Cysteines 153 and 154 of transmembrane transforming growth factor-alpha are palmitoylated and mediate cytoplasmic protein association.

Transforming growth factor-α (TGF-α) is synthesized as a transmembrane protein with a highly conserved, short cytoplasmic domain that is rich in cysteines. TGF-α is a prototype of a large family of growth factors involved in cell-cell communication. We have shown previously that transmembrane TGF-α associates with a kinase activity and two proteins of 106 and 86 kDa. In this study, we have used site-directed mutagenesis of the cytoplasmic domain of TGF-α to define the structural requirements for these protein interactions. Whereas the cytoplasmic domain of TGF-α was not essential for association with transmembrane p106, deletion of the C-terminal 8 amino acids, including a cysteine pair, abolished the interaction with p86 and greatly reduced the kinase activity associated with transmembrane TGF-α. Replacement of these 2 cysteines by serines similarly reduced the association of p86 with transmembrane TGF-α. Using a combination of mutational analysis and direct microsequencing, we have determined that this cysteine pair was palmitoylated. We therefore conclude that these cysteines play a critical role in the interaction of TGF-α with associated proteins and in the function of this protein complex. The palmitoylation of these cysteines suggests a possibly dynamic role of fatty acid modification in the integrity and function of the transmembrane TGF-α complex.

Transforming growth factor-␣ (TGF-␣) is synthesized as a transmembrane protein with a highly conserved, short cytoplasmic domain that is rich in cysteines. TGF-␣ is a prototype of a large family of growth factors involved in cell-cell communication. We have shown previously that transmembrane TGF-␣ associates with a kinase activity and two proteins of 106 and 86 kDa. In this study, we have used site-directed mutagenesis of the cytoplasmic domain of TGF-␣ to define the structural requirements for these protein interactions. Whereas the cytoplasmic domain of TGF-␣ was not essential for association with transmembrane p106, deletion of the C-terminal 8 amino acids, including a cysteine pair, abolished the interaction with p86 and greatly reduced the kinase activity associated with transmembrane TGF-␣. Replacement of these 2 cysteines by serines similarly reduced the association of p86 with transmembrane TGF-␣. Using a combination of mutational analysis and direct microsequencing, we have determined that this cysteine pair was palmitoylated. We therefore conclude that these cysteines play a critical role in the interaction of TGF-␣ with associated proteins and in the function of this protein complex. The palmitoylation of these cysteines suggests a possibly dynamic role of fatty acid modification in the integrity and function of the transmembrane TGF-␣ complex.
Transforming growth factor-␣ (TGF-␣) 1 is a mitogenic growth factor, which is secreted by a large variety of cells and is synthesized as a 160-amino acid transmembrane glycoprotein (1)(2)(3). The transmembrane form has a short extracellular domain that can be specifically cleaved to release the soluble factor. This factor, which contains a characteristic epidermal growth factor (EGF) motif consisting of 6 disulfide-bonded cysteines in a defined spacing pattern, interacts with the EGF/ TGF-␣ tyrosine kinase receptor and thus induces a signaling cascade culminating in mitosis and cell proliferation (2,4).
However, the transmembrane form often remains uncleaved and represents the predominant TGF-␣ form in various normal cell types (5). During cell-cell contact, this transmembrane form can also interact with the receptor and elicit receptor responses (6,7). The cytoplasmic domain of transmembrane TGF-␣ is only 39 amino acids long and lacks any enzymatic motif.
TGF-␣ is synthesized by a large variety of cells, primarily of ectodermal origin (8). Thus, most epithelia, including skin, gastrointestinal tract, and mammary epithelia, as well as several neuronal cell types and macrophages express TGF-␣ (for review, see Ref. 5 and references within), with skin as a major source of TGF-␣. Many tumor cell types, especially carcinomas, express TGF-␣ at levels that may be up-regulated in comparison with the normal cell types (9). TGF-␣ expression is also up-regulated in psoriasis of the skin (10) and Ménétrier's disease of the gastric mucosa (11,12).
