Selective CD28 blockade attenuates acute and chronic rejection of murine cardiac allografts in a CTLA-4-dependent manner

Abstract

Selective blockade of CD28 is a promising therapy to inhibit pathogenic alloimmunity. However, evaluation of this approach in transplantation has been very limited. Using a novel non-activating single-chain Fv-based reagent (α28scFv), we have investigated the role of CD28 and cytotoxic T lymphocyte antigen 4 (CTLA-4) in a murine cardiac transplant model. Blockade of CD28 for two weeks after engraftment promoted allograft survival, and significantly attenuated chronic rejection when combined with transient CD154 blockade or calcineurin inhibition. Graft acceptance was associated with decreased alloantibody production, increased proportion of early graft infiltration by regulatory T cells and increased expression of regulatory dendritic cell genes. Blockade of CTLA-4 during α28scFv-based treatments led to prompt rejection in all animals and inhibited expression of forkhead box P3 (Foxp3), programmed death (PD)-1 and 2,3-indoleamine dioxygenase (IDO) in the graft. These results show that CD28 signaling during the first weeks after transplant is a pivotal mediator of pathogenic alloimmunity, and that selective CD28 blockade prolongs graft acceptance by at least two immunomodulatory mechanisms. Selective CD28 inhibition while sparing CTLA-4 is thus a promising approach to inhibit pathogenic alloimmunity.

Introduction

Immune responses mediated by the CD28 family of costimulatory molecules determine the balance between regulatory and pathogenic effector mechanisms after initial antigen exposure. Constitutively expressed CD28 binds B7 to provide a stimulatory signal important for sustaining T cell proliferation and augmenting pro-inflammatory responses (1–3). CTLA-4, another B7 ligand induced on T cells subsequent to high affinity TCR ligation, delivers anti-proliferative (4–6) and/or tolerogenic signals to T cells, and to B7-bearing antigen presenting cells (APCs), in which it triggers increased 2,3-indoleamine dioxygenase (IDO) (7). Targeting the CD28/B7 pathway by use of CTLA4-Ig reagents (Belatacept, Abatacept), which directly bind B7 molecules, is a promising alternative to prevent autoimmunity (8, 9) and part of a calcineurin-free maintenance immunosuppressive regimen in renal transplantation (10–12).

However, B7-directed blocking strategies may deprive the evolving immune response of CTLA-4-driven signals crucial to development of antigen-specific peripheral regulatory T cells (Tregs). In support of this hypothesis, blocking the CD28/B7 pathway by ligation of B7, using either a CTLA-4 analogue (11, 12) or antibodies against B7 family members (13, 14), does not reproducibly induce tolerance across a full MHC mismatch in rodents or primates. CTLA-4/B7 interactions are required for the induction of peripheral T cell tolerance to soluble antigens (15–18), tumors (5) and allografts (19, 20). Further, selective agonistic ligation of CTLA-4 attenuates in vivo T cell responses and prevents development of autoimmunity (21, 22), while the absence (6, 23) or blockade (24–26) of CTLA-4 is associated with autoimmunity.

Based on these mechanistic considerations, selective inhibition of CD28 should prevent maturation of pathogenic effectors while preserving CTLA-4-dependent immune regulation. In this regard, several lines of agonistic anti-CD28 antibody have been generated and have demonstrated beneficial effects on prevention of allograft rejection and graft-versus-host disease (27–33). In nonhuman primate kidney and heart transplants, we recently reported that non-cross-linking selective recombinant CD28 antagonists decrease the allogeneic immune response and augment the pool of antigen-specific Tregs (34). Here we present additional evidence that supports this approach and reveal its mechanistic basis using a similar non-activating, non-cross-linking anti-CD28 monovalent receptor antagonist in a murine heart transplant model.

Material and Methods

Reagents

α28scFv was developed from the hamster antibody clone PV1.17.10 as described in Supplementary Information. Anti-CD154 (clone MR1) and blocking anti-CTLA-4 (clone UC10-4F10-11) (19) antibodies were purchased from Bioexpress (West Lebanon, NH). Hamster anti-CD28 antibodies (clones PV1 and 37.51) were from Southern Biotech (Birmingham, Alabama) and eBioscience (San Diego, CA) respectively. Mouse anti-mouse CD28-specific superagonist antibody (clone D655) was from AbD Serotec (Raleigh, NC). Murine CTLA4-Ig was from Chimerigen (Allston, MA, Cat#MF120A4). Abatacept (Orencia) was purchased from Bristol-Myers Squibb Company (Princeton, NJ). Purified hamster IgG, hamster IgG Fab, and rat IgG were from Bioexpress.

