Disruption of Coiled-coil Domains in Fer Protein-tyrosine Kinase Abolishes Trimerization but Not Kinase Activation*

The protein-tyrosine kinase Fer and the highly homologous proto-oncoprotein Fps/Fes are implicated in signaling from a variety of growth factor and cytokine receptors. Here we examine the molecular basis of Fer kinase activation with an emphasis on the role of oligomerization. We show that Fer forms trimers in vivo and that disruption of either the first or second coiled-coil domain abolishes oligomerization, suggesting a co-operative interaction between these two domains. Although Fps/Fes also forms homotypic oligomers, probably via homologous coiled-coil domains, no heterotypic interactions were observed between Fer and Fps/ Fes. Incorporation of catalytically inactive Fer peptides into the oligomeric complex caused only mild reduction of wild type Fer kinase activity, suggesting that kinase-inactive Fer would not behave as a potent dominant negative. Although oligomerization of Fer can potentiate autophosphorylation in trans at three major phosphorylation sites, these residues can likely also be phosphorylated in cis . In contrast, the testis-specific FerT isomer does not oligomerize and is able to autophosphorylate in cis at two of the same three residues autophosphorylated in Fer. These results suggest that although oligomerization potentiates autophosphorylation in trans , this is apparently not necessary for Fer activation. Chemical Cross-linking— Transfected COS-1 cells were in X-linking (20 m M Hepes-KOH (pH 7.5), 150 m M NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate) and clarified by centrifugation at 14,000 rpm for 10 min at 4 °C. Extracts were incubated on ice in the absence or presence of 1 m M ethylene glycol-bis(succinic acid N -hy- droxysuccinimide ester) (EGS) for 15 min (EGS from Sigma was prepared as a 100 m M stock solution in Me 2 SO). The reactions were terminated by the addition of 1 m l of 50 3 TE (500 m M Tris-HCl (pH 8), 50 m M EDTA) and 50 m l of SDS-loading buffer. Samples were heated at 100 °C for 4 min, followed by SDS-PAGE, and prepared for Western blotting as described above. Tryptic Phosphopeptide Mapping— COS-1 cells were transfected and harvested, and immune complex kinase assays performed as described above. The sources of native Fer and FerT, were mouse liver and testis, respectively, which were homogenized in KLB. Fer was immunoprecipi- tated with anti-Fer polyclonal antiserum (5 m l), while FerT was immunoprecipitated with anti-Fps/Fer polyclonal antiserum (5 m l), and Myc epitope-tagged proteins were immunoprecipitated with monoclonal antibody 1-9E10. the kinase reactions, samples were subjected to SDS-PAGE and transferred to a polyvinylidene difluoride membrane, and autoradiography was performed. Portions of the membrane con- taining the radiolabeled proteins were excised, in separate 1.5-ml tubes containing 1 ml of 0.5% polyvinyl pyrolidone-360 (in 100 m M acetic acid), and incubated for 30 min at 37 °C. Following aspira-tion, the membranes were washed five times with

The protein-tyrosine kinase Fer and the highly homologous proto-oncoprotein Fps/Fes are implicated in signaling from a variety of growth factor and cytokine receptors. Here we examine the molecular basis of Fer kinase activation with an emphasis on the role of oligomerization. We show that Fer forms trimers in vivo and that disruption of either the first or second coiledcoil domain abolishes oligomerization, suggesting a cooperative interaction between these two domains. Although Fps/Fes also forms homotypic oligomers, probably via homologous coiled-coil domains, no heterotypic interactions were observed between Fer and Fps/ Fes. Incorporation of catalytically inactive Fer peptides into the oligomeric complex caused only mild reduction of wild type Fer kinase activity, suggesting that kinaseinactive Fer would not behave as a potent dominant negative. Although oligomerization of Fer can potentiate autophosphorylation in trans at three major phosphorylation sites, these residues can likely also be phosphorylated in cis. In contrast, the testis-specific FerT isomer does not oligomerize and is able to autophosphorylate in cis at two of the same three residues autophosphorylated in Fer. These results suggest that although oligomerization potentiates autophosphorylation in trans, this is apparently not necessary for Fer activation.
Fer is a cytoplasmic protein-tyrosine kinase (PTK) 1 with close structural similarity to the product of the fps/fes protooncogene (1). Indeed, Fer was first observed as one of two PTKs detected in myeloid cell lines using antibodies raised against viral Fps/Fes peptides. Peptide mapping analysis revealed these to be distinct, yet closely related PTKs (2,3). Human and rat cDNAs encoding Fer were subsequently cloned, and a comparison with the genes encoding cellular and viral Fps/Fes proteins confirmed the close structural similarity between p92 Fps/Fes (hereafter referred to as Fps) and p94 Fer (4 -6). Although dozens of novel tyrosine kinases have been identified in recent years, Fps and Fer remain the only two known members of a distinct subclass within the nonreceptor PTK family. They each consist of a C-terminal tyrosine kinase domain, a central Src homology 2 (SH2) domain, and an N-terminal region, which contains three predicted coiled-coil (CC) motifs (7,8). Unlike members of the Src family, Fps and Fer do not possess Nterminal myristoylation sites, SH3 domains, or C-terminal negative regulatory tyrosine phosphorylation sites. It has been suggested that interactions between the SH2 and catalytic domains of Fps might serve a regulatory role (9,10); however, a clearer understanding of this awaits detailed structural determination.