TGF-␣ is considered as a prototype factor for a larger family of growth factors, which are all synthesized as transmembrane proteins and have the characteristic 6-cysteine motif in their extracellular domain. The first member of this family to be isolated was EGF, which has an extracellular domain that is much longer than that of the other family members and includes multiple cysteine-rich repeats (13)(14)(15)(16). This family also includes amphiregulin (17,18), heparin-binding EGF (HB-EGF) (19), betacellulin (20), and vaccinia virus growth factor (21), all of which bind to the EGF/TGF-␣ receptor. Finally, the differentially spliced forms of heregulins (22)(23)(24)(25), lin 3 in Caenorhabditis elegans (26), and spitz (27) and gurken (28) in Drosophila also belong to this family. Aside from the conserved 6-cysteine EGF motif in the extracellular domain and the general features of transmembrane proteins, the different members of the TGF-␣ family have no structural similarity and the sequences of their cytoplasmic domains are unrelated.
The 39-amino acid cytoplasmic domain of transmembrane TGF-␣ contains 7 cysteines and is highly conserved among species (1, 3, 29 -31), suggesting a defined biological function. As a result, transmembrane TGF-␣ may have additional functions aside from the release of soluble ligand to the receptor. In the case of HB-EGF, the transmembrane form allows regulation of receptor activation, since physical association of another transmembrane protein, CD9, with transmembrane HB-EGF potentiates the mitogenic effect of this growth factor (32). However, no functions have been defined for the cytoplasmic domains of the members of the TGF-␣ family. A possible function of the cytoplasmic domain of TGF-␣ might be its involvement in the cleavage of the extracellular domain. This is suggested by the decrease in cleavage efficiency of transmembrane TGF-␣ when the C-terminal 2 valines are deleted (33). Since the cyto-plasmic domain of TGF-␣, or of any other members of the TGF-␣ family, does not contain any known functional motifs, interacting proteins might be instrumental in any activity of the cytoplasmic domain. In a previous study, we identified two proteins that associate with transmembrane TGF-␣ (34). One of these proteins has a molecular mass of 106 kDa and is a tyrosine-phosphorylated transmembrane protein. It associates with both the full-length and the cytoplasmic truncated forms of transmembrane TGF-␣, suggesting that the cytoplasmic domain of TGF-␣ is not essential for p106 association. We also detected an 86-kDa cytoplasmic protein that interacts with the cytoplasmic domain of TGF-␣. Deletion of the 31-amino acid C-terminal sequence abolishes the interaction of p86 with transmembrane TGF-␣. Finally, the protein complex that associates with transmembrane TGF-␣ displays kinase activity toward exogenous substrates and requires the presence of the cytoplasmic domain of transmembrane TGF-␣. The sequences required for the interaction of the cytoplasmic domain of TGF-␣ with p86 or the kinase activity are unknown. However, the cysteine-rich nature of this cytoplasmic domain suggests a possible involvement of cysteines similar to the cysteine-mediated interaction of p56 lck with the short cytoplasmic domains of CD4 and CD8 (35)(36)(37).
We have shown previously that the cytoplasmic domain of transmembrane TGF-␣ is post-translationally modified by the addition of palmitate, a 16-carbon fatty acid chain (38). Palmitoylation is a dynamic process that results in modification of a wide variety of membrane and cytoplasmic proteins participating in signal transduction pathways (see, for review, Refs. 39 -41). Palmitoylation may serve to target proteins toward the plasma membrane as in the case of Ras (42,43), or regulate protein function by mediating membrane-cytosol translocation of proteins via palmitoylation-depalmitoylation cycles as in the case of G ␣ (44). This modification most commonly occurs on cysteines (see, for review, Refs. 39, 45, and 46), although it has also been found on serine (47), threonine (48), and lysine residues (49). The cysteine residues in the cytoplasmic domain of TGF-␣ are potential sites of this type of lipid modification of transmembrane TGF-␣.
In the current study, we have used site-directed mutagenesis to define critical residues involved in the interaction of the cytoplasmic domain of transmembrane TGF-␣ and p86 as well as the kinase activity. Furthermore, we have also defined the sites of palmitoylation. Our results indicate the importance of a cysteine doublet in both palmitoylation and interaction with p86 and the kinase activity.
