B7 Expression on T Cells Down-Regulates Immune Responses through CTLA-4 Ligation via R-T Interactions¹

Productive T cell activation, proliferation, and development of full effector function require two signaling events. The first, Ag recognition, is the engagement of the Ag-specific TCR with the peptide Ag/MHC complex on the surface of the APC. This initial activation event is generally insufficient to result in T cell expansion, cytokine production, or cytolytic capacity but results in the up-regulation of B7-1 (CD80) and B7-2 (CD86) on APCs that provide costimulation (signal 2) to the T cell (1). Although B7 molecules on APCs can either costimulate T cell responses by the ligation of CD28 or inhibit T cell responses by the ligation of CTLA-4 on T cells (2, 3, 4, 5), the function of constitutive and activation-induced B7 expression on T cells has been less well studied. These studies were undertaken to determine the functional consequences of B7 up-regulation that occurs on T cells during an in vivo alloresponse.

Our data indicate that T cells from B7-1/B7-2-deficient (B7 double-knockout (DKO)⁴) mice resulted in significant acceleration of alloresponses compared with wild-type (WT) control T cells as measured by graft-vs-host disease (GVHD) mortality. Conversely, T cells obtained from mice that constitutively overexpress B7-2 at high levels on all T cells resulted in dramatically reduced GVHD mortality. Data further indicate that down-regulation by T cell-associated B7 occurred through CTLA-4 ligation via a T-T interaction. These studies establish an important role for B7 expression on T cells in the in vivo down-regulation of immune responses in a relevant animal model of disease. We hypothesize that the up-regulation of B7 on T cells may be a significant component of normal in vivo T cell immune responses that functions in concert with other negative regulators, thereby preventing uncontrolled T cell proliferation and autoaggressive immune responses. Moreover, the finding that B7 expression on T cells is a negative regulator of immune responses is potentially clinically relevant. The therapeutic targeting of the CD28/CTLA-4:B7 costimulatory pathway by anti-B7 Abs may be complicated by the finding that B7 expression on T cells down-regulates T cell responses, and as such, blockade might interfere with a clinically desirable down-regulatory process.

B10.BR (H2^k) mice were purchased from The Jackson Laboratory (Bar Harbor, ME). C57BL/6 (H2^b) (termed B6) and BALB/c (H2^d) WT mice were purchased from the National Institutes of Health (Bethesda, MD). B7 DKO mice (6) were backcrossed >10 generations to B6 or BALB/c mice. B7.2 transgenic (B7-2 Tg) mice (7) were backcrossed >5 generations to B6 mice. The B7-2 transgene in this transgenic line is expressed only on T cells and not on other cell populations (e.g., B cells). Littermates not expressing the B7-2 transgene were used as WT controls. Donor and recipient mice were maintained in our specific pathogen-free colony and used at 2–3 mo of age.

For plate-immobilized anti-CD3 stimulation, anti-CD3ε mAb (hybridoma 145-2C11; hamster IgG; BD PharMingen, San Diego, CA) was precoated on 96-well round-bottom microtiter plates (Costar, Cambridge, MA) by overnight incubation (10 μg/ml). Lymph node (LN) T cells from B6 WT and B7 DKO were purified by column purification (R&D Systems, Minneapolis, MN; and Cedarlane, Hornby, Ontario, Canada), CD25-depleted as described below, and plated in well-rinsed anti-CD3 mAb-coated plates. Soluble anti-CD28 mAb (hybridoma 37.52; hamster IgG; BD PharMingen) was added to indicated wells at 1.0 μg/ml. Replicates of six wells were pulsed with tritiated thymidine (1 μCi/well; Amersham Life Science, Arlington Heights, IL) on the indicated days for 20–24 h before harvesting and counted in the absence of scintillant amplification on a beta-plate reader (Packard Instrument, Meriden, CT).

