CTLA-4–B7 Interaction Is Sufficient to Costimulate T Cell Clonal Expansion

C57BL/6 mice were purchased from the National Cancer Institute (Rockville, MD). CD28-deficient mice (26), backcrossed to C57BL6j for six generations, were provided by Dr. T.W. Mak (University of Toronto, Canada).

Anti–CTLA-4 mAb 4F10 (9) and anti-CD3 mAb 2C11 (27) were provided by Dr. J.A. Bluestone (University of Chicago, Chicago, IL). Anti–B7-1 mAb 16.10A1 (28) was provided by Dr. H. Reiser (Dana-Farber Cancer Institute, Boston, MA). Another anti–B7-1 mAb 3A12 was produced in this laboratory and has been described (29). Anti-CD45RB mAb C363.16A (30) was provided by Dr. K. Bottomly (Yale University, New Haven, CT), anti-CD11c mAb HB224 (31) was a gift from Dr. R. Steinman (Rockefeller University, New York). Intact mAbs were purified by a protein G column, while the Fab fragments were prepared using a Fab kit (catalog no. 44885) from Pierce Chemical Company (Rockford, IL) according to the manufacturer's instructions. The purity of both intact and Fab fragments was verified by 10% SDS-PAGE.

Construction and production of fusion proteins, murine CTLA-4Ig, and CD28Ig have been fully described (32).

COS cells were transfected with wild-type and mutant B7-1 molecules by DEAE–dextran methods, as has been described (33). The procedure for Chinese hamster ovary (CHO)¹ cell transfection has also been described (34).

Expression of wild-type and mutant B7-1 was determined by flow cytometry using anti–B7-1 mAbs, either 16.10A1, or 3A12, as primary antibody, and FITC-labeled goat anti–hamster IgG was used as second-step reagent, as has been described. Fusion proteins, either CD28Ig or CTLA-4Ig, were used as first-step reagents. The binding to wild-type and mutant B7-1 was determined by using FITC-labeled goat anti–mouse IgG as second-step reagents.

CD4 T cells were purified from spleen cells as described (35). The purity of CD4 T cells isolated was >95%. Given numbers of the CD4 T cells were stimulated with anti-CD3 mAb (0.25 μg/ml) using mitomycin C (100 μg/ml, 37°C for 1 h)–treated CHO cells transfected with either FcR (CHOFcR) or FcR plus murine B7-1 (CHOFcRB7). The cultures were maintained for 42 h, and were pulsed with [³H]TdR (1.25 μCi/ well) for an additional 6 h. The cultures were then harvested and the [³H]TdR incorporated into the genomes was determined by a betaplate counter. The blocking mAbs were added at the beginning of the culture. The data presented are means of duplicates with variation <15% of the mean.

In some experiments, CD4 T cells were separated based on their cell surface expression of CD45RB. In brief, CD4 T cells were incubated with anti-CD45RB mAb C363.16A (5 μg/ml) for 1 h at 4°C; the unbound antibodies were washed away. After adding DYNAL beads coated with goat anti–rat Ig for 1 h, the cells that bound to C363.16A were collected by magnet. The cells bound to the beads were eluted by adding 2 mg/ml of normal rat Ig at 37°C for 1 h and used as naive cells. The cells that did not bind DYNAL beads in the first place were depleted once more with DYNAL beads and used as memory cells. The CD45RB profiles were determined by flow cytometry using either C363.16A (CD45RB) or medium (control) as the first-step reagent and FITC-labeled mouse anti–rat Ig as second-step reagent.

