Transient Blockade of Delta-like Notch Ligands Prevents Allograft Rejection Mediated by Cellular and Humoral Mechanisms in a Mouse Model of Heart Transplantation Free

Immune-mediated rejection limits the success of organ transplantation in patients. Acute rejection causes morbidity and mortality, as well as a need for urgent retransplantation in selected patients. Despite current immunosuppressive strategies, chronic allograft rejection occurs in a majority of recipients, limiting the life span of transplanted organs. Alloreactive conventional T cells play a central role in the rejection process and represent the main target of existing interventions, whereas regulatory T cells (Tregs) have protective effects (1). Alternative pathogenic mechanisms are increasingly recognized in both acute and chronic rejection, including a central role for donor-specific Abs and complement deposition (2–6). New therapeutic interventions are needed to better preserve allografts from these different forms of immune-mediated damage.

Notch signaling was first recognized for its requirement at early stages of T cell development in the thymus (7, 8). Subsequently, other effects of Notch signaling were discovered in the regulation of T cell differentiation and function, as well as in selected B cell subsets and innate lymphoid cells (9–11). Notch signals are mediated by the interaction of cell surface Notch receptors (Notch1–4) with agonistic delta-like [(Dll)1/4] or Jagged (Jagged1/2) ligands (12). Notch ligand–receptor binding triggers regulated proteolysis of the receptor, leading to the release of intracellular Notch (13). Intracellular Notch migrates into the nucleus where it interacts with the DNA-binding transcription factor CSL/RBP-Jk and a member of the mastermind-like (MAML) family of transcriptional coactivators (14–16). Truncated N-terminal MAML fragments with potent and specific dominant-negative activity (DNMAML) block transcriptional activation downstream of all Notch receptors (17, 18). DNMAML expression represents a powerful approach to capture the overall effects of canonical Notch signaling in specific cell types (17, 19–23). In addition, targeted inhibition of specific Notch ligands and receptors can identify the unique effects of individual family members in vivo and provide new therapeutic opportunities (21, 24, 25).

Major regulatory effects of Notch signaling in alloreactive T cell immunity were recently discovered in mouse models of allogeneic bone marrow transplantation (21, 23, 26). Inhibition of all Notch signals in donor T cells led to potent protection from acute graft-versus-host disease (GVHD) (21, 23). Notch1/2 receptors and Dll1/4 Notch ligands accounted for all of the effects of Notch signaling in GVHD, with dominant roles for Notch1 and Dll4 (21). Transient inhibition of Dll1/4 in the peritransplant period led to prolonged GVHD control. Notch blockade markedly reduced the production of inflammatory cytokines, while increasing Treg expansion. Notch-deprived alloreactive T cells showed features of acquired hyporesponsiveness, suggesting that Notch should be considered a new major regulator of alloreactivity and tolerance (21, 26, 27). In organ rejection, initial work using exposure of T cells to overexpressed Notch ligands showed a potential role for Notch in tolerance induction (27–30). However, because of the artificial nature of this experimental system, no definitive information could be gathered about the role of endogenous Notch signals in transplant rejection. Riella et al. (31) targeted Dll1 Notch ligands with mAbs in a mouse model of heart transplantation. In combination with B7/CD28 blockade, they observed a significant, although modest, protective effect of Dll1 blockade associated with STAT6-dependent Th2 polarization. In contrast, Jagged2-mediated agonism mediated increased rejection (32). These observations are consistent with a tolerogenic effect of Notch inhibition during graft rejection. However, they may markedly underestimate the full impact of the Notch pathway because the study focused only on blocking a single Notch ligand, and only partial inhibition of Notch signaling was achieved, as evidenced by the persistence of Dll1-dependent marginal zone B cells in this model (31, 33). In addition, the mechanisms of protection may differ from those seen with more efficient methods of Notch inhibition across multiple ligands or receptors.