B10.BR or BALB/c (WT and B7 DKO) mice were lethally irradiated with 8.0- or 6.0-Gy total body irradiation, respectively, by x-ray on day −1. On day 0, 20 × 10⁶ B6 WT or B7 DKO donor BM cells were T cell depleted by incubation with anti-Thy-1.2 (hybridoma 30H-12; rat IgG2b) plus rabbit complement (Nieffenegger, Woodland, CA) and administered i.v. to recipients. To induce GVHD, supplemental T cells consisting of either whole splenocytes or purified LN whole T cells or CD4⁺ T cells were coinfused with BM. To obtain whole T cells or CD4⁺ T cells, LN cells were depleted of NK cells (hybridoma PK136; mouse IgG2a) and CD8⁺ T cells (for CD4⁺ T cells) (hybridoma 2.43; rat IgG2b) by incubation with mAb, followed by passage through a goat anti-mouse and goat anti-rat Ig-coated column (Cedarlane). Purified cells were ≥94% whole T cells or CD4⁺ T cells. Cells were depleted of CD25⁺ cells by incubation with anti-CD25 mAb (hybridoma 3C7; rat IgG2b; BD PharMingen) and sheep anti-rat Dynabeads (Dynal, Great Neck, NY). To more rigorously deplete contaminating APCs, some experiments used anti-Ly-6G (Gr-1) mAb (hybridoma RB6-8C5; rat IgG2b; BD PharMingen), anti-CD11b mAb (hybridoma M1/70; rat IgG2b; BD PharMingen), and anti-I-A/I-E mAb (hybridoma M5/114.15.2; rat IgG2b; BD PharMingen) in conjunction with anti-CD25 mAb and sheep anti-rat Dynabeads. This was followed by incubation with a biotinylated Ab to CD11c (hybridoma HL3; hamster IgG; BD PharMingen) and CELLection Biotin Binder Dynabeads (Dynal) resulting in >99% whole CD25-depleted T cells or CD4⁺CD25⁻ T cells. Mice were weighed twice weekly and monitored daily for GVHD lethality. In all experiments, control recipients receiving only T cell-depleted BM with no supplemental T cells survived the observation period and were deemed GVHD-free by examination of weight curves and clinical appearance. Where indicated, mice received anti-CTLA-4 mAb (hybridoma UC10-4F10-11; hamster IgG) i.p. at 400 μg from day −1 to +5, then 200 μg at three times per week to day 28, or anti-B7-2 mAb (hybridoma GL1; rat IgG2a) i.p. at 250 μg from day −1 to +1, then 125 μg three times per week to day 28. Control mice received irrelevant hamster or rat IgG (Rockland, Gilbertsville, PA) at the indicated dose and schedule. Survival data were analyzed by life-table methods, and actuarial survival rates are shown. Group comparisons were made by log-rank test statistics. A value of p ≤ 0.05 was considered significant.

Cannulae were inserted in the thoracic duct of mice 6 days after transplantation with allogeneic BM and spleen as described previously (8, 9). We have determined previously that this is the time of peak expansion of donor alloreactive T cells. Thoracic duct lymphocytes from normal nontransplant control mice and transplanted mice undergoing GVHD were collected overnight and incubated with anti-B7-1 FITC, anti-B7-2 FITC, anti-CD4 PE, and anti-CD8 biotin or irrelevant isotype control fluorochromes and streptavidin-PerCP (BD PharMingen) and analyzed on a FACSCalibur with CellQuest software (BD Biosciences, Mountain View, CA). Individual mice from two separate experiments were analyzed. Quadrants were determined by irrelevant isotype control staining. Representative data are shown.

To determine the level of B7 expression on naive and alloreactive T cells, thoracic duct lymphocytes were harvested from control nontransplanted mice and lethally irradiated recipients undergoing GVHD after transplantation with MHC class I- and II-disparate T cells (Fig. 1). Although B7-1 was virtually absent on resting naive control CD4⁺ T cells, B7-1 was expressed on 34.9% of donor alloreactive lymphocytes retrieved 6 days after allogeneic transplantation, which is the time of peak donor T cell expansion (Fig. 1 and data not shown). Unlike B7-1, B7-2 was constitutively expressed on 12.6% of resting CD4⁺ T cells and was further up-regulated to ∼36% of alloreactive T cells during a GVHD response (Fig. 1 A). Similar findings were seen for CD8⁺ T cells (data not shown).

