EphrinB1 Is Essential in T-cell-T-cell Co-operation during T-cell Activation*

Eph kinases are the largest family of receptor tyrosine kinases, and their ligands are ephrins (EFNs), which are also cell surface molecules. We have very limited knowledge about the expression and function of these kinases and their ligands in the immune system. In this study we investigated the effect of EFNB1 on mouse T-cells. EFNB1 and the Eph kinases it interacts with (collectively called EFNB1 receptors (EFNB1R)) were expressed on T-cells, B cells, and monocytes/macrophages. Some T-cells were double positive for EFNB1 and EFBB1R. Solid phase EFNB1 in the presence of suboptimal TCR ligation augmented T-cell responses in terms interferon-γ secretion, proliferation, and cytotoxic T lymphocyte activity but not interleukin-2 production. After T-cell receptor (TCR) ligation, EFNB1R congregated to TCR caps, and then both of them translocated to raft caps. This provides a morphological basis for EFNB1R to enhance TCR signaling. Further downstream of the signaling pathway, EFNB1R stimulation led to increased LAT (linker for activation of T-cells) phosphorylation and p44/42 and p38 MAPK activation. Similar to CD28 costimulation, EFNB1R costimulation was insensitive to cyclosporin A inhibition. On the other hand, unlike the former, EFNB1R costimulation failed to activate Akt, which is essential in triggering interleukin-2 production. Our study suggests that EFNB1 is pivotal in T-cell-T-cell costimulation and in reducing T-cell response threshold to antigen stimulation.

EphA9) and EphBs (EphB1 to EphB6) (1). 1 Ephrins (EFNs) 2 are ligands of Eph kinases; they are cell surface molecules as well and can be classified into A and B subfamilies. EFNAs (EFNA1 to EFNA6) are glycosylphosphatidylinositol-anchored proteins and bind to EphA members with loose specificity; EFNBs (EFNB1 to EFNB3) are transmembrane proteins and bind to EphBs, again with loose specificity (1). 1 An exception is EphA4, which can bind to EFNB2 in addition to EFNA members (2). EFNBs can also function as reciprocal receptors for EphB molecules and transduce reversely signals into cells (3). Most Eph kinases or ephrins have likely already been identified because the sequences from the human genome project have revealed 14 Eph entries and 8 EFN entries (4).
Because Eph kinases and their ligands are all cell surface molecules, they can only interact with each other when expressed on adjacent cells. Not surprisingly, the clearly demonstrated function of these receptors and ligands is to control accurate spatial patterning and cell positioning in the central nervous system (5,6) and during angiogenesis (7).
Some of the Eph kinases and their ligands are expressed on immune cells (8 -11); limited knowledge about their function in immune responses is available and is described as follows. We have previously reported that a kinase-defective Eph family member, EphB6, is capable of transducing signals into T-cells, probably through adaptor molecules (notably Cbl, Grb2, and CrkL) associated with its intracellular tail (12). Activation of EphB6 with solid phase anti-EphB6 mAb results in Jurkat cell apoptosis (12) or augmentation of normal human T-cell response to Ag stimulation (13). However, probably due to functional overlap of Eph kinases, no abnormal phenotype in thymocyte structure and subpopulation composition was found in EphB6 Ϫ/Ϫ mice (14). Munoz et al. (15) have reported that a few soluble recombinant EphAs and EFNAs interfere with T-cell development in thymic organ culture. Sharfe et al. (16) have reported that several EFNs, notably EFNA-1, inhibit T-cell chemotaxis. In this study, we investigated the expression of EFNB1 and its receptors in the immune system and the role of EFNB1 in modulating T-cell responses.

EXPERIMENTAL PROCEDURES
In Situ Hybridization-A 534-bp cDNA fragment of mouse EFNB1 cDNA from positions 336 -890 (accession number U12983) was fetched with PCR from a mouse embryonic tissue cDNA library and cloned into pGEM-4Z (Invitrogen). The resulting construct, pGEM-4Z-mB1, served to transcribe antisense probes with SP6 RNA polymerase or to transcribe sense probes with T7 RNA polymerase using digoxigenin RNA-labeling kits (Roche Diagnostics). In situ hybridization was carried out according to instructions from the kit manufacturer.
Generation of Mouse EFNB1-Fc-The coding sequence of the extracellular domains of mouse EFNB1 from positions 255-803 (accession number U12983) was cloned in-frame upstream of the human IgG 1 -Fccoding sequence in an expression vector pCMVhFc. The constructs and pcDNA3 were then transfected into CHO/dhfr Ϫ cells with Lipofectamine. The cells were cultured in selection medium (␣-minimum essential medium without ribonucleosides and deoxyribonucleosides containing 5% dialyzed fetal calf serum, 0.01 mM methotrexate, 0.8 mg/ml G418, and 0.1 mg/ml gentamycin). After 2 weeks of culture, stably transfected clones were handpicked. Fusion proteins were isolated from supernatants by protein A columns and verified by N-terminal peptide sequencing (Sheldon Biotechnology Center, McGill University, Montreal, Canada).
