most likely reflects the phosphorylation of a nuclear protein(s) that differs from Stat1, Stat2, Stat3, and p51 ferT .
The Unique N-terminal Sequences of p51 ferT Interfere with Its Ability to Phosphorylate Stat3-The fact that truncated nuclear forms of the somatic p94 fer led to the phosphorylation of Stat3 (Fig. 5A), whereas the nuclear meiotically expressed p51 ferT failed to do so, raised the possibility that specific sequences in the p51 ferT protein impair the ability of this meiotic enzyme to induce phosphorylation of Stat3. To test this hypothesis, the unique 43-aa N-terminal region of p51 ferT (Fig. 1) was removed, together with further downstream 15 aa. The new truncated form of p51 ferT (Fig. 1, fer⌬1-427) was tyrosinephosphorylated in vivo (Fig. 11B, upper panel, lane 1 and Fig. 10C). However the truncated p51 ferT did not gain the ability to induce phosphorylation of Stat3 (Fig. 11A, top panel,  lane 1). This could be because of the newly introduced truncation point in p51 ferT , which is proximal to the SH2 domain (starting at aa 460 in p94 fer ) of the kinase (Fig. 1, fer⌬1-427). To further understand the effects of the unique sequences in the N terminus of p51 ferT , the 43 N-terminal aa of p51 ferT were replaced with the parallel 36 unique aa of p94 fer (Fig. 1, fer⌬1-376). In this modified enzyme, the unique N-terminal tail of p51 ferT was replaced by a fragment extending from aa 376 to 412 in p94 fer . This fragment is 36 aa in length, and its position in p94 fer parallels the position of the unique N-terminal tail in p51 ferT (Fig. 1). The replacing p94 fer fragment includes the last 11 aa of the p94 fer coiled-coil domain III, and it was attached to the p94 fer /p51 ferT divergence point (Fig. 1, fer⌬1-376). Thus the modified enzyme, fer⌬1-376, differs from the native p51 ferT by its replaced N-terminal sequences. Like p51 ferT , fer⌬1-376 accumulated mainly in the cell nucleus and induced a nuclear tyrosine phosphorylation signal (Fig. 10B). However, unlike p51 ferT , the transient expression of fer⌬1-376 enabled prominent phosphorylation of Stat3 (Fig. 11A, upper panel, lane 3). Thus, it seems that the N-terminal tail of p51 ferT plays a role in the inability of this enzyme to induce phosphorylation of Stat3. Replacement of the tail with parallel N-terminal p94 fer sequences endowed the meiotic enzyme with the ability to induce Stat3 phosphorylation. DISCUSSION The FER tyrosine kinases are members of the Fes/FER family that belong to the wider group of non-receptor tyrosine kinases. However, unlike most other known tyrosine kinases, Cells were double stained with monoclonal ␣HA antibody, which detects the ectopic HA-p51 ferT , the HA-tagged fer variants (green), and with polyclonal ␣PT (red). A, HA-p51 ferT localization (a); induced tyrosine phosphorylation in the nucleus (b); the merged images (c). B, HA-fer⌬1-376 localization (a); induced tyrosine phosphorylation mainly in the nucleus (b); the merged images (c). C, HA-fer⌬1-427 localization (a); the nuclear-and perinuclear-induced tyrosine phosphorylation (b); the merged images (c). These photographs represent stacked confocal laser sections taken 1 m apart. Experiments were repeated more than ten times, and more than 100 cells were scored in each experiment. Representative photographs are shown. Scale bar, 50 m. the FER tyrosine kinases include both somatically and meiotically expressed members, whose cellular roles are not well understood. To further explore the function of these enzymes, immunofluorescent staining approaches accompanied by Western blotting and immunoprecipitation were employed. Double immunofluorescence was found to be a most valuable tool in characterizing the in vivo phosphorylation activity of a defined tyrosine kinase and for identification of its substrate(s).