Animals

Six- to 10-week-old C57BL/6 (H-2^b) and BALB/c (H-2^d) male mice were obtained from The Jackson Laboratory (Bar Harbor, ME). Animals were housed under conventional conditions and used according to the guidelines of the Institutional Animal Care and Use Committee (IACUC) of the University of Maryland Medical School. Protocols approved by the IACUC were carried out in compliance with the Guide for the Care and Use of Laboratory Animals (HHS, NIH Publication 86-23, 1985).

Isolation of cell populations

Single cell suspensions were prepared from the spleen and lymph nodes by mincing with forceps and passage of the resulting cell suspension through nylon mesh of 100 μm pore size, followed by red blood cell lysis. In addition, in selected murine allograft recipients at days 10–12 (d10-12), graft infiltrating lymphocytes (GILs) were isolated by mincing the graft and incubating the resulting fragments for 30 min in medium containing 1 mg/ml collagenase type 4 (Worthington Biochemical, Freehold, NJ), 1 mg/ml soybean trypsin inhibitor (Sigma-Aldrich, St. Louis, MO), and 0.1 mg/ml DNase (Roche, Indianapolis, IN) as previously described (35). Lymphocytes were separated from cardiac parenchymal cells by Ficoll-gradient centrifugation.

In Vitro Cytokine assays

Supernatants were assayed for cytokines by sandwich enzyme-linked immunosorbent assay (ELISA), using paired antibodies obtained from PharMingen. Immulon II plates (Dynatech, Chantilly, VA) were coated overnight with 5 μg/ml purified anti-interleukin(IL)-2 or anti-interferon(IFN)γ, as appropriate. Plates were blocked for 2 h with PBS/0.5% Casein at 37°C. Triplicate samples were added and incubated 2 h at RT. Plates were washed with Tris/NaCl/NP-40 (TNN) using a Skanwasher 300 (Skatron Instruments, Sterling, VA). After washing, plates were incubated with biotin-coupled anti-IL-2 or anti-IFNγ (100 ng/well), for 1 h at RT. Plates were washed and incubated an additional hour with Avidin-HRPO. Enzyme substrate (2,2′-azino-di[3-ethyl]-benzthiazoline sulfonate, Kirkegaard & Perry Laboratories, Inc., Gaithersburg, MD) was added and the reaction was allowed to develop for 5 min. OD405 was read on a Vmax, automated ELISA Plate reader (Molecular Devices, San Diego, CA). OD405 values for samples were compared to appropriate cytokine standards. Data were reported as pg/ml of cytokine.

Polyclonal T cell Activation

Murine splenocytes were activated using anti-CD3ε (145 2C11, PharMingen, San Diego, CA), at a concentration of 0.1–10 μg/ml. Cells were cultured in 96 well round bottom plates (Costar, Cambridge, MA) at 37°C for 72 h. After 48 h of culture, 100 μl of culture supernatant was harvested and assayed for cytokine production by ELISA as described in Supplementary Information. Fresh medium was added to culture wells and plates were returned to incubator for 24 additional hours. Proliferation was measured by adding [³H]-Thymidine (NEN, Cambridge, MA) to these cultures for the final 6 h of culture. Cells were harvested on a Tomtec harvester (Wallac, Inc., Gaithersburg, MD) and [³H]-Thymidine incorporation was counted in an LSC counter (Wallac Microbeta). α28scFv was added to cultures on d0 in three fold serial dilutions.

Mixed lymphocyte reaction (MLR)

Splenocytes from naive BALB/c and C57BL/6 mice were used as responder and stimulator cells respectively. Mouse responder cells were cocultured with irradiated stimulator cells (3×10⁵ each/well) in RPMI containing 10% FBS, gentamycin and 2-βmercaptoethanol in 96 round bottom plates. Antibodies (anti-CD28, anti-CD154, or irrelevant IgG) were added at indicated concentrations. At 96 h, cultures were pulsed with [³H] thymidine and harvested 16 h later. MLR results were expressed as counts per minute (cpm). Data represent means ± SEM of triplicate wells and are representative of 2–6 independent experiments. Purified hamster IgG was used as specificity control.

