The unique domain as the site on Lyn kinase for its constitutive association with the high affinity receptor for IgE.

Aggregation of the high affinity receptor for IgE (FcepsilonRI) leads to the phosphorylation of tyrosines on the beta and gamma chains of the receptor by the Src family kinase Lyn. We have studied the interaction between Lyn and the FcepsilonRI in vivo using a transfection-based approach. FcepsilonRI were stably transfected into Chinese hamster ovary cells. The small amount of endogenous Src family kinase was sufficient to phosphorylate receptor tyrosines upon extensive aggregation of FcepsilonRI but not after addition of dimers of IgE. Upon stable co-transfection of Lyn kinase into the cells, dimers were now able to stimulate receptor phosphorylation and the response to more extensive aggregation was enhanced. In contrast, co-transfection with catalytically inactive Lyn inhibited the aggregation-induced phosphorylation by the endogenous kinase, and a quantitatively similar inhibition was observed in cells transfected with the SH4-containing unique domain of Lyn. Consistent with the results of others using alternative approaches, our additional studies using a yeast two-hybrid system detected a direct interaction between intact Lyn or its unique domain and the C-terminal cytoplasmic domain of the beta chain but not with the receptor's other cytoplasmic domains.

The family of proteins known as the "multichain immune recognition receptors" includes the antigen receptors on B and T-lymphocytes and Fc receptors including the receptor with high affinity for IgE (Fc⑀RI) 1 (1). Highly homologous in structure, all these receptors utilize, at least in part, a common mechanism to initiate cellular responses; multivalent interactions with antigen lead to aggregation of the receptors and is followed by enhanced phosphorylation of tyrosines (in the "ITAM" motifs within the cytoplasmic domains) of the receptor itself by a receptor-associated Src family kinase (2). For Fc⑀RI, we recently presented direct evidence for a "transphosphorylation" mechanism that accounts for the earliest events (3,4). The data showed that a small fraction of receptors are consti-tutively associated with the Src family kinase Lyn (4,5) and that the enhanced phosphorylation that follows aggregation of the receptors is likely to result simply from the apposition of the kinase with its substrate. We have also shown that when the kinase available to the receptor is limited, shuttling of the enzyme between individual aggregates can regulate the intensity of the signal (6). 2 The experiments described in this paper mainly explored the sites of interaction between Lyn kinase and Fc⑀RI. For the most part, the prior studies of others explored the interaction between Lyn and isolated portions of the receptor (7)(8)(9)(10). The yeast two-hybrid system (11) used in some of our studies is an analogous approach. We also employed transfection techniques, which allowed us to examine the kinase-receptor interactions in a more physiological setting. The latter experiments also allowed us to test the effect of varying the level of Lyn on the responsiveness of the receptors to discrete stimuli, and thereby to test certain quantitative predictions made by the current model.

EXPERIMENTAL PROCEDURES
Materials-The yeast strains (CG1945 and Y187) and cloning vectors (pAS2-1 and pACT) were obtained from CLONTECH (Palo Alto, CA); the expression vectors pBlueBac, pCDM8, and pZeo, as well as a baculovirus MAXBAC expression kit from Invitrogen (Carlsbad, CA); polyacrylamide gels used for electrophoresis (PAGE) from NOVEX (San Diego, CA); the antibiotics (G418, zeocin) from Life Technologies, Inc. and Invitrogen, respectively; and plasmid DNA purification kits from Qiagen (Santa Clarita, CA).
Antibodies-Monoclonal anti-phosphotyrosine (anti-Tyr(P)) antibodies were obtained conjugated to horseradish peroxidase from Transduction Laboratories  or Upstate Biotechnology, Inc. (4G10). Polyclonal antibodies to human Src family kinases Lyn and Fyn were purchased from Upstate Biotechnology, Inc.; antibodies to c-Src and c-Yes were from Santa Cruz Biotechnology (Santa Cruz, CA). Mouse monoclonal anti-DNP IgE (12) and rat IgE (of unknown specificity) (13) were purified as described previously (14,15) and labeled with carrierfree 125 I using chloramine T (16). Goat anti-mouse IgE was purchased from ICN (Costa Mesa, CA); rabbit anti-rat IgE was purified as described (17). Covalently cross-linked IgE oligomers were prepared and analyzed as described (6).