Cell Lines and Transfections-Chinese hamster ovary (CHO) cells were cultured in F-12 Ham's medium (Life Technologies) with 10% fetal calf serum (HyClone Laboratories, Logan, UT), 100 units/ml penicillin, and 100 g/ml streptomycin. Cotransfections of CHO cells with a TGF-␣ expression plasmid and pRSV-Neo, which confers neomycin resistance, were done using the calcium phosphate method (50). Transfected clones were selected and maintained in the presence of 400 g/ml G418 (Life Technologies, MD). TGF-␣ expression was assessed by immunofluorescence using the anti-TGF-␣ monoclonal antibody TGF-␣1 (38) as primary antibody and rhodamine-conjugated rabbit anti-mouse IgG (Jackson Immunoresearch Laboratories) as secondary antibody. At least two different isolated clones were evaluated in the study of each TGF-␣ mutant.
293 cells were maintained in Dulbecco's modified Eagle's H-16 medium (Life Technologies) with 10% fetal calf serum (HyClone), 100 units/ml penicillin, and 100 g/ml streptomycin. Transient transfection of 293 cells were performed using the calcium phosphate method. Assays were performed 48 -72 h after transfection.
Metabolic Labeling and Immunoprecipitation-To detect TGF-␣-associated proteins, cells were grown to 80% confluence and labeled overnight with 160 Ci/ml [ 35 S]cysteine/methionine protein labeling mix (Amersham Corp, IL) in cysteine-and methionine-free F-12 Ham's medium with 10% dialyzed fetal calf serum. On the next day, cells were washed twice with cold PBS and treated with 2 mM dithiobis(succinimidyl)propionate (DSP) (Pierce) in PBS for 30 min at 4°C. Cells were then washed again twice in Ca 2ϩ /Mg 2ϩ -free PBS, after which they were lysed for 30 min at 4°C in 50 mM Tris, pH 7.4, 100 mM NaCl, 2 mM EDTA, 1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, 5 g/ml aprotinin, and 5 g/ml leupeptin, sheared three times with a 25-gauge needle, and centrifuged for 30 min at 14,000 rpm at 4°C. The supernatant was then collected for immunoprecipitation using the anti-TGF-␣ monoclonal antibody TGF-␣1 and protein A-Sepharose (Pharmacia Biotech Inc.) as described (34). The proteins were analyzed by SDS-PAGE. The gels were fixed, enhanced, dried, and exposed for autoradiography using standard procedures.
To visualize palmitoylation, cells were labeled overnight with 1 mCi/ml [ 3 H]palmitic acid (Amersham) in medium containing 10% dialyzed fetal calf serum, 10% tryptose phosphate, 5 mM pyruvate, 1% Me 2 SO. On the next day, cells were washed twice with cold PBS, lysed in immunoprecipitation buffer, and subjected to immunoprecipitations using the TGF-␣1 monoclonal antibody. The samples were analyzed under non-reducing conditions. To assay for palmitoylation of cysteines, the hydroxylamine-sensitive thioester bonds were disrupted by washing the fixed gel twice in water for 10 min and incubation in 1 M hydroxylamine for 1 h. In control treated samples, the hydroxylamine treatment was replaced by incubation in 1 M Tris, pH 7.0. Following two additional washes in water, the gel was processed for autoradiography.
In Vitro Kinase Assay and Densitometry-For in vitro kinase assays, the immunoprecipitates linked to protein A-Sepharose beads were washed in cold kinase buffer (25 mM Hepes, pH 7.4, 10 mM MgCl 2 ) and then incubated with 100 ng/ml histone 2B substrate (Boehringer Mannheim) in kinase buffer containing 1 Ci of [␥-32 P]ATP (Amersham) for 30 min on ice. The reaction was terminated by adding 20 l of gel sample buffer and heated at 95°C for 5 min. The products were analyzed by SDS-PAGE and autoradiography.
To quantitate TGF-␣ expression and phosphorylation, densitometry was performed on autoradiograms using NIH Image 1.44 (National Institutes of Health, Bethesda, MD). Since TGF-␣ is processed into multiple species, its expression level was determined by summation of the density of the individual bands. The kinase activities were determined by determining the density of phosphorylated histone 2B, relative to the kinase activity associated with full-length TGF-␣.
Peptide Microsequencing-TGF-␣ immunoprecipitated from [ 3 H] palmitic acid-labeled cells was digested with 10 g of trypsin in 100 mM ammonium bicarbonate, 0.1% Tween 20 for 12 h at 37°C. The resulting peptides were applied to a C18 reversed phase column in an HPLC model LC 1090 system (Hewlett Packard, Palo Alto, CA) and eluted using a linear gradient of acetonitrile in 0.1% trifluoroacetic acid over 60 min. Fractions containing the radioactivity peaks were subjected to Edman degradation using a protein sequencer, model 492 (Perkin-Elmer). Each sequencing cycle was measured for release of radioactivity.