Flow cytometry analysis of α28scFv binding

To detect peripheral lymphocytes bearing α28scFv, samples of peripheral blood, splenocytes or lymph nodes were harvested from mice injected ip with single-chain antibody or control Fab (50 μg), 2 h previously. Cells (1 X 10⁷/ml) were blocked with anti-CD16/32, then stained with anti-CD3-A647, CD19-FITC (all purchased from PharMingen, San Diego, CA), and anti-6HIS-PE (R&D Systems, Minneapolis, MN). Stained samples were then washed with PBS-0.5%BSA and analyzed by flow cytometry. Dead cells were excluded using Hoechst 33258 (1 μg/ml). For peripheral blood samples, 100 μl of blood was blocked and stained with CD3, CD19, and anti-6HIS, and then fixed using BD FACSLyse reagent (BD Biosciences, San Diego, CA). Samples were washed with PBS-0.5%BSA and then analyzed.

In vivo cytokine production

C57BL/6 mice received α28scFv, anti-CD28 superagonist antibody or control rat IgG as a single dose of 100 μg. Mice were sacrificed 3 h after injection, and serum levels of IL-6, IFN-γ, KC, and TNF-α were measured using a luminex assay with a Milliplex kit (Millipore, Billerica, MA) and a Luminex 200 equipped by MasterPlex software (MiraiBio, Inc., Alameda, CA).

Transplant procedures

Vascularized heterotopic hearts from C57BL/6 and BALB/c donors were transplanted into the abdomen of BALB/c recipients using the microsurgical technique of Corry et al. (36). Graft survival was monitored by daily palpation. Rejection was defined as complete cessation of the palpable heartbeat and was confirmed histologically. In initial dosing experiments, recipients were treated with α28scFv at 200 μg on d0-2, 2, 2–4, or 0–13; or 50 μg on d0-13 (SI Figure 4). Subsequent recipients received α28scFv at 100 μg twice daily on d0-13. Anti-CD154 mAb (MR1) was dosed at 250 μg on d0, and cyclosporine A (CsA) at 400 μg on d0-3. α28scFv plus MR1 or CsA were given in combination at these doses and durations. Additional recipients received αCTLA-4 mAb (500 μg on d0; 250 μg on d1-10) during α28scFv-based combined treatment. Abatacept was administered at 500 μg every other day on d 0–12. Control allograft and isograft recipients were untreated.

Graft histology

At the time of explant, cardiac grafts were trisected. The apex was immediately snap-frozen for molecular analysis. The basal part of the heart was fixed in 10% buffered formalin, embedded in paraffin, sectioned, and stained with H&E and Verhoeff’s elastin according to standard procedures (37). Elastin-stained sections were used to assess transplant arteriosclerosis. The incidence (proportion of vessels affected) and grade (severity) of arteriosclerosis were scored as follows: 0 represents a normal artery; 1, 1%–20% occlusion; 2, 21–40% occlusion; 3, 41–60% occlusion; 4, 61–80% occlusion; and 5, >80% occlusion), as described (37–39). Grafts that failed within 60 days in animals treated with CsA, MR1, or anti-CD28 monotherapy exhibited Grade 4 acute cellular rejection (SI Figure 4). The middle part of the heart was frozen in optimal cutting temperature (OCT) compound.

Intracellular Foxp3, anti-donor alloantibody and ELISPOT assays

Intracellular Foxp3 and donor-reactive antibodies were measured by flow cytometry; cytokine production in splenocytes was measured by ELISPOT. Detailed experimental procedures for these assays are described in Supplementary Information.

Intra-graft gene expression

Real-time reverse transcription (RT)-polymerase chain reaction (PCR) was performed as previously reported (40) and described in Supplementary Information. The expression of each gene was normalized to the housekeeping rpL32 using the ΔΔC_T calculation and mRNA levels were expressed as relative fold increase over native unmanipulated C57BL/6 heart tissue.

Statistical analysis

Graft survival time was expressed as mean±standard deviation (SD), graphed with the Kaplan-Meier method. The log-rank test was used to compare survival time between groups. Continuous variables were expressed as the mean±standard error of the mean (SEM) unless otherwise indicated and compared using the Mann-Whitney non parametric test. P<0.05 were considered statistically significant. All statistical analyses were performed with SPSS (SPSS, Chicago, IL, USA) or GraphPad InStat (San Diego, CA, USA).