Lymphocyte Preparation and Culture-Cells were flushed out from the BALB/c mouse spleen, and red blood cells were lysed with 0.84% NH 4 Cl, as described elsewhere (17). The resulting cells were referred to as spleen cells. Spleen T-cells were purified by deleting mouse IgG (HϩL)-positive cells from spleen cells with T-cell columns according to the manufacturer's instructions (Cedarlane, Hornby, Ontario, Canada); the T-cell purity was about 90%. In some experiments the spleen cells were fractionated into CD4 and CD8 cells using Miltenyi magnetic beads (Miltenyi, Biotec, Auburn, CA), and the purity of CD4 and CD8 cells was about 95%. The cells were cultured in RPMI 1640 medium supplemented with 10% fetal calf serum, L-glutamine, and penicillinstreptomycin. Solid phase EFNB1 and anti-CD3 were prepared by coating 96-well Costar 3595 plates overnight with anti-mouse CD3 (clone 2C11) in PBS at 4°C followed by incubating EFNB1-Fc or normal human IgG (as a control, Southern Biotechnology, Birmingham, AL) of different concentrations at 37°C for 2 h. The plates were finally incubated on ice for 1-2 additional h before use.
Flow Cytometry-Flow cytometry was employed for measurement of EFNB1 receptor expression as well as EFNB1 expression in different cell populations after overnight culture. BALB/c spleen cells were stained with biotinylated EFNB1-Fc/streptavidin-PE or with goat anti-EFNB1 (R&D, Minneapolis, MN)/donkey anti-goat IgG-PE (Cedarlane); for the second color, anti-CD3-FITC (Clone 2C11), anti-B220-FITC (Clone RA3-6B2), anti-F4/80-FITC (Clone CI: A3-1) was used. For EFNB1, EFNB1 receptor (EFNB1R), and CD3 3-color staining, the spleen cells were first reacted with goat anti-EFNB1; after washing, phosphatidylethanolamine-donkey-anti-goat IgG, EFNB1-Fc, and biotin-anti-CD3 were reacted with the cells; after extensive wash again, the cells were finally incubated with FITC-goat-anti-human IgG and Quantum Red-streptavidin. In some experiments, CD4 and CD8 cells were first purified from spleen cells with Miltenyi magnetic beads (Miltenyi Biotech, Auburn, CA). Normal human IgG was used as a negative control for EFNB1-Fc in EFNB1R staining, and normal goat IgG was used as a negative control for EFNB1 staining. All the gatings in flow cytometry were set according to these negative controls. All these mAbs were from Pharmingen unless indicated otherwise.
[ 3 H]Thymidine Uptake Assay-T-cells were cultured in 96-well plates (Costar 3595) coated with different mAb or recombinant proteins, the concentrations of which were indicated in the figure legends. [ 3 H]Thymidine uptake was measured as described previously (18).
Cytotoxic T-cell Assay-The assay was performed as detailed earlier using spleen cells from TCR transgenic 2C mice (19). T-cells in 2C mice (H-2 b ) are predominantly L d -specific CD8 cells (more that 85% Thy1.2 ϩ cells are CD8 cells, and less than 3% Thy1.2 ϩ cells were CD4 ϩ cells) (20). Because of such a high frequency of L d -specific precursors, this assay has high sensitivity and a high signal versus noise ratio and was routinely used to assess CTL activity in our laboratory (17,19,21). 2C spleen cells (0.4 ϫ 10 6 cells/well) were stimulated with an equal amount of mitomycin C-treated BALB/c mouse spleen cells (H-2 d ) in flat-bottomed 96-well plates, which were pre-coated with EFNB1-Fc or normal human IgG (both at 10 g/ml). The cells were cultured in the presence of 10 units/ml IL-2 for 6 days. On day 6, cells receiving the same treatment were pooled and counted, and their CTL activity was measured by a standard 4-h 51 Cr-release assay, using 51 Cr-labeled P815 cells (H-2 d ) as targets at different effector/target ratios. The lysis percentage of the test sample was calculated as follows: % lysis ϭ (cpm of the test sample Ϫ cpm of spontaneous release)/(cpm of maximal release Ϫ cpm of spontaneous release).
Cytokine Measurement-Culture supernatants of T-cells placed in anti-CD3-and/or EFNB1-Fc-coated wells were harvested 1-3 days after initiation of culture. IL-2, IL-4, and IFN-␥ in the supernatants were quantified by enzyme-linked immunosorbent assay (R & D Systems) according to the manufacturer's instructions.
Confocal Microscopy-Five million BALB/c T-cells were reacted on ice for 30 min with 5 g of EFNB1-Fc and 1 g of biotinylated anti-CD3 (clone 2C11, hamster mAb). After washing, the cells were incubated with goat anti-hamster IgG (5 g/sample) for 30 min on ice. The cells were washed with cold PBS and transferred to warm PBS to start the cross-linking process at 37°C. The cells were then immediately fixed with 3.7% formalin. For TCR and EFNB1 receptor staining, the cells were reacted with streptavidin-Alexa Fluor 594 (1 g/sample) and goat anti-human IgG-Alexa Fluor 488 (1 g/10 6 cells) on ice for 30 min. For raft and EFNB1 receptor staining, the procedure was similar as described above, but cholera toxin-Alexa Fluor 594 (0.5 g/sample) was used in place of streptavidin-Alexa Fluor 594. The slides were examined under a confocal microscope.