By using the combined in situ and biochemical methods, we showed that p94 fer induced the phosphorylation of Stat3 but not of Stat1 or Stat2. We were unable to determine whether p94 fer also activates Stat5 because of the low levels of that transcription factor in COS1 cells. Native p94 fer activated Stat3 when expressed in the cytoplasm, yet truncated forms of p94 fer that were expressed in the nucleus did lead to increased tyrosine phosphorylation levels of Stat3. Thus truncated nuclear FER variants most probably induce the tyrosine phosphorylation of Stat3 molecules that underwent dephosphorylation in the cell nucleus (35). It is unlikely that the residual cytoplasmic levels of the nuclear variants of p94 fer led to the tyrosine phosphorylation of Stat3. This is supported by the fact that nuclear FER variants such as fer⌬1-299 induce higher phosphorylation levels of Stat3 than the native p94 fer , which is mainly cytoplasmic (Fig. 5A). Tyrosine-phosphorylated Stat3 accumulated in most cells expressing the ectopic p94 fer (Fig. 7) and was detected in both the cytoplasm or in the nucleus of HA-p94 fer -expressing cells (Fig. 7). This could reflect different stages in the activation pathway of Stat3. During this process, tyrosine phosphorylation of Stat3 by p94 fer in the cytoplasm is most probably followed by the translocation of the activated transcription factor to the cell nucleus (33). Preliminary results indicated that both processes take place during the G 1 phase of the cell cycle. 2 We noted that in cells exhibiting a prominent nuclear tyrosine phosphorylation signal, which represents the activated Stat3, there was barely any PT signal detected in the cytoplasm (Fig. 2B and Fig. 7B). Thus, translocation of downstream effectors of p94 fer to the cell nucleus is most probably coupled to the dephosphorylation of p94 fer , which resides in the cytoplasm. Interestingly, we detected several p94 fer -transfected cells that did not show any induced tyrosine phosphorylation activity. This could reflect our preliminary observation indicating that the kinase activity of p94 fer is cell cycle regulated. 2 Does p94 fer directly phosphorylate Stat3 or does it lead indirectly to the tyrosine phosphorylation of this transcription factor? The specific association of the endogenous p94 fer with Stat3 but not with Stat1 strongly suggests the direct activation of Stat3 by p94 fer . This could also explain the inability of p94 fer to activate Stat1 in the COS-1 cell overexpression system (Fig. 5). One cannot exclude, however, the possibility that another tyrosine kinase also associates with the p94 fer -Stat3 complex, thus leading to the phosphorylation of Stat3. The direct phosphorylation of Stat3 by c-fes, which is another member of the Fes/FER family (29,36), could however suggest that p94 fer directly phosphorylates Stat3. In addition, it should be noted that the tyrosine phosphorylation levels of Jak1, another potent activator of Stat3 (31), were similar in cells overexpressing HA-p94 fer and in non-transfected COS-1 cells (data not shown). These data support the notion that p94 fer leads directly to the tyrosine phosphorylation of Stat3 and could thus serve as a bona fide activator of Stat3.
Stat3 was shown to be activated in several human tumors (37)(38)(39)(40)(41) and can act as an oncogene (42). The constitutive activation of Stat3 was found to be essential for cellular transformation by v-src (43). Several oncogenic tyrosine kinases including v-Abl (44), v-Fps (45), Bcr-Abl (46), and v-Eyk (47) also activate Stat3. Together these findings link the function of Stat3 to cellular proliferation. This is supported by the fact that Stat3 Ϫ/Ϫ mouse embryos implant but fail to grow (48). Thus, the specific activation of Stat3 by p94 fer directly links this tyrosine kinase to regulation of cell growth, with Stat3 as a downstream effector of p94 fer . This coincides with the activation of p94 fer by growth factors and its association with their receptors (19). The involvement of p94 fer in growth-promoting processes may also explain the ability of Drosophila Fer to transform mammalian cells (16) and the association of p94 fer expression with the proliferation of prostatic cancer cells (49). p94 fer could thus be a novel mediator for the activation of Stat3 by growth factors like platelet-derived growth factor and epidermal growth factor (50, 51). As mentioned above, the other somatic member of the Fes/FER family, the tyrosine kinase c-fes, was also shown to phosphorylate and activate Stat3 (29,36). Stat3 thus seems to be a common effector of the somatic members of the Fes/FER family. Interestingly, Stat3 was not activated in our assay by c-abl (data not shown), another cytoplasmic and nuclear tyrosine kinase, a fact that stresses the specificity of the cellular link between the Fes/FER tyrosine kinases and Stat3.