Results

α28scFv inhibits lymphocyte activation

A novel monovalent recombinant single-chain antibody fragment containing the variable region of a high-affinity anti-CD28 clone (α28scFv) was developed to block murine CD28 as described in SI Methods. We found that α28scFv effectively inhibited T cell receptor (TCR)-dependent lymphocyte activation and proliferation in vitro (Figure 1A). IL-2 production was more sensitive to costimulation blockade than cell proliferation (IC₅₀ 10 pg/ml for IL-2 secretion and 97 pg/ml for cell proliferation). α28scFv also suppressed allogeneic lymphocyte proliferation (Figures 1B and 1C), as previously reported for anti-human CD28 antagonists (34, 41). In contrast, intact anti-CD28 antibodies dramatically enhanced T cell proliferation (Figure 1B, top), as expected (42). Finally, whereas CD154 blockade minimally affected mouse T cell proliferation (Figure 1C, top), an additive effect was seen with additional α28scFv relative to either treatment alone (Figure 1C, bottom).

Figure 1. Selective CD28 blockade inhibits T cell activation in vitro.

(A) Inhibition of polyclonal T cell proliferation (top) and IL-2 cytokine secretion (bottom) by α28scFv. Murine splenocytes were stimulated with 10 μg/ml anti-CD3ε. Cytokine production was measured in culture supernatant by ELISA and cell proliferation by thymidine incorporation after 48 and 72 h respectively. (B) Differential effect of α28scFv and αCD28 antibodies in a mixed lymphocyte culture reaction (MLR). α28scFv (top), CTLA4-Ig and CsA (bottom) inhibit T cell proliferation, whereas anti-CD28 (top) and anti-CTLA-4 (bottom) IgG antibodies enhance cell proliferation. All indicated reagents were added at 20 μg/ml on d0 of culture. Control, normal hamster IgG; PV1, 37.51, hamster anti-CD28 IgG; PMA, phorbol myristate acetate; CsA, cyclosporine A. (C) α28scFv is synergistic with CD154 blockade. Physiologically relevant concentrations of α28scFv inhibit allogenic T cell proliferation in a dose dependent manner (top) and are synergistic with CD154 blockade (MR1, 20 μg/ml) (bottom). Results were expressed as the mean ± SEM of triplicate wells, and are representative of 2–6 independent experiments. *P<0.05 vs. α28scFv at the same concentration; **P<0.01 vs. untreated control.

α28scFv is readily detectable on peripheral T cells

We injected α28scFv to mice and analyzed binding in blood and lymphoid tissues by flow cytometry. Peripheral blood T cells harvested 2 h after injection revealed >98% of circulating T cells staining with α28scFv (SI Figure 1). Peripheral blood B cells did not stain with α28scFv. Examination of secondary lymphoid tissues demonstrated staining on spleen and lymph node T cells. The protein half-life was 10 h (SI Figure 2).

In vivo cytokine secretion in response to α28scFv

A fully humanized CD28 superagonist monoclonal antibody TGN1412 induced an unexpected cytokine release syndrome during a pilot human trial in 2006 (43). Therefore, we analyzed serum cytokine levels after injection of α28scFv. Serum levels of IL-6, KC, and TNF-α were increased by a mouse anti-mouse CD28 superagonist (SA) antibody as previously shown (44). In contrast, cytokine levels did not increase significantly compared to control IgG after injection of α28scFv (SI Figure 3).

CD28 blockade promotes murine cardiac allograft engraftment

Induction monotherapy with α28scFv significantly prolonged cardiac allograft survival (Figure 2A and SI Figure 4). Whereas untreated BALB/c recipients rejected C57BL/6 allografts within 10 days (mean survival time (MST), 9 d, n=10), grafts in recipients treated with α28scFv (200 μg/day) for 14 days rarely rejected during therapy, and had significantly prolonged graft survival (MST, 27 d; n=5; P<0.05). All allografts rejected within 51 days, demonstrating that this regimen does not induce tolerance across this full MHC disparity.

Figure 2. Selective CD28 inhibition prolongs cardiac allograft survival and prevents chronic rejection.

