T Cell Proliferation-Augmenting Activities of the Gene 3 Protein Derived from a Phage Library Clone with CD80-Binding Activity¹

Tcells require two signals; one is delivered from TCR and the other is generated through costimulatory molecules such as CD28, CTLA4, CD80, and CD86 to proliferate and produce cytokines (1, 2). CD28 and CTLA4 are expressed on T cell surface (3). CD80 and CD86 are expressed on APC (4, 5). All of these molecules belong to members of the Ig super family. The kinetic study on the expression of these molecules suggested that CD28 may be responsible for CD86, and CTLA4 is responsible for CD80, respectively, although CD28 and CTLA4 cross-reacted not only with CD86 but also with CD80 (2, 6, 7, 8). Site-directed mutagenesis analysis revealed the distinct structural requirement of CD28 or CTLA4 for binding to CD80 or CD86 (9, 10, 11, 12, 13, 14, 15, 16, 17). CD80 and CD86 bind to CD28 with low avidity, resulting in transducing a positive signal for T cell activation such as IL-2 production (1, 11). On the other hand, they also bind the high affinity ligand CTLA4, which is preferentially present on activated T cells (12). The blockade of CTLA4 engagement with its natural ligands by Fab fragments of anti-CTLA4 mAbs enhanced T cell proliferation both in vitro and in vivo (2, 18, 19). Cross-linking CTLA4 with mAbs inhibited T cells stimulated through CD3 and CD28 (20). CTLA4 knockout mice demonstrate a massive lymphoproliferative response (21). Thus, CTLA4 appears to deliver a negative signal, suggesting the important roles of costimulatory molecules in the immunoregulation.

The information about the critical residues of receptors and counter-receptors for their binding and functions is essential to develop antagonists or agonists for the regulation of these biological reactions. To do this, site-directed mutagenesis is a powerful approach. Photoaffinity labeling is also useful to know the topological relationship between critical residues of ligand and receptor molecules. However, since the binding groove in general is formed by the three-dimensional (3D)³ folding structure which determines the function of the molecule, these residues scatter in alignment of amino acid sequences. Therefore, it is often difficult to illustrate the whole image of 3D structure of antagonists or agonists judging from short amino acid stretches of ligand and receptor molecules (22, 23). To overcome this difficulty, antireceptor Abs or antiligand Abs, which inhibit the binding of these molecules, may be useful to illustrate the 3D structure motif which may properly fit in the binding groove. Phage display library is powerful to identify the binding epitopes recognized by mAbs (22, 24, 25, 26, 27, 28).

In our previous studies, we isolated a phage clone, F2, by panning with anti-CTLA4 conformation recognizing mAb (29). The F2 clone showed binding ability to CD80 and costimulation of the T cell proliferation in vitro (24). The unique sequence of 15 amino acids (F2 motif) was inserted in the gene 3 protein (g3p) molecules of the F2 phage clone. To characterize the molecular nature of the F2 motif, we purified the g3p from the F2 phage clone by HPLC and analyzed the relationship between CD80-binding activity and its immunoregulatory activity.

We show here that F2-g3p bound to CD80 but not to CD86 by ELISA and surface plasmon resonance (SPR). F2-g3p inhibited the binding of CTLA4 with CD80 but not with CD86, whereas F2-g3p weakly inhibited the binding of CD28 with CD80. When hen egg lysozyme (HEL)-primed lymphocytes were stimulated with HEL in the presence of F2-g3p in vitro, the cell proliferation was highly potentiated by F2-g3p as observed in the F2 phage clone but not in other phage proteins such as F2-g8p. T cells did not proliferate by F2-g3p alone, indicating the costimulatory nature of F2-g3p. The T cell augmenting activity of F2 clone or F2-g3p was abolished when they were preincubated with CD80-Ig before addition to cultures. These results suggested that F2-g3p preferentially blocked the negative signaling through the interaction of CTLA4 with CD80.