Cell Lines-Rat basophilic leukemia (RBL-2H3) cells were maintained as described previously (18). Chinese hamster ovary cells (CHO) were grown in stationary flasks at 37°C in a humidified atmosphere containing 5% CO 2 in Iscove's modified Eagle's minimum essential medium, 10% fetal calf serum, 25 mM HEPES, and the appropriate antibiotics to maintain expression of the transfected genes. Spodoptera frugiperda (Sf9) insect cells were maintained in spinner culture at 27°C as described previously (19).
DNA Sequencing-The nucleotide sequence of each expression construct was confirmed by automated DNA sequencing using a dye terminator kit obtained from Applied Biosystems (Foster City, CA).
Isolation of Rat Lyn Kinase cDNAs-A 5Ј-stretch cDNA library was prepared from mRNA isolated from RBL cells. Two separate priming reactions with either oligo(dT) or random primers were performed to generate the first strand. The reactions were pooled prior to second * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AF000300 and AF000301.
strand synthesis. The cDNA library was then prepared in the expression vector pCDM8 (20). Probes were prepared by restriction digestion of human Lyn A-pSVL. Probes representing the N terminus (amino acid residues 1-298) and the C terminus (residues 163-512) were purified. The library was plated, and colony lifts were hybridized with either probe. Positive colonies went through secondary and tertiary screening. The nucleotide sequences of two clones, designated N14 (2052 bp) and C18 (2316 bp), were determined by primer walking and DNA sequencing of both strands. The Wisconsin package from the Genetics Computer Group, Inc. was used to assemble and analyze the nucleotide sequences of the isolated clones. N14 contained an open reading frame of Lyn A, beginning with ATG from bp 80 to bp 1616, while clone C18 encoded Lyn B beginning with ATG between bp 236 and bp 1709. The sequence of Lyn A in the coding region was identical to a previously published sequence (21); the sequence of rat Lyn B lacks an "insert" of 21 amino acids found in the A form of the kinase at a position identical to that previously shown for human and murine Lyn (22,23) but is otherwise identical to Lyn A. Therefore, it differs somewhat from the previously published sequence for rat Lyn B (24). 3 CHO cells were transiently transfected with the Lyn-pCDM8 plasmids by electroporation, harvested 48 -72 h later, and a lysate of the whole cells was prepared using SDS. After separation by PAGE and transfer, Western blotting with anti-human Lyn confirmed that the expressed proteins had the expected the size for Lyn A (56 kDa) and Lyn B (53 kDa) (data not shown).
Yeast Two-hybrid Fusion Constructs-To generate DNA binding domain fusion proteins, the N-terminal (1-58) and C-terminal (201-243) cytoplasmic domains of the rat Fc⑀RI␤ were amplified by PCR from the full-length cDNA and cloned into the EcoRI/BamHI sites of pAS2-1. The cytoplasmic domain of rat Fc⑀RI␥ (residues 27-68), as a PCR fragment, was cloned into NcoI/BamHI site of pAS2-1, to generate pAS2-1-␥C. To create activation domain fusion proteins, the full-length Lyn A and Lyn B and various deletion mutants were amplified by PCR and cloned into BamHI/XhoI sites of pACT (Fig. 1).
Yeast Two-hybrid Co-transformation, Selective Growth, and ␤-Galactosidase Assays-Plasmid constructs were introduced into yeast cells by lithium acetate, following the protocol provided by CLONTECH. Transformants were plated on synthetic medium containing 5 mM 3-amino-1,2,4-trizole and lacking leucine, tryptophan, and histidine (SD-3) to detect the His phenotype, or synthetic medium lacking leucine and tryptophan (SD-2), to measure transformation efficiency. The ␤-galactosidase activity of transformants was measured in a filter assay with 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside as substrate or in a liquid assay with an o-nitrophenyl-␤-D-galactoside substrate according to the CLONTECH protocol.
Mammalian Transfection Constructs-The 2.3-kilobase pair Lyn B XhoI-digested insert was isolated from pCDM8 and subcloned into the XhoI site of pZeo. The unique domain construct was generated by PCR amplification using internal sense and antisense primers for the unique domain: (5Ј-CGGGCGGCTCGATGGGATGTATTAAATCAAAAAGGA-AAG-3Ј, and 5Ј-CGGCGGCTCGAGCTAGTCCCCTTGCTCCTCTGGA-TC-3Ј, respectively).