Association of p106 and p86 with the Cytoplasmic Domain of
Transmembrane TGF-␣-We have shown previously that two proteins, p106 and p86, associate with transmembrane TGF-␣ (34). Whereas p106 is a transmembrane protein, our inability to detect p86 by cell surface biotinylation suggests an intracellular localization. Both proteins could be chemically crosslinked using the reversible cross-linker DSP, and efficient coimmunoprecipitation of these proteins with transmembrane TGF-␣ required this stabilization.
To evaluate the involvement of specific sequences of amino acids of the cytoplasmic domain of TGF-␣ in the interaction with p106 and p86, we generated expression plasmids for a set of cytoplasmic truncations and mutations of transmembrane TGF-␣ ( Fig. 1). Several C-terminal deletions of increasing length were generated. Whereas ⌬122 had its entire cytoplasmic domain removed up to amino acid 122, ⌬158 had only the last 2 amino acids removed. The rationale behind the latter deletion was that it had been shown previously that removal of the C-terminal 2 valines of transmembrane TGF-␣ strongly reduced the proteolytic cleavage of the TGF-␣ ectodomain (33). In addition, two mutations focused specifically on a possible involvement of the cysteine pairs of the cytoplasmic domain of transmembrane TGF-␣ in the interaction with associated proteins. Indeed, previous studies have demonstrated the involvement of cysteine pairs in intracellular protein-protein associations, e.g. between CD4 or CD8 and p56 lck (35)(36)(37). Furthermore, the cytoplasmic domain of transmembrane TGF-␣ has a high cysteine content (i.e. 7 out of 39 amino acids) and all cysteines are conserved among many different species (1, 3, 29 -31), thereby suggesting biological significance.
Expression plasmids for the wild-type and mutated forms of transmembrane TGF-␣ were transfected into CHO cells, and stable clones expressing transmembrane TGF-␣ were established. Cells were selected with similar expression levels of transmembrane TGF-␣, as assessed by 35 S labeling and subsequent gel electrophoretic analysis. To further calibrate our analyses, we monitored the expression levels of transmembrane TGF-␣ by densitometry of the three forms of 35 S-labeled cell surface-associated TGF-␣. These three forms result from differential glycosylation and proteolysis of the ectodomain at the two cleavage sites that flank the prosegment (38). All three forms, which in the case of wild-type TGF-␣ migrate as 15-30-kDa species, were apparent for all cytoplasmically mutated forms of TGF-␣, albeit with different molecular weight values consistent with the extent of the deletions (Fig. 2). Finally, because of the large number of cysteines in the cytoplasmic domain which was the target of the mutagenesis, the relative densitometric values of TGF-␣ were corrected for the total number of cysteines to more accurately reflect the expression levels, since the incorporated 35 S radioactivity during metabolic labeling was on cysteine residues.
The recombinant cell lines expressing the various mutations of transmembrane TGF-␣ were subjected to chemical crosslinking on intact cells using DSP and immunoprecipitation with an anti-TGF-␣ monoclonal antibody (Fig. 3). This analysis has previously resulted in the identification of p106 and p86 as transmembrane TGF-␣-associated proteins (34). Accordingly, p106 and p86 were detected in association with wild-type transmembrane TGF-␣. Nonspecific protein bands were detected in these coimmunopricipitation analyses, but they are also found in control experiments (34). Furthermore, p106 was also associated with the ⌬122 mutant lacking the cytoplasmic domain as well as all other mutant forms of transmembrane TGF-␣. The level of detectable associated p106 was similar among the different forms of transmembrane TGF-␣. We therefore conclude that the cytoplasmic domain of transmembrane TGF-␣ is not required for association with p106, which is consistent with its nature as a transmembrane protein (38). However, this does not exclude a possible role of the cytoplasmic domain in stabilizing the association of these two transmembrane proteins, especially since the level of p106 was higher in the full-size than in cytoplasmic truncated form of transmembrane TGF-␣.