Site-directed mutagenesis of murine B7-1 was achieved by PCR using oligonucleotides carrying the desired mutations. B7Y that carries a mutation from Y >A at position 201 has been described (32). Mutant B7W (W >A at position 88) was created as follows. A B7-1 fragment containing the desired mutation was produced by PCR using GCTCGAGAAGCTTATGGCTTGCAATTGTCAG as forward primer, and CTTAGCCTCGGGCGCCAC TTTTAGTTTCCC as reverse primer. After digestion by AvaI, the PCR product was ligated to the AvaI–XbaI fragment of murine B7-1 (32). B7L (L109 >A) was created by a two-step PCR. First, a mutant B7-1 fragment containing the desired mutation (L >A) is generated using GCTCGAGAAGCTTATGGCTTGCAATTGTCAG as forward primer, and CGACGCAGCTGTAGGTGCCCCGGTCGCTAGCGACCAGGC as reverse primer; another fragment covering the remaining sequence of B7-1 open reading frame is amplified using CACCTACAGCTGCGTCGTTCAAAAGAAGGA as forward primer and CGAATTCTAGAACTAAAGGAAGACGGTCT as reverse primer. Second, a mixture of the two B7-1 fragments were used as template to amplify the full-length B7-1 mutant; GGCTCTAGATTCCTGGCTTTCCCCATCATG was used as forward primer and GTCAGCCATCTCGAGTTTTTCCCAGGTGAAGTC was used as reverse primer. The PCR condition has been described (32).

To examine the function of CD28 and CTLA-4, we compared wild-type and CD28-deficient T cells in their response to costimulation by B7 in vitro. We produced CHO cells transfected with either FcR (CHOFcR), which cross-links anti-CD3 mAb on T cell surface, or FcR in conjunction with murine B7-1 (CHOFcRB7) (23), and tested their costimulatory activity for clonal expansion of CD4 T cells from wild-type and CD28(−/−) mice. As expected, wildtype CD4 T cells respond to costimulation by B7-1 (Fig. 1,a). Surprisingly, T cells from CD28-deficient mice also mount a positive response to B7-1, although the response is approximately fivefold lower than that of wild-type T cells (Fig. 1,b). The critical role of B7-1 is confirmed, as an anti– B7-1 mAb 3A12 (29) significantly blocks responses of wildtype and CD28-deficient T cells. In addition, CD4 T cells from wild-type and CD28-deficient mice bearing naive or memory markers both respond positively to costimulation by B7-1 (Fig. 2). These results demonstrate that B7-1 can costimulate T cell clonal expansion by a CD28-independent mechanism.

To verify the contribution of CTLA-4, we tested whether an anti–CTLA-4 mAb affects the function of B7-1. We prepared Fab of the mAb by papain digestion and depletion of Fc-containing immunoglobulin. As shown in Fig. 3, in SDS-PAGE under reducing conditions, intact mAb preparation consists of two peptides of ∼50 kD and ∼28 kD, the respective molecular masses of the heavy and light chain of IgG. Fab preparation run as one band of ∼28 kD, which indicate that the Fab used for the study is free of detectable intact Ig. As shown in Fig. 4, the anti–CTLA-4 mAb 4F10 completely blocks the function of B7-1. This blocking does not require cross-linking of CTLA-4, as the Fab fragment of the mAb also blocked the function of T cells. Surprisingly, both intact and Fab fragment of anti–CTLA-4 mAb inhibit response of CD28(+/+) T cells. The ability of both intact and Fab of anti–CTLA-4 mAb 4F10 to block the function of wild-type T cells indicates that CTLA-4 is critically involved in B7-1–mediated costimulation for wildtype T cells.

A potential caveat of this interpretation is that anti– CTLA-4 may negatively signal T cells and thus prevent T cell clonal expansion regardless of the presence of B7-1. To rule out this possibility, we tested whether anti–CTLA-4 mAb inhibits T cell response to costimulation by CD44H, which we recently identified as a CD40L-induced costimulatory molecule that promotes T cell clonal expansion by a CD28-independent mechansim (36). As shown in Fig. 5, anti–CTLA-4 mAb 4F10 does not inhibit clonal expansion of T cells when CHO cells expressing murine CD44H are used as costimulator. These results confirm that anti– CTLA-4 mAb is a specific inhibitor of T cell response to costimulation by B7-1.