Immunoblotting-MAPK and Akt activation was assayed using BALB/c spleen T-cells. Twelve-well plates were coated overnight with anti-CD3 (0.8 g/ml, 500 l/well) at 4°C. After washing, the wells were incubated with EFNB1-Fc, anti-CD28, or normal human IgG (NHIgG, both at 10 g/ml, 500 l/well) at 37°C for 1-2 h and then at 0°C for another 2 h. BALB/c spleen T-cells were seeded in these precoated plates at 5 ϫ 10 6 cells/well, and the plates were centrifuged at 228 ϫ g for 5 min to achieve rapid contact between the cells and the bottom of the culture wells. The cells were then cultured at 37°C for 30 min to 120 min, as indicated, before being harvested. The remainder of the immunoblotting procedure was detailed in our previous publication (17). Phosphorylated MAPK or Akt was detected with rabbit anti-phospho-p38 MAPK Ab or rabbit anti-phospho-Akt, respectively; total MAPK or Akt was revealed with rabbit anti-p44/42 MAPK Ab or rabbit anti-Akt Ab (all Abs were from New England BioLabs, Mississauga, ON, Canada). LAT (linker for activation of T-cells) phosphorylation was measured in BALB/c spleen T-cells (10 ϫ 10 6 cells/sample) , which were incubated with anti-CD3 (clone 2C11 hamster mAb; 1 g/ml) plus EFNB1-Fc, anti-CD28, or NHIgG (both at 5 g/ml) on ice for 1 h followed by 1 wash in cold PBS. Anti-hamster IgG and anti-human IgG (both at 1 g/ml) were then added, and the cells were incubated for 30 min on ice; cross-linking (2 min) took place by transferring the cells to 37°C. The cleared cell lysates were resolved by 10% SDS-PAGE and blotted onto polyvinylidene difluoride membranes. Phosphorylated LAT was detected with polyclonal anti-phospho-LAT (Cell Signaling, Mississauga, Ontario, Canada); total LAT was assayed with anti-LAT mAb (Clone 45, BD Transduction Laboratory, Mississauga, Ontario, Canada). All signals were revealed by enhanced chemiluminescence. Expression of EFNB1 and its receptor(s) at the protein levels on various leukocytes was investigated by flow cytometry. EFNB1 expression on isolated CD4 T-cells and CD8 cells were similar ( Fig. 2A) at about 33% according to anti-EFNB1 staining. B cells (B220 ϩ ) or monocytes/macrophages (F4/80 ϩ ) also expressed EFNB1 at 24.8 and 9.5%, respectively (Fig. 2B). Activation of T-cells (CD3 ϩ ) by solid phase anti-CD3 for 24 h did not increase but moderately reduced the EFNB1 expression (Fig. 2C). The EphBs that EFNB1 bound were collectively referred to as EFNB1Rs (EFNB1 can bind to more than one EphB subfamily member), and their expression was investigated using EFNB1-Fc staining. EFNB1R had similar levels of expression on CD4 and CD8 cells (28.3 and 32.0%, respectively; Fig. 2A). B cells expressed more EFNB1R than macrophages/ monocytes (30.8 and 10.2%, respectively; Fig. 2B). After activation by solid phase anti-CD3, CD3 ϩ T-cells moderately re-duced their EFNB1R expression (Fig. 2C). Because EFNB1 and EFNB1R are both expressed on T-cells, we wondered whether the same T-cell could express both. Fig. 2D showed the EFNB1 and EFNB1R expression on CD3 ϩ cells; about 13% of the T-cells are EFNB1 and EFNB1 double-positive, with about 8 and 38% T-cells single positive for EFNB1 and EFNB1R, respectively. It is to be noted that the whole T-cell population is a continuum for EFNB1 and EFNB1R expression with each T-cell expressing different level of EFNB1 and/or EFNB1R.

EFNB1 and EFNB1R Expression in
These results revealed that among the leukocytes examined, a significant percentage of T and B cells and small percentages of macrophages/monocytes expressed EFNB1 ϩ and EFNB1R ϩ . This provides a molecular basis for EFNB1R on T-cells to receive EFNB1 signaling from other T-cells as well as from non-T-cells such as B cells and monocytes/macrophages. T-cell activation did not lead to enhanced expression of EFNB1 at the protein level; the expression of EFNB1 mRNA in the dense spleen white pulp and peri-arteriole lymphoid sheath after the F1 spleen cell transfusion most likely reflected increased density of leukocytes in these regions.