To comprehensively evaluate the role of Notch signaling in transplant rejection, we combined a genetic approach to block all canonical Notch signals in host T cells and a biochemical strategy to achieve potent systemic inhibition of Dll1/4 Notch ligands. This approach allowed us to capture the overall effects of Notch signaling in alloreactive T cells, while investigating both T cell–intrinsic and -extrinsic consequences of systemic Dll1/4 inhibition. Notch blockade in T cells increased the survival of heart allografts, especially when combined with transient CD8⁺ T cell depletion in the recipients, suggesting that alloreactive CD4⁺ T cells are particularly sensitive to the effects of Notch inhibition. Dll1/4 blockade induced superior protection from rejection compared with inhibition of Notch signaling only in T cells. Importantly, transient Dll1/4 inhibition was sufficient to induce long-term graft acceptance and inhibit pathogenic T cells, while decreasing the numbers of germinal center B cells and plasmablasts, as well as the production of donor-specific alloantibodies. This represents a different, broader, and more efficient immune intervention than reported previously only with partial Dll1 inhibition. Our data identify potent immunobiological effects of Notch signaling in multiple pathogenic aspects of alloreactivity. These observations suggest new Notch-based strategies to control organ rejection that could be considered for human interventions.

BALB/c (H-2^d) and C57BL/6 (B6) (H-2^b) mice were obtained from Charles River Laboratories (Raleigh, NC). Cd4-Cre⁺ × ROSA^DNMAMLf mice on a B6 background (B6-DNMAML) were described to express the DNMAML-GFP pan-Notch inhibitor in all mature CD4⁺ and CD8⁺ T cells (21–23, 26). DNMAML expression allowed for efficient blockade of Notch transcriptional activation downstream of all Notch receptors in T cells, without interference with Notch signaling at early stages of T cell development. All mice were housed under specific pathogen–free conditions in the Unit for Laboratory Animal Medicine at the University of Michigan. Experiments were performed according to National Institutes of Health guidelines and approved by the University of Michigan’s Committee on the Use and Care of Animals.

Heterotopic transplantation of intact allogeneic BALB/c hearts into B6 or B6-DNMAML recipients was performed as described (34, 35). Briefly, the aorta and pulmonary artery of donor hearts were anastomosed end-to-side to the recipient’s abdominal aorta and inferior vena cava, respectively. Upon perfusion with the recipient’s blood, the transplanted heart resumes contraction. Graft function was monitored by abdominal palpation. Rejection was scored based on the cessation of heart contraction.

The hybridoma-secreting anti-CD8 mAb (clone 2.43) was obtained from American Type Culture Collection (Manassas, VA). Anti-CD8 Abs were purified and resuspended in PBS by Bio X Cell (West Lebanon, NH). Where indicated, allograft recipients received 1 mg anti-CD8 mAb i.p. on days −1, 0, and 7 relative to the day of transplantation. The efficiency of CD8 depletion was verified in pilot experiments (typically near-complete depletion for ∼2 wk, followed by a gradual return to ∼50% of normal levels by day 50) (36, 37).

Humanized IgG1-neutralizing mAbs specific for the Dll1 or Dll4 extracellular domain were described previously (21, 24). An irrelevant human IgG1 Ab specific for HSV gD protein (anti-GD) was used as isotype control in selected experiments. Abs were administered i.p. (5 mg/kg). The potency and specificity of each batch of Abs were verified by in vivo administration and assessment of Dll1-dependent marginal zone B cells and Dll4-dependent thymocytes, as described (21). Abs were administered on day 0 after surgery and again on days 3, 7, and 10 after transplantation.

Allografts were recovered either at the time of rejection or at prespecified time points, fixed in formalin, and embedded in paraffin. When prespecified time points were used, graft survival data from these particular recipients were censored. Sections were stained with H&E to assess myocyte viability (presence of cross striation and myocyte nuclei) and the nature, intensity, and localization of graft-infiltrating cells (GICs). For isolation and quantification of GICs, portions of the transplanted hearts were weighed, pooled, minced, and digested with 1 mg/ml Collagenase A (Roche Diagnostics, Indianapolis, IN) for 30 min at 37°C. After tissue debris settled, suspended GICs were harvested. RBCs were lysed by hypotonic shock, and GICs were passed through a 30-μm nylon mesh. Viable cells were enumerated by trypan blue exclusion and/or assessed by flow cytometry.

