Ligand for EPH-related Kinase (LERK) 7 Is the Preferred High Affinity Ligand for the HEK Receptor*

HEK is a member of the EPH-like receptor tyrosine kinase family, which appear to have roles in development and oncogenesis. Recently, we purified a soluble HEK ligand which is also a ligand (AL1) for the HEK-related receptor EHK1. Promiscuity appears to be a characteristic feature of interactions between the EPH-like receptors and their ligands, termed ligands for EPH-related kinases (LERKs). This prompted us to analyze the interactions between the HEK exodomain and fusion proteins comprising candidate LERKs and the Fc portion of human IgG1 (Fc) or a FLAG™-peptide tag by surface plasmon resonance, size exclusion high performance liquid chromatography, sedimentation equilibrium, and transphosphorylation. Our results indicate that AL1/LERK7 is the preferred high-affinity ligand for HEK, forming a stable 1:1 complex with a dissociation constant of 12 nm. As expected the apparent affinities of bivalent fusion proteins of LERKs and the Fc portion of human IgG1 had significantly reduced dissociation rates compared with their monovalent, FLAG™-tagged derivatives. High-avidity binding of monovalent ligands can be achieved by antibody-mediated cross-linking of monovalent ligands and with LERK7 results in specific phosphorylation of the receptor. By extrapolation, our findings indicate that some of the reported LERK-receptor interactions are a consequence of the use of bivalent ligand or receptor constructs and may be functionally irrelevant.

The EPH-like branch of receptor tyrosine kinases (RTKs) 1 has more than 28 members described in several vertebrate species (1)(2)(3). All were identified prior to the characterization of their cognate ligands by methods independent of a biological assays or specific physiological activity (2). As a consequence little is known about their specific functions. However, expression patterns of several EPH-like RTKs in embryogenesis, in particular in the nervous system, suggests a role in development (4,5). Overexpression of some family members including HEK, EPH, ERK, and ECK in tumor-derived cell lines, tumor specimens, and transfected cells implicate these receptors in oncogenesis (6 -10).
Recently we identified HEK on the cell surface of a pre-B acute lymphoblastic leukemia cell line, LK63, using the IIIA4 monoclonal antibody (mAb) (7). Immunofluorescence studies with IIIA4 revealed expression of HEK in blood samples from patients with acute leukemia, but not in normal adult tissues or blood cells (7,11). In embryos, the expression patterns of the murine and chicken HEK homologues MEK 4 and CEK 4, and their recently identified ligand ELF1 (12) and RAGS (13), respectively, suggest a role in the development of the retinotectal projection map. We isolated a soluble HEK ligand from human placenta-conditioned medium using a biosensor-based affinity detection approach (14). The HEK ligand was identified by sequence homology as a soluble form of AL-1 (15), a member of the ligands for EPH-Related Kinases (LERKs) family (16,17) and for consistency with other members will be referred to as LERK 7. This family of transmembrane or membrane-associated proteins were isolated as potential ligands for EPH-like RTKs through their interactions with recombinant EPH receptor family exodomains (15, 18 -21). In most cases, bivalent Fc fusion proteins of either the receptor or the ligand were used for detecting potential binding partners. A requirement for membrane association of the ligand and the ability of more than one ligand to bind the same receptor with comparable affinity appeared to be characteristic of many of these ligands (21)(22)(23)(24)(25). However, soluble ligands for both ECK (26) and HEK (14) have been isolated from biological sources using receptor affinitybased protocols. Functional assays with the natural ligands leave some ambiguity about the absolute requirement for membrane association of the ligands for biological function (9,15,26,27). In addition to binding LERK7, a bivalent HEK fusion construct was shown to bind with significant affinity to LERK 1, 2 (18), LERK 3,4 (28), and LERK 5 (24).
Here we report the results of studies on the interaction between HEK and bivalent Fc-fusion proteins of LERKs 1 to 5 and LERK 7 or monovalent LERK3 and LERK7 produced as Flag-epitope tagged proteins. We use BIAcore technology, SE-HPLC, sedimentation equilibrium centrifugation, and functional assays which show that LERK 7 is the principal HEK ligand. Although most of the apparent affinity constants obtained from our analysis of LERK-Fc binding were well within the range reported in the literature, our experiments indicate that the use of bivalent Fc constructs gives misleading estimates of the likely affinities for the physiological LERK-HEK interactions.

Production of LERK 3 and LERK 7 (AL-1) Expression Constructs
The 5Ј-LERK7 oligonucleotide (GTAGTCTAGACAGGACCCGGGCT-CCAAGGC) was based on the N-terminal amino acid sequence (QDPG-SKA) of the mature protein, with a 5Ј-tag sequence and XbaI site preceding the coding nucleotides. The 3Ј-LERK7 oligonucleotide (GTA-GTCTAGATCAGTTCTCGCCGCGGGATGGC) was based on the predicted C terminus of the glycosylphosphatidylinositol-linked form (PSR-GEN), the coding bases were followed by a stop signal, an XbaI site, and a spacer sequence. The polymerase chain reaction was performed using an aliquot of a placental cDNA library (kindly provided by Dr. Tracy Wilson, Walter & Eliza Hall Institute) and Taq EXTEND (Boehringer Mannheim). A 490-base pair product was detected on a 1.4% TAE, agarose gel. This was excised and the DNA purified using GeneClean II (BIO 101, Inc., Vista, CA). The polymerase chain reaction product and the IL-3 sig-FLAG-pEFBOS vector (29) were digested with XbaI and the vector treated with calf intestinal alkaline phosphatase to prevent religation. After ligation correctly oriented clones were detected and verified by automated DNA sequencing (Model 373, Applied Biosystems, Inc., automated DNA sequencer). Fig. 1B depicts the LERK7-FLAG construct by comparison with the native LERK 7 molecule.