By analogy with the receptor PTKs, activation of nonreceptor kinases may also be stimulated by oligomerization followed by autophosphorylation in trans. It has recently been suggested that homotypic CC interactions may mediate oligomerization of both Fer (7), and Fps (8). The conserved putative CCs in these kinases could also mediate heterotypic interactions between Fer and Fps. This raises the possibility of functional or regulatory interactions between these two kinases. Alternatively, the CC domains could mediate interactions of these kinases with other CC domain-containing proteins. In this regard, the N-terminal domain of Fer has been shown to mediate association with the catenin family member, p120CAS (7). In addition to the central armadillo repeats, p120CAS also contains a predicted N-terminal CC domain (11,12); however, it is not known which domain confers interaction with Fer and whether this interaction is direct (13).
Fer has also been detected in association with the activated EGF and PDGF receptors in fibroblasts (7), as well as the Fc ⑀ RI receptor in mast cells (14). In each of these cases, ligand stimulation of the receptor resulted in elevated Fer kinase activity. Associations of Fps with a number of cytokine receptors have also been described, including those for interleukin-3 (15), granulocyte-macrophage colony-stimulating factor (15,16), interleukin-4 (17), interleukin-6, leukemia inhibitory factor, oncostatin M, ciliary neurotrophic factor, and interleukin-11 (18). Due to the close similarity between Fps and Fer, many of the antisera currently in use are unable to distinguish between these two PTKs. This raises the possibility that the PTKs seen in association with some of these cytokine receptors may have been Fer rather than Fps. Alternatively, both kinases could be involved, just as different members of the Jak kinase family may be involved in signaling from some cytokine receptors (19).
The expression patterns of Fps and Fer are quite distinct. Fps is more restricted, with relatively high levels seen in a limited subset of cell types including myeloid, vascular endothelial, and some epithelial and neuronal cells (20,21). On the other hand, Fer is widely expressed (4,5), and levels comparable with that of Fps in myeloid cells are observed in most tissues. A further distinctive characteristic of fer is the expression of a testis-specific isoform, FerT, which likely arises from an internal tissue-specific promoter and alternative splicing (22). Murine FerT lacks most of the N-terminal domain of p94 Fer and contains a unique 43-amino acid N terminus, followed by the common SH2 and kinase domains. ferT mRNA accumulates transiently in primary spermatocytes during the pachytene stage of meiotic prophase (23). In Drosophila melanogaster, cDNAs encoding both the larger DFer and the shorter DFerT isoforms have been isolated (24). In Caenorhabditis elegans, a putative ferT homolog was identified on chromosome III (25). However, there do not appear to be additional upstream coding sequences that would produce a larger Fer gene product, suggesting that sequences encoding the N-terminal domain of Fer had not yet evolved in C. elegans.
There is currently little information regarding the molecular function of Fer, although a role in signaling downstream of activated platelet-derived growth factor, epidermal growth factor, and Fc ⑀ RI receptors has been proposed (7,14). Recent studies suggest roles for Fer in regulation of the actin cytoskeleton via interaction with cortactin (13) and control of cell adhesion (26). While fps was originally identified as a retroviral oncogene (27,28), the involvement of fer in malignancy or other diseases is not yet established. However, the human fer locus maps to chromosome 5q21, in a region which is frequently deleted or rearranged in myeloid leukemia (29), and recently the in vitro transforming potential of Fer was demonstrated (24). These observations suggest a role for Fer in cell signaling and growth control and a potential involvement in myeloid leukemia or other diseases.
Here we examine the role of oligomerization in Fer kinase activation. We show that disruption of either the first or second CC domain completely abolishes oligomerization, and chemical cross-linking indicates that Fer forms trimers. Although both Fer and Fps exist as oligomers in vivo and they share conserved CC motifs, no heterotypic association between Fer and Fps was observed. Furthermore, we show that autophosphorylation of Fer probably involves both cis and trans phosphorylation mechanisms. Tryptic phosphopeptide mapping revealed differences in autophosphorylation sites between Fer and FerT that may reflect differences in signal transduction between these isoforms.

MATERIALS AND METHODS
Plasmid Constructions-Fer proteins were produced in COS-1 cells using the vector pXN1, which is a modified version of pECE (30) in which the XbaI cloning site was converted to a NotI site by digestion with XbaI, and ligation with an annealed NotI adapter oligonucleotide (XNot, 5Ј-CTAGGCGGCCGC-3Ј). The 2.7-kilobase pair murine fer cDNA from the pF2c.8 plasmid was excised with ApaI and DraI, blunted with Klenow, and subcloned into the SmaI site of pECE. The resulting pECEmFer plasmid was kindly provided by K. Letwin and T. Pawson. The construction of wild type and Fer D743R expression constructs was as described previously (31).