EFNB1 Enhances T-cell Proliferation-Because about 20 -30% T-cells were EFNB1R ϩ in culture, it was logical to assume that EFNB1 might modulate T-cell function. To examine this possibility, both anti-CD3 mAb and EFNB1-Fc were anchored on solid phase. Anti-CD3 (at a suboptimal concentration) or EFNB1-Fc (at an optimal concentration) alone caused negligi- Two-color flow cytometry was used to assess EFNB1 or EFNB1R expression on T-cells, B cells, and monocytes/macrophages. T-cells were stained with anti-CD3, B cells were stained with anti-B220, and monocytes/macrophages were stained with F4/80. One-color flow cytometry was used to assess EFNB1 or EFNB1R expression on magnetic beadpurified CD4 or CD8 cells. All the cells were cultured overnight in RPMI 1640 medium before staining. A, EFNB1 and EFNB1R expression on magnetic beadpurified CD4 ϩ or CD8 ϩ cells. B, EFNB1 and EFNB1R expression on B220 ϩ B cells and F4/80 ϩ monocytes/macrophages. C, EFNB1 and EFNB1R expression on CD3 ϩ T-cells cultured in the absence (Medium) or presence of solid phase anti-CD3 (Activated). D, simultaneous staining of EFNB1 and EFNB1R on CD3 ϩ cells. The percentages of positive cells after deduction of background staining (NHIgG as control for EFNB1-Fc; goat IgG as anti-EFNB1 control) are indicated. All the experiments described in this figure were performed more than twice, and representative data are shown.
ble T-cell proliferation (Fig. 3C), but EFNB1-Fc dose-dependently induced the proliferation in the presence of suboptimal anti-CD3 (Fig. 3A). Next, T-cells were cultured in wells coated with an optimal amount of EFNB1-Fc and various amounts of anti-CD3. As shown in Fig. 3B, EFNB1-Fc but not NHIgG (a negative control for EFNB1-Fc) augmented T-cell proliferation when anti-CD3 was used at different concentrations. This result suggests that EFNB1R cross-linking reduces the T-cell response threshold and that cell surface EFNB1 on B cells, monocytes/macrophages, and T-cells might be able to enhance T-cell responses to antigens. We also compared the costimulation by EFNB1-Fc with that by anti-CD28 mAb (Fig. 3C). The costimulation mediated by EFNB1R was in the same order of magnitude as that mediated by the classical costimulatory molecule CD28. CD4 and CD8 cells responded to EFNB1 costimulation with similar magnitude (Fig. 3D), and this is consistent with the similar levels of EFNB1R expression on these two populations of cells. It is to be mentioned that the peak of EFNB1-Fc-enhanced proliferation was at day 3 after the culture, similar to that stimulated by optimal anti-CD3 (data not shown), indicating that observed EFNB1-costimulation was not due to a kinetics shift of proliferation. The results from this section suggest that EFNB1 receptors can enhance T-cell activation to antigen stimulation.
EFNB1 Enhances Lymphokine Production and CTL Activity-We next examined the effect of EFNB1 on T-cell effector functions, such as lymphokine production and CTL activity. Again, T-cells were stimulated with solid phase anti-CD3 alone (at a suboptimal concentration) or in combination with EFNB1-Fc or anti-CD28 (both at optimal concentrations and on solid phase). Three representative lymphokines, IL-2, IL-4, and IFN-␥, were measured from day 1 to day 3 after the stimulation. Anti-CD3 with or without NHIgG did not trigger lympho-kine production. Anti-CD28 costimulation drastically induced IL-2, IL-4, and IFN-␥, as expected (Fig. 4A); the IL-2 level declined since day 2 and the IL-4 level, since day 3, reflecting consumption of the lymphokines by the cells. EFNB1-Fc costimulation moderately increased IL-4 secretion and resulted in IFN-␥ secretion at a level comparable with that achieved by anti-CD28 costimulation; however, it did not stimulate IL-2 production at all.
The lack of IL-2 production was not due to a shift of secretion kinetics during EFNB1 costimulation because no production of IL-2 was observed during any time between days 1 and 3 after culture. This again demonstrated the qualitative difference between costimulation mediated by CD28 and EFNB1R.
When mixed lymphocyte reaction was elicited in EFNB1-Fccoated wells, CTL development was greatly enhanced (Fig. 4B), whereas control normal human IgG had no such effect. The enhancement was due to better T-cell activation but not due to better effector cell function or effector-target cell interaction because the solid phase EFNB1 was only present during the 5-day stimulation stage but not in the 4-h 51 Cr release assay. The results of this section demonstrate that EFNB1 can selectively enhance T-cell functions such lymphokine secretion and CTL activity.
Signaling Events in T-cells Stimulated by EFNB1-To understand the mechanism of EFNB1 costimulation, we first examined translocation of EFNB1R and TCR on the cell surface and their relationship to membrane lipid rafts immediately after TCR-cross-linking, which was achieved using biotinylated anti-CD3 followed by goat anti-hamster IgG. The TCR complex was stained by streptavidin-Alexa Fluor 594, in red; EFNB1R, was stained by EFNB1-Fc followed by antihuman IgG-Alexa Fluor 488, in green; and the lipid rafts in the T-cell membrane was stained by cholera toxin-Alexa Fluor 594, FIG. 3. Solid phase EFNB1 enhances the T-cell proliferation upon TCR stimulation. The affix "anti-" used in conjunction with mAbs is simplified as ␣ in this and all the other figures of this article. All the concentrations indicated in the figure legends represent those deployed during the coating procedure. NHIgG served as a control for EFNB1-Fc. In some cases the wells were only coated with anti-CD3 followed by PBS incubation (PBS) and were used as additional blank controls. BALB/c T-cells were cultured in wells coated with a suboptimal amount of anti-CD3 (0.8 g/ml) and different amounts of EFNB1-Fc (A), a fixed optimal amount of EFNB1-Fc (10 g/ml) and different amounts of anti-CD3 (B), or a suboptimal amount of anti-CD3 (0.8 g/ml) along with optimal amounts of anti-CD28 or EFNB1-Fc (both at 10 g/ml) (C). Magnetic bead-purified CD4 ϩ or CD8 ϩ cells were similarly stimulated with solid phase anti-CD3 (suboptimal) along with solid phase EFNB1-Fc, NHIgG, or anti-CD28 (D), as described above. The cells were cultured for 48 h, and their [ 3 H]thymidine uptake in the last 16 h was measured. The means Ϯ S.D. of the cpm from triplicate samples are shown. All the experiments described in this figure were performed more than three times, and representative data are shown.