The LERK 3 cDNA was derived in a similar fashion and from the same source using a sense oligonucleotide based on the N-terminal sequence (NRHAVYW) preceded by an XbaI site and a spacer (GTAG) sequence (GTAGTCTAGAAACCGGCATGCGGTGTACTG) and the antisense oligonucleotide (GTAGTCTAGATGCTCTTCTCAAGCTTTGG) based on the C-terminal sequence (PKLEKSI) followed by a stop signal, an XbaI site, and a tag sequence.

Expression and Purification of Recombinant Receptor and
Ligand Proteins sHEK-Recombinant soluble HEK protein (sHEK) was purified from culture supernatants generated from a Chinese hamster ovary (CHO) cell transfectant and tested for conformational integrity using the BIAcore with the sensor chip-immobilized anti-HEK mAb IIIA4 as described previously (14).
Transfection of cells with LERK 3 and LERK 7 DNA-Purified LERK 7-pEFBOS or LERK 3-pEFBOS DNA was transfected into CHO cells. Briefly, 2 ϫ 10 7 cells were suspended in 500 l of PBS and 10 g of LERK-pEFBOS DNA and 1 g of pSV2neo DNA added. After mixing and transfer to a 0.4-cm electroporation cuvette (Bio-Rad), the cells were electroporated at 270 V and 960 microfarads and the cells centrifuged through an fetal calf serum underlayer. Transfectant clones were selected in medium containing 600 g/ml G418. Individual clones were isolated and samples (5 l) of CHO cell supernatants from confluent cultures were dotted onto a nitrocellulose membrane, air-dried, and re-hydrated in blocking buffer (5% skim milk powder, 0.1% Tween 20 in PBS) prior to incubation with M2 anti-FLAG antibody at 1:1000 dilution. After washing the blot was incubated with horseradish peroxidase-conjugated rabbit anti-mouse Ig antibody (1:1000, Dako) in blocking buffer and following further washes developed with the ECL detection system (Amersham). Positives clones, indicated by signals above background, were retested by analysis on SDS-PAGE (12%) and Western blots prepared and probed as described above to confirm the presence of FLAG proteins of the expected molecular size.
Purification of LERK 7 and LERK 3 from CHO Cell Supernatants-FLAG-tagged fusion proteins were purified from CHO cell supernatants by affinity extraction on anti-FLAG mAb-agarose according to the manufacturers' protocol (IBI Kodak, New Haven, CT) followed by Mono Q (Pharmacia Biotech Inc.) ion exchange chromatography in 20 mM Tris, pH 8.5, 0.02% Tween 20 at 1 ml/min using a linear 40-min gradient from 0 to 600 mM NaCl and SE-HPLC (Superose 12, 10/30, Pharmacia Biotech) in 20 mM Tris, pH 7.4, 150 mM NaCl, 0.02% Tween 20 at 0.25 ml/min. The homogeneity, concentration, and identity of the purified proteins was confirmed by reverse phase-HPLC, SDS-PAGE, amino acid analysis, and N-terminal amino acid sequence analysis as described (14,30).
Fc Fusion Proteins of LERK 1, LERK 2, LERK 3, LERK 4, LERK 5, and LERK 7-Ligand fusion proteins, comprising soluble forms of LERK 1, 2, 3, 4, 5, and LERK 7 generated by fusion of the extracellular domain of each ligand with the hinge region and CH2 and CH3 regions of human IgG 1 were purified from supernatants of transfected cells as described previously (16,18,28). Fig. 1C depicts the LERK7-Fc construct as compared with native LERK 7 (Fig. 1A). The apparent protein concentration was estimated by BCA protein assay (Bio-Rad) and the purity of the fusion constructs assessed by SDS-PAGE and silver staining. Apparent molecular weights of the fusion proteins, determined by SDS-PAGE, were used to calculate apparent molar concentrations of the ligand-Fc protein stock solutions. Disulfide bond-mediated homodimerization of the fusion proteins was confirmed by analysis of the purified proteins by reducing and nonreducing SDS-PAGE.

Size Exclusion (SE)-HPLC Analysis of LERK⅐sHEK Complexes
The interaction between sHEK and LERK3-FLAG or LERK7-FLAG was analyzed by fractionation on a Superose 12 column (3.2 ϫ 300 mm, Pharmacia Biotech) at 50 l/min in 20 mM phosphate-buffered saline, pH 7.4 (PBS), in the presence or absence of 0.02%, Tween 20 (Pierce) using a Waters Model 600 Protein Purification system equipped with a Model 996 photo diode array detector (Waters, Box Hill, Australia).
The flow path of the LC system was modified to minimize the preand post-column volume to 1.0 l using 0.005 inch internal diameter PEEK tubing and a modified flow cell of the detector (generous gift from Dr. P. M. Young, Waters, MA) which allowed a direct connection of the column via a 80-mm length tubing. A zero dead-volume column holder to adapt the Superose 12 column was prepared from premoulded PEEK components in the workshop of the institute.