Expression plasmids encoding Myc epitope-tagged wild type and Fer K592R are as described previously (31). Mutations within the CC domains were generated using the unique site elimination method (32) as recommended by the manufacturer (U.S.E. mutagenesis kit; Amersham Pharmacia Biotech). The selection primer (5Ј-AGATCTAAGCT-TCTCGAGCACAGTGTGGAGG-3Ј) destroys a unique 5Ј SalI site in pESNmFerMyc and confers a XhoI restriction site. Both CC1-and CC2-mutagenic primers insert a proline codon and a StuI restriction site (KL134RP, 5Ј-GAATTGGAGAGGCCTAAATCCAGCTATCG-3Ј; ML322RP, 5Ј-CAGACCCAGCAGAGGCCTTTACACAAGGAGGC-3Ј). The Fer-N expression construct (encoding the N-terminal 462 amino acids of Fer) was generated by digestion of pECE-Fer with KpnI, followed by ligation of the released 1.5-kilobase fragment into the KpnI site of pECE (30). Thus, the translation termination codon arises from vector sequences. The N-terminal deletion mutant Myc-Fer⌬N was constructed as follows. Polymerase chain reaction was carried out with a forward primer that conferred a BglII site and a Kozak consensus translation initiation site at methionine 412 (5Ј-GAAGATCTACCATG-GAAAGAAAGGAGAGG-3Ј) and a reverse primer (5Ј-CTTGCAGCTTT-TAATGGCAACAGG-3Ј). Both the resulting polymerase chain reaction product and the plasmid pESNmFerMyc were digested with BglII and KpnI (the latter digestion released a 1.5-kilobase fragment encoding the N-terminal CC domain of Fer), and following gel purification were ligated together. The resulting plasmid allows for initiation of translation at methionine 412 and the production of an N-terminally truncated Fer protein (Myc-Fer⌬N) that resembles the testis-specific FerT.
Fps protein was expressed in COS-1 cells using the expression plasmid pEF4, which was generated by the insertion of an EcoRI fragment encoding the human fps cDNA from pLF5.13 (33) into the EcoRI site of pECE (30). The expression plasmid for Myc-Fps was constructed by replacing the green fluorescence protein-encoding portion of Fps-green fluorescence protein (20) with six copies of the epitope recognized by the anti-Myc monoclonal antibody 1-9E10, as described above for Fer.
Transfection-COS-1 cells were seeded onto 60-mm tissue culture dishes (Falcon) at approximately 25% confluence for transfections the following day. DNA-calcium phosphate precipitates were prepared in 4-ml polystyrene tubes by mixing plasmid DNA (see figure legends for amounts) in 600 l of 0.25 M CaCl 2 with 600 l of 2ϫ BBS solution (50 mM N,N-bis(2-hydroxyethyl)-2-aminoethanesulfonic acid, 280 mM NaCl, 1.5 mM Na 2 HPO 4 , pH 7.06). DNA precipitates were allowed to form for 60 min at room temperature and then resuspended and added dropwise to cells cultured in 4 ml of Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. After overnight incubation, the medium was replaced, and cells were cultured an additional 24 h before harvesting.
Immune Complex Kinase Assays-Cells were washed with ice-cold TBS-V (10 mM Tris-HCl (pH 7.5), 150 mM NaCl, 100 M vanadate) and harvested by scraping into KLB (20 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1% (v/v) Nonidet P-40, 0.5% (w/v) sodium deoxycholic acid, 10 g/ml aprotinin, 10 g/ml leupeptin, 100 M sodium orthovanadate, 100 M phenylmethylsulfonyl fluoride). Cell lysates were clarified by centrifugation at 14,000 ϫ g for 10 min at 4°C. A fraction of the soluble cell lysates (ϳ15 g) was retained for direct loading, while the remainder (ϳ300 g) was added to 25 l of 50% (v/v) GammaBind-Sepharose (Amersham Pharmacia Biotech), which had been preincubated with 200 l of 1-9E10 anti-Myc hybridoma culture supernatant (34). After mixing on a nutator platform for 2 h at 4°C, immune complexes were collected by brief centrifugation, and washed five times with KLB followed by one wash with KRB (20 mM Tris-HCl (pH 7.5), 10 mM MnCl 2 , 100 M sodium orthovanadate). Kinase reactions were performed by resuspending the washed immune complex with 30 l of KRB supplemented with 10 Ci of [␥-32 P]ATP and incubating for 20 min at 30°C. For kinase assays in which enolase was added, enolase was first denatured as follows. For 10 samples, 50 g of enolase was diluted to 25 l in 25 mM Hepes-KOH (pH 7), 1 mM dithiothreitol, 1 mM MnCl 2 , 50% glycerol and mixed with 25 l of 50 mM acetic acid. Following incubation at 30°C for 15 min, denatured enolase was diluted with 25 l of water and 25 l of 50 mM acetic acid. Immune complexes were resuspended in 20 l of KRB-E (100 mM Hepes-KOH (pH 7.5), 20 mM MnCl 2 ) containing 10 Ci of [␥-32 P]ATP, and 10 l of denatured enolase. The reactions were terminated by the addition of 30 l of 2ϫ SDS sample buffer and heating at 100°C for 5 min. Proteins were then resolved on SDS-PAGE gels and either transferred to Immobilon-P membrane (Millipore Corp.) using a semidry apparatus (Bio-Rad) or incubated in 1 M KOH at 50°C for 2 h and then stained with Coomassie Blue and dried on a gel dryer (Bio-Rad). Kinase activity was determined on both the dried gel and membrane by autoradiography. Quantification was performed on an Instant Imager (Packard).