The following Abs were used: anti-CD4, -CD8α, -CD19, -CD45.2, -CD95, -CD138, -B220, –I-A/I-E, –TCR-β, -Thy1.2, -GL7 (BioLegend, San Diego, CA) and anti-Foxp3 (eBioscience, San Diego, CA). For intracellular Foxp3 staining, we used a fixation/permeabilization kit, according to the manufacturer’s instructions (eBioscience). Dead cell exclusion was performed though the addition of DAPI or Ghost Violet (Tonbo Biosciences, San Diego, CA). Samples were evaluated on a BD Fortessa analyzer. In selected cases, sorting was performed on BD FACSAria III (Becton Dickinson, San Jose, CA). Flow cytometry files were analyzed using FlowJo software (TreeStar, Ashland, OR).

Paraffin-embedded allografts were sectioned, deparaffinized, and processed for Ag retrieval in Trilogy (Cell Marque, Rocklin, CA). As described previously (38), sections were incubated with rabbit anti-mouse C4d (kindly provided by Dr. William Baldwin, Cleveland Clinic, Cleveland, OH) at a 1:500 dilution, followed by detection and diaminobenzidine development using the SuperPicture Polymer Detection Kit (Life Technologies, Carlsbad, CA) and counterstaining with hematoxylin.

For enumeration of alloreactive cytokine-producing cells, ELISPOT assays were performed as previously described (39, 40). Splenocytes were cultured in RPMI 1640 supplemented with 2% FCS, 1 mM sodium pyruvate, 100 U/ml penicillin, 100 μg/ml streptomycin, 1.6 mM l-glutamine, 10 mM HEPES buffer (Life Technologies), 0.27 mM l-asparagine, 0.55 mM l-arginine HCl, 14 μM folic acid, and 50 μM 2-ME (Sigma-Aldrich, St. Louis, MO). Capture and detection Abs were as follows: IFN-γ (R4-6A2, XMG1.2), IL-4 (11B11, BVD6-24G2), and IL-17 (TC11-18H10, TC11-8H4.1) (BD Biosciences, San Jose, CA). Polyvinylidene difluoride–backed microtiter plates (Millipore, Billerica, MA) were coated with unlabeled mAb and blocked with 1% BSA in PBS. Irradiated (1000 rad) BALB/c splenocytes (4 × 10⁵) and 1 × 10⁶ recipient splenocytes were added to the plates for 24 h. After washing, a 1:1000 dilution of anti-biotin alkaline phosphatase conjugate (Vector Laboratories, Burlingame, CA) was added to IFN-γ and IL-17 plates, and a 1:2000 dilution of HRP-conjugated streptavidin (Dako, Carpinteria, CA) was added to IL-4 plates. Plates were washed, and spots were visualized by addition of Nitroblue Tetrazolium (Bio-Rad, Hercules, CA)/3-bromo-4-chloro indolyl phosphate (Sigma-Aldrich) to IFN-γ and IL-17 plates or 3-amino-9-ethylcarbazole (Pierce) to IL-4 plates. Color development continued until spots were visible and was stopped by adding H₂O. Plates were dried, and spots were quantified with an Immunospot Series 1 ELISPOT analyzer (Cellular Technology, Shaker Heights, OH).

For RNA extraction, cardiac allografts were homogenized in TRIzol reagent (Life Technologies), followed by extraction according to the manufacturer’s protocol. cDNA was prepared with Superscript II (Life Technologies). Real-time PCR was performed in triplicate for each sample with SYBR Green PCR Master Mix and analyzed on a Mastercycler ep realplex (Eppendorf). Relative transcript abundance was calculated using the ΔΔCt method after normalization with Cd3 to account for the abundance of T cells in the graft. Primer sequences were obtained from the PrimerBank (http://pga.mgh.harvard.edu/primerbank). For analysis of Dtx1 Notch target gene expression, RNA was extracted from sorted lymphocyte populations and processed as above, but using TaqMan primers with normalization to Hprt (Life Technologies).