Sedimentation Equilibrium Analysis
The molecular weights of sHEK, LERK7-FLAG, and SE-HPLC-purified sHEK⅐LERK7-FLAG complexes were evaluated by sedimentation analysis of purified proteins or protein complexes which had been buffer-exchanged into PBS using a Fast Desalting column (3.2 ϫ 100 mm, Pharmacia Biotech). Soluble receptor-ligand complexes were purified on a Superose column as described above. Protein solutions (approximately 0.1 mg/ml) were subjected to sedimentation equilibrium in a Beckman XL-A analytical ultracentrifuge at 20°C and the protein concentration distribution recorded spectrophotometrically at 230 nm as described (31).

Analysis of the Interaction between sHEK and Its Proposed Ligands
The binding of various recombinant LERK fusion proteins to sHEK was analyzed on the BIAcore optical biosensor (Pharmacia Biosensor, Sweden) using purified sHEK or LERK7-FLAG derivatized CM 5 sensor chips. The immobilization of sHEK onto the sensor chip surface was performed essentially as described (14), LERK7-FLAG (47 g/ml in 20 mM sodium acetate, pH 4.5) was coupled at 5 l/min onto N-hydroxysuccinimide (0.05 M), N-hydroxysuccinimide-N-ethyl-NЈ-(diethylaminopropyl)carbodiimide (0.2 M)-activated sensor chips (45 l, 2 l/min) to yield an increase in the response level of 2500 -3000 response units. The affinity surface was regenerated between subsequent injections of samples with a 35-l injection of 50 mM 1,2-diethylamine, 0.1% Triton X-100, followed by two washes with BIAcore running buffer (Hepesbuffered saline, 0.005% Tween 20).
The interaction kinetics of LERK-binding to immobilized sHEK was analyzed from raw data of the BIAcore sensorgrams suitable for analysis using linear and nonlinear kinetic models included in the BIAevaluation software (32). All results recorded in this report were within the typical dynamic ranges of BIAcore measurements (k a , 10 3 -10 6 M Ϫ1 s Ϫ1 ; k d , 10 Ϫ5 -10 Ϫ1 s Ϫ1 , BIAcore catalogue 1996/97) and the BIAevaluation software. Single component kinetics was derived from the equations below.
(Eq. 1) R 0 is the response (R) at time t 0 , R eq the steady state response level (not necessarily reached in the sensorgram), and C the molar concentration of the analyte. The two component dissociation was derived from, Apparent affinities of LERKs 3, 4, 5, and 7 were also derived from equilibrium responses according to, where R eq and R max are the equilibrium and maximum response levels, respectively (32). In addition to the analysis of ligand binding to sensor chip-immobilized sHEK, the interaction between LERK 3 and LERK 7 with sHEK was studied in solution. A constant ligand concentration was incubated with increasing concentrations of the soluble receptor. The free ligand concentration (F LERK ), estimated from the BIAcore response of a known LERK sample (32) according to Ward et al. (31), and Scatchard transformation yielded equation 7.

Immunoprecipitation and trans-Phosphorylation Assays
LK63 cells (33) were maintained in Roswell Park Memorial Institute medium (RPMI 1640) supplemented with 10% fetal calf serum. The cells were passaged into fresh medium 24 h before the experiments. Following three washes with PBS, 0.5-ml samples of 2 ϫ 10 7 LK 63 cells/ml were incubated with 5 g/ml LERK3-FLAG, LERK7-FLAG, or with preformed complexes (30 min at room temperature) of either LERK3-FLAG (2 g) or LERK7-FLAG (2 g) and anti-FLAG mAb (2 g) for 1 h on ice. The cell suspension was warmed to 37°C for 10 min, then washed (ice-cold PBS) and the cell pellet lysed in 750 l of TBS, pH 7.5, containing 1.0% Triton X-100, 1 mM orthovanadate, 4 mM sodium fluoride, and 1 M chymostatin, leupeptin, antipain, and pepstatin. HEK was immunoprecipitated fron the lysates using IIIA4-Trisacryl beads (Sepracor/IBF, Villeneuve la Garenne, France) as described previously (7). The immunoprecipitates were subjected to 7.5% SDS-PAGE and electroblotted onto Hybond ECL nitrocellulose membranes (Amersham, Australia). The Western blots were probed using an anti-phosphotyrosine mAb (PY20, ICN Immuno-Biologicals, Costa Mesa, CA) and developed using the ECL Western blot analysis kit (Amersham, Aus-tralia). The membranes were stripped and re-probed with a rabbit polyclonal antiserum raised against sHEK (rabbit anti-HEK). In each case the precursor protein is depicted with an arrow leading to the final processed form. The original precursor protein is processed to remove the signal sequence and the hydrophobic glycosylphosphatidylinositol linkage sequence (cleavage site indicated by arrowhead) yielding the final glycosylphosphatidylinositol (GPI)linked form (Fig. 1A). The LERK7-FLAG is engineered to stop before the hydrophobic tail and the native N-terminal signal sequence is replaced with the IL-3 signal peptide and the FLAG epitope (B). C illustrates the LERK 7-Fc construct where the hydrophobic tail of the native sequence is replaced by the Fc and hinge regions of human IgG 1 . After processing, this yields the disulfide-linked homodimer (16). Relative responses obtained during the initial part of the association phase (characterized by a linear function, ln(dRU/dT) and of the dissociation phase (characterized by a linear function, ln(R 0 /R) were used to calculate kinetic rate constants according to linear interaction models.