The fd phage display library was constructed as described (30). The purification of phage proteins was performed as described previously (31). Briefly, the fd phages (7 mg/ml) were incubated with 1% SDS at 37°C for 20 min. The DNA and coat proteins were separated by size-exclusion chromatography on a HiLoad Superdex 200 (26/60) column (Pharmacia, Uppsala, Sweden). Elution was conducted by 50 mM Na₂SO₄ containing 5 mM citric acid and 0.1% SDS (pH 3.0) at the flow rate of 800 μl/min. The peaks were monitored at 280 nm by LC-10AD liquid chromatograph (Shimazu, Tokyo, Japan), equipped with SPD-10A UV-VIS detector (see Fig. 1,A). SDS-PAGE was performed as described previously (24). The g3p fraction was resolved as a single band at 44 kDa, whereas the g8p fraction was resolved as a broad band by SDS-PAGE (see Fig. 1 B). The g8p fraction might contain a minute amount of the other phage proteins. The fractions were extensively dialyzed with PBS (pH 7.3) before use.

The recombinant molecules were murine costimulatory molecules fused with human IgG Fc. The CD80-Ig, CD86-Ig, CD28-Ig, and CTLA4-Ig were prepared as described previously (24). These are divalent molecules.

The hamster anti-mouse CTLA4 mAb (UC10–4F10–11; PharMingen, San Diego, CA), the hamster anti-mouse CD28 mAb (37.51; PharMingen), human IgG Fc fragment (hFcγ; Jackson ImmunoResearch, West Grove, PA), alkaline phosphatase (AP)-conjugated streptoavidin (Leinco Technologies, Ballwin, MO), AP-conjugated goat anti-human IgG (Zymed, San Francisco, CA), and AP-conjugated goat anti-hamster IgG (Zymed) were purchased. Anti-fd phage serum was prepared by hyperimmunizing BALB/c mice i.p. with 10 μg of K7 phage clone (control clone with the unrelated motif) in Freund’s incomplete adjuvant.

ELISA was performed as described previously (24, 29). The binding activities of phage proteins were as follows: plastic plates (Nunc, Copenhagan, Denmark) were coated with the phage proteins (30 ng/40 μl/well) in 50 mM Tris-HCl (pH 7.5) and 150 mM NaCl (TBS) containing 0.02% NaN₃ at 4°C for 1 h. Blocking was done by using 350 μl of 1% BSA (Sigma, St. Louis, MO). The plates were washed five times with TBS containing 0.05% Tween-20 (TBS/Tween) and once with TBS. Varying doses of mAb or fusion proteins were added, followed by detection with AP-conjugated goat anti-hamster IgG or by AP-conjugated goat anti-human IgG (H+L) at a dilution of 1:250. The substrate (85 μl) consisted of 1 mg/ml p-nitrophenylphosphate (Wako, Osaka, Japan) and 10% diethanolamine (Wako) in TBS. Absorbence was read at 405 nm by microplate photometer (InterMed NJ-2011, Tokyo, Japan).

CTLA4-Ig or CD28-Ig was coated (10 ng/40 μl/well) on 96-well plastic plate. After blocking plates with BSA, biotinylated anti-CTLA4 mAb or biotinylated CD80-Ig (10 ng) premixed with varying concentrations of phage proteins at 4°C for 1 h was added to wells of plates. The binding was monitored by AP-conjugated streptoavidin.

The kinetic studies of the binding between CTLA4-Ig and anti-CTLA4 mAb were done by SPR using BIAcore 2000 system (Pharmacia, Piscataway, NJ) at 25°C. CTLA4-Ig (10 μg/ml) was immobilized onto the sensor chip CM5 according to amine coupling protocol, and the unreacted sites were blocked with 1 M Tris-HCl (pH 8.5). The association reaction was initiated by injecting 5, 10, 20, or 30 nM anti-CTLA4 mAb. The analyte injection was performed at a flow rate of 5 μl/min. The dissociation reaction was done by washing with PBS containing 0.005% Tween-20. At the end of the cycle, the sensor chip surface was regenerated for 30 s with 0.1 M glycine-HCl buffer (pH 2.5). The association (M⁻¹s⁻¹) and dissociation constants (s⁻¹) were calculated using BIAcore system software (BIAevaluation, Version 2.1).