The final PCR product was digested with XhoI and cloned back into the pZeo XhoI site. The catalytically inactive Lyn B(K279R)-pZeo construct was prepared using the Altered Sites in vitro mutagenesis system from Promega (Madison, WI) as follows. A 2.3-kilobase pair XbaI fragment of Lyn B from Lyn-pCDM8 was ligated into the XbaI site of pAlter-1. Mutagenesis followed the manufacturer's protocol using the ampicillin repair primer provided in the kit and a Lyn single mutation antisense oligonucleotide: (5Ј-GCCAGGCTTGAGGGTCCTTACAGC-CACTTTTGTGC-3Ј) to convert TTC (Lys) 3 TCC (Arg). The mutant Lyn B was digested with XhoI and BstBI and ligated back into pZeo.
Transfection of CHO Cells-Using Lipofectin reagent-mediated transfection (Life Technologies, Inc.), pSVL constructs of the ␣, ␤, and ␥ subunits of rat Fc⑀RI along with pSV2neo had been previously introduced into CHO cells, and a clone expressing a high number of receptors was frozen. 4 After thawing, expression of receptors decreased rapidly with time in culture, so we recloned the culture by incubating it with fluorescein-conjugated IgE and sorting on a fluorescence activated cell sorter. The 1% of cells expressing the highest number of receptors were resorted on 96-well plates at 0.5 cell/well. Fifty surviving clones were screened for expression of receptors, by growing the cells to confluence and sensitizing them with 125 I-labeled mouse IgE. The washed adherent cells were solubilized with boiling SDS sample buffer and the IgE in the extract quantitated by ␥-counting. Of five high expressing clones, one (CHO-B12) proved highly stable and was used for all subsequent studies. CHO-B12 cells were cryopreserved by freezing in 5% dimethyl sulfoxide, 95% growth medium; higher concentrations of dimethyl sulfoxide caused a rapid decline in Fc⑀RI. The cells were electroporated (0.4-cm gap cuvettes, 200 V, 500 microfarads) in the presence of one of several rat Lyn-pZeo constructs or empty pZeo vector, which had been linearized by digestion with Eco57I. To select resistant clones, the medium was supplemented with 250 g/ml zeocin 72 h post-transfection.
Baculovirus Expression of Human Lyn B-The human Lyn B cDNA (1.5 kilobase pairs) was excised from pSVL by XbaI digestion and ligated into the homologous NheI site of pBlueBac. Sf9 cells were co-transfected with wild type AcMNPV DNA and the Lyn construct to generate recombinant Lyn baculoviruses. Adherent Sf9 cells were infected with plaque-purified baculovirus at a multiplicity of infection of 0.4 and, after 48 h, lysed in 0.1% Nonidet P-40 buffer containing protease and phosphatase inhibitors. Western blotting with anti-Tyr(P) indicated that the Lyn B protein was phosphorylated on tyrosine as it was produced in the insect cells (data not shown).
Stimulation of Cells-CHO cells to be stimulated with antigen were sensitized overnight with 125 I-labeled mouse anti-DNP mouse IgE, washed three times in buffer A (150 mM NaCl, 5 mM KCl, 25 mM Pipes, pH 7.2) plus 0.1% (w/v) gelatin and 5.4 mM dextrose, and resuspended at 1 ϫ 10 7 cells/ml. DNP 6 -BSA was added as a 5-fold stock solution to 5 ϫ 10 6 cells at 37°C for the times indicated. CHO cells stimulated with IgE oligomers were incubated with the indicated concentrations at 37°C for the times indicated.
Solubilization and Immunoprecipitation-After stimulation, the receptors were solubilized in 0.05% Triton X-100 (3). For immunoprecipitation, anti-mouse or anti-rat IgE antibody was prebound to 30 l of protein A-Sepharose beads overnight in borate-buffered saline, pH 8, containing 0.1% gelatin. The beads were recovered by centrifugation and combined with the lysates ("precleared" with 100 l of protein A-Sepharose beads overnight) for 2 h. After recentrifugation the immunoprecipitates were washed four times as described previously (3), and the bound proteins released by boiling in SDS sample buffer for 5 min.