Binding of Various LERK-Fc Fusion Proteins to Sensor
analysis also detected the interaction of various LERKs with HEK we compared the binding of bivalent, Fc-fusion proteins of LERK (1 to 5) and AL1/LERK 7 (Fig. 1C) to sensor chipimmobilized sHEK. Each ligand construct was injected at concentrations between 0.1 and 10 g/ml (approximately 0.8 -80 nM) across the sensor chip. A sample containing 10 g/ml of the recombinant human Fc fragment was used as a control. The relative binding response units of various samples at 10 g/ml are illustrated in Fig. 2, indicating comparable responses for LERK 3, 4, 5, and 7 which were considerably greater than the responses of LERK 1 and LERK 2. Apparent dissociation constants derived from equilibrium responses (Equation 4) at the four highest concentrations suggested a decreasing order of nanomolar affinities AL1/LERK 7 Ͼ LERK 3 Ͼ LERK 4 Ͼ Ͼ LERK 5 (data not shown). The interactions with LERKs 1 and 2 did not reach equilibrium responses in our experiments and hence precluded estimation of dissociation constants. Only background binding was seen with the control recombinant Fc construct on its own.
Interaction Kinetics of LERK 3-Fc and LERK 7-Fc Binding to sHEK-The similar binding of the LERK 3, LERK 4, and LERK 7-Fc fusion proteins to sHEK (Fig. 2) prompted us to study the binding kinetics of two of the proposed HEK ligands, LERK 3 (28)) and AL1/LERK 7 (14) in more detail. sHEK-BIAcore signals to serial dilutions of bivalent, LERK 3 or AL1/LERK 7-Fc fusion proteins are shown in Fig. 3 and were used to derive the kinetics of the interactions. As suggested (32), the early parts of the dissociation phase (530 -650 s, Fig. 2) were chosen for analysis. Examination of the raw data on the basis of a pseudofirst order reaction revealed a concentration-correlated change (not shown) in the apparent dissociation rate constants from 0.1 to 1.1 ϫ 10 Ϫ3 s Ϫ1 for LERK 3 (5-80.0 nM) and from 7.2 to 8.6 ϫ 10 Ϫ4 s Ϫ1 for AL1/LERK 7 (5-80.0 nM) and an increasingly inferior fit ( 2 ) to the assumed dissociation model (Table  I). Apparent association rate constants of k a ϭ 1.3 ϫ 10 5 M Ϫ1 s Ϫ1 for LERK 3-Fc and of k a ϭ 9.6 ϫ 10 5 M Ϫ1 s Ϫ1 for LERK 7-Fc (Table I) were estimated from the rate change of the BIAcore signal (Equations 2 and 3) with an improved fit to a singlecomponent association model (Table I). Again, an early part of the association phase was selected for the analysis. Nonlinear least square analysis performed on the same regions of the sensorgrams gave similar values (data not shown). Estimation of apparent equilibrium responses from the association data (Equation 3) and their analysis in a Scatchard format (Equation 5) yielded curvilinear plots (not shown) suggesting devia-tion of the LERK-Fc/HEK interaction from linear, single component kinetics.
The deconvolution of BIAcore raw data on the basis of a two-component dissociation (Equation 4) improved the correlation with the kinetic model substantially and a comparison of the two approaches by t test unambiguously favored the two component model (not shown). Dissociation rate constants (Table I) for LERK 3-Fc of 0.05 s Ϫ1 (fast) and 5.7 ϫ 10 Ϫ4 s Ϫ1 (slow) and for LERK 7-Fc of 0.04 s Ϫ1 (fast) and 5.7 ϫ 10 Ϫ4 s Ϫ1 (slow) were estimated. With the association rate constants suggested above, apparent low and high affinity constants were derived (K D ϭ k d /k a , Table I) for LERK 3-Fc (K D1 ϭ 3.8 ϫ 10 Ϫ7 M, K D2 ϭ 4.4 ϫ 10 Ϫ9 M) some 5-10-fold lower than the corresponding affinity constants for LERK 7-Fc (K D1 ϭ 4.1 ϫ 10 Ϫ8 M, K D2 ϭ 5.9 ϫ 10 Ϫ10 M). Although this evaluation undoubtly harbors some ambiguity with respect to the association kinetics, the identical treatment of the data for both LERK-Fc constructs justifies the use of these dissociation constants in a qualitative comparison of the LERK-Fc interactions with sHEK.