The up-regulation of B7 on activated alloreactive donor T cells during a GVHD response suggested that T cell-associated B7 might play a role in the regulation of alloresponses, although it was not clear whether B7 expression would amplify or dampen the response. Initial experiments focused on comparing in vitro proliferative responses of purified T cells from either WT or B7 DKO mice in response to immobilized anti-CD3 mAb, a strong TCR signal. B7 DKO T cells had a substantially higher proliferative response to immobilized anti-CD3 mAb compared with WT T cells, suggesting that B7 expression on T cells can down-regulate T cell responses under some in vitro culture conditions (Fig. 1,B). When soluble anti-CD28 mAb was added to cultures, WT and B7 DKO T cells had equivalent proliferative responses, indicating that CD28 signals could restore WT proliferative responses to B7 DKO levels (Fig. 1 C).

To determine the in vivo functional consequences of B7 up-regulation that occurs during an alloresponse, experiments were performed with GVHD-inducing allogeneic T cells from donor mice that are incapable of expressing B7 Ags (6). B10.BR (H2^k) mice were lethally irradiated and transplanted with B6 (H2^b) T cell-depleted BM and 5 or 25 million splenocytes from either B6 WT or B6 B7 DKO mice. GVHD mortality, an unequivocal indicator of in vivo alloresponses, was significantly accelerated in recipients of B7 DKO vs WT spleen (Fig. 2,A). The median survival time (MST) for recipients of 5 × 10⁶ DKO or WT splenocytes was 24 or 37.6 days, respectively (p = 0.0014; Fig. 2,A). Recipients of 25 × 10⁶ DKO or WT splenocytes had a MST of 12.7 or 25 days, respectively (p = 0.0023; Fig. 2 A). Survival in mice receiving 5 × 10⁶ B7 DKO splenocytes was the same as in mice receiving 5-fold more WT splenocytes.

Although the potent acceleration in GVHD mortality by DKO T cells suggested that T cell B7 expression down-regulates alloresponses, a caveat of these data was that B7 DKO mice have reduced numbers of immune regulatory CD4⁺CD25⁺ cells. CD4⁺CD25⁺ cells comprise ∼10% of CD4⁺ T cells from the spleens or LNs of B6 WT mice and are reduced to 2–4% of CD4⁺ T cells in DKO mice (Ref. 10 and data not shown). CD4⁺CD25⁺ immune regulatory cells have been shown to down-regulate alloresponses in both mice and humans (11, 12, 13, 14, 15, 16, 17). Thus, the acceleration of GVHD in recipients of B7 DKO cells could have been due to the reduced numbers of CD4⁺CD25⁺ cells in the donor T cell inoculum.

To address this issue, experiments were performed with CD25-depleted T cells to adjust for the differences in this critical regulatory cell population in WT vs DKO donor mice. Lethally irradiated B10.BR mice were transplanted with B6 BM and 5 or 15 million CD25-depleted splenocytes from B6 WT or DKO mice. GVHD lethality was accelerated in recipients of CD25-depleted DKO splenocytes compared with recipients of CD25-depleted WT splenocytes at both cell doses (5 × 10⁶, MST of 22 vs 28 days; p = 0.046; 15 × 10⁶, MST of 9 vs 16 days; p = 0.003) (Fig. 2,B). The finding that T cell-associated B7 molecules play a role in the inhibition of alloresponses and GVHD generation was not unique to a single strain combination. Lethally irradiated BALB/c mice were transplanted with B6 BM and CD25-depleted CD4⁺ T cells from either B6 WT or DKO mice (Fig. 3,A). B7 DKO CD4⁺ T cells resulted in a potent acceleration of GVHD lethality as compared with WT CD4⁺ T cells with a reduction in MST from 35 to 13 days (Fig. 3 A; p = 0.043). Thus, T cell-associated B7 down-regulated alloresponses independently of CD4⁺CD25⁺ immune regulatory T cells.