The meiotic p51 ferT is exceptional in this respect. Despite the fact that it shares the same kinase and SH2 domains with p94 fer , these two highly related kinases do not elicit the same phosphorylation events in fibroblastic cells. Although overexpression of p94 fer led to elevated tyrosine phosphorylation of Stat3 but not of Stat1 or Stat2, p51 ferT failed to induce the phosphorylation of any of these proteins. This difference between p94 fer and p51 ferT did not result from the different sub-2 S. Priel-Halachmi and U. Nir, unpublished results.  , lower panel). The membranes were exposed to ␣Stat3 antibody. Arrows on the left indicate the migration distances of the tyrosine-phosphorylated Stat3 (p.Stat3) and non-phosphorylated Stat3. B, the immunoprecipitates from A (upper panel) and their corresponding untreated lysates (lower panel) were exposed to ␣HA antibody in a Western blot analysis. IP, immunoprecipitations; IB, immunoblotting; delHAfer, tagged modified forms of p94 fer ; p.delHAfer, tyrosine-phosphorylated modified forms of HA-p94 fer . cellular distribution profiles of the two kinases, because truncated forms of p94 fer , which like p51 ferT exhibit constitutive nuclear localization profiles, did induce the increased tyrosine phosphorylation level of Stat3. p94 fer and p51 ferT thus appear to have different substrate specificity.
We found that p51 ferT increases the phosphotyrosine levels of 80-and 105-kDa proteins in COS1 cells (data not shown), which prove not to be Stat1, Stat2, or Stat3. One of these proteins could be related to the 66-kDa nuclear protein that is phosphorylated by p51 ferT in CHO cells (27). The inability of p51 ferT to activate Stat3 may suggest that unlike p94 fer , p51 ferT could induce growth-suppressive pathways, or it could be linked to growth-promoting processes that do not involve Stat3. The impairment imposed by p51 ferT on S phase progression in transfected CHO cells (10), indicates the possible involvement of p51 ferT in growth-inhibitory processes in meiotic cells.
The inability of p51 ferT to phosphorylate Stat3 seems to result from an inhibitory effect that is imposed by its unique 43-aa-long N-terminal tail. This was demonstrated by showing that replacement of these 43 aa with a parallel N-terminal sequence from p94 fer (Fig. 1, fer⌬1-376) restored the ability of p51 ferT to phosphorylate Stat3 (Fig. 11). One can not exclude, however, the possibility that the unique tail of p51 ferT does not inhibit Stat3 phosphorylation but rather lacks some positive signal that is present in the 36 N-terminal aa of fer⌬1-376 and that is essential for the interaction of the FER kinases with Stat3. Interestingly, the regulatory effect of the N-terminal tail of p51 ferT is specific and does not prevent the interaction of that enzyme with other cellular substrates. It should be noted, however, that the unique N-terminal sequences also interfere with the autophosphorylation activity of p51 ferT in COS1 cells (Fig. 11). 3 This could be related to the inability of p51 ferT to phosphorylate Stat3. The unique N-terminal tail of p51 ferT may thus be involved in the formation of a structure that dictates the defined substrate specificity of the meiotic kinase. It should be mentioned that we did not identify any post-translational modifications that could be linked to the modulatory effect of these N-terminal 43 aa in p51 ferT (data not shown).
The FER kinases constitute a unique subgroup of tyrosine kinases in which two related enzymes share identical kinase and SH2 domains but are each linked to a different N-terminal tail. This directs different substrate specificity of these enzymes and most probably leads to their different cellular roles, which are linked to key regulatory processes of cell growth.