In contrast with p106, the association of p86 with transmembrane TGF-␣ depended on critical structural features of the cytoplasmic domain. p86 was apparent in immunoprecipitates of full-size TGF-␣, albeit at lower intensity than p106, but not in the ⌬122 mutant lacking the cytoplasmic domain of trans- The arrow to the right marks p106, which associates with the wild-type and all mutant forms of transmembrane TGF-␣. As verified by quantitative densitometry, p86 (arrowheads) was detectably associated with wild-type TGF-␣ and the ⌬154 and ⌬158 truncations. p86 was not detected in cells expressing ⌬122 and ⌬141 form of TGF-␣, and its level was greatly reduced to only 16% and 18% of control values in ⌬152 and ⌬CC160 cells. Parental CHO cells (lane 1) were negative for both p106 and p86. The positions of molecular size markers are shown on the left. membrane TGF-␣. In addition, p86 did not detectably associate with the ⌬141 and ⌬152 truncations of transmembrane TGF-␣ either, strongly suggesting an involvement of the C-terminal 8 amino acids in the association of p86 and transmembrane TGF-␣. In the ⌬158 mutant, the deletion of the 2 C-terminal valines did not affect the association of p86. In addition, p86 also detectably associated with the ⌬154 mutant of TGF-␣ which has the last 6 amino acids deleted, yet has the most distal cysteine pair retained. This association of p86 was apparent in spite of the fact that this transmembrane TGF-␣ mutant was expressed at only 25% of wild-type TGF-␣. Taken together, the results on the association of p86 with transmembrane TGF-␣ support the notion that the cysteine pair at positions 153 and 154 might play a critical role in the interaction of p86 with transmembrane TGF-␣, although the 4 amino acids distal from this cysteine pair might stabilize this interaction. To test this hypothesis, we generated the TGF-␣ mutant ⌬CC160, which is identical to the full-size, 160-amino acid TGF-␣, except that the distal cysteine pair was replaced by 2 serines. As is apparent from Fig. 3, p86 was not or was barely detectably associated with transmembrane TGF-␣. We therefore conclude that this cysteine pair is necessary for the interaction of p86 with transmembrane TGF-␣.
Association of a Kinase Activity with the Cytoplasmic Domain of Transmembrane TGF-␣-We have previously demonstrated the association of a kinase activity with transmembrane TGF-␣. This kinase activity, which results in substrate phosphorylation on serine, threonine, and tyrosine, was detected by immunoprecipitating a cross-linked complex of transmembrane TGF-␣ with its associated proteins, followed by an in vitro kinase reaction using an exogenous substrate such as histone 2B, enolase, and casein. Using the panel of mutants of the cytoplasmic domain of TGF-␣, described above, we evaluated whether the association of the kinase activity could be attributed to the presence of specific sequences in the cytoplasmic domain. Similar to the original assay (34), the wild type and mutant transmembrane TGF-␣ forms were, following chemical cross-linking with DSP, immunoprecipitated from the transfected CHO cells and assayed for in vitro kinase activity using histone 2B as substrate (Fig. 4). As with the association of p86 with TGF-␣, the kinase activities measured as the intensity of phosphorylated histone 2B were densitometrically normalized for expression levels of the different forms of transmembrane TGF-␣. Furthermore, each kinase experiment was carried out using two stably transfected CHO cell lines expressing the same mutant transmembrane TGF-␣ (Fig. 5). Whereas wild type transmembrane TGF-␣ had a considerable kinase activity associated with it, deletion of most of the cytoplasmic domain, as in the ⌬122 mutant, resulted in a kinase level which was only marginally above the nonspecific background level of untransfected CHO cells. Thus, the cytoplasmic domain is required for efficient association of the kinase activity with transmembrane TGF-␣. Furthermore, the normalized level of kinase activity associated with the ⌬152 mutant was significantly higher than with the ⌬122 mutant (p Ͻ 0.05 by t test), whereas the ⌬141 value was also clearly higher. In addition, the kinase activities of all three truncation mutants were significantly lower than with full-size transmembrane TGF-␣. In contrast, the kinase activity associated with the ⌬154 and ⌬158 mutants were similar to wild type TGF-␣. To evaluate the possible involvement of the distal cysteine pair in the association of the kinase activity with TGF-␣, we also tested the ⌬CC160 mutant in the same assay. No significant decrease of the kinase activity was found in association with this mutant form, when compared with wild type transmembrane TGF-␣. Taken together, our data indicate that the cytoplasmic domain is required for association of the kinase activity and that the efficiency of this association is considerably lower when the C-terminal 8 amino acids are deleted. Addition of the cysteine pair following amino acid 152, thus resulting in the ⌬154 mutant, increases the efficiency of kinase association but replacement of these cysteines in ⌬CC160 argues against an essential role of these 2 cysteines in this association. Nevertheless, these cysteines may play a secondary role in the conformation of the cytoplasmic domain of TGF-␣, since there was a trend in increased association of the kinase activity with decreased length of the Cterminal truncation. Even though the difference in associated kinase activity between each individual step of truncation from ⌬152 to ⌬122 was not statistically different (p Ͻ 0.05), our data nevertheless indicate an involvement of the C-terminal segment, especially the last 8 amino acids, in mediating and stabilizing this association.