Western Blotting-Fer proteins were detected by incubation at room temperature for 1-2 h with 1:500 dilutions of rabbit polyclonal anti-FerLA antiserum (raised against a glutathione S-transferase fusion protein containing murine Fer amino acids Leu 97 to Ala 382 (20)). To detect both Fer and Fps, a 1:500 dilution of anti-Fps/Fer antiserum was used (anti-Fps/Fer, also known as anti-FpsQE was raised against a glutathione S-transferase fusion protein containing murine Fps amino acids Gln 381 to Glu 563 and is cross-reactive to both Fps and Fer (20)). Myc-tagged Fer proteins were detected with 1:20 dilutions of 1-9E10 hybridoma culture supernatants. Anti-phosphotyrosine blotting was carried out with 1:1000 dilutions of monoclonal antibody PY99 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). After washing with TBS-T (10 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.5% (v/v) Tween-20), membranes were incubated with a 1:10,000 dilution of horseradish peroxidase-conjugated goat anti-rabbit IgG (for anti-Fer or anti-Fps/Fer) or goat anti-mouse IgG (for 1-9E10 and PY99) secondary antibodies (Amersham Pharmacia Biotech) in TBS-T for 1 h at room temperature. After washing with TBS-T, immune complexes were detected using enhanced chemiluminescence (NEN Life Science Products). Before reprobing of membranes with other antibodies, they were stripped by incubation in 0.2 N NaOH for 5 min followed by washing in TBS-T.
Chemical Cross-linking-Transfected COS-1 cells were harvested in X-linking buffer (20 mM Hepes-KOH (pH 7.5), 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate) and clarified by centrifugation at 14,000 rpm for 10 min at 4°C. Extracts were incubated on ice in the absence or presence of 1 mM ethylene glycol-bis(succinic acid N-hydroxysuccinimide ester) (EGS) for 15 min (EGS from Sigma was prepared as a 100 mM stock solution in Me 2 SO). The reactions were terminated by the addition of 1 l of 50ϫ TE (500 mM Tris-HCl (pH 8), 50 mM EDTA) and 50 l of SDS-loading buffer. Samples were heated at 100°C for 4 min, followed by SDS-PAGE, and prepared for Western blotting as described above.
Tryptic Phosphopeptide Mapping-COS-1 cells were transfected and harvested, and immune complex kinase assays performed as described above. The sources of native Fer and FerT, were mouse liver and testis, respectively, which were homogenized in KLB. Fer was immunoprecipitated with anti-Fer polyclonal antiserum (5 l), while FerT was immunoprecipitated with anti-Fps/Fer polyclonal antiserum (5 l), and Myc epitope-tagged proteins were immunoprecipitated with monoclonal antibody 1-9E10. Following the kinase reactions, samples were subjected to SDS-PAGE and transferred to a polyvinylidene difluoride membrane, and autoradiography was performed. Portions of the membrane containing the radiolabeled proteins were excised, placed in separate 1.5-ml tubes containing 1 ml of 0.5% polyvinyl pyrolidone-360 (in 100 mM acetic acid), and incubated for 30 min at 37°C. Following aspiration, the membranes were washed five times with 1 ml of water and once with 50 mM NH 4 HCO 3 . Ten micrograms of diphenylcarbamyl chloride-treated trypsin (Sigma) in 200 l of buffer (50 mM NH 4 HCO 3 (pH 8.3)) was added to each sample and incubated overnight at 37°C. Following vortexing, reactions were supplemented with 1 g of trypsin for 1 h at 37°C; this step was then repeated. Water was added (300 l), and samples were centrifuged at 10,000 rpm for 5 min and then lyophilized. Oxidation was performed as follows. Pellets were dissolved in 100 l of formic acid, and then 25 l of methanol and 40 l of performic acid (made fresh by combining 0.9 ml of formic acid with 0.1 ml of 30% H 2 O 2 for 1 h at room temperature) were added; samples were incubated at 0°C for 2 h. Each sample was divided into two 1.5-ml tubes, and 1.4 ml of water was added prior to lyophilization. Pellets were resuspended in 10 l of pH 1.9 buffer (50 ml of formic acid (88%, w/v), 156 ml of glacial acetic acid, 1794 ml of water) and spotted on 20 ϫ 20-cm cellulose TLC plates (Selecto Scientific) by the repeated addition of 0.5 l. TLC plates were electrophoresed using a Hunter Thin Layer Electrophoresis System (C.B.S. Scientific Co.) in pH 1.9 buffer for 25 min at 1000 V. Following evaporation of the buffer, plates were rotated 90°and subjected to chromatography in phosphochromatography buffer (150 ml of n-butanol, 100 ml of pyridine, 30 ml of glacial acetic acid, 120 ml of water). Phosphopeptides were visualized by autoradiography.