BALB/c recipients received C57BL/6 heterotopic cardiac allografts or BALB/c isografts. Allograft recipients were treated with α28scFv (200 μg, d0-13), MR1 (250 μg, d0), CsA (400 μg, d0-3) or combinations. (A) α28scFv prolongs graft survival, an effect significantly augmented when combined with CD154 blockade (MR1) or calcineurin inhibition (CsA). Addition of αCTLA-4 mAb resulted in rejection within 10 days in all treated animals. Color coding corresponds to treatment groups. *P<0.01 vs. untreated control; **P<0.001 vs. α28scFv or α28scFv+CsA+αCTLA-4; ***P<0.05 vs. MR1 and P<0.001 vs. α28scFv or α28scFv+MR1+αCTLA-4. (B) Representative arteries in surviving grafts over 100 days after transplant, demonstrating the effect of CD28 blockade on chronic rejection [Verhoeff’s elastin staining, original magnification ×200]. An MR1-treated cardiac allograft shows grade 3 CAV (>50% luminal occlusion) with severe intimal thickening (arrow) and a mild-moderate perivascular and neointimal cellular infiltrate. In contrast, grafts treated with α28scFv plus MR1 or CsA show absence of neointimal proliferation (arrows). (C) Incidence and severity of CAV measured as described in SI Methods. Beyond 88 days after cessation of treatment, α28scFv plus CsA or MR1 was associated with markedly less neointimal thickening characteristic of CAV relative to MR1 alone. *P<0.05 vs. MR1 alone.

We next investigated whether graft acceptance is facilitated by transiently attenuating either calcineurin- or CD154-dependent adaptive immune pathways. Eight of 11 animals treated with α28scFv combined with anti-CD154 mAb (MR1) had indefinite (>100 days) graft survival (P<0.05 vs. anti-CD28 or MR1 monotherapy). Similarly, α28scFv combined with a short course of CsA significantly prolonged graft survival, with 9 of 12 allografts surviving >100 days (Figure 2A). Histological analysis indicated decreased lymphocyte infiltration and preserved myocardial architecture in surviving grafts, consistent with an inhibition of cell-mediated rejection (SI Figure 5). Importantly, severe lesions of cardiac allograft vasculopathy (CAV) were observed in long-term (100 days) surviving grafts from recipients treated with MR1 alone, whereas CAV was significantly attenuated with α28scFv plus either MR1 or CsA (P<0.05 vs. MR1 alone) (Figures 2B and 2C).

CD28 blockade modulates humoral and cellular responses

The production of anti-donor IgG subclasses associated with T helper (Th)1 (IgG2a) and Th2 (IgG1) immune responses was evaluated (Figure 3A). At d10-14, abatacept-treated recipients had barely detectable levels of donor-reactive alloantibody. Both IgG1 and IgG2a alloantibody were consistently observed in all other groups, although recipients treated with α28scFv plus either CsA or MR1 had significantly lower alloantibody levels. At d100, all recipients with surviving grafts exhibited high levels of IgG1, but IgG2a alloantibody levels were significantly lower than those observed in animals within 2 weeks after transplant.

Figure 3. Mechanism of immune modulation by selective CD28 blockade.

(A) Th2 (IgG1) and Th1 (IgG2a) alloantibody production measured early (d10-14) and after d100 following transplantation. Addition of α28scFv to CsA or MR1 significantly inhibited initial elaboration of both Th1 and Th2 antibodies relative to α28scFv alone, an effect which was sustained only for Th1 alloantibodies. (B) Frequency of alloantigen-specific cytokine-producing splenocytes by ELISPOT early (d10-14) or >100 days after transplant. Data expressed as the average of triplicate wells ± SEM. (C) The proportion of CD4⁺Foxp3⁺ T cells at d10-12 is increased in graft infiltrating cells (GILs) isolated from recipients treated with α28scFv plus MR1 or CsA relative to native heart (naïve), acutely rejecting grafts without treatment (No Rx), or grafts with either treatment alone. No increase in the proportion of CD4⁺Foxp3⁺ cells was observed at d10-12 with abatacept treatment as compared to no treatment or α28scFv monotherapy. Results are expressed as the proportion of CD4⁺Foxp3⁺ cells among graft infiltrating cells. Top: Representative FACS profile; Bottom: Each dot represents an individual graft, the bar displays the group mean, and the box displays the 25^th and 75^th quartiles. *P<0.05 vs. untreated control; **P<0.01 vs. untreated control; ⁺P<0.05 vs. α28scFv; ⁺⁺P<0.01 vs. α28scFv; ^#P<0.05 vs. MR1 or CsA alone; ^εP<0.05 vs. abatacept; and ^†P=0.05 vs. α28scFv; ^††P=0.067 vs. CsA alone.

The frequency of donor-reactive splenocytes expressing Th1 and Th2 cytokines was assessed by ELISPOT (Figure 3B). Although treatment with α28scFv plus MR1 or CsA was associated with fewer IFN-γ- and IL-2-producing splenocytes relative to no treatment or α28scFv monotherapy at d10-14, donor-reactive cells producing IFN-γ and IL-2 were consistently present at increased levels relative to naïve animals and isograft (Iso) recipients at d100. Th1/Th2 ratios with α28scFv+CsA, but not α28scFv+MR1 showed a trend toward early Th2 and late Th1 immune bias (SI Figure 6). Thus, induction of graft acceptance by either CD154 costimulation blockade or calcineurin inhibition combined with α28scFv was associated with less prolific expansion of donor-reactive splenocytes than with ineffective monotherapy regimens, but with distinct patterns of late cytokine skewing.