in red. In resting T-cells, rafts, TCR, and EFNB2 were evenly distributed throughout the cell surface (Fig. 5). After 10-min of cross-linking with anti-CD3, TCR rapidly polarized and formed a cap in one end of the cell. EFNB1R also congregated, and they co-localized with TCR. Such co-capping lasted more than 20 min (data not shown). Control human IgG (in place of EFNB1-Fc) followed by anti-human IgG-Alexa Fluor 488 detected no signals in these cells (data not shown). After CD3 cross-linking, rafts underwent congregation and formed caps, but this process was slower than TCR capping and became obvious at 20 min. EFNB1R congregation preceded raft congregation, but eventually at 20 min, EFNB1R translocated into the raft caps. Taken together, these data indicate that TCR and EFNB1R first cocap and then both congregate to a raft cap on the cell surface after TCR-cross-linking. This provides a morphological base for EFNB1 to enhance TCR signaling, since now both TCR and EFNB1R are closely associated and located in aggregated rafts, which are scaffolds accommodating many signaling molecules.
We next examined a couple of signaling molecules involved in T-cell activation. LAT is a transmembrane protein, which is phosphorylated after T-cell activation, and its phosphorylation enables it to recruit additional signaling molecules (22). We found that EFNB1R cross-linking resulted in augmented LAT phosphorylation in spleen T-cells over that caused by TCR crosslinking (Fig. 6A). As expected, anti-CD28 costimulation also enhanced LAT phosphorylation but with stronger strength. Total LAT protein in each treatment showed no change, suggesting that EFNB1R could enhance the LAT function.
MAPK activity is modulated in other cell types when some Eph kinases are activated (22)(23)(24)(25), and this kinase is also an important downstream signaling event in T-cell activation. The MAPK activity in the presence and absence of solid phase EFNB1 costimulation in spleen T-cells was assessed. Because the solid phase anti-CD3 and EFNB1 depended on contact of T-cells in culture wells, the process is a much slower one than cross-linking TCR with antibodies in solution. As a consequence, phosphorylation of p38 MAPK and p44/42 MAPK only peaked around 2 h after putting cells into the wells. As shown in Fig. 6B, a combination of solid phase EFNB1-Fc and suboptimal anti-CD3 stimulation for 2 h led to increased p38 MAPK and p44/42 MAPK phosphorylation, a sign of their activation, whereas anti-CD3 at suboptimal concentration had little effect according to immunoblotting. The membranes were reprobed with anti-p38 MAPK Ab or anti-p44/42 MAKP Ab, respectively. The total p38 MAPK and p44/42 MAPK protein levels of in the third lane (protein from samples stimulated with both anti-CD3 and EFNB1) were equal (in the case of p38MAPK) or even lower (in the case of p44/42 MAPK, due to protein loading) than that in other control lanes. Therefore, these MAPKs were activated after EFNB1R cross-linking. To test whether such MAPK activation was relevant and necessary in EFNB1 costimulation, we used p38 and p44/42 MAPK-specific inhibitors to treat T-cells to be costimulated by EFNB1. Both inhibitors but not their nonfunctional structural analogue inhibited EFNB1 costimulation in terms of proliferation (Fig. 6C). This indicates that p38 and p44/42 MAPK activation is an integral and necessary part of the EFNB1R-signaling pathway.
Similarities and Differences of Signaling in Anti-CD28 and EFNB1 Costimulation-The vigorous IL-2 secretion in anti-CD28-costimulated but not in EFNB1-costimulated T-cells indicates that these two types of costimulation are not identical. We attempted to discover the similarities and differences in their signaling. First, we tested whether these two types of costimulation were additive. When anti-CD28 and EFNB1-Fc . Anti-CD28 (10 g/ml) and NHIgG were used as positive and negative controls, respectively. The culture supernatants were harvested from days 1 to 3, and lymphokines in the supernatants were measured by enzyme-linked immunosorbent assay. The means ϩ S.D. of duplicate samples are shown. B, effect of solid phase EFNB1-Fc on CTL development. 2C mouse spleen cells were mixed with an equal amount of mitomycin C-treated BALB/c mouse spleen cells and seeded in flat-bottomed 24-well plates, which were precoated with EFNB1-FC or NHIgG (both at 10 g/ml) or not coated (Medium). After 6 days CTL activity in the stimulated cells was measured by a standard 4-h 51 Cr-release assay using P815 cells as targets. The experiments were repeated at least three times, and data from a representative experiment are shown. The samples were tested in triplicate, and means Ϯ S.D. of the percentage of target cell lysis are presented. All the experiments described in this figure were performed more than three times, and representative data are shown.