To address ambiguities which may affect the kinetic analysis of BIAcore data, including rebinding of dissociating ligand and heterogeneity of the receptor derivatized sensor surface (35), we examined the LERK-Fc interactions with sHEK in solution and estimated affinities by Scatchard analysis of free and receptor-bound ligand at equilibrium (31). Sensor chip-immobilized sHEK was used to monitor the abundance of free ligand in samples of 2 nM LERK 3-Fc (Fig. 4A) a Dissociation rate constants for a single-site interaction model were estimated according to R ϭ R 0 e Ϫk d (tϪt 0 ) , where R 0 is the response (R) at the start of the dissociation (t 0 ). b Dissociation rate constants for two parallel dissociation reactions were estimated according to R ϭ R 1 e Ϫk d1 (tϪt 0 ) ϩ (R 0 Ϫ R 1 )e Ϫk d2 (tϪt 0 ) , where R 1 and R 0 Ϫ R 1 are the contributions to R 0 from component 1 and 2, respectively. c Association rate constants were derived from R ϭ r 0 /k s (1 Ϫ e Ϫk s (tϪt 0 ) ), where r 0 is the initial binding rate using a secondary linear plot of k s ϭ k a C ϩ k d at various concentrations C, yielding a correlation for LERK 3-Fc of R ϭ 0.980 and for LERK 7-Fc of R ϭ 0.998. d Estimation of association rate constants for competing interactions of two (or more) ligands to the same immobilized receptor require knowledge of individual ligand concentrations. In the absence of defined actual concentrations of monomeric and dimeric ligands apparent equilibrium constants K D1 and K D2 for the low and high affinity interactions, respectively, were derived from the association rate constants k a and the dissociation rate constants k d1 and k d2 . order kinetics. The apparent affinity derived from this experiment for the LERK 3-Fc/HEK interaction with a K D ϭ 3.6 ϫ 10 Ϫ8 M (Fig. 4A) was significantly lower than the affinity of the LERK7-Fc/HEK interaction (K D ϭ 5.7 ϫ 10 Ϫ9 M, Fig. 3B).
Interaction Kinetics of LERK3-FLAG and LERK7-FLAG Binding to sHEK-The above discussed kinetic analyses indicated that reactions of both LERK-Fc proteins with sensor chip-immobilized sHEK apparently involved low-and highaffinity components. To evaluate if the bivalency of Fc-ligand constructs was a likely cause for the observed heterogeneity of the kinetics we performed binding experiments with monovalent forms of LERK 3 and LERK 7. Corresponding FLAG peptide-tagged fusion proteins were expressed in CHO cells and purified to homogeneity from culture supernatants of selected clones by anti-FLAG affinity chromatography and ion exchange HPLC. 2 The identity of the recombinant ligand proteins was confirmed by N-terminal amino acid sequence analysis.
A qualitative comparison of the BIAcore data, illustrating binding of increasing amounts of LERK3-FLAG (Fig. 5A) and LERK7-FLAG (Fig. 5B) to a HEK sensor chip, reveals marked differences in the kinetics of the two interactions. The LERK3-sHEK interaction is characterized by extremely fast on-and off-rates and comparable responses of LERK 3 or LERK 7 binding to sHEK were found only at approximately 30-fold higher LERK 3 concentrations in the applied sample. Kinetic analysis of the association and dissociation phases using a single component model yielded apparent association and dissociation rate constants of k a ϭ 4.8 Ϯ 0.13 ϫ 10 5 M Ϫ1 s Ϫ1 and k d ϭ 6.1 Ϯ 0.8 ϫ 10 Ϫ3 s Ϫ1 for LERK 7 and k a ϭ 3.7 Ϯ 0.9 ϫ 10 5 M Ϫ1 s Ϫ1 and k d ϭ 0.26 Ϯ 0.06 s Ϫ1 for LERK 3. Apparent dissociation constants K D ϭ 1. (Ϯ0.4) ϫ 10 Ϫ7 M for LERK3-FLAG were estimated, which compare well with the estimates of the low-affinity components of the corresponding LERK-Fc constructs (see Table I). Analysis of the raw data revealed good fits to linear, "one to one" interactions, yielding 2 values of 0.64 Ϯ 0.18 and 0.47 Ϯ 0.06 for the LERK 7 and LERK 3 reactions, respectively. The apparent equilibrium affinity constant for LERK 7 was substantiated by Scatchard analysis of the "in-solution" interaction (31), yielding an identical dissociation constant of K D ϭ 1.2 ϫ 10 Ϫ8 M (Fig.  5C). On the other hand, the affinity of the LERK3-FLAG interaction was too low to obtain reliable data by in-solution analysis.
Cross-linking of LERK-FLAG Proteins with Anti-FLAG MAb Alters Interaction with sHEK-We next addressed the notion that the differences observed in the binding of either FLAGtagged and Fc-tagged LERKs to sensor chip-immobilized HEK were due to increased avidity of the divalent, Fc-tagged ligands. To qualitatively examine this effect in situ, we assembled bivalent ligand-mAb complexes before or during BIAcore experiments by cross-linking FLAG-tagged LERK 7 (Fig. 6, A and B) and LERK-3 (Fig. 6, C and D) with the anti-FLAG mAb, M2.
The interactions of preformed LERK-FLAG⅐M2 mAb complexes (Fig. 6, B and D, graph e) with a sHEK sensor chip resulted in 3-5-fold increased BIAcore responses and markedly reduced off-rates of the ligand-antibody complexes compared with the non-complexed LERK-FLAG proteins (Fig. 6, A and C,  graph c), reflecting the increased size, and indicating an altered avidity, of the interacting complexes. To confirm this notion, we injected FLAG peptide, competing with the LERK-FLAG proteins for M2 mAb-binding sites, into the dissociation phase of LERK-FLAG⅐M2 mAb complex (Fig. 6B, graph f, 22Љ). A dramatically increased off-rate in this experiment confirmed that the suggested increase in avidity was dependent upon M2 mAb-mediated cross-linking of the LERK-FLAG proteins. Furthermore, injection of the M2 mAb into dissociating LERK-FLAG proteins at the end of the first injection cycle (22Љ) resulted in a pronounced rise of the BIAcore signals, likely due to binding of newly formed ligand-mAb complexes (Fig. 6, B  and D, graph d). The increase of the responses above the levels observed with the monovalent ligands in the first part of the sensorgram (between 21Љ and 22Љ), presumably reflects the increased size of the interacting ligand-mAb complexes. On the other hand, amplitude and slope of the response curve are also determined by the abundance and affinity of the ligand available for complex formation at the time of mAb injection. Since injection of equimolar amounts of LERK 3-Fc or LERK 7-Fc are expected to yield the same ligand concentrations at the end of the first injection cycles, differences in the amplitude of the response following mAb injection 22Љ (compare graph d in Fig.  6, B and D) must portray primarily the different affinities of the LERK-FLAG⅐M2 mAb complexes. In support of this notion, the dissociation curves and the response levels of pre-formed (graph e) and in situ formed (graph d) ligand⅐mAb complexes at the end of the second injection cycle (after 1090 s) were found identical (Fig. 6D) or very similar (Fig. 6B).