RESULTS
Fer is a 94-kDa protein that contains three predicted Nterminal CC motifs (CC1, CC2, and CC3), an SH2 domain, and a C-terminal protein-tyrosine kinase domain (Ref. 6; shown schematically in Fig. 1). In order to explore the role of the SH2 and N-terminal domains in Fer kinase regulation, we generated a series of expression constructs including two separate missense mutations within either catalytic subdomain II (K592R) or subdomain IX (D743R); mutations at either of these highly conserved residues completely abolished Fer kinase activity (31). Also, wild-type and mutant Fer proteins were expressed with a C-terminal Myc epitope fusion (Myc-Fer), which encodes six copies of the epitope recognized by the anti-Myc monoclonal antibody 1-9E10 (34). To assess the role of the N terminus of Fer in oligomerization, two missense mutations were engineered that position a proline residue within either CC1 (KL134RP) or CC2 (ML322RP). The MultiCoil program (35) predicts mostly trimeric interactions for CC1 and CC2 but no multicoil interactions for CC3; the two mutations described above are predicted to abolish formation of multiple CC interactions. Two truncated proteins were also expressed including a C-terminal deletion of the SH2 and kinase domains that retains the CC domains (Fer-N) and an N-terminal deletion that encodes a Myc epitope-tagged protein containing only the SH2 and kinase domains of Fer (Myc-Fer⌬N). This truncated protein resembles the FerT protein found in testis that arises from usage of an internal testis-specific promoter located in fer (22).
Fps is a 92-kDa protein that is closely related to Fer and possesses a similar domain structure (1,8). The N-terminal CC domain of human Fps exhibits 37 and 35% identity to human and mouse Fer, respectively, while the SH2 and kinase domains of Fps and Fer display 67 and 68% sequence identity, respectively (6). Fps was expressed both as an untagged protein and with a C-terminal Myc epitope fusion (Myc-Fps).
Homotypic but Not Heterotypic Oligomerization of Fer and Fps in Vivo-Since both Fps and Fer possess N-terminal CC domains that are implicated in oligomerization in vitro (7,8), we wished to determine whether Fps and Fer form homotypic or heterotypic oligomers in vivo. COS-1 cells were transfected with plasmids expressing Fer, Fps, Myc-Fer, and Myc-Fps. At 48 h post-transfection, soluble cell lysates were collected; a fraction of each lysate was retained for Western blotting, while immunoprecipitations with anti-Myc were performed on the remainder of the lysates. Western blotting with antiserum that recognizes both Fps and Fer (anti-Fps/Fer), and in vitro kinase assays allowed for detection of Fps and Fer in the immunoprecipitates (Fig. 2). Western blotting of soluble cell lysates from mock transfected COS-1 cells identified the endogenous simian Fer protein (top panel, lane 1), which co-migrated with the overexpressed murine Fer protein (lane 2). Fps protein migrated slightly faster than Fer (compare lanes 3 and 2), consistent with its slightly smaller size. The 120-kDa Myc-Fer protein was detected when co-expressed with either Fer (lane  7), which is consistent with a previous study (8). Immune complex kinase assays were also performed, and the results were consistent with those described above. No kinase activities were observed in anti-Myc immunoprecipitates from mock transfected cell lysate (Fig. 2, bottom panel, lane 1 . Therefore, Fer and Fps form homotypic but not heterotypic oligomers in vivo. It is also noteworthy that the amount of 32 P incorporated into Fps was lower than that for Fer (compare lanes 4 and 5 with lanes 6 and 7). Fer is also more highly phosphorylated than Fps in vivo, as determined by anti-phosphotyrosine blotting (data not shown), suggesting either that Fps has a lower specific activity than Fer or that Fer has more autophosphorylation sites than Fps.
Disruption of Coiled-coil Domains Abolishes Oligomerization but Not Autophosphorylation-Next, we wished to address the role of oligomerization in activation of Fer. COS-1 cells expressing various Fer proteins (Fig. 3,  Thus, disruption of either CC1 or CC2 domains completely abrogates co-immunoprecipitation of Fer, suggesting that both ␣-helical domains participate in oligomerization. Kinase activity or phosphorylation is not required for oligomerization, since we have observed strong association between Myc-Fer K592R and Fer D743R in similar experiments (data not shown).
Immune complex kinase assays were performed in the presence of denatured enolase, which is an in vitro substrate of Fer. Myc-Fer autophosphorylation (Fig. 3, bottom panel, lane 3) was similar when complexed with either Fer (lane 4) or Fer-N (lane 5). In contrast, enolase phosphorylation was higher in the presence of two active kinases (Myc-Fer and Fer) compared with Myc-Fer alone (lanes 3 and 4). phorylation was slightly reduced (30 -50%). Therefore, oligomerization is not a prerequisite for Fer autophosphorylation in vitro, but it can potentiate phosphorylation in trans (i.e. with one subunit contributing to the phosphorylation of another).
Oligomerization-independent Autophosphorylation of Fer in Vivo-To assess the state of autophosphorylation of oligomeric and monomeric Fer proteins in vivo, Myc-tagged Fer proteins expressed in COS-1 cells were immunoprecipitated with anti-Myc monoclonal antibody 1-9E10, followed by Western blotting with an anti-phosphotyrosine monoclonal antibody (Fig. 4, top  panel). Myc-Fer displayed higher phosphotyrosine levels than Myc-FerKL134RP and Myc-FerML322RP (compare lanes 2-4); however, this coincided with reduced protein levels for the CC mutants (bottom panel, compare lanes 2-4). Likewise, the truncated protein Myc-Fer⌬N, which lacks all CC domains, was tyrosine-phosphorylated in vivo (top panel, lane 5). Thus, oligomerization is not obligatory for autophosphorylation of Fer in vivo.