were both used at optimal concentrations (both at 10 g/ml for coating), EFNB1 costimulation did not further enhance the anti-CD28-triggered costimulation in terms of proliferation measured on 3 consecutive days (Fig. 7A). We wondered whether the lack of additive effect between anti-CD28 and EFNB1-Fc was because proliferation was already driven to the maximal rate by anti-CD28. We, therefore, used a suboptimal concentration of anti-CD28 (5 g/ml, Fig. 7B) in the presence of an optimal EFNB1-Fc concentration; the latter still did not enhance the proliferation triggered by the former. This shows that EFNB1R costimulation cannot override CD28 costimulation, suggesting that the two types of costimulations are not

FIG. 6. LAT phosphorylation and p42/44 MAPK activation of spleen T-cells after EFNB1R cross-linking.
A, the phosphorylation of LAT. BABL/c spleen T-cells were first incubated on ice with anti-CD3, anti-CD28, EFNB1-Fc, or NHIgG as indicated, the cells were then reacted on ice with anti-hamster IgG and anti-human IgG, and cross-linking was conducted at 37°C for 2 min, which was the peak of LAT phosphorylation according to our pilot test. The phosphorylation of LAT was detected with anti-phospho-LAT Ab. The membrane was reprobed with anti-LAT Ab to detect the total LAT protein, which indicates similar loading in each lane. The experiment was repeated three times, and representative data are shown. B, immunoblotting of p38 and p44/42 MAPK. BALB/c spleen T-cells were added to wells coated with EFNB1-Fc or NHIgG (both at 10 g/ml) in the presence of a suboptimal amount of anti-CD3 (0.8 g/ml). The cells were harvested after 2 h and analyzed by immunoblotting. Arrows indicate signals of p38 phospho-MAPK and total p38 MAPK of the same membrane and signals of p44/42 phospho-MAPK and total p44/42 MAPK of the same membrane. The experiment was repeated three times and are representative data are shown. C, p38 and p44/42 inhibitors specifically inhibits EFNB1-costimulated T-cell proliferation. BALB/c spleen T-cells were preincubated for 1 h in complete culture medium containing the p38 MAPK-specific inhibitor SB203580, p44/42-specific inhibitor PD98059, its non-functional structural analog SB272474 (all at 10 M), or vehicle Me 2 SO (dimethyl sulfoxide (DMSO), 0.1%). The cells were then transferred to wells coated with EFNB1-Fc (10 g/ml), anti-CD3 mAb 2C11(0.8 g/ml), or both and cultured for 48 h. [ 3 H]Thymidine was added to the culture for the last 8 h, and thymidine uptake by the cells was measured. Results of a representative experiment from three similar ones are shown. totally independent. For T-cell activation, CD28 activates a Ca 2ϩ -independent pathway, which is insensitive to cyclosporin A (CyA) (26,27). We compared CD28 and EFNB1R costimulation in their sensitivity to CyA. As shown in Fig. 7C, the proliferation of T-cells activated by PMA and ionomycin was drastically suppressed by CyA, whereas that by anti-CD3 and anti-CD28 was not, as expected. Interestingly, EFNB1-Fc-costimulated T-cells proliferation was also resistant to CyA, suggesting that like CD28, ENFB1R also uses the Ca 2ϩ -independent pathway for T-cell activation. Despite the similarity and overlap between CD28 and EFNB1R signaling as revealed above, CD28 and EFNB1R costimulation must have differences that result in IL-2 production in the former and the lack of production in the latter. We, therefore, focused attention to events related to IL-2 production. Akt is a kinase in the phosphatidylinositol 3-kinase pathway, which is involved in CD28 costimulation and IL-2 production (28,29). We coated wells with anti-CD3 (suboptimal concentration) plus anti-CD28 or EFNB1-Fc (both at the optimal concentration of 10 g/ml) and incubated spleen T-cells in the wells after a brief centrifugation to allow cells to interact with the well surface. As shown in Fig.  7D, CD28 costimulation significantly augmented Akt phosphorylation, but EFNBR failed to do so, showing a distinct difference between the two costimulation pathways. DISCUSSION In this study, we examined the expression and function of EFNB1 and EFNB1R in leukocytes, with special attention FIG. 7. Similarities and differences of CD28 and EFNB1R costimulation. A and B, EFNB1-Fc did not further enhance anti-CD28 costimulation. BALB/c spleen T-cells were stimulated with suboptimal solid phase anti-CD3 (0.8 g/ml) along with various concentrations of solid phase anti-CD28 (10 g/ml in panel A; 5-20 g/ml in panel B); optimal concentrations of EFNB1-Fc (10 g/ml) on the solid phase was always present. [ 3 H]Thymidine was added to the culture for the last 16 h, and thymidine uptake by the cells were measured on days 2, 3, or 4 (panel A) or on day 3 (panel