Taken together, this strictly qualitative analysis demonstrates that M2 mAb cross-linked LERK-FLAG dimers bind sHEK with increased avidity due to decreased dissociation rates. The resulting response curves are qualitatively very similar to the sensorgrams of the corresponding LERK-Fc fusion proteins (Fig. 3), suggesting that avidity plays a major role in the interaction kinetics of these ligand constructs.
Analysis of LERK-FLAG Binding to sHEK by SE-HPLC and SDS-PAGE-A key issue in interpreting the biological response to ligand binding is the stoichiometry of receptor-ligand complexes. Since kinetic considerations above strongly implied a one to one interaction we further examined receptor-ligand complexes by SE-HPLC, SDS-PAGE, and equilibrium sedimentation. To facilitate complete conversion of the monomeric ligands into ligand-receptor complexes, solutions of LERK3-FLAG or LERK7-FLAG were incubated for 2-12 h with a 10-fold molar excess of the receptor exodomain prior to analysis by micropreparative SE-HPLC in a physiological buffer. The absorption of eluting proteins was monitored at 215 nm and proteins in manually collected fractions analyzed by SDS-PAGE and silver staining (Fig. 7). The chromatogram and corresponding SDS-PAGE profile (Fig. 7A) indicate complete conversion of monomeric LERK7-FLAG into a sHEK⅐LERK7-FLAG complex which elutes as a shoulder (lane 1) on the ascending part of the major HEK peak (Fig. 7A, É, lane 2). No free LERK7-FLAG is detected at the corresponding elution position (Fig. 7A, å), confirmed by SDS-PAGE/silver staining of this sample (Fig. 7A, lane 3).
By contrast, SE-HPLC and SDS-PAGE profiles of the LERK3-FLAG containing sample (Fig. 7B) do not reveal a stable, high-molecular weight LERK 3-FLAG⅐sHEK complex. A protein peak eluting between sHEK (Fig. 7B, É) and LERK3- A and B) or LERK3-FLAG (panels C and D) with (sensorgrams e and f) or without addition (sensorgrams c) of cross-linking M2 anti-FLAG mAb (5 g/ml) were injected across a sHEK-derivatized sensor surface (21) followed by an subsequent injection (22) of buffer (sensorgrams c and e), M2 mAb (5 g/ml, sensorgrams d), or FLAG peptide (25 g/ml, sensorgrams f). For comparison, injections of buffer (21) followed (22) by M2 mAb or FLAG peptide (sensorgrams a and b, respectively) were performed in parallel experiments.

FIG. 6. Characterization of bivalent ligand binding by generation of ternary sHEK⅐LERK-FLAG⅐M2-mAb complexes in situ. Solutions (5 g/ml) of purified LERK7-FLAG (panels
FLAG and containing both, ligand and receptor (Fig. 7B, lane  2), as well as some apparently non-complexed LERK3-FLAG in the adjacent fraction corresponding to its original elution position (Fig. 7B, å, lane 3) suggest only a weak interaction between the two components.
However, the addition of cross-linking M2 antibody to the receptor/ligand-FLAG mixture results in the formation of a high-molecular weight complex, eluting as a shoulder in front of the elution position of the LERK3-FLAG⅐M2 complex (Fig.  7C, å). SDS-PAGE and silver staining suggests that this fraction contains M2-mAb, sHEK, and LERK3-FLAG (Fig. 7C,  lanes 1 and 2), whereas no free LERK3-FLAG is detected in the fraction corresponding to free ligand (lane 4).
Equilibrium Sedimentation Analysis-Equilibrium sedimentation was used to analyze the receptor-ligand complexes in detail and substantiate our findings from BIAcore and SE-HPLC experiments, inferring a stable one to one interaction between sHEK and LERK7-FLAG, but not LERK3-FLAG. To achieve a complete conversion of sHEK into a receptor-ligand complex suitable for sedimentation analysis, a 10-fold molar excess of LERK 7 was incubated with the soluble receptor for 1 h and the complex purified by SE-HPLC (inset to Fig. 8A). Rechromatography of this material confirmed a stable interaction and yielded a purified receptor-ligand complex (Fig. 8A) suitable for analysis by analytical ultracentrifugation. The linear dependence of the logarithm of the concentration upon the square of the radial distances for ligand, receptor, and ligandreceptor complexes is shown in Fig. 8B and suggests homogeneity with respect to the molecular weights. By contrast, we were unable to isolate a stable LERK 3-sHEK complex under identical conditions (not shown). Results illustrated in Fig. 8B and Table II reveal molecular sizes of 27,300 and 67,900 daltons for LERK7-FLAG and sHEK, respectively, in very good agreement with the apparent molecular masses derived from SDS-PAGE (LERK 7, 28,000, sHEK, 68,000, Fig. 7) and SE-HPLC (14). Furthermore, an unambiguous molecular mass assignment of 89,400 daltons for the receptor-ligand complex for sHEK (f), and 89,400 for the ligand-receptor complex (å). The theoretical correlations (indicated by solid lines) were derived for single species with the molecular weights given in Table II. implies a one to one interaction between LERK7-FLAG and sHEK, confirming our results from BIAcore and SE-HPLC studies.