BALB/c mice were immunized s.c. with 50 μg HEL in Freund’s incomplete adjuvant 2 wk before or immunized i.p. with 10 μg of WT phage in PBS 4 wk before use. Cell cultures were conducted as described (24). Briefly, lymph node cells (1.5 × 10⁵/0.2 ml/well) were stimulated with 3 μg/ml HEL in the presence or absence of varying doses of various phage clones or phage proteins (g3p or g8p) in round-bottom 96-well plates (Iwaki Glass, Tokyo, Japan). In the case of WT-primed mice, spleen cells (1.5 × 10⁵/0.2 ml/well) were stimulated with or without 5 μg/ml of WT in the absence or presence of phage proteins in flat-bottom 96-well plates. Three days later, T cell proliferation was monitored by pulsing cells with 0.5 μCi [³H]thymidine (Amersham, Arlington Heights, IL) for the final 18 h.

To characterize the CD80-binding activity of the F2 motif, phage proteins were purified by gel permeation chromatography (Fig. 1,A). The second peak was the g3p fraction, which was resolved as a single band with m.w. of 44 kDa by SDS-PAGE (Fig. 1 B). The third broad peak consisted of the g8p as a major protein containing a trace of other phage proteins (31).

The F2 motif of 15 amino acids was inserted in the g3p of the F2 phage clone. First, we tested the reactivity of the g3p and the g8p derived from the F2 phage clone (F2-g3p and F2-g8p) to anti-CTLA4 mAb (UC10–4F10–11) by ELISA. F2-g3p (30 ng/well) was coated on the plates and tested on the reactivity to anti-CTLA4 mAb or anti-CD28 mAb. As shown in Fig. 2,A, anti-CTLA4 mAb bound to F2-g3p. Neither anti-CD28 mAb nor normal hamster IgG bound to F2-g3p. The binding of anti-CTLA4 mAb to F2-g3p was dependent on the doses of F2-g3p (Fig. 2,B). F2-g8p was not recognized with anti-CTLA4 mAb (Fig. 2 C). The anti-CTLA4 mAb did not bind to the g3p or g8p derived from control phage (K7 or L4) that were randomly selected from the phage library (data not shown). These results indicated that F2-g3p bore the specific motif (F2 motif) that was recognized by anti-CTLA4 mAb.

The anti-CTLA4 mAb recognized the CTLA4 conformation at the binding site to CD80 (24). This result suggested that F2-g3p might be a CTLA4 mimic with the binding activity to CD80 or CD86 molecules. Therefore, we tested this possibility by ELISA. First, the binding activity of the CD80-Ig to F2-g3p or F2-g8p was examined. The CD80-Ig (200 ng/well) were added to plastic wells that had been coated with varying doses of F2-g3p (Fig. 3,A) or F2-g8p (Fig. 3,B). Following the addition of AP-conjugated anti-human IgG and substrates, the development at A₄₀₅ was recorded at various reaction times. The CD80-Ig bound to F2-g3p in a dose-dependent fashion (Fig. 3,A) but showed no binding activity to F2-g8p (Fig. 3,B). The binding activity of the CD86 to F2-g3p or F2-g8p was examined by the same protocol (Fig. 3 C). The CD86-Ig showed no binding activity to F2-g3p, F2-g8p, K7-g3p, or K7-g8p.

We attempted to characterize the binding reaction of F2-g3p to CD80 by SPR. To examine the binding of CTLA4-Ig to anti-CTLA4 mAb, the CTLA4-Ig (10 μg/ml) was immobilized onto the sensor chip CM5 (Fig. 4,A). The association reaction was initiated by injecting the varying concentration of anti-CTLA4 mAb. The analyte injection was performed at a flow rate of 5 μl/min. The binding of CTLA4-Ig (Fig. 4,B) or F2-g3p (Fig. 4,C) to the immobilized CD80-Ig was also examined under identical experimental conditions. The gel permeation chromatography showed that the preparation of F2-g3p as well as CTLA4-Ig and CD80-Ig contained significant amounts of high m.w. aggregates. To demonstrate the binding activity of F2-g3p, we tentatively calculated the binding parameters using a simple one site-binding model from the response curve. Therefore, each binding parameter might be overestimated in terms of affinity (see Discussion). These results were summarized in Fig. 4 D. The bindings of CTLA4-Ig/CD80-Ig and the anti-CTLA4 mAb/CTLA4-Ig showed the almost equal K_d values (0.31 and 0.48 nM) respectively. In contrast to these reactions, the F2-g3p preparation associated to CD80-Ig at a slower rate (k_on, 0.63 vs 2.4 × 10⁵s⁻¹M⁻¹) and it dissociated from CD80-Ig at a faster rate (k_off, 2.7 vs 0.76 × 10⁻⁴s⁻¹), resulting in a lower avidity (K_d = 4.29 nM) in comparison with the CTLA4-Ig preparation.