Fer Protein Forms Trimers in Vivo-To address the stoichiometry of Fer oligomerization, chemical cross-linking was performed on lysates from COS-1 cells expressing various Fer proteins (Fig. 5, indicated at the top). Soluble cell lysates were either loaded directly (lanes 1-7) or following incubation with the cross-linking reagent EGS (lanes 8 -14). Following SDS-PAGE, Western blotting was performed sequentially with anti-Fer (top panel) and anti-Myc antibodies (bottom panel). Fer-N migrated as a ϳ65-kDa protein in the absence of EGS (top panel, lane 2), while upon cross-linking, a portion of the protein migrated at ϳ190 kDa (lane 9, indicated by the arrow at the right). An additional ϳ170-kDa complex was present in lower amounts and may be due to partial proteolysis of Fer-N. Taken together, these results indicate that the N-terminal domain of Fer forms a trimer in vivo and are consistent with a previous in vitro study (7). Cross-linking of Myc-Fer resulted in a shift in mobility from ϳ120 kDa (lanes 3 and 4), to ϳ350 kDa (lanes 10 and 11, indicated by the upper arrow at the right). In the presence of both Myc-Fer and Fer-N, three additional crosslinked species were observed (lane 11), including the aforementioned ϳ190-kDa product (probably arising from trimers of Fer-N) and two species migrating slightly faster than the ϳ350-kDa product (indicated by two arrows at the right). The latter probably correspond to incorporation of either one or two Fer-N subunits in the Myc-Fer oligomers. In agreement with the results described above, disruption of either CC1 or CC2 abolished oligomerization (compare lanes 5 and 6 with lanes 12  and 13). Taken together, these results indicate that both CC1 and CC2 domains mediate trimerization of Fer in vivo. These results do not exclude additional protein-protein interactions, for example via the SH2 domain and phosphotyrosine residues. We addressed this issue using an N-terminal deletion construct (Myc-Fer⌬N), which retains the SH2 and kinase domains of Fer and resembles the testis-specific FerT (22). Myc-Fer⌬N was not recognized with anti-Fer antibodies (Fig. 5, upper  panel, lane 7) because these were raised against the missing N-terminal portion of Fer; however, anti-Myc antibody detected the ϳ65-kDa Myc-Fer⌬N protein (bottom panel, lane 7). No difference in mobility was observed following the incubation of Myc -Fer⌬N with EGS (lane 14), suggesting that Myc-Fer⌬N is monomeric although it is phosphorylated in vivo, as determined by anti-phosphotyrosine blotting (Fig. 4, lane 5). Furthermore, co-expression of wild type Fer with Myc-Fer⌬N, followed by immune complex kinase assays (using anti-Myc monoclonal antibody), indicated that Fer and Myc-Fer⌬N do not associate (data not shown). Therefore, oligomerization of Fer occurs exclusively via the N-terminal CC domains.
Do Inactive Fer Proteins Have Dominant Negative Activity?-We next wished to address the role of oligomerization on Fer activity and the effect of inactive Fer proteins on the activity of associated wild type Fer. COS-1 cells were transfected with various Fer expression plasmids (Fig. 6, as indicated at the top), and lysates were prepared for immunoprecipitation with anti-Myc antibody, followed by either Western blotting or immune complex kinase assays. Western blotting of soluble cell lysates with anti-Fer or anti-Myc antibodies revealed that all Fer proteins were expressed (first and second panels, lanes [2][3][4][5][6][7][8][9][10][11][12][13][14]. Western blotting of the immunoprecipitates (third and fourth panels) revealed increasing amounts of Myc-Fer when expressed alone (lanes 4 -7) or when co-expressed with either Fer-N (lanes 8 -10) or Fer D743R (lanes [11][12][13]. Myc-Fer⌬N was readily detected in the immunoprecipitate (lane 14, fourth panel).
Immune complex kinase assays were performed in the presence of denatured enolase (bottom panel; phosphorylated enolase is indicated on the right). Specific activities of Myc-Fer and Myc-Fer⌬N were calculated for both autophosphorylation and substrate phosphorylation via quantification of incorporated 32 P in Fer and enolase proteins relative to the amounts of Myc-Fer and Myc-Fer⌬N. The specific activity of Myc-Fer when expressed alone (lanes 4 -7) was 27.4 Ϯ 5.9 for autophosphorylation and 20.0 Ϯ 2.7 for enolase phosphorylation (the error given is the S.D.). When in a complex with Fer-N (lanes 8 -10), Myc-Fer autophosphorylation was 19.6 Ϯ 5.1 and enolase phosphorylation was 14.1 Ϯ 4.5. In the presence of Fer D743R , autophosphorylation of Myc-Fer was 14.5 Ϯ 4.5, and enolase phosphorylation was 13.5 Ϯ 3.8. Therefore, the specific activity of Myc-Fer was reduced slightly (25-40%) when complexed with inactive Fer proteins. The N-terminal deletion mutant Myc-Fer⌬N had substantially lower specific activity for autophosphorylation than Myc-Fer (4-fold) but showed much higher enolase phosphorylation (2.5-fold). This would suggest that although Myc-Fer⌬N does not oligomerize (Fig. 5, and data not shown), it retains robust kinase activity toward substrate in vitro. However, the greatly reduced autophosphorylation signal suggests that efficient autophosphorylation of Fer may require oligomerization.