Quantitation of Phosphorylation of Receptors-Immunoprecipitated receptors were separated by electrophoresis in SDS on 10% polyacrylamide gels equilibrated with Tricine and the phosphorylated proteins detected with an anti-Tyr(P) antibody and an enhanced chemiluminescent detection system (ECL, Amersham) (25). Autophotographs of Western blots were quantitated by computerized densitometry (Molecular Dynamics, Sunnyvale, CA). Three steps were taken to ensure equal numbers of receptors were being compared in those studies in which cells co-transfected with inactive forms of Lyn were compared with cells that had not been co-transfected. First, the cells were incubated with IgE that had been labeled with 125 I and equal numbers of counts were loaded per lane. Second, one lane on each gel was loaded with the same amount of phosphorylated human Lyn B to correct for differences in transfer, antibody staining, washing, etc. Third, the primary anti-Tyr(P) blots were stripped and reprobed with an antibody (JRK) to the ␤ chain of the receptor (26), and densitometric analysis was repeated. The densitometric values from the primary anti-Tyr(P) blots were then corrected for any differences in anti-Tyr(P) staining or loading of receptors. In separate experiments, the linearity of antibody staining (anti-Tyr(P), anti-␤) was verified by loading increasing amounts of an appropriate protein extract and quantitating the band intensity.
Quantitation of Lyn-To quantitate the relative amounts of Lyn, whole cell lysates containing either 7 ϫ 10 4 or 1.6 ϫ 10 5 cell eq were prepared with SDS for each transfectant. Depending on which molecules had been transfected, the samples were separated on 8% (Lyn B, RK Lyn), 10% (CHO-B12, pZeo), or 4 -20% (unique Lyn A) Tris-glycine gels and blotted with an anti-Lyn antibody and an HRP-conjugated anti-rabbit secondary antibody. One (central) lane on each gel was loaded with a fixed amount of human Lyn B (above). The densitometric readings for the bands corresponding to Lyn were normalized relative to the human Lyn B standard.
Quantitation of Fc⑀RI --CHO cells were suspended at 5 ϫ 10 6 cells/ml and incubated with 5 g/ml 125 I-labeled IgE for 1 h at 37°C. Nonspecific binding was evaluated by preincubating the cells with a 10-fold excess of unlabeled IgE for 30 min at 37°C. Cells were separated 3 The open reading frame of our Lyn B cDNA is identical to the Lyn A except for the missing 21-residue "insert." The differences in the sequence originally reported (24) probably resulted from sequencing errors or errors introduced during amplification with PCR. Two further clones, a partial Lyn A cDNA and a full-length Lyn B cDNA isolated from an RBL library, confirmed our sequence. 4 C. Pucillo, unpublished studies.
from unbound IgE by pelleting through phthalate oil (15,27). Subcellular Fractionation-CHO cells were sonicated, and the 140,000 ϫ g supernatant (cytosolic fraction) and pellet (membrane fraction) were prepared from the post-nuclear supernatant as described previously (28). Membrane proteins were solubilized in 0.5% Triton X-100, for 30 min at 4°C. Each subcellular fraction was treated with an equal volume of boiling 2 ϫ SDS sample buffer for 5 min prior to gel electrophoresis.
Other Procedures-Coupled in vitro transcription-translation reactions were conducted with [ 35 S]Cys according to the manufacturer's recommendation (T3 TnT ® coupled reticulocyte lysate system, Promega).

Yeast Two-hybrid Studies
Initial identification of potentially interacting domains was conducted by co-transforming constructs containing the cytoplasmic domains of the Fc⑀RI fused to the binding domain of the Gal4 transcription factor with constructs containing Lyn or various mutated forms of Lyn fused to the activation domain of Gal4 (Fig. 1).
The nucleotide sequence coding for the N-and C-terminal cytoplasmic domains of the ␤ subunit of the rat IgE receptor, ␤ N and ␤ C , were subcloned into pAS2 to generate Gal4 DNA binding domain fusion proteins. Unfortunately, both fusion proteins autonomously activated the reporter genes. This is presumably due to the acidic hemagglutinin epitope located between the Gal4 DNA binding domain and the inserted proteins (29). How-ever, the fusion protein containing the cytoplasmic domain of the ␥ subunit was not autonomously active. Therefore, we subcloned nucleotide sequences coding for ␤ N and ␤ C into the newly developed vector pAS-2-1, which is similar to pAS2, but has the acidic hemagglutinin epitope removed. Neither pAS2-1-␤ C or pAS2-1-␤ N were autonomously active.