To examine the inhibitory activities of F2-g3p on the interaction of costimulatory molecules, binding of CD80 to CTLA4 or to CD28 was assessed by ELISA in the presence or absence of F2-g3p or control phage molecules. The biotinylated CD80-Ig was preincubated with varying concentrations of F2-g3p, F2-g8p, or anti-CTLA4 mAb for 1 h at 4°C and then added to CTLA4-Ig-coated plates (Fig. 5,A). Anti-CTLA4 mAb strongly inhibited the binding of CD80-Ig to CTLA4-Ig. F2-g3p also inhibited the binding of CD80-Ig to CTLA4-Ig and showed complete inhibition at 50 ng/well, whereas F2-g8p showed no inhibitory activity on this binding. Neither F2-g3p nor F2-g8p showed any inhibition on the binding of CD86-Ig to CTLA4-Ig-coated plates while the anti-CTLA4 mAb completely inhibited this interaction (Fig. 5,B). In contrast, F2-g3p showed only partial inhibitory activity on the binding of CD80-Ig to CD28-Ig-coated plates (Fig. 5,C). At 16 ng/well of F2-g3p, CD80/CTLA4 interaction was strongly inhibited, but CD80/CD28 interaction was the least effected. In CD28 binding systems (Fig. 5, C and D), anti-CD28 mAb (37.51) did not exhibit the complete inhibition even at 50 ng/well. It is uncertain whether this result may be due to the characteristics of the 37.51 clone or CD28-Ig. F2-g8p showed no blocking effect. CD86-Ig to CD28-Ig binding was not inhibited by either F2-g3p or F2-g8p, but it was inhibited with anti-CD28 mAb (Fig. 5 D). These results suggested a distinct structural requirement for CD80 binding domain to either CTLA4 or CD28, in which F2-g3p could distinguish each other.

The inhibitory activity of F2-g3p on the interaction of CD80 with CTLA4 suggested that F2-g3p might affect the T cell activation stimulated with Ags. We have previously shown that the F2 phage clone highly augmented the proliferative response of WT-primed spleen cells in comparison with the responses induced with WT or control phage clones (24). To examine whether the augmenting activity of the F2 phage is due to costimulatory nature, the HEL-primed lymph node cells were stimulated with 3 μg/ml of HEL in the absence or presence of various phage clones including the F2 phage clone. The WT lacks g3p molecules due to the frame-shift of g3p codon. The L4 clone was randomly selected from the phage library and bore the unrelated peptide motif in L4-g3p. As shown in Fig. 6 A, the F2 phage (5 μg/ml) strongly augmented the HEL-induced T cell proliferation approximately four times higher than the control responses. In contrast, L4 or WT showed no effects in comparison with the response stimulated with HEL alone.

To confirm that this effect was induced with F2-g3p but not with the other phage proteins, WT-primed spleen cells were stimulated with 5 μg/ml of WT in the presence or absence of various phage proteins including F2-g3p in vitro (Fig. 6 B). The cells stimulated with WT alone exhibited a response of 18,000 dpm. The background responses without WT were around 4000 dpm. In the absence of antigenic stimulation, F2-g3p and F2-g8p showed no effect in comparison with the background response. In contrast, F2-g3p strongly augmented the T cell proliferation induced by the stimulation of WT. This effect was not induced with F2-g8p, K7-g3p, or K7-g8p.

To test the timing of the F2 addition relative to antigenic stimulation, various phage clones were added on day 0, 1, or 2 after the HEL stimulation (Fig. 6 C). The augmentation was induced only by the simultaneous addition of both HEL and F2 phage clone. This effect was not induced by the addition of L4 or WT phage clone. The experiments using F2-g3p also gave the same results (data not shown). Thus, the F2 motif inserted in g3p molecules showed an immunoregulatory activity which is characteristic for the responses regulated with the T cell costimulatory molecules.