As described (38, 41), P815 (H-2^d) cells (American Type Culture Collection) were stained for flow cytometric analysis using a 1/50 dilution of sera obtained from cardiac allograft recipients as the primary Ab, followed by FITC-conjugated rabbit anti-mouse IgG Ab (Life Technologies) used at a 1/50 dilution. Data are reported as the mean channel fluorescence determined on a Becton Dickinson FACScan.

Allograft survival curves were analyzed using a log-rank test. Significance of ELISPOT and alloantibody results was determined by an unpaired t test with Welch correction. All data were analyzed using GraphPad Prism v. 6.0. The p values ≤ 0.05 were considered statistically different.

To evaluate the role of Notch signaling in host T cells during allograft rejection, we first used transplantation of MHC-mismatched BALB/c hearts into B6 recipient mice. This model triggers an acute form of cellular rejection dominated by a Th1 pattern of cytokine release as well as production of donor-reactive alloantibodies (37, 42–44). To block Notch signaling in T cells, we studied mice expressing the pan-Notch inhibitor DNMAML in all mature CD4⁺ and CD8⁺ T cells (Cd4-Cre⁺ × ROSA^DNMAMLf or B6-DNMAML mice) (21–23, 26, 45). DNMAML expression blocks transcriptional activation downstream of all Notch receptors, an efficient strategy to capture all of the effects of canonical Notch signaling in T cells (17, 21–23, 26). In wild-type (WT) B6 recipients, hearts were rejected after a median observation of only 7 d, consistent with past observations (Fig. 1A) (42–44). In B6-DNMAML recipients, allograft survival was doubled, with a median rejection occurring at day 14. DNMAML expression in T cells led to decreased numbers of GICs when assessed at the time of rejection (Fig. 1B, 1C). B6-DNMAML mice had lower numbers of alloreactive IFN-γ–producing cells in the spleen, without an increase in IL-4–producing cells (Fig. 1D). Despite this effect on cytokine production, Notch inhibition in T cells did not prevent the appearance of donor-reactive IgG Abs (Fig. 1E). Thus, Notch blockade in T cells increased allograft survival in this acute rejection model, without providing long-term protection.

To evaluate the impact of Notch inhibition specifically in CD4⁺ T cells, we depleted CD8⁺ T cells in WT and B6-DNMAML mice (Fig. 2). This inductive strategy eliminated CD8⁺ T cells from the periphery for ≥2 wk, with a gradual return to ∼50% of levels observed in nondepleted mice by day 50 after treatment (36, 37; data not shown). CD8⁺ T cell depletion is accompanied by a decrease in IFN-γ–producing cells and an increase in IL-4–producing cells in this model (43). In WT recipients, transient Ab-mediated depletion of CD8⁺ T cells in the peritransplant period provided modest prolongation of allograft survival, as previously reported (Fig. 2A compared with Fig. 1A) (43). In contrast, rejection was markedly delayed to a median of 45 d in CD8-depleted B6-DNMAML recipients. DNMAML expression decreased the number of GICs at days 14 and 30 (Fig. 2B, 2C). ELISPOT assays in the spleen showed a trend toward decreased IFN-γ–producing cells and a significant decrease in IL-4–producing cells in B6-DNMAML recipients compared with B6 recipients (Fig. 2D). DNMAML expression also decreased the titers of donor-specific alloantibodies at day 14, although not at later time points (Fig. 2E). Altogether, alloreactive CD4⁺ T cells appeared particularly sensitive to Notch inhibition in the setting of CD8 depletion. Pan-Notch inhibition in CD4⁺ T cells decreased both Th1 and Th2 cytokines, as well as T cell help to alloantibody-secreting cells.