Analysis of Autophosphorylation Sites within Monomeric and Oligomeric Fer Proteins-To identify the number of autophosphorylation sites in Fer, and the effect of oligomerization on Fer phosphorylation, tryptic phosphopeptide mapping was performed. Soluble cell lysates from COS-1 cells expressing Fer or co-expressing Myc-Fer K592R and Fer were subjected to immunoprecipitation with anti-Fer antibody, followed by in vitro kinase reactions. Myc epitope-tagged proteins were immunoprecipitated with anti-Myc antibody, followed by in vitro kinase reactions. The radiolabeled products were subjected to SDS-PAGE, transferred to a polyvinylidene difluoride membrane, and visualized by autoradiography. Portions of the membrane corresponding to autophosphorylated Fer, Myc-Fer KL134RP , Myc-Fer ML322RP , Myc-Fer⌬N, and Myc-Fer K592R (which was phosphorylated in trans by Fer), were excised and digested with trypsin. Peptides that were released from the membrane were oxidized and then spotted on cellulose plates for electrophoresis followed by thin layer chromatography (Fig. 7). Subsequent autoradiography resulted in the identification of three tryptic phosphopeptides for Fer (Fig. 7A). The peptide corresponding to spot 1 was most highly labeled, followed by spots 2 and 3. Three major tryptic phosphopeptides were also described for Fps (36), although the mobilities of the phosphopeptides are distinct from those observed here for Fer. The ability of Fer to phosphorylate an inactive Fer protein in trans (as shown in Fig. 3, lane 6) allowed for analysis of the sites within Myc-Fer K592R that were phosphorylated by Fer in trans (Fig.  7B). Again, three major phosphopeptides were detected that co-migrated with those described for Fer (labeled 1-3). However, the stoichiometry of phosphorylation was quite different, in that spot 1, which corresponds to the major phosphopeptide in Fer (Fig. 7A), became the minor phosphopeptide in Myc-Fer K592R (Fig. 7B). The monomeric Myc-Fer KL134RP and Myc-Fer ML322RP mutants also displayed all three tryptic phosphopeptides; however, the stoichiometry was distinct from wild type Fer in that spot 1 was the minor phosphopeptide (Fig. 7, C  and D, respectively). This would suggest that all of the major autophosphorylation sites in Fer can be phosphorylated in cis but that spot 1 is less efficiently phosphorylated in the monomeric mutants. Since Myc-Fer⌬N does not oligomerize (Fig. 5,  lane 14) and displayed reduced autophosphorylation (Fig. 6, lane 14), some differences in the phosphorylation state were expected. Indeed, the tryptic phosphopeptide map of Myc-Fer⌬N revealed only two major phosphorylation sites, which correspond to spots 2 and 3 of Fer (Fig. 7E). In contrast, there was no evidence of spot 1. The loss of this phosphorylation site might indicate that an N-terminal phosphorylation site is lost with this deletion; however, Fer-N was not phosphorylated when complexed with Fer (Fig. 3, lane 5, and Fig. 6, lanes  8 -10). We are currently attempting to identify the autophosphorylation site that corresponds to spot 1.
To determine whether the phosphorylation sites in Fer protein overexpressed in COS-1 cells were similar to that of native Fer, tryptic phosphopeptide mapping was performed on Fer that was immunoprecipitated from mouse liver extracts using anti-Fer polyclonal antiserum and labeled in vitro (Fig. 7F). Three major phosphopeptides were observed with mobility similar to that described for Fer expressed in COS-1 cells (Fig. 7A). We performed a similar analysis of native FerT, which was immunoprecipitated from mouse testis (Fig. 7G). Two major phosphopeptides were observed corresponding to spots 2 and 3. However, no phosphopeptide corresponding to spot 1 was detected. Since autophosphorylation of Myc-Fer⌬N resembled that of native FerT (Fig. 7, compare E and G), the unique N-terminal sequence of FerT, which is not present in Myc-Fer⌬N, probably plays no role in kinase activation. Overall, these results suggest that oligomerization promotes kinase activation by potentiating autophosphorylation in trans. However, since monomeric Fer mutants (or the naturally occurring FerT) can autophosphorylate in cis, oligomerization is clearly not required for kinase activation. Finally, the differences in phosphorylation sites between Fer and FerT may reflect distinct roles in signal transduction between the ubiquitous Fer and the testis-specific FerT. DISCUSSION The Fer protein-tyrosine kinase and the highly homologous oncoprotein Fps both contain an N-terminal domain with three predicted CC motifs that are thought to promote oligomerization (7,8). In this study, we have examined the requirements of these CC motifs for oligomerization and Fer kinase activation.