HEL-primed lymph node cells showed the augmented T cell response when stimulated with HEL in the presence of F2-g3p or CD80-Ig alone in vitro (Fig. 7,A). K7-g3p or human IgG did not induce this effect. To determine the involvement of CD80-binding activity of F2 motif in the augmentation of T cell response, HEL-primed cells were stimulated with HEL in the absence or presence of F2 phages (0.5 μg/ml) preincubated with concentrations of CD80-Ig or hFcγ ranging from 0.1 to 1.0 μg/ml. CD80-Ig strongly inhibited the F2-mediated augmentation of T cell response, whereas control hFcγ or CD86-Ig (data not shown) had no effects (Fig. 7 B). This effect was also shown using human CD80-Ig (data not shown). These results indicated that the binding of F2-g3p to CD80-Ig neutralized each other’s augmenting activity and suggested the involvement of CD80-binding activity of F2 in the augmentation of T cell response.

The F2 phage clone was isolated by anti-CTLA4 mAb from a phage display library (29). We demonstrated that F2 phage strongly augmented the T cell proliferation of spleen cells derived from WT-immunized BALB/c mice (24). As the peptide motif is inserted in the g3p molecule of phage, we studied here whether the g3p molecule derived from the F2 phage clone was responsible for the augmenting effect through the regulation of T cell costimulatory signals. In accordance with the characteristics of F2 phage clone, F2-g3p bound to CD80 but not to CD86 (Fig. 3). The SPR analysis also showed the strong binding of F2-g3p to CD80. The binding valency of the molecule affects the strength and stability of the binding interaction (15, 32, 33). Greene et al. (32) reported that the CTLA4-Ig as a divalent molecule showed the complex kinetics with the high (K_d = 1 nM) and low avidity (K_d = 150 nM) phases to the immobilized CD80-Ig in comparison with the monovalent CTLA4 (K_d = 211 nM), which might be due to the different valencies of the complex. They demonstrated that the the association reaction of the monovalent or divalent CTLA4-Ig to CD80-Ig was very rapid with k_on value of 4.4 × 10⁵ or 8.9 × 10⁵ M⁻¹S⁻¹, respectively. This value was similar to the k_on value (2.4 × 10⁵ M⁻¹S⁻¹) of CTLA4-Ig preparation used here (Fig. 4). Morton et al. (15) reported that the k_on value was estimated to be 1.4 × 10⁶ M⁻¹S⁻¹. In our experimental condition, F2-g3p associated to CD80 at a slower rate than CTLA4-Ig. Greene et al. (32) also determined the dissociation reaction of monovalent or divalent CTLA4-Ig from the immobilized CD80-Ig as 91 × 10⁻³ or 4 × 10⁻³ S⁻¹ of k_off, respectively. In our results, the k_off of the CTLA4-Ig preparation from CD80-Ig was calculated to be 7.6 × 10⁻⁵ S⁻¹ (Fig. 4). This slow rate may be due to the aggregated form of the CTLA4-Ig preparation and rebinding in dissociation reaction (32). Morton et al. (15) reported that the k_off value was estimated to be 5.8 × 10⁻⁴ S⁻¹. The K_d values on the binding of divalent CTLA4-Ig to CD80-Ig were estimated to be comparable between Greene’s (1 nM), Morton’s (0.4 nM), and ours (0.3 nM). The k_off of the F2-g3p preparation from CD80-Ig was calculated to be 2.7 × 10⁻⁴ S⁻¹, indicating that F2-g3p dissociated from CD80-Ig at a faster rate than the CTLA4-Ig preparation used here. The K_d value on the binding to CD80 was eventually one order lower than that of the binding of CTLA4-Ig to CD80-Ig or CTLA4-Ig to anti-CTLA4 mAb (Fig. 4 D).