We previously showed dominant collective effects of Dll1 and Dll4 Notch ligands in the pathogenesis of acute GVHD after allogeneic bone marrow transplantation (21). Thus, we compared the effects of DNMAML-mediated pan-Notch inhibition in T cells with those of systemic inhibition of Dll1/4 Notch ligands during allograft rejection. To achieve potent and specific Dll1/4 blockade, we injected humanized neutralizing mAbs, starting immediately after heart transplantation and on days 3, 7, and 10 posttransplant (21, 24, 25). DNMAML was expressed in the vast majority of CD4⁺ T cells, as shown by detection of the DNMAML-GFP fusion protein in these cells at day 14 after transplantation (Fig. 3A), as well as in residual CD8⁺ T cells (data not shown). Both DNMAML expression and systemic Dll1/4 blockade profoundly decreased the abundance of Dtx1 Notch target gene transcripts in T cells, indicating efficient Notch inhibition (Fig. 3A). Transient Dll1/4 blockade had superior activity in reducing the number of GICs at early and late time points after transplantation (Fig. 3B). T cell–specific DNMAML expression and systemic Dll1/4 blockade both decreased the numbers of IFN-γ– and IL-4–producing cells at day 14 (Fig. 3C). Interestingly, the effects of Dll1/4 blockade were more persistent than were those of DNMAML expression at days 30 and 50, even though anti-Dll1/4 Abs were administered only transiently for 10 d after transplantation.

To further understand the effects of Notch inhibition, we assessed expression of Ifng and key mediators of T cell cytotoxicity (Fig. 4A). Ifng, Prf1, and Gzmb transcripts were decreased in the heart allografts of B6-DNMAML and anti-Dll1/Dll4–treated WT recipients. In contrast, Foxp3 expression was relatively increased upon Notch inhibition. Consistent with these findings, flow cytometric analysis showed a relative increase in the frequency of Foxp3-expressing graft-infiltrating CD4⁺ T cells in B6-DNMAML and anti-Dll1/4–treated transplant recipients, with no change among CD4⁺ splenocytes (Fig. 4B). In terms of absolute numbers, graft-infiltrating Foxp3⁻ conventional T cells (Tconvs) were decreased in both B6-DNMAML and anti-Dll4–treated mice, whereas absolute numbers of Foxp3⁺ Tregs were not significantly changed (Fig. 4C). The Tconv/Treg ratio was decreased in the graft, but not in the spleen, of B6-DNMAML or anti-Dll1/4–treated mice (Fig. 4D). Thus, both methods of Notch inhibition decreased the numbers of conventional effector T cells in the graft while increasing the relative frequency of Tregs.

To evaluate whether systemic Dll1/4 inhibition had effects outside of the T cell lineage that could contribute to decreased rejection, we assessed the impact of Dll1/4 blockade on alloantibody production (Fig. 5). At early time points after transplantation (day 14), both DNMAML expression in T cells and systemic Dll1/4 inhibition decreased serum titers of donor-reactive IgG Abs, but with more potent inhibition upon Dll1/4 blockade (Fig. 5A). At later time points (days 30 and 50), WT recipients were not available for comparison as a result of prior allograft rejection, whereas B6-DNMAML recipients had rising titers of donor-reactive alloantibodies (Figs. 2A, 5A). The emergence of donor-reactive Abs was associated with eventual graft loss in B6-DNMAML mice (Fig. 2A). In contrast, alloantibody titers remained very low at days 30–50 after systemic Dll1/4 inhibition (Fig. 5A). Next, we compared the effects of DNMAML expression and systemic Dll1/4 inhibition on CD138⁺ plasmablasts in the spleen (Fig. 5B). DNMAML-mediated Notch inhibition in T cells induced a significant, but modest, decrease in plasmablasts, presumably through an indirect effect on T cell help. Anti-Dll1/4 Abs markedly reduced plasmablast numbers to the same extent as their effects on alloantibody titers (Fig. 5A, 5B). Numbers of splenic B220⁺CD95⁺GL7⁺ germinal center B cells also were profoundly decreased in anti-Dll1/4–treated recipients but not in B6-DNMAML recipients (Fig. 5C). Finally, systemic Dll1/4 inhibition, but not DNMAML expression in T cells, eliminated marginal zone B cells (data not shown), consistent with the Dll1 dependence of this population (33), further revealing the direct impact of Dll1/4 Notch ligands on B cell populations. At the termination of the experiment, immunohistochemistry showed complement deposition within the heart allografts in B6-DNMAML recipients but not in anti-Dll1/4–treated mice (Fig. 5D). These findings suggest that systemic blockade of Dll1/4 Notch ligands decreased the number of B lineage cells differentiating into Ab-producing cells, controlling the accumulation of pathogenic alloantibodies and complement deposition in this rejection model.