Induction of HEK Phosphorylation in Ligand-treated Cell Cultures-In addition to the kinetic analysis of the LERK-sHEK interaction we compared the ability of either LERK 3 or LERK 7 to mediate transphosphorylation of HEK in LK63 cells which have been shown to express the receptor constitutively (7). LK63 cell cultures were incubated with buffer or solutions containing LERK3-FLAG, LERK7-FLAG, or pre-formed complexes of these ligands with anti-FLAG mAb, M2. In the latter samples the concentrations of LERK-FLAG proteins and M2 mAb were adjusted to provide divalent ligand constructs by occupancy of both binding domains of the mAb with ligand-FLAG construct. The HEK receptor was then immunoprecipitated from the cells and analyzed by Western blot analysis. The results in Fig. 9 illustrate the analysis with PY20 antiphosphotyrosine antibody (panel A) followed by reprobing the stripped blots with rabbit anti-HEK antibody. Phosphotyrosine analysis shows no significant differences between control (lane 5), LERK3-FLAG (lane 4), or LERK7-FLAG (lane 3)-treated samples. In contrast, incubation of cells with LERK3-FLAG⅐M2 complex induced a small but significant increase (lane 2) and with LERK7-FLAG⅐M2 complex (lane 1) gave a dramatic increase in phosphotyrosine content of HEK. Corresponding bands on the rabbit anti-HEK probed blots (panel B), show no significant difference in total HEK protein between the experimental groups.

DISCUSSION
The apparent "cross-talk" between various members of the LERK family and the EPH-like receptor HEK (18,22,24,28) including LERK 7 (14), prompted us to study receptor-ligand interactions between HEK and its proposed ligands, employing BIAcore and other analytical technologies. In our experiments we sought to (a) define the highest affinity ligand for the HEK RTK; (b) define the kinetics and stoichiometry of the receptorligand complex formation; (c) determine the biological response to ligands either in monomeric or dimeric forms.
Most of the studies of EPH-like receptors and their ligands carried out to date have been performed with divalent (Fc fusion) constructs of either ligand or receptor (Refs. in 4 and 37). We compared the binding of different LERK-Fc fusion proteins to HEK sensor chips and confirmed the suggested (18,28) cross-reactivity of all the tested LERK-Fc constructs with HEK (Fig. 2). In accord with these reports, the interaction between HEK and Fc constructs of LERKs 1 and 2 was distinctively weaker than binding of LERKs 3 and 4, which yielded in our experiments similar BIAcore responses to LERK 7. On the other hand, while the previously published affinities of LERKs 1, 2, and LERK 5-Fc for HEK are very similar (18,43, and 23 nM, respectively (18,24)), we could estimate apparent dissociation constants only from equilibrium responses of LERKs 3, 4, 5, and LERK 7-Fc (K D values of 5, 6, 24, and 3 nM, respectively, data not shown), whereas binding of LERKs 1 and 2 was too weak for a kinetic analysis. In addition, biphasic binding was reported previously only for the interaction between HEK-Fc with LERK 2, where a low affinity constant of 430 nM was found (18). Our comparative analysis of the association and dissociation phases of two candidate HEK ligands, LERK 3 (28) and LERK 7 (14), indicated a concentration-dependent increase of the apparent dissociation rate constants (not shown) and an increasingly poor fit to the assumed one-component dissociation model (Table I).
This significant deviation of the divalent LERK-Fc kinetics from a linear, single component interaction, suggesting a highaffinity interaction at low and a low-affinity interaction at high ligand concentrations confirms earlier studies by Hogg et al. (38) and Posner et al. (39) which demonstrate that kinetic models based on a one to one stoichiometry do not adequately describe the dissociation of bivalent solutes from surface-bound receptors.
The use of different approaches for the kinetic analysis of HEK/LERK-Fc interactions could explain the differences between the published data and our findings. A direct evaluation of kinetic data from BIAcore progress curves is likely to be more sensitive to changes in kinetic rate constants than "indirect Scatchard analysis" which relies on the use of labeled mouse anti-human IgG antibodies to detect receptor-Fc fusion proteins bound to ligand-transfected cells (15,18,22,24,28,37). Competitive binding experiments of the LERK-Fc/HEK interaction in solution (Fig. 4), which are not affected by immobilization artifacts and/or rebinding of dissociating ligand (31, 40) but rely on an "indirect" estimation of bound ligand or receptor (see "Materials and Methods"), gave no direct indication for biphasic kinetics from the slope of the Scatchard plots but yielded negative [B Lerk ]/[F Lerk ] values at low sHEK concentrations, thus indicating artifactually high responses in these samples. The interaction of bivalent LERK 7-Fc, containing only a single bound sHEK, via the remaining free LERK 7 moiety to the sHEK sensor surface, is a likely explanation for this artifact and confirms the concentration-dependent bivalency of the LERK-Fc/HEK interaction.