To exclude the possibility that the increased mortality by B7 DKO T cells was the result of recognition of host B7 as a neoprotein Ag, the experiment in Fig. 3,A was modified to use BALB/c DKO mice as transplant recipients of B6 BM and CD25-depleted CD4⁺ T cells from either B6 WT or DKO mice (Fig. 3,B). As expected, GVHD mortality was delayed in DKO recipients, suggesting that host B7 was required for optimal donor T cell activation and GVHD generation (Fig. 3, compare A with B). However, B7 DKO T cells still resulted in an acceleration of GVHD in a B7-deficient host indicating that the increased alloresponsiveness of DKO T cells was not due to the recognition of host B7 as a neoantigen.

These data were reproduced in a confirmatory experiment in which BALB/c DKO recipients received B6 DKO BM and highly purified (>99%) CD4⁺CD25⁻ T cells from either B6 WT or B6 DKO donors. This created a transplant scenario in which constitutive and inducible B7 expression was limited to WT donor T cells. Again, GVHD was substantially accelerated in recipients of DKO T cells (Fig. 3 C). At 1 mo after transplantation, the survival rate in recipients of DKO cells was only 30 vs 100% in recipients of WT cells (p < 0.001). These data confirmed that accelerated GVHD mortality in recipients of DKO T cells was not the result of recognition of B7 as a neoantigen. Moreover, the use of highly purified donor T cells (rigorously depleted of APCs) excluded the possibility that indirect allorecognition mediated by contaminating APCs in the donor T cell inoculum was playing a significant role in GVHD generation. Additionally, if indirect presentation of host Ag by contaminating APCs in the donor T cell inoculum was playing a significant role in GVHD generation, the costimulatory-deficient B7⁻ APCs in the B7 DKO T cell inoculum would have been expected to result in reduced GVHD or even tolerance induction.

Because B7-deficient T cells resulted in accelerated GVHD mortality and the only two known ligands for B7 molecules, CD28 and CTLA-4, are present on T cells, we hypothesized that B7 up-regulation on WT T cells may result in the attenuation of T cell activation by a T-T interaction with CTLA-4. Although CD28 is constitutively expressed on T cells and CTLA-4 is expressed on cell surface only after T cell activation, CTLA-4 has a higher binding affinity for B7-1 and B7-2 than does CD28 and functions to down-regulate T cell responses (2, 18). The importance of CTLA-4 as a negative regulator in T cell homeostasis is evidenced by the profound and fatal phenotype of the CTLA-4-deficient mouse (19, 20). Although B7-1 and B7-2 are well-characterized ligands on APCs for engagement of CTLA-4 expressed on activated T cells, it is possible that T cell-associated B7 also regulates in vivo immune responses via engagement of CTLA-4 from T-T interaction.

To test this hypothesis, BALB/c DKO mice were transplanted with B6 DKO BM and B6 WT highly purified CD25-depleted T cells. Anti-CTLA-4 mAb or control hamster IgG was administered to recipients. B7 was not present on the host or BM-derived donor cells but was expressed and up-regulated only on WT donor T cells as they became activated by host alloantigen. Thus, anti-CTLA-4 mAb was expected to block only the interaction between CTLA-4, induced on activated donor T cells, and T cell-associated B7, also induced on activated WT donor T cells. If T cell-associated B7 served to down-regulate alloresponses, then the administration of anti-CTLA-4 mAb was expected to result in an acceleration of GVHD mortality. Consistent with the hypothesis, the administration of anti-CTLA-4 mAb resulted in a modest acceleration in GVHD mortality (Fig. 3 D; MST, 32.5 vs 26.5 days for irrelevant vs anti-CTLA-4 mAb; p = 0.0185). These data indicate that B7 expression on T cells down-regulated alloresponses via a T-T interaction with CTLA-4.