Palmitoylation of the Distal Cysteine Pair at Positions 153 and 154 -We have shown previously that transmembrane TGF-␣ is post-translationally modified by palmitoylation (38), but the sites of palmitoylation were not determined. Palmitoylation occurs most commonly on cysteine residues, but also on threonine, serine, and lysine. Cysteine palmitoylation can be distinguished from the others by its sensitivity to hydroxylamine at pH7 (51). Accordingly, we have shown previously that TGF-␣ palmitoylation was sensitive to hydroxylamine treatment, indicating that cysteine residues were the target of this type of lipid modification (38). To define the sites of palmitoy- . The levels of kinase activity associated with the ⌬122, ⌬141, and ⌬152 TGF-␣ mutants were significantly reduced when compared with wild-type values, whereas those of the ⌬154, ⌬158, and ⌬CC160 TGF-␣ forms were comparable to controls. lation, we combined an analysis of the mutated transmembrane TGF-␣ forms with direct protein sequencing.
Stably transfected CHO cell lines that individually expressed the different TGF-␣ forms at comparable levels ( Fig. 2) were labeled with [ 3 H]palmitic acid and subjected to immunoprecipitation using the TGF-␣ monoclonal antibody (Fig. 6A). The three different transmembrane forms resulting from differential ectodomain processing of the full-size precursor incorporated [ 3 H]palmitate indicative of the palmitoylation of the cytoplasmic domain. In contrast, deletion of the cytoplasmic domain (⌬122 mutation) resulted in lack of [ 3 H]palmitate incorporation. Furthermore, the ⌬CC129 mutant with its cysteine pair at positions 123 and 124 also lacks palmitoylation, strongly suggesting lack of palmitoylation of these cysteines. Furthermore, the ⌬141, ⌬152, and ⌬154 mutant TGF-␣ forms had only a low level of [ 3 H]palmitate incorporation compared with the full-size cytoplasmic domain. These data suggest that the cysteines at positions 153 and 154 may be targets of palmitoylation, and that the downstream amino acids may be involved in the substrate recognition of the cysteine pair by palmitoyl transferase.
To further explore the palmitoylation of the 2 distal cysteines, we compared the palmitoylation of the full-size, wild type transmembrane TGF-␣ with the ⌬CC160 mutant in which the 2 cysteines are replaced by serines. Transiently transfected cells expressing similar levels of transmembrane TGF-␣ (Fig.  6B) were labeled using [ 3 H]palmitic acid. Immunoprecipitation of transmembrane TGF-␣ revealed the palmitoylation of the different wild type transmembrane TGF-␣ forms. The radiolabeling of the ⌬CC160 mutant was at a considerably lower level, suggesting that the cysteine pair is a major palmitoylation site. This background labeling of the ⌬CC160 mutant was not removed by hydroxylamine treatment, indicating that there was no palmitoylation of cysteines. In contrast, hydroxylamine treatment reduced the high level of [ 3 H]palmitate labeling of wild type transmembrane TGF-␣ to the same background level as in the ⌬CC160 mutant.