Increased proportion of Foxp3⁺ cells in cardiac allografts during α28scFv-based immunomodulation

Our evidence suggests that complete deletion, anergy, or ignorance are not associated with graft prolongation in this system since donor-specific splenocytes producing IFN-γ, IL-2, and IL-10 were readily detected at late follow-up in animals with accepted grafts. To determine whether induction of cardiac allograft acceptance was associated with regulatory mechanisms, the frequency of intra-graft Tregs was assessed by flow cytometry at d10-12 after transplant. Recipients treated with α28scFv+MR1 had a significantly increased proportion of CD4⁺Foxp3⁺ T cells among graft infiltrating cells (GILs) compared to recipients that received either α28scFv or MR1 alone (Figure 3C). In addition, recipients treated with α28scFv+CsA showed a trend toward an increased proportion of Tregs vs. α28scFv (P=0.05) or CsA (P=0.067) alone. A similar trend was observed with absolute numbers of Tregs (SI Figure 7A), but the proportion of Foxp3⁺ among CD4⁺ cells was mostly unchanged (SI Figure 7B). Foxp3 expression was minimal on CD4 negative cells (Figure 3C). Thus α28scFv-based therapies were associated with an increased proportion of early graft infiltration by CD4⁺Foxp3⁺ T cells.

In abatacept-treated recipients, no increase in the proportion and absolute number of CD4⁺Foxp3⁺ T cells was observed compared with allografts from α28scFv-treated recipients (Figure 3C and SI Figure 7). These results demonstrate that although multiple high doses of abatacept promote indefinite graft survival (>35, >100, and >100 days, MST > 78 days, n=3) (Figure 2A), they do not facilitate the infiltration of Tregs.

Gene expression profiles during cardiac allograft acceptance

Relative to normal mouse hearts or isografts, expression of Th1 and Th2 cytokine genes, transforming growth factor (TGF)-β, tumor necrosis factor (TNF)-α, inducible nitric oxide synthase (iNOS) and Granzyme B was increased to a similar degree in allografts with long-term graft acceptance (α28scFv with MR1 or CsA) or chronic rejection (MR1) (SI Figure 8A and 8B). Foxp3, CTLA-4, Fas ligand (FasL), and PD-1, remained increased at d100 with a trend towards increased FasL with α28scFv+CsA (P=0.06) and increased PD-1 with α28scFv+MR1 (P=0.25) relative to MR1. IDO mRNA was particularly enriched with α28scFv+CsA (P=0.03) and tended to be increased (P=0.17) with α28scFv+MR1, compared to MR1 alone.

Critical requirement for CTLA-4 in the induction of allograft acceptance

To establish whether the B7/CTLA-4 pathway mediates cardiac engraftment associated with selective CD28 blockade, additional recipients received a non-activating anti-CTLA-4 mAb. CTLA-4 blockade during induction with α28scFv-based therapies led to prompt rejection in all α28scFv-treated animals (Figure 2A and SI Figure 9). Moreover, targeting of CTLA-4 increased anti-donor antibodies with α28scFv+MR1 (P<0.05), and to a lesser extent with α28scFv+CsA (Figure 4A). Real-time RT-PCR revealed decreased intra-graft Foxp3, PD-1, and IDO with α28scFv+MR1+αCTLA-4 (P<0.05), and a trend towards decreased Foxp3 (P=0.05), PD-1 and IDO levels with α28scFv+CsA+αCTLA-4 (Figure 4B), whereas IL-2 and Granzyme B levels were relatively similar (data not show). Therefore, induction of graft acceptance by α28scFv+MR1 or α28scFv+CsA is associated with Foxp3, PD-1, and IDO, and requires CTLA-4.

Figure 4. Effect of CTLA-4 blockade on alloantibody production and intra-graft gene expression.

Treatment with αCTLA-4 mAb resulted in elevated serum alloantibody production (A) and attenuated intra-graft Foxp3, PD-1, and IDO (B) relative to therapies without CTLA-4 targeting, demonstrating that upregulation of these genes is CTLA-4-dependent. Each dot represents an individual animal, the bar displays the group mean and the box displays the 25^th and 75^th quartiles. *P<0.05; ^†P=0.05.