The finding that B7 expression on T cells down-regulates GVHD also was verified by the use of T cells obtained from B6 mice that constitutively overexpress B7-2 at high levels on all T cells but not B cells (B7-2 Tg). Lethally irradiated B10.BR recipients were transplanted with B6 BM and highly purified CD25-depleted T cells from either B6 B7-2 Tg or WT littermate control mice (Fig. 4,A). Recipients of WT T cells died from GVHD by 20 days after transplantation. In contrast, 38% of recipients of B7.2 Tg T cells survived long term (Fig. 4 A).

We have shown that the costimulatory blockade of T-APC interactions by anti-B7.1 and anti-B7.2 mAbs resulted in amelioration of GVHD, whereas individual mAbs were less effective (21). Both T cell-associated B7-2 and APC-associated B7-2 would be blocked by the administration of anti-B7-2 mAb. We hypothesized that APC-associated B7-2 would dominate the in vivo alloresponse in the case of WT T cells; thus, the administration of anti-B7.2 mAb to recipients of WT T cells would ameliorate GVHD by CD28:B7-2 costimulatory blockade of T-APC interactions. In contrast, the high expression of T cell-associated B7-2 might dominate the in vivo response in the case of B7-2 Tg donor T cells. Consistent with this hypothesis, the administration of anti-B7.2 mAb as a single agent to recipients of allogeneic WT T cells resulted in a significant increase in median survival at two different T cell doses (Fig. 4,B). In contrast, the administration of anti-B7.2 mAb to recipients of allogeneic B7-2 Tg T cells resulted in a substantial acceleration of GVHD lethality presumably by precluding the engagement of B7-2 with CTLA-4 via T-T interactions (Fig. 4 C; p = 0.0189). These data are additional evidence that T cell-associated B7 down-regulates immune responses through the ligation of CTLA-4 via T-T interaction.

Although other studies have noted the constitutive or inducible expression of B7 molecules on resting or activated murine or human T cells, scant data are available on the functional in vivo consequences of B7 expression on T cells (22, 23, 24, 25). One study found that B7 was expressed on activated human T cells in vitro and hypothesized that T cell-associated B7 may play a role in the avoidance of anergy (22). However, Greenfield et al. (26) noted that constitutively expressed B7-2 on freshly isolated murine T cells preferentially bound CTLA-4 and hypothesized that B7 expression on T cells served to limit their response. Hollsberg et al. (27) further found that T cells express a hypoglycosylated form of cell membrane B7-2, resulting in undetectable binding to CD28 but partial preservation of CTLA-4 binding. Our data extend these studies by providing the first definitive in vivo data illustrating the importance of T cell-associated B7 as a negative regulator of immune responses in a clinically relevant murine model of rapidly lethal GVHD. Data indicate that in vivo down-regulation by T cell-associated B7 occurred through ligation with CTLA-4 via T-T interaction. Although positive and negative regulation of T cell responses is generally thought of in terms of costimulation via T-APC interactions, these data also highlight the importance of T-T interactions in the regulation of alloresponses and GVHD generation. These data support the hypothesis that the up-regulation of B7 on T cells may be an important component of normal in vivo T cell homeostasis functioning in concert with other important negative regulators of T cell responses.

We believe these data may have important ramifications in many areas of clinical immunology. The clinical targeting of the CD28:B7 costimulatory pathway by anti-B7 Abs may be complicated by the finding that B7 expression on T cells down-regulates T cell responses because blockade might interfere with a clinically desirable down-regulatory T-T interaction. Therefore, whereas anti-B7 mAbs may block auto- and alloaggressive CD28:B7 interactions and ameliorate disease, they also may block inhibitory T-APC and T-T CTLA-4:B7 interactions, resulting in a less efficacious therapeutic outcome.