Finally, we pursued a direct radioactive microsequencing approach to confirm the sites of palmitoylation. Using this method, tryptic peptides derived from [ 3 H]palmitate-labeled transmembrane TGF-␣ were subjected to Edman degradation, whereby the pattern of 3 H label released with each sequencing cycle reveals the position of the palmitoylated amino acids. The predicted sequences of the expected tryptic peptides localized cysteines 130 and 133 at the second and fifth positions of peptide 1, cysteine 138 at the fourth position of peptide 2, and cysteines 153 and 154 in the third and fourth positions of peptide 3 (Table I). Finally, cysteines 123 and 124 should be located at positions 26 and 27 of a long tryptic peptide, thus making it difficult to localize them by Edman degradation. In three independent experiments, the radiolabeled, HPLC-purified tryptic peptides were subjected to radioactive sequencing, as shown in Table II. The first microsequencing cycle often yielded a higher level of radioactivity than subsequent cycles, a commonly encountered artifact of the procedure due to washoff of peptide. In addition, the third and fourth cycles consistently showed peaks of radioactivity. Accordingly, the radioactivity in these cycles confirm and are consistent with the palmitoylation of cysteines 153 and 154. The radioactivity in the fourth cycle could in principle also be due to labeling of cysteine at position 138, but this position is not palmitoylated based on our in vivo labeling results of TGF-␣ mutants. Therefore, taken together with our results from the different analyses, we conclude that the distal cysteine pair at positions 153 and 154 constitutes the major sites of palmitoylation of the cytoplasmic domain of transmembrane TGF-␣. DISCUSSION TGF-␣ is synthesized as a transmembrane precursor that can undergo subsequent proteolytic cleavage of the ectodomain, thus resulting in the release of soluble TGF-␣. Whereas the extent of this secretion of TGF-␣ is variable and the cell surface-associated proteolysis can be regulated, the transmembrane form is usually the most common form of TGF-␣. Its expression at the cell surface allows transmembrane TGF-␣ to stimulate the EGF/TGF-␣ receptor, thus allowing a paracrine or juxtacrine mechanism of receptor activation associated with cell-cell contact (6,7). The short cytoplasmic domain of TGF-␣ is extremely conserved to an extent exceeding that of the extracellular ligand domain. This remarkable sequence conservation suggests a biologically important function, but no known enzymatic or structural motifs have been identified in this domain. Instead, the most remarkable feature is that this short sequence is rich in cysteines. Indeed, in the 39-amino acid sequence, 7 conserved residues are found with an additional residue located in the transmembrane region. Short cysteinerich sequences are also found in the cytoplasmic domains of CD4 and CD8, and p56 lck has been shown to non-covalently associate with the cytoplasmic domains of these two proteins through interactions between two cysteine pairs (35)(36)(37). Furthermore, the interaction of polyoma middle T with pp60 c-src , protein phosphatase 2A, and phosphatidylinositol 3-kinase is dependent on a single cysteine residue, and its mutation abolishes these interactions and its biological activities (52). Thus, the cysteines in the cytoplasmic domain of transmembrane TGF-␣ might similarly play a role in protein-protein associations.
Considering the lack of recognizable motifs in the cytoplasmic domain of transmembrane TGF-␣, any role associated with its cytoplasmic domain is likely to be mediated through interacting proteins. Accordingly, we have evaluated the possible association of proteins with transmembrane TGF-␣. Using reversible, chemical cross-linking and immunoprecipitation analyses, we identified another transmembrane protein p106 that interacts with transmembrane TGF-␣ as well as a cytoplasmic p86 protein that requires the cytoplasmic domain for association. Furthermore, a kinase activity is also associated with the cytoplasmic domain of TGF-␣. Using a series of mutations and truncations in the cytoplasmic domain of transmembrane TGF-␣, we have now explored the structural requirements of the interactions of these proteins.