Discussion

The results reported here show that selectively blocking CD28 using a monovalent non-activating scFv reagent significantly modulates the immune response to MHC alloantigens in mice. Induction monotherapy with α28scFv attenuated the pace of acute cardiac allograft rejection, and durably attenuated pathogenic alloimmunity by an additional short course of peritransplant CD154 or calcineurin inhibition. Protection from allograft injury was associated with decreased alloantibodyproduction, an increased proportion of early graft infiltration by Tregs and diminished chronic rejection long after discontinuation of treatment. Finally, graft acceptance required CTLA-4, confirming for the first time the expectation that selective CD28 blockade leads to a CTLA-4-dependent immunomodulation.

In contrast to primate models, where similar efficacy was observed with respect to prevention of pathogenic alloimmunity and inhibition of chronic rejection, the murine model permits elucidation of associated mechanisms. We show that, unlike agonistic anti-CD28 antibodies (42), α28scFv inhibits the activation of naïve T cells in vitro and does not induce significant cytokine release in vivo. In addition, we previously showed that α28scFv inhibits IL-2 production by memory CD4 T cells (45). Anti-donor antibody isotype profile and the pattern of cytokine gene expression in the graft show that the mechanism of initial graft protection and subsequent acceptance was not primarily due to a Th2 bias, as is the case in several other models of peripheral tolerance (46–48) and might be expected in conditions of a decreased “signal 2” (49–52). Moreover, Th bias or IL-10 or TGF-β expression did not obviously account for protection from CAV, since splenic ELISPOT cytokine profiles after in vitro challenge with donor antigen, intra-graft gene expression phenotypes and alloantibody titers were similar in MR1-treated animals with severe graft CAV and in animals treated with either αCD28-based combined treatment, which exhibited relatively mild CAV.

Production of Th2 IgG1 isotype alloantibodies at d100 suggest that robust immune tolerance to donor antigens was not induced using the regimen and dosing used. Modifications of α28scFv to increase its half-life and development of alternative, continuous, drug delivery modalities may further improve the in vivo dosing and efficacy of α28scFv. Alternatively, extended α28scFv therapy after transplantation or combination with targeting of alternative costimulatory pathways (ICOS, CD70, 4-1BB) may further downsize the pool of alloreactive effector T cells and promote regulatory mechanisms to induce stronger tolerance remains to be established.

Foxp3 is a transcription factor important in the development and function of murine CD4⁺CD25⁺ Tregs (53). With α28scFv combined to MR1 or CsA, we observed increased Foxp3 gene expression at d100 and increased proportions of Foxp3⁺ cells at d10-12 within the graft. However neither Foxp3 expression nor a panel of other Tregs genes (CTLA-4, TGF-β, IL-10, IL-2RA) individually distinguished accepted from chronically rejecting grafts at 100 days after transplant. Rather, addition of α28scFv to MR1 or CsA was associated with a trend towards enhanced expression of FasL, IDO and PD-1 in the graft relative to MR1 alone. Our results suggest that CD28 blockade promotes coordinated evolution of both T cell and dendritic cell (DC) protective mechanisms within the graft.

Tregs constitute an important mechanism of peripheral tolerance that is increasingly recognized as a mechanism to control immune responsiveness after organ transplantation. Therefore, manipulating costimulatory signals such as CD28/B7 interactions toward maintaining the development and survival of Tregs may hold promise as a powerful component of immunotherapies. Monotherapy with murine CTLA4-Ig leads to long-term (>100 days) cardiac graft acceptance in a related (BALB/c to C57BL/6) model (54). Here, using the same mouse strains as for α28scFv therapies (C57BL/6 to BALB/c), we found that intense abatacept monotherapy prolonged cardiac graft survival in a similar fashion (>78 days), prevented alloantibody production (Figure 3A) and dramatically reduced graft infiltration (data not shown). However, abatacept did not increase intra-graft Tregs. The differential effects of B7 vs. CD28 targeting approaches may be related to the pharmacokinetic properties of these biologics. The half-life of CTLA4-Ig is much longer than that of α28scFv (55). Alternatively, these findings may point to mechanistic differences between CTLA4-Ig and selective CD28 blockade in protecting graft injury. CD28-mediated costimulation is required for the survival of natural Tregs (56–58). However, the generation of antigen-stimulated adaptive Tregs can occur in situations of suboptimal costimulation (59–61) or in the absence of CD28 signals (62). In addition, selective engagement of CTLA-4 can induce adaptive Tregs with specificity for alloantigen (61). Therefore, a CD28-independent adaptive Treg mechanism might be involved in the action of selective CD28 blockade.