The activities of the His and LacZ reporter genes in CG1945 yeast transformants expressing Lyn and ␤ N , ␤ C or ␥ C were tested as described (see "Experimental Procedures"). Both the full-length and unique domain of both Lyn A and Lyn B interacted directly with ␤ C (data not shown). However, the interaction was much weaker than the interaction detected between the p53 and SV40 fusion proteins used as a positive control. Thus, per microgram of DNA, co-transformation with p53 and SV40 resulted in more colonies on His-deficient medium (SD-3) and rapid growth into large colonies. All of the colonies containing p53 and SV40 rapidly turned blue. In contrast, cotransformation with the Lyn and ␤ C constructs resulted in fewer colonies and slower growth on His-deficient medium and only the large colonies turned blue. No interaction was detected between ␤ N or ␥ C with any forms of Lyn in this assay.
To quantitate the interaction between Lyn and ␤ C or ␤ N , we measured the ␤-galactosidase activity of these co-transformants in yeast strain Y187 in a liquid assay. In addition to the full-length Lyn and the construct containing only the unique domain, we tested a series of Lyn mutants based on the results of Pleiman et al. (30) and Timson Gauen et al. (31). The negative control in this experiment was the 40 amino acid residues (from 27 to 66) of Lyn fused out-of-frame to the Gal4 activation domain (pACT-27-66-OOF). As shown in Fig.  2, the activity of the LacZ reporter gene from co-transformants with the unique domain of either Lyn A or B was as high as that from co-transformants with the full-length form of either Lyn. These values are 3-fold higher than those than from the negative control pACT-27-66-OOF. Consistent with the result from CG1945 strain, the interaction between Lyn and ␤ C is weaker (on the basis of the ␤-galactosidase activity, only 1% as strong) than that between p53 and SV40. Cotransformants containing Lyn residues 1-10, 1-27, or 27-66 produced only slightly higher amounts of ␤-galactosidase than the negative control. Again, no interaction between Lyn and ␤ N was detected.

Characterization of Transfected Fc⑀RI in CHO Cells
A clone of transfected CHO cells that stably expressed Ϸ170,000 receptors/cell (CHO-B12) ( Table I) was further characterized. When immunoblotted with anti-human Lyn antibody, extracts of these cells, like those of the untransfected CHO cells, show a weakly reactive component at Ϸ58 kDa, i.e. slightly greater than the apparent molecular mass of 53 and 56 kDa observed for rat Lyn (Fig. 3A). There was no reactivity with a panel of antibodies to human c-Src, Fyn, or c-Yes (data not shown). Cells from the B12 clone were incubated with anti-DNP-specific mouse IgE and after solubilization with detergent, the bound (unaggregated) receptors were immunoprecipitated with goat anti-mouse IgE. Upon Western blotting with anti-Tyr(P), no evidence for phosphorylation was observed (Fig. 3B, lane 1). When the cells were incubated with multivalent antigen (DNP-BSA) prior to solubilization, phosphorylation of tyrosines on the ␤ and ␥ subunits of the transfected receptors was observed (lane 2). Disaggregation of the receptors in vivo by addition of hapten (DNP-caproic acid) after the exposure to DNP-BSA led to the complete reversal of the antigen-induced phosphorylation of receptor tyrosines within Յ1 min (data not shown).
RBL cells can be stimulated either by aggregating receptorbound monomeric IgE with antigen or by incubating the cells with preformed dimers of IgE (Fig. 3B, lane 8). In contrast, incubation of the CHO-B12 cells with dimeric IgE failed to induce detectable phosphorylation of the receptors (Fig. 3B,  lane 4). These results are consistent with a limiting amount of protein-tyrosine kinase being associated with the receptors in these cells (see "Discussion").

Correlation between Total Lyn and Phosphorylation of Fc⑀RI
A series of stable transfectants of the CHO-B12 cells with rat Lyn were isolated. The relative ratios of full-length Lyn/receptor of six clones (A6 through D8) are shown in the upper part of column 5 of Table I. Subcellular fractionation of the transfected cells indicated that the transfected full-length Lyn was expressed as a membrane-associated protein (Fig. 3A), as expected for a Src family kinase (32).
The various transfectants were stimulated either with IgE dimers or with monomeric IgE and then antigen, to examine the relationship between the total cellular content of Lyn and the responsiveness of the cells. Care was taken to ensure equal numbers of receptors were being compared (see "Experimental Procedures").