The association of p106 with transmembrane TGF-␣ does not depend on the cytoplasmic domain of TGF-␣ since p106 still associates with a cytoplasmic truncated form of TGF-␣. Accordingly, the association of these two transmembrane proteins is likely to be mediated via their extracellular and/or transmembrane domains. However, the level of p106 associated with the truncated TGF-␣ is frequently lower than that of p106 with full-size transmembrane TGF-␣, thus suggesting a stabilizing role of the cytoplasmic domain of transmembrane TGF-␣ in this interaction. In contrast with p106, the association of p86 requires the cytoplasmic domain of transmembrane TGF-␣. Our mutation analysis of the cytoplasmic domain strongly suggests that the cysteine pair at positions 153 and 154 plays a primary role in this association. Indeed, replacement of these 2 cysteines by serines strongly reduces the level of p86 association. However, the sequence immediately downstream from the cysteines might also play a role since the association of p86 to the ⌬154 mutant was lower than that with full-size transmembrane TGF-␣. Unlike the defined involvement of the distal cysteine pair in the interaction of the cytoplasmic domain of TGF-␣ with p86, no clear requirement of specific amino acids could be delineated for the interaction of the kinase activity with the cytoplasmic domain of TGF-␣. Instead, the analyses of the TGF-␣ mutants suggested that the associated kinase activity decreased as the C-terminal deletion increased. Whereas the replacement of the distal cysteine pair by serines did not greatly affect the association of the kinase activity, deletion of the C-terminal 8 amino acids significantly reduced the level of associated kinase activity. Based on mutation analysis of truncated TGF-␣, it has previously been concluded that the Cterminal 2 valines of the cytoplasmic domain of TGF-␣ are required for efficient proteolytic cleavage of the ectodomain (33). Whereas this could be explained by an involvement of these amino acids in an interaction with associated proteins, we did not find any differences in association of p86 or kinase activity between full-size transmembrane TGF-␣ and the ⌬158 mutant, which lacks the C-terminal 2 residues. Furthermore, deletion of the C-terminal 31 amino acids of transmembrane TGF-␣ did not greatly decrease the proteolytic processing of the ectodomain in our experiments (Ref. 34; Fig. 2 and data not shown).
Whereas the cysteines at positions 153 and 154 appear to be important for protein-protein interactions, their exact role is as yet unclear. One possibility would be that they play a role in direct protein-protein interactions, similar to the cysteine pair in the CD4 and CD8 cytoplasmic domains that mediates an interaction with p56 lck (35)(36)(37). In addition, replacement of 2 cytoplasmic cysteines in the cytoplasmic domain of the insulin receptor also results in impaired receptor signaling, although the role of these cysteines is unknown (53). However, another function, which is not necessarily mutually exclusive, is related to the possible fatty acylation of these cysteines. Accordingly, our analyses strongly suggest that the distal cysteine pair is    3 H radioactivity in dpm collected from each Edman degradation cycle for the first six cycles of each microsequenced tryptic peptide derived from [ 3 H]palmitate-labeled transmembrane TGF-␣, as determined in three independent experiments. The radioactivity in the first cycle is a common artifact of peptide washoff during the sequencing procedure. Radioactivity was consistently detected in the third and fourth positions. Correlation of these data with the predicted peptide sequences in Table I  palmitoylated and represents the major if not the only site of palmitoylation. This type of fatty acylation occurs with a variety of membrane-associated proteins and plays a role in membrane attachment. For example, several Src family members are palmitoylated, and this palmitoylation serves both for protein localization into caveolae and association with GPI anchors of a defined class of surface-associated extracellular proteins (54 -57). Furthermore, increasing evidence indicates that palmitoylation is a reversible, dynamic process that can be regulated. Indeed, regulation of palmitoylation-depalmitoylation cycles may play an important role in the signaling of trimeric G-proteins (58). The palmitoylation of the distal cysteine pair and the involvement of these cysteines in the association of transmembrane TGF-␣ with cytoplasmic protein(s) strongly suggests that palmitoylation may be required for formation of the transmembrane TGF-␣-associated protein complex. The palmitoylation is then likely to confer membrane association of the cytoplasmic domain in a defined configuration, which allows protein associations. It is furthermore likely that the association of cytoplasmic proteins could be subject to regulation by reversible palmitoylation of the distal cysteine pair. Such regulation could be induced by interaction of transmembrane TGF-␣ with the EGF/TGF-␣ receptors, which dimerize as a consequence of ligand occupation, most likely resulting in dimerization of transmembrane TGF-␣ and associated proteins as well. In addition, proteolytic cleavage of the ectodomain of transmembrane TGF-␣ by the cell surface-associated protease can be induced by 12-O-tetradecanoylphorbol-13-acetate (59 -61), an activator of protein kinase C signaling, and by ligand-induced activation of the EGF receptor signaling cascades (62). These signaling mechanisms could regulate the association and functions of proteins that associate with the cytoplasmic domain. The function and regulation of the proteins that constitute the transmembrane TGF-␣-associated complex and their relevance for TGF-␣-associated signaling await the identification and characterization of these proteins.