CTLA-4 was required for graft acceptance and increased intra-graft expression of Foxp3, PD-1 and IDO. Our findings strongly support the general hypothesis that interaction between CTLA-4 and its various DC ligands is pivotal to Treg expansion in response to allogeneic stimulation as previously suggested in other models (15–20). Membrane-bound CTLA-4, induced or upregulated on T cells after T cell receptor ligation and partial costimulation through other available pathways (e.g. VCAM-1, ICAM-1) (63) ligates B7 receptors on DC to induce up-regulation of indoleamine 2,3-dioxygenase (IDO), a tryptophan-catabolizing enzyme associated with immunosuppressive activity (64). We postulate that CTLA-4/B7 molecular interactions mediate improved allograft survival in mouse recipients treated with α28scFv by increasing IDO transcription and thus regulatory function in graft DCs (65–67). Moreover, CTLA-4 controls Tregs function (5, 34), and natural Tregs critically require CTLA-4 to suppress immune responses by affecting the potency of antigen-presenting cells to activate other T cells (23). Like CTLA-4, PD-1 negatively regulates T cell activation (21, 68); we find that PD-1 expression tends to be increased in accepted grafts relative to those with chronic rejection, and requires CTLA-4. Thus CTLA-4 and PD-1 may cooperate to induce graft acceptance in our model as recently demonstrated in other models (69).

In summary, based on our current results and the findings of others, our working model holds that CTLA-4 plays a pivotal role to promote graft acceptance and intra-graft CD4⁺ Foxp3⁺ Treg cells when CD28 is selectively blocked. The efficacy of anti-CD28 with conventional immunosuppression to inhibit chronic rejection in mice as well as primates (34) is a novel observation, one with potentially important clinical implications. While further work will be required to more completely dissect the mechanisms underlying CD28 blockade, optimize dosing, and directly develop clinical regimens, our studies confirm our initial hypothesis that non-activating CD28 blockade using a single-chain antibody fragment can modulate pathogenic alloimmunity by an active process, one that is strictly dependent on CTLA-4. More studies are warranted to determine whether selective monovalent CD28-directed therapy or alternative CD28 blockade approaches (70) have significant practical advantages relative to B7 blockade (11, 12) as theoretical models predict.

Supplementary Material

Supp Figure S1-S9

Supplemental Figure S1. α28scFv is readily detectable on peripheral T cells.

Supplemental Figure S2. Pharmacokinetic evaluation of α28scFv in vivo.

Supplemental Figure S3. In vivo cytokine release.

Supplemental Figure S4. Cardiac allograft survival time in dose finding studies in mice.

Supplemental Figure S5. Representative histological analysis of mouse cardiac allografts two weeks after transplant.

Supplemental Figure S6. Th1/Th2 cytokine ratios.

Supplemental Figure S7. Quantification of CD4⁺Foxp3⁺ T cells in naïve heart and surviving allografts at day 10–12 after transplantation.

Supplemental Figure S8. Intra-graft gene expression.

Supplemental Figure S9. Effect of CTLA-4 blockade on cardiac allograft survival.

Abbreviations

APCs

antigen presenting cells

CAV

cardiac allograft vasculopathy

CsA

cyclosporine A

d

day

CTLA-4

cytotoxic T lymphocyte antigen 4

DC

dendritic cells

ELISA

enzyme-linked immunosorbent assay

FACS

fluorescence activated cell sorting

FasL

Fas ligand

Foxp3

forkhead box P3

GIL

graft infiltrating lymphocytes

IACUC

Institutional care and use committee

IDO

2,3-indoleamine dioxygenase

IFN

interferon

IL

interleukin

iNOS

inducible nitric oxide synthase

Iso

isograft

MLR

mixed lymphocyte reaction

MR1

anti-CD154 monoclonal antibody

MST

mean survival time

OCT

optimal cutting temperature

PMA

phorbol myristate acetate

PD-1

programmed death-1

RT-PCR

reverse transcription-polymerase chain reaction

scFv

single-chain fragment variable

SD

standard deviation

SEM

standard error of the mean

SI

stimulation index

SI

supporting information

Tregs

regulatory T cells

TGF

transforming growth factor

Th

T helper

TNF

tumor necrosis factor

Tx

transplantation