As shown in Fig. 4, there was a good correlation between the amount of Lyn expressed and the amount of receptor tyrosine phosphorylation seen on both the ␤ and ␥ subunits upon aggregation of the receptors with antigen. Furthermore, all of the cells expressing transfected Lyn now responded to dimers of IgE. More extensive phosphorylation was observed in those cells whose receptors were aggregated with antigen rather than with dimers. However, the stimulation by dimers was more sensitive to the amount of Lyn expressed as can be seen by comparing the slopes of the two response "curves" (Fig. 4).
One clone, A11, in which the relative Lyn/receptor ratio was exceptionally high, showed a significant degree of phosphorylation of the receptors even without stimulation. Western blotting of A11 lysates revealed a phosphorylated component with an apparent molecular mass of 53 kDa (presumably Lyn) but To control for differences in tyrosine phosphorylation that may have arisen due to zeocin resistance alone, CHO-B12 cells were transfected with pZeo vector and resistant colonies isolated and expanded. Upon stimulation with 0.5 g/ml trimeric IgE from 5 to 30 min, the six zeocin-resistant clones tested showed no significant differences in phosphorylation of the ␤ and ␥ subunits of the receptor compared with CHO-B12 cells (Fig. 5A). In a similar experiment, the responses to varying doses of antigen (25-300 ng/ml) of three zeocin-resistant clones were compared with CHO-B12 cells. A similar dose dependence of phosphorylation of the receptors was observed (Fig. 5B). No differences were noted in either the magnitude or pattern of total cellular proteins that became tyrosine phosphorylated. By Western blotting the level of endogenous Lyn was also unchanged. Since the number of Fc⑀RI on the pZeo transfectants varied between 80,000 and 150,000 (clones Z1-Z6, Table I), the degree of phosphorylation was found to be independent of the number of receptors under the conditions used in this study.

Mapping the Site of Lyn-Fc⑀RI Interaction by Competition
The presence in the CHO-B12 cells of an endogenous kinase (presumably Lyn) capable of phosphorylating the receptor, permitted us to probe the site on Lyn interacting with the Fc⑀RI by a competition protocol. We transfected the cells with domains of Lyn that would potentially interact with the receptor but that were themselves catalytically inactive. We compared the responsiveness of such transfectants to aggregation of their Fc⑀RI either to CHO-B12 or to cells co-transfected with the "empty" pZeo vector.
Catalytically Inactive Lyn Kinase-A full-length, catalyti-cally inactive Lyn B kinase was prepared by mutating Lys 279 to Arg (RK Lyn). As shown in Fig. 6A, in a coupled in vitro transcription-translation reaction, the wild type Lyn was autophosphorylated 5 whereas the mutant Lyn was not. Three stable transfectants expressing substantial amounts of the mutant Lyn were isolated and assessed (clones RK17, RK21, and RK26; Table I). The catalytically inactive Lyn was expressed largely or exclusively as a membrane anchored protein (Fig. 3A). On a per receptor basis, such stable RK Lyn-Fc⑀RI transfectants showed 20 -75% less antigen-induced phosphorylation of receptor tyrosines than cells transfected with the vector alone (Fig. 6B). Therefore, a single point mutation converted a construct that stimulated phosphorylation of tyrosines on Fc⑀RI to one that inhibited it (cf. Figs. 4 and 6B).
Unique Domain of Lyn Kinase-Prompted by our results from the yeast-two hybrid studies, we transfected the unique domain of Lyn A kinase into receptor-containing cells (clones B5, C6, U7, and U8, Table I). The isolated unique domain was also expressed largely or exclusively in a membrane-anchored form (Fig. 3A). Figs. 7 and 8 show comparisons between the responses to two different stimuli of the transfectants and CHO-B12 cells not transfected with Lyn. Upon stimulation with multivalent antigen, a partial inhibition of phosphorylation of receptor tyrosines was observed (Fig. 7A). A comparison of two clones expressing increasing levels of the unique domain protein showed that increasing amounts of the competing domain led to increasing inhibition (Fig. 7B). With a weaker stimulus (IgE trimers), complete inhibition of phosphorylation of the ␤ and ␥ chains was observed at early time points (Fig. 8A) and at low concentrations of stimulant (Fig. 8B).