Anti–CTLA-4 mAbs show different effects on T cell proliferation depending whether costimulation is provided by anti-CD28 or by antigen-presenting cells (9, 10, 20–24). To avoid any question that may be associated with antibody blocking studies, we have carried out a systematic site-directed mutagenesis of B7-1 molecules in both IgC-like (32) and IgV-like domains in search of a mutant B7 that retains binding for CTLA-4 but not CD28. As shown in Fig. 6,a, saturating amount of CTLA-4Ig binds B7W, which carries a mutation from W to A at position 88 of B7-1 (number starts at the first methionine encoded by B7-1 cDNA), at levels comparable to those of wild-type B7, although careful titration of CTLA-4Ig reveals that it takes about ninefold more CTLA-4Ig to achieve such saturation for B7W (Fig. 6,c). In contrast, B7W does not bind CD28Ig (Fig. 6 b). Mutant B7 L109 >A (B7L), which binds CTLA-4Ig less well than B7W, binds CD28 even better than wild-type B7-1. On the other hand, mutant B7Y, which has a mutation at position 201 from Y to A, does not show detectable binding to either CD28Ig or CTLA-4Ig, as we have reported (32). These results demonstrate that B7-1 is recognized by the two receptors asymmetrically, and the selective loss of CD28 binding activity in B7W cannot be accounted for simply on the basis of low avidity of CD28–B7 interaction. The preferential effect of B7W >A mutation on CD28 binding is consistent with an earlier mutagenesis analysis involving human B7-1 expressed on fibroblasts (37), although it is apparently at variance with mutagenesis analysis using recombinant human B7-1 fusion protein (38), perhaps owing to differential glycosylation of B7-1.

Therefore, we transfected B7-1, B7W, and B7Y into CHOFcR cells, and produced stable cell lines expressing comparable levels of FcR and wild-type or mutant B7-1 (Fig. 7,a). These cell lines were used to determine the receptors responsible for the costimulatory activity of B7-1. As shown in Fig. 7,b, both wild-type and CD28-deficient T cells respond to costimulation by wild-type B7-1 and B7W, but not to that by B7Y. Thus, CTLA-4 binding activity is sufficient to costimulate T cell proliferation. Moreover, wild-type T cells respond to B7-1 significantly better than to B7W, as ∼20-fold more CHO cells expressing B7W are required to achieve a level of T cell proliferation similar to that costimulated by wild-type B7-1. This difference is diminished when CD28(−/−) CD4 T cells are used, thus confirming the inability of B7W to stimulate CD28. The function of B7W is blocked by anti–B7-1 and anti–CTLA-4 mAbs (Fig. 8).

CTLA-4 plays an important role in regulating T cell function, as evidence by the phenotypes of CTLA-4–deficient mice (17, 18) and by the effect of anti–CTLA-4 mAbs (9–11). The biological consequences of B7–CTLA-4 interactions have not been studied in detail. Here we seek to measure directly the effect of B7–CTLA-4 interaction on T cell clonal expansion. Our results demonstrate that B7–CTLA-4 interaction is sufficient to induce T cell clonal expansion in vitro.

First, T cells from CD28-deficient mice react to costimulation by B7-1, although the proliferative response is generally 5 to 10-fold lower than wild-type T cells. This response is mediated by CTLA-4 on T cells, because anti-CTLA-4 mAb, either intact or Fab, blocks B7-1–dependent T cell proliferation. Surprisingly, in numerous experiments, anti– CTLA-4 mAb also completely inhibit B7-1–mediated proliferation of wild-type T cells. Because this mAb does not interfere with T cell proliferation when CD44H is used as costimulator (Fig. 5), it is unlikely that this mAb inhibits T cell response to B7-1 by negative signaling. These results suggest that B7–CD28 interaction may not be sufficient to promote T cell proliferation. This is paradoxical, given the fact that anti-CD28 mAb used as the prototypic costimulator for T cell clonal expansion (39–43), although it has not been demonstrated whether B7–CD28 interaction is sufficient to costimulate T cell proliferation. Because anti-CD28 mAb 9.3 has ∼100-fold higher avidity than B7-1Ig for CD28 (7), this paradox can be reconciled by higher avidity of the anti-CD28 mAb. It should be noted that in an earlier report, Green et al. (25) failed to detect CD28-independent function of B7-1 in promoting T cell clonal expansion in the presence of PMA. This difference can be explained by possible lack of expression of CTLA-4 in the CD28(−/−) cells used by Green et al. Alternatively, our assay based on copresentation of TCR ligand and costimulator B7-1 on the same cells (44) is ∼20-fold more sensitive than the earlier study; the difference can be reconciled on the basis of sensitivity of the assays.