B). Results of a representative experiment from three similar ones are shown. C, T-cell proliferation triggered by CD28 and EFNB1R costimulation was insensitive to CyA inhibition. BALB/c spleen T-cells were stimulated by PMA (20 ng/ml) plus ionomycin (1 mg/ml), by solid phase anti-CD3 (0.8 g/ml) plus anti-CD28 (10 g/ml), or by solid phase anti-CD3 (0.8 g/ml) plus EFNB1-Fc (10 g/ml). CyA (250 nM) was present or absent as indicated. [ 3 H]Thymidine was added to the culture for the last 16 h of culture, and thymidine uptake by the cells were measured on day 3. Results of a representative experiment from three similar ones are shown. D, EFNB1R costimulation failed to phosphorylate Akt during T-cell activation. BALB/c spleen T-cells were stimulated with solid phase anti-CD3 (0.8 g/ml), anti-CD28 (10 g/ml), and EFNB1-Fc (10 g/ml) as indicated for 1 h at 37°C. The phosphorylated Akt (upper panel) was analyzed by immunoblotting. The same membrane was reprobed with Ab against total Akt (lower panel). Results of a representative experiment from two similar ones are shown. directed to T-cells. After TCR ligation, T-cell responses in terms of proliferation, lymphokine production, and CTL activity were augmented by solid phase EFNB1. At the molecular level, EFNB1R congregated to TCR and raft caps after TCR ligation, providing a morphological basis for EFNB1 to costimulate Tcells. EFNB1R cross-linking significantly augmented LAT tyrosine phosphorylation, creating possible docking sites to recruit additional SH2 domain-containing signaling molecules for full T-cell activation. Further downstream, MAPK activity was augmented. EFNB1R and CD28 costimulation shared similarities in that both used the Ca 2ϩ -independent pathway and were both resistant to CyA inhibition. However, they also differed in that the former failed to activate Akt and to produce IL-2, but the latter was robust in triggering these two events. Our results have so far revealed undocumented functions of EFNB1 and EFNB1R in T-cell biology.
The expression pattern of EFNB1 and EFNB1R on leukocytes has similarities compared with that of other EFNBs and EFNBRs we studied so far. Like EphB6 (13), EFNB2 and EFNB2R (25) and EFNB3 and EFNB3R (30), both EFNB1 and EFNB1R had significant expression on T-cells. The expression of both the ligand and receptor on the T-cells suggests that T-cells could receive costimulation from fraternal T-cells. The T-cell-T-cell collaboration is not a well investigated topic in immunology, but recently it has gained certain prominence. Wang et al. (31) have demonstrated that T-cell surface LIGHT, which is a ligand belonging to the tumor necrosis factor superfamily, can enhance T-cell response via its receptor HveA on T-cells in a pure T-cell culture system; this is an indication of T-cell-T-cell cooperation. We have demonstrated that in a pure T-cell culture system, soluble EphB6 and EFNB2, both of which are expressed on T-cells (12,25) and are the receptor and ligand to each other, can repress T-cell activation; this suggests that T-cell-T-cell interaction via EphB6 and EFNB2 is essential in the process (12,25). We have shown here that EFNB1 and EFNB1R are another pair of molecules involved in the T-cell-T-cell cooperation. In lymphoid organs, T-cells are tightly packed together with fraternal T-cells. It is conceivable that there is constant interaction between EFNB1R on the T-cells and EFNB1 on their neighboring T-cells to lower the T-cell response threshold, as our in vitro data suggested; naturally, the neighboring cells in this case could well include EFNB1-expressing B cells and monocytes/macrophages.
We have demonstrated that solid phase EFNB1 could costimulate T-cells. Which receptor did it use to achieve such an effect? As a general rule, EFNBs have loose specificity to EphB subfamily receptors. For EFNB1, its association with EphB1, -2, and -3 has been documented (32). Although EFNB1 does not interact with EphB6 directly (33), Freywald et al. (34) has reported that EphB1 dimerizes with EphB6, and EFNB1 indirectly activates EphB6 via EphB1. At least EphB2, -3, and -6 have expression in thymus and/or spleen (10,35,36). Therefore, in theory, these EphB members can mediate the observed T-cell costimulation by EFNB1. Indeed, similar to EFNB1, solid phase anti-EphB6 mAb can stimulate T-cell activation and proliferation (13); such similarity suggests that some of the EFNB1 effect might be mediated by EphB6. Because EFNBs are rather promiscuous in binding EphB receptors (37)(38)(39)(40) and both EFNB2 and EFNB3 have very similar effect on T-cells (25,30), it is conceivable that EFNB1, -B2, and -B3 achieve their effect through a repertoire of overlapping EphB family receptors.