DISCUSSION
Interactions in the Yeast Two-hybrid System-Several groups have studied the interaction between Fc⑀RI and Lyn kinase by a variety of techniques. Consistent with previous findings (7-10), the results from our studies in the yeast twohybrid system indicate a direct interaction between the kinase and the C-terminal cytoplasmic extension of the receptor's ␤ chain. No interaction was detected between Lyn and ␤ N or ␥ C . As judged by the relative activity of a reporter gene, the interaction is very weak (Fig. 2). This is consistent with the difficulty in demonstrating co-immunoprecipitation of Lyn with unphosphorylated Fc⑀RI in the absence of chemical cross-linking (4). Our results extend those of previous workers in showing that the Lyn A and Lyn B behave equivalently (5). This result is also consistent with our previous finding that the two forms of Lyn become equivalently attached to the receptor after chemical cross-linking (4). Furthermore, we demonstrated that the unique domain alone interacts with the receptor as effectively as the full-length kinase, but the weakness of the interaction makes problematic any attempt to define the site of interaction more narrowly by this method (Fig. 2).
It is conceivable that in this experimental system the receptor component is phosphorylated, but this seems unlikely because in the natural setting, dephosphorylation of the receptor is strongly favored over phosphorylation in the absence of aggregation (28,33). Therefore, the interactions we observed probably mimic the constitutive association of Lyn with the receptor rather than the interaction of recruited Lyn with the phosphorylated receptor (3,4).

CHO Transfection Studies: Quantitative Aspects of Fc⑀RI Aggregation-induced Phosphorylation of Tyrosine-The results
with the cells transfected with active Lyn provide strong evidence that the amount of Lyn available to the receptor determines the capacity of the system to initiate signaling. The results with the cells transfected with catalytically inactive forms of Lyn (below) provide strong evidence that an equilibrium exists between receptor-associated and non-receptor-associated Lyn. The molecular mechanism we currently envision predicts that the capacity of small aggregates of receptors to initiate a response will be particularly sensitive to the amount of Lyn per receptor (3). 2 The slopes of the lines in Fig. 4 indicate that, indeed, cells stimulated with dimers of IgE are more

FIG. 3. Expression, distribution, and activity of Lyn in CHO transfectants and RBL cells. A, expression and membrane association of Lyn in CHO transfectants. CHO cells previously transfected with
Fc⑀RI (CHO-B12) were stably co-transfected with either intact Lyn B (clone D1), the unique domain from Lyn A (C6), the catalytically inactive RK mutant of Lyn B (RK26), or the empty pZeo vector (Z5). SDS lysates of intact cells (Lys) or of the membrane (Mem) or cytosolic (Cyt) fractions of sonicated cells were prepared. The proteins from 1.6 ϫ 10 5 cell eq of each lysate were separated by PAGE and blotted with antihuman Lyn antibody (see "Experimental Procedures"). The apparent molecular mass of the principal component is shown at right. B, phosphorylation of receptor tyrosines in CHO-B12 and RBL cells. Six million cells were incubated with 5 g/ml 125 I-labeled mouse anti-DNP-specific IgE for 1 h at room temperature. The cells were washed, and duplicate samples were then incubated at 37°C with (Antigen ϩ) or without (Antigen Ϫ) 100 ng/ml DNP-BSA, for 2 min more. Other samples were incubated for 15 min with either 0.5 g/ml 125 I-labeled monomeric rat IgE (Dimer Ϫ) or equivalent amounts of chemically cross linked dimers of rat IgE (Dimer ϩ). Fc⑀RI were immunoprecipitated from the detergent lysate of the cells with anti-IgE, and the samples were blotted with anti-Tyr(P) (PY-20; see "Experimental Procedures"). Equal numbers of receptors (based on the cpm of bound 125 I-IgE) were loaded for each immunoprecipitate. One experiment representative of the two conducted is shown. sensitive to the concentration of Lyn than those stimulated with antigen.
Mapping of Sites of Interaction by Competition for Binding to the Fc⑀RI-Catalytically inactive Lyn (RK Lyn) consistently inhibited signaling by the receptor compared with control cells (Fig. 6B). The effect was even more dramatic using trimers of IgE in cells transfected with the unique domain, and complete inhibition was detected at early time points (Fig. 8A). Again, this is consistent with the prediction that small aggregates would be more sensitive to the ratio of active Lyn:receptor. The inhibitory effect of the catalytically inactive Lyn indicates that the interaction between Lyn and the receptor is not dependent on an intact catalytic site on the kinase.