Second, mutant B7-1, B7W, which lacks detectable CD28 binding activity, promotes proliferation of both wild-type and CD28-deficient T cells. For better quantitation, we measured CD28 binding using fusion protein CD28Ig. However, our preliminary data indicated that CD28 expressed on CHO cells as a transmembrane protein also fail to mediate adhesion to B7W (data not shown), much like human B7-1 with a mutation of the corresponding amino acid (37). Furthermore, while mutant B7W stimulates wild-type T cells less well than wild-type B7-1, it is comparable to wild-type B7-1 if CD28(−/−) T cells are used as responder. Thus, the major functional difference between B7W and wildtype B7-1 is in their interaction with CD28. Our functional analysis of B7-1 mutant indicates that B7–CTLA-4 binding is sufficient to costimulate wild-type and CD28deficient T cells. The use of B7-1 mutants allows us to dissect the function of CTLA-4 without the caveats of anti– CTLA-4 mAb.

A recent study suggests that another yet unidentified receptor for B7 may exist on NK cells (45). An intriguing question is whether the CD28-independent response to B7-1, as reported here, is mediated by a yet unidentified receptor for B7-1. Although it is impossible to rule out a possible contribution of other unknown receptors, CTLA-4 is at least an essential part of the signaling complex, because the CD28-independent function of B7-1 is completely blocked by anti–CTLA-4 mAb.

Taken together, our results presented in this report demonstrate that B7–CTLA-4 interaction promotes T cell clonal expansion. Moreover, optimal costimulation by B7 requires both CD28 and CTLA-4, as has been proposed by Linsley et al. (11). The mechanism of the cooperation between CD28 and CTLA-4 is still unclear at the present. B7–CD28 interaction may be critical for enhancing IL-2 production. Thus, it has been reported that CD28 deficiency leads to a drastic reduction in IL-2 production (25, 26, 37). Our analysis of IL-2 production in CD28-deficient T cells (data not shown) also supports this concept. In addition, mutant B7W, which bind CTLA-4 but not CD28, promotes T cell clonal expansion but barely enhances IL-2 production (data not shown). Therefore, unlike CD28-B7 interaction, CTLA-4–B7 interaction may not enhance T cell response by increasing the production of IL-2. Our results are inconsistent with the notion that CTLA-4 is a negative regulator for T cell activation, although it is conceivable that CTLA-4 may downregulate T cell activation on other circumstances. Several important factors, such as expression and cellular localization of CTLA-4 may influence the function of CTLA-4. Although it is clear that the expression and cellular expression of CTLA-4 is under stringent control (47), the regulatory mechanisms are not well understood. This lack of knowledge has made it difficult to predict the function of CTLA-4 under physiological conditions.

Several recent studies have shown that anti–CTLA-4 mAbs can be potent enhancers of immune response to defined foreign antigen, self-antigens, and tumors (22–24). Although these results were interpreted in the light of blocking negative signal from CTLA-4, it is also possible that these results are achieved by enhancing positive signal transduction from CTLA-4. The mutant B7-1 molecules described in this study may help to resolve the function of CTLA-4 in vivo. The answer to this question will be critical for immune intervention targeted at CTLA-4, as it will determine whether signal transduction initiated by CTLA-4 should be amplified or inhibited.

We thank Dr. Tak Mak for CD28-deficient mice, Dr. Jeff Bluestone for anti–CTLA-4 monoclonal antibody, Dr. Stan Vukmanovic for critical reading of the manuscript, and John Hirst for assistance in flow cytometry.