The costimulation triggered by EFNB1 was qualitatively different from that of anti-CD28 in terms of T-cell activation marker expression and lymphokine secretion profile. Notably, it stimulated IFN-␥ but not IL-2 production. The failed IL-2 production was not due to consumption of IL-2 by the activated T-cells, as we measured IL-2 from day 1 to day 3, and at no time points was IL-2 detectable in the supernatants. This seems to be a norm rather than exception in Eph-mediated costimulation, as we have demonstrated that anti-EphB6 mAb (13), EFNB2-Fc (25), EFNB3-Fc (30) similarly fail to stimulate IL-2, but T-cell proliferation is robust. The drastic T-cell proliferation in the absence of IL-2 production is not unique to the Eph system. For example, signals through an inducible co-stimulator costimulate T-cell proliferation but do not lead to IL-2 production (36). CD2 (41), CD5 (42), CD9 (43), and CD44 (44) have the ability to costimulate T-cells. In the presence of an immobilized sub-mitogenic dose of anti-CD3, mAbs against these molecules all induce activation of naive T-cells, and the [ 3 H]thymidine incorporations after these co-stimulations are comparable with that of anti-CD28. Remarkably, the co-stimulation by these molecules produce limited amounts of IL-2. It is obvious that CD28 delivers a signal different from that derived from some of other costimulatory molecules including EFNB1. These studies also suggest that IL-2 is not obligatory for T-cells proliferation; its function in enhancing T-cell proliferation might be substituted by other cytokines in its absence.
In our effort to identify mechanisms responsible for the lack of IL-2 production in EFNB1-cosimulated T-cells, we found that Akt activation in EFNB1, but not anti-CD28, costimulated T-cells was compromised. Akt (also called protein kinase B) is a serine/threonine kinase that is recruited to the plasma membrane when its pleckstrin homology domain binds phospholipid products of phosphatidylinositol 3-kinase. Antibody cross-linking of CD28 induces Akt phosphorylation and activation even in the absence of signaling through the TCR complex-associated CD3 proteins (28,29), and such activation is phosphatidylinositol 3-kinase-dependent. After phosphorylation by phosphoinositide-dependent kinase 1 and a second kinase, which may be an integrin-linked kinase, at the membrane, Akt is activated (45); it then phosphorylates downstream targets, such as the inhibitory kinase glycogen synthase kinase 3␤, which is inactivated by Akt. Constitutively, activation of Akt is able to stimulate IL-2 production in mature CD28-deficient T-cells, but these cells are not proliferative (29). Thus, Akt seems to be pivotal in IL-2 production, but CD28 triggers other signals needed for T-cell proliferation. It is possible that the failure of Akt activation in EFNB1-costimulated T-cells is a reason for the compromised IL-2 production, whereas other EFNB1-trggered signaling events (including LAT phosphorylation and MAPK activation) are sufficient for T-cell proliferation. It will be interesting to investigate whether there is failed Akt activation in all other costimulation system where IL-2 production is lacking.
We have demonstrated that after TCR ligation, EFNB1R aggregated and formed caps within 10 min. This process preceded the aggregation of rafts, suggesting that EFNB1R, unlike Src kinases, LAT, and Ras (46,47), is not normally embedded in the lipid rafts but is translocated into rafts after TCR ligation. Such co-localization of EFNB1R and rafts provides a molecular basis for EFNB1R to enhance T-cell signaling, because rafts are rich in signaling molecules, upon which TCR depends for signal transduction to achieve T-cell activation. Indeed, we have demonstrated that LAT phosphorylation was significantly enhanced after EFNB1R cross-linking, and this finding is consistent with the translocation of EFNB1R to raft where LAT is located. Conceivably, LAT phosphorylation creates additional docking sites for interaction with other signaling molecules, and T-cell activation is, thus, enhanced. How EFNB1R connects to LAT phosphorylation remains to be investigated. Further downstream, Ras-and Rac-signaling path-ways need to be mobilized for full T-cell activation. Ras activation leads to activation of p44/42 MAPK kinases, which in turn results in the synthesis and activation of various transcription factors. On the other hand, activation of Rac and Cdc42 small G proteins leads to p38 MAPK activation, which is essential for cytoskeleton reorganization (24,48). Such reorganization is now known to be critical for the T-cell signaling (49). We have found that the activities of both p44/42 and p38 MAPK are enhanced in the presence of EFNB1 costimulation, and this is consistent with the roles of these MAPK in T-cell activation. It has been reported that EphB2 activation results in inhibition of p44/42 MAPK in neuronal cells (50) and that EphA activation leads to inhibition of this kinase in several cell lines of endothelial and epithelial origin (51). Obviously, these reports deal with cell types different from those in our study. The consequence is different as well. In neuronal cells, endothelial cells, and epithelial cells, Eph activation does not induce cell proliferation, whereas in T-cells it does. Further comparative studies on EFNB1R signaling in immune and nonimmune cells will be interesting.
We have reported here that EFNB1 could costimulate T-cells through cell surface Eph kinases, the largest family of receptor tyrosine kinases. There are several lines of evidence showing that our finding in this study is not an in vitro artifact. 1) We have generated EphB6 Ϫ/Ϫ mice and revealed that these mice had compromised delayed type hypersensitivity reaction, which is a T-cell-mediated response (data to be published elsewhere); because EphB6 can associate with EFNB1 (33), this finding corroborated our conclusion that EFNB1 is important in T-cell biology. 2) More directly, we have generated EFNB1 knockdown mice, and preliminary results showed that these mice were significantly compromise in both T-cell development and function (data not shown). The important roles of Eph kinases and their ligands, EFNs, in the immune system are emerging; more studies in this new dimension of T-cell biology are warranted.