The comparative evaluation of all our binding data suggests that the bivalent, high-affinity interaction of two covalently linked binding domains of the ligand/Fc fusion protein with two adjacent, sensor chip-immobilized receptor molecules will compete at saturating ligand/Fc concentrations with a low-affinity monovalent interaction of a single binding domain with a single receptor molecule. Similar effects have been described for the analysis of mAb-antigen interactions (40 -45) and for the interaction of dimeric interleukin (IL) 6 with the sensor chipimmobilized IL-6 receptor-exodomain (31, 46).  a The values for M (1 Ϫ ) were derived from sedimentation equilibrium data (Fig. 8B).
b The values of were estimated as described previously (31) from sedimentation equilibrium data, assuming a value of 0.61 ml/g for bound carbohydrate and using values of 0.7267 ml/g and 0.728 ml/g calculated from the amino acid composition of sHEK and LERK 7, respectively. The value of for the complex was calculated with the assumption of zero volume change on association and equimolar ratios of the components. In other studies of EPH receptor-LERK interactions, the effect of solute bivalency has not been addressed. The necessity of ligand clustering for efficient receptor activation (15,22,24) seemed to warrant the use of bivalent receptor or ligand constructs. Such constructs were also used most recently in whole embryo in situ staining to confirm kinetic experiments performed with the same receptor-Fc constructs (37). On the other hand, it remains to be demonstrated that the interaction between membrane-bound ligands (or receptors) and Fc-tethered, bivalent receptors (or ligands) is a suitable system to study kinetics of physiological interactions of membrane-bound ligands and receptors (4). Our experiments indicate that the artificial bivalency of the ligand constructs obscures an unambiguous analysis of the reaction kinetics. In agreement with a report on the kinetics of the cell adhesion molecule CD 2 and its glycosylphosphatidylinositol-anchored ligand, CD 48 (47), we find that very low affinity, due to fast ligand dissociation is apparently increased by high avidity-binding of multimeric ligand aggregates.
By analyzing the binding of monovalent LERKs to sHEK, either in solution (Figs. 7 and 8) or using the (sensor) surfaceimmobilized receptor (Figs. 5 and 6), we were able to characterize the receptor-ligand interaction in detail. In situ crosslinking of the monovalent ligands with a mAb during BIAcore experiments (Fig. 6) and prior to SE-HPLC analysis of LERK3⅐sHEK complexes (Fig. 7) demonstrated qualitatively the effect of avidity on the interaction and confirmed the apparent higher affinities of bivalent ligand constructs. Differences in the dissociation phases of specific LERK interactions were concealed by the higher avidity of divalent binding components (Figs. 3 and 6) but have a major impact on the affinities of the monovalent ligands (Figs. 5 and 6). Due to an extremely fast off-rate, the interaction of monovalent LERK3-FLAG with the immobilized receptor is very weak (Fig. 5B), an observation confirmed in solution which indicated an unstable, transient LERK3-FLAG⅐sHEK complex (Figs. 6 and 7). By contrast, binding of LERK7-FLAG to sHEK was characterized by a 40 times lower off-rate and resulted in a stable receptorligand complex (Fig. 5, A and C) which was confirmed by kinetic analysis of sHEK binding to sensor chip-immobilized LERK 7, yielding an apparent K D of 7.2 ϫ 10 Ϫ8 M (data not shown). The dissociation rate of the LERK7-FLAG/sHEK reaction was low enough to allow purification of the ligand-receptor complex from solution (Figs. 7 and 8) and to facilitate its characterization by equilibrium sedimentation analysis (Fig.  8B). The demonstration of a 1:1 stoichiometry confirms our findings from BIAcore and SE-HPLC experiments, indicating that HEK has a single binding site for LERK 7 and explaining the necessity of ligand cross-linking for receptor activation and transphosphorylation demonstrated in this study (Fig. 9) and reported by others (21,22,48).
We attempted to produce a LERK 7-dependent cell line by transfection of FDCP-1 and Ba/F3 cells with HEK at several receptor densities. The inability of monomeric or complexed LERK 7 to rescue transfected cells after intermittent or complete IL-3 removal suggests that the HEK cytoplasmic domain does not mediate signals to induce proliferation or that cytosolic components for HEK-specific signaling are absent from these cells. On the other hand, this finding could also suggest that EPH-type RTKs regulate cell movement rather than cell growth.
In summary, our results clearly identify LERK 7 as the best candidate for a physiological HEK ligand. Despite very similar apparent affinity constants for the LERK 3 and LERK 7-Fc fusion proteins the interaction between their monovalent analogues and sHEK differ substantially by a markedly higher dissociation rate of LERK3-FLAG protein. Cross-linking of the dissociating ligands with anti-FLAG mAb decreases the dissociation rates and results in similar interaction kinetics for both ligands. Our results could suggest, that the reported interactions between some of the LERKs and HEK are influenced by the choice of the ligand construct. Extrapolating our observations to the in vivo situation, it seems likely that LERK 3 would function as an effective ligand only at very high receptor and ligand densities on opposing cell membranes, whereas a stable LERK7⅐HEK complex will persist at much lower receptor and ligand numbers. The demonstration of ligand and receptor gradients of HEK and AL1 homologues during neural development (13,49) would support a notion whereby the hierarchy of ligand affinities and characteristic ligand gradients would provide a subtle regulation of the migratory behavior of HEKpositive cells.