F2-g3p inhibited the binding of CTLA4 to CD80 but not to CD86. It is to be noted that F2-g3p inhibited the interaction of CD80 with CTLA4 completely, but with CD28 only partially (Fig. 5). Furthermore, F2-g3p did not inhibit the interaction of CD86 with CTLA4 or with CD28. Site-directed mutagenesis analysis demonstrated that the exon 3-encoded IgC domain of CD80 is essential for CD28/CTLA4 binding and suggested that CD28 and CTLA4 bind to the same site on CD80 (10). In this context, our results suggested that F2-g3p finely distinguished the conformational structural differences of the binding site of CD80 to CTLA4 from that of CD80 to CD28. Furthermore, it was reported that the IgV domain of human CD86 is sufficient to costimulate T lymphocytes (16). Although our results were obtained from murine molecules, it is conceivable that the binding site of F2-g3p to CD80 may be located to IgC domain but not IgV domain because F2-g3p showed no interference on the binding of CD86 to CTLA4/CD28.

F2-g3p as well as the F2 phage clone showed remarkable augmenting activity on the T cell proliferation in the presence of antigenic stimulation (Fig. 6). The augmentation was induced only by the simultaneous addition on day 0 of both HEL and F2 phage in vitro (Fig. 6,C). F2-g3p or F2 phage did not induce T cell proliferation without antigenic stimulation, indicating their T cell costimulatory characteristics. CTLA4-Ig inhibits T cell activation because CTLA4-Ig binds CD80 as well as CD86. On the other hand, F2-g3p binds CD80 but not CD86. It is conceivable that F2-g3p directly stimulates T cells through their CD80 molecules. However, it is unlikely because CTLA4-Ig as well as F2-g3p should augment the T cell proliferation if the direct stimulation through CD80 molecules is possible. We prefer an alternative hypothesis that F2-g3p may selectively block the negative signal through CTLA4/CD80 interaction but not through CTLA4/CD86. The binding of CD80 to CD28 was slightly blocked at 16 μg/ml of F2-g3p (Fig. 5, A and C) while the binding of CD80 to CTLA4 was strongly inhibited, indicating that F2-g3p interfered only partially with the interaction of CD80 to CD28. Furthermore, F2-g3p did not inhibit the interaction of CD86 to CD28 (Fig. 5,D). Thus, the positive signal via CD80/CD28 and CD86/CD28 might become dominant in comparison with the negative signal, eventually resulting in the augmentation of T cell proliferation. Not only F2-g3p but also CD80-Ig (Fig. 7) and CD86-Ig (data not shown) enhanced the T cell proliferation induced with HEL in vitro. These effects may also be attributed to the inhibition of CTLA4 binding for negative regulation. F2-g3p was relatively more effective in the enhancing activity than CD80-Ig. This might be due to the concentration of natural counter receptor molecules (CD28) in the microenvironment of early response phase, i.e., high concentration of CD28⁺ cells competitively reduce the chance for the binding of cell surface CTLA4 to CD80-Ig. In contrast, CD80 expression is very low at early response phase and could not work as an effective competitor for the binding of F2-g3p to CTLA4.

The blockade of CTLA4 engagement with its natural ligands by Fab fragments of anti-CTLA4 mAbs enhanced T cell proliferation both in vivo and in vitro (8, 18, 20). The magnitude is 3 to 4 times higher than the response without the anti-CTLA4 mAb (20). In the T cell proliferation system induced with suboptimal dose (0.6 μg/ml) of anti-CD3 mAb, F2-g3p enhanced the response at a comparable degree with the response induced with anti-CTLA4 mAb (data not shown). In the case of F2-g3p inhibition but not anti-CTLA4 mAb inhibition, CD86 could interact with CTLA4 to deliver the negative signal even in the presence of F2-g3p. Thus, these results suggested that the negative signal through CD86/CTLA4 seemed to be less effective.

Phage library has been used to determine the continuous epitope (sequential determinant) mapping, and the synthetic peptides of these motifs were examined for their biological activities (22, 26, 28). However, we demonstrated that CTLA4-like 3D motifs grafted onto a few surface molecules of a phage particle strongly potentiated the immune response against the phage. Most viruses/phages have highly organized structures with repetitive surface molecules and influence the elicited immune response (34). This study suggests that the F2 sequence may be useful in developing the virus vaccine.