Our results show that both CC1 and CC2 are required and probably cooperate in trimerization of Fer in vivo (Figs. 3 and 5). It is also apparent that oligomerization is not required for autophosphorylation in vitro (Figs. 3 and 6) or in vivo (Fig. 4). So what is the role of CC-mediated trimerization of Fer? It clearly distinguishes the ubiquitous p94 Fer from its testisspecific isoform FerT, which is monomeric. However, it would appear that activation of Fer, and probably Fps, differs from many protein kinases that require oligomerization for activation (37). Few cytoplasmic PTKs have been shown to oligomerize, with Fer and Fps being notable exceptions. Activation of most cytoplasmic PTKs involves interactions with oligomeric proteins such as activated growth factor, cytokine, or immune recognition receptors (37). These interactions may promote activation through autophosphorylation in trans and induced conformational changes (38,39). Src family kinases also possess a negative regulatory C-terminal tyrosine phosphorylation  site that must be dephosphorylated for kinase activation (40). While Fer and Fps do not undergo this mode of negative regulation, intramolecular interactions have been proposed between the SH2 domain and a phosphotyrosine residue in Fps (10,41). Preliminary evidence indicates that Fer is constitutively oligomeric and that growth factor stimulation has no effect on formation of trimers (data not shown). Another potential role for the CC domains of Fps and Fer may involve their subcellular localization. It will be of great interest to determine whether disruption of CC1 or CC2 causes aberrant localization of Fer. It is also worth noting that a number of serine/threonine kinases possess CC oligomerization domains including p160 ROCK (42), TOUSLED (43), and myosin heavy chain kinase A (44). Fer and Fps may have retained CC domains throughout evolution from their putative serine/threonine kinase ancestor. There are other examples of PTKs that oligomerize by virtue of chromosomal translocations that juxtapose oligomerization domains with tyrosine kinase domains of ABL (45,46), RET (47), MET (48), and TRKA (49). All of these fusion proteins are hyperactive kinases that possess oncogenic activities.
Although oligomerization is not required for kinase activation, it does promote autophosphorylation in trans (Fig. 3). The significance of this observation is not clear, since incorporation of inactive proteins into the trimer has only mild inhibitory effects on activation of wild type Fer (Figs. 3 and 6), which argues against a prominent role for trans phosphorylation. Also, the ability of monomeric Fer proteins to autophosphorylate (Figs. 3 and 4), suggests that an additional cis-mediated autophosphorylation mechanism probably exists. The relative contributions of these two mechanisms to Fer activation in vivo is not known and is difficult to address experimentally. Our findings contrast with previous results for Fps, in which a marked reduction in Fps activity was observed when Fps was immunoprecipitated with inactive Fps proteins (8). Also, a recent study suggested a dominant negative effect of overexpressed Fer K592R on the activity of endogenous Fer (13). However, in both studies a molar excess of the inactive Fer or Fps proteins were required for this effect, whereas in our experiments the amounts of Fer D743R or Fer-N were closer to equimolar with wild type Fer (Fig. 3, lane 5, and Fig. 6, lanes 8 -13). In contrast, dominant negative mutations were identified in the Kit PTK that produce a marked reduction in pigmentation in heterozygous mice compared with that in heterozygotes with a null allele (50). Therefore, a true dominant negative effect such as that described for Kit is not expected for Fer based on our biochemical analysis of Fer. To further elucidate the role of Fer in vivo, we have generated mice harboring the D743R inactivating mutation. 2 The heterozygous mice are being interbred to determine whether Fer kinase activity is essential for development.
Despite similar predictions of trimeric interactions for Fps and Fer using the program MultiCoil (35), we found no evidence for heterotypic interactions between Fer and Fps (Fig. 2). However, Fps and Fer do form homotypic oligomers in vivo, consistent with previous in vitro results (7,8). The close structural relationship between Fps and Fer suggests that these two kinases may engage in similar biochemical functions. However, our results indicate that if there is any redundancy in function between Fer and Fps, it probably occurs downstream of oligomerization and kinase activation. The observed association of Fps and Fer with a variety of different cytokine and growth factor receptors suggests these kinases may serve a general role in cell signaling and that Fps may have evolved more recently to provide this function in more specialized cell types, such as those of the myeloid lineages or the vascular endothelium. We have recently found that mice lacking Fps kinase are viable with only mild effects in hematopoietic lineages. 3,4 This raises the possibility that Fps and Fer may provide redundant biochemical functions in tissues with overlapping expression. The apparent higher specific activity of Fer compared with Fps that we observed (Fig. 2) may allow Fer to effectively compensate for a loss of Fps activity. This higher activity of Fer compared with Fps may also explain its ability to cause malignant transformation of cells (24). In contrast, Fps requires the addition of a myristoylation sequence to efficiently cause cell transformation (36). Compound knock-outs of fer and fps will allow us to address the potential redundancy of function between these highly related kinases.
The observation that oligomeric Fer and monomeric FerT differ in one autophosphorylation site in vitro (Fig. 7, A, F, and  G, spot 1), suggests that this site may be located in the Nterminal region. This autophosphorylated residue may provide a binding site for SH2-or phosphotyrosine-binding domaincontaining proteins. Therefore, differences in the number of autophosphorylated tyrosine residues and the absence of the N-terminal CC domain in FerT imply potentially distinct interactions with effectors or substrates between the ubiquitous Fer and its testis-specific counterpart FerT. This may result in differences in signals transduced by these two isoforms. A strong candidate for one of the major phosphorylation sites in Fer is tyrosine 715, which is homologous to the major autophosphorylated residue in Fps, tyrosine 713 (51), which is located in the activation loop of the kinase (52). Mutation of the homologous residue in v-Fps (tyrosine 1073) indicates that phosphorylation of this residue regulates the phosphoryl transfer reaction but does not greatly affect accessibility of the active site (53). We have preliminary evidence that mutation of tyrosine 715 in Fer reduces autophosphorylation in vitro and in vivo and also impairs substrate phosphorylation. 5 We are currently attempting to identify all of the autophosphorylated residues in Fer, which will be of utmost importance in delineating binding sites for substrates and/or effectors of Fer.