Notably, the unique domain alone was about as effective as  Table I). The number of Fc⑀RI per cell and the ratios of transfected Lyn per Fc⑀RI for each clone are described in Table I the catalytically inactive Lyn in inhibiting the interaction between the receptor and the wild type endogenous Src family kinase in the CHO cells. This was not necessarily predictable for the following reason. We previously demonstrated that after aggregation, an initial phosphorylation of the receptor by the constitutively bound Lyn kinase is required for the recruitment of additional molecules of Lyn to the activated receptor (3). Because direct studies in vitro have shown that the SH2 domain of Lyn can interact with the phosphorylated ITAM of the ␤ subunit (10), it is reasonable to think that the recruitment occurs through the interaction of the SH2 of the Lyn kinase with the phosphorylated receptor. Therefore, high expression of the catalytically inactive Lyn, which retains its SH2 domain, might have affected the level of phosphorylation differently than the unique domain alone. For example, in addition to blocking constitutive association of the endogenous kinase, it might have prevented the recruitment of further kinase to the phosphorylated receptors. Alternatively, the inactive Lyn might have protected the phosphotyrosine(s) from dephosphorylation (34). Because these would lead to opposing effects on the level of phosphorylation, we cannot rule out the possibility that fortuitously the two effects quantitatively canceled each other out. A more likely explanation is that the SH2 region of Lyn does not play a major role in regulating the level of phosphorylation of the receptor and may not be the basis of the recruitment to the phosphorylated receptors.
It appears likely that the principal interaction the competition experiments are assessing is the constitutive interaction between Lyn and the receptor. For example, in vitro experiments on the antigen receptor of B lymphocytes have demonstrated an interaction of the unique domain of Lyn and the related Src family kinase Fyn with the unphosphorylated but not the phosphorylated ITAMs from Ig␣ (35). Earlier studies on the association of Fyn with the subunits of the CD3 complex of the T-cell receptor (36) and of Lck with CD4 (37) have implicated the analogous region in those kinases. Resh et al. have noted the highly homologous sequences within the first 10 residues of the Src family kinases and have presented evidence that this region, which she dubbed SH4, is critical for the targeting of the kinases to membranes (32). Citing unpublished data, Lin et al. implicate the same region in the interaction of Lyn and the ␤ subunit (8). Timson Gauen et al. have studied the targeting of p59 fyn to membranes and its interaction with chimeric constructs of the T cell receptor CD3 ⑀-chain using mutational analysis (38). They concluded that four residues within the SH4 region, i.e. Gly 2 , Cys 3 , Lys 7 , and Lys 9 , were required for both interactions. It should be noted however, that in their analysis, the interaction of p59 fyn with the ⑀ chain might well have been influenced by the interaction of p59 fyn with the plasma membrane. Lyn shares these critical N-terminal residues with Fyn, and it is likely, therefore, that this region plays a homologous role in this kinase's interaction with the Fc⑀RI.
Alternative constructs of Lyn could be used to probe further the nature of this interaction, but such studies would have to be very extensive to obtain any more insight than the present studies provide. Furthermore, such additional studies could only provide rather indirect evidence about which structures in Lyn are important. Rather than pursuing such intermediate results, what the field really needs is structural information at the atomic level of resolution, and we are turning our experimental strategy in that direction. More detailed analyses must also control for the possibility that these interactions may be occurring in the context of specialized membrane domains (39).
As already noted, our experimental findings are consistent with the prediction of the current model that the ability of small aggregates to initiate a response should be particularly sensitive to the fraction of receptors constitutively associated with kinase. No such enhanced sensitivity is predicted for the recruitment of further molecules of kinase to the phosphorylated receptors. The interpretation that it is the constitutive interaction that is affected is also consistent with one of the findings reported by Wilson et al. (9). They observed that a chimeric construct bearing the ␤ c domain when transfected into RBL cells failed to become phosphorylated but inhibited both base-line and aggregation-induced phosphorylation of the endogenous Fc⑀RI. This result likely reflects competition by the transfected ␤ chain for limiting amounts of constitutively associated kinase. Thus their experiment is in effect the mirror image to the ones we describe.