Based on the superior effect of transient Dll1/4 blockade on surrogate end points of alloreactivity, we tested the overall impact of this strategy on allograft survival and the individual contribution of Dll1 and Dll4 ligands in CD8-depleted recipients (Fig. 6). Either Dll1 or Dll4 inhibition alone markedly prolonged allograft survival and decreased the number of GICs, indicating that both Notch ligands contributed to promote rejection (Fig. 6A, 6B). Remarkably, peritransplant inhibition of both Dll1 and Dll4 allowed 100% long-term graft survival during the entire observation period (Fig. 6A). This is superior to graft survival in CD8-depleted B6-DNMAML recipients (Fig. 2A). Combined Dll1/4 blockade decreased graft infiltration more profoundly than did Dll1 or Dll4 inhibition alone (Fig. 6B), while also blocking the emergence of donor-specific alloantibodies (Fig. 6C). Thus, when combined with CD8 depletion, transient systemic Dll1/4 blockade induced long-term acceptance of MHC-mismatched heart allografts.

Given its profound effects in the CD8-depleted mice, we assessed the impact of combined Dll1/4 inhibition on rejection of MHC-mismatched allografts in the absence of T cell depletion. In contrast to the modest protective effects of T cell–specific DNMAML expression, systemic Dll1/4 blockade in WT recipients markedly prolonged allograft survival to a median duration of 37 d (Figs. 1A, 7A). Rejection in these animals correlated with the presence of serum alloantibodies. When combined with DNMAML expression, peri-transplant Dll1/4 blockade led to long-term graft acceptance for ≥50 d after transplantation, even in the absence of CD8 depletion (Fig. 7A). Anti-Dll1/4 Abs decreased graft infiltration and production of inflammatory cytokines more profoundly than did DNMAML expression alone (Fig. 7B, 7C). The combination of DNMAML expression and anti-Dll1/4 Abs was most effective at decreasing the number of cytokine-producing cells and donor-reactive alloantibodies, correlating with the absence of rejection in non-CD8–depleted mice (Fig. 7C, 7D). In this model with an intact CD8⁺ T cell compartment, our findings suggest that long-term T cell–specific inhibition was important to promote the maximum degree of prolonged graft acceptance, indicating that the temporal effects of Notch signaling in CD4⁺ and CD8⁺ alloreactive T cells may differ. However, systemic peritransplant Dll1/4 blockade was required for maximal protection (Figs. 1A, 7A), suggesting that non-T cell effects of Dll1/4 inhibition play an important role.

Notch signaling is emerging as a powerful context-specific regulator of Ag-driven immune responses (10, 46, 47). Understanding these effects is essential to characterize the versatile immunobiological effects of Notch signaling, as well as to discover new Notch-based therapeutic interventions with translational potential. Our findings identify a major pathogenic role for Notch signaling driven by two Dll Notch ligands (Dll1/4) in cellular and humoral rejection of mouse heart allografts. These effects are consistent with the protective effects of Dll1 inhibition that were reported previously, although those were modest and only were identified upon concomitant inhibition of B7/CD28 signaling (31). Past observations were obtained using systemic Abs that only partially inhibited Dll1-mediated signals, as evidenced by the persistence of Dll1-dependent marginal zone B cells (31, 33). In contrast, we used a genetic approach to fully inhibit Notch signaling downstream of all Notch receptors in T cells, as well as systemic administration of potent anti-Dll1 and anti-Dll4 mAbs to achieve near-complete inhibition of both Dll ligands in vivo. Using this strategy, we observed a high degree of protection, even without interfering with CD28-mediated signals. Thus, our findings suggest a much broader impact of Notch inhibition than previously observed with systemic Dll1 blockade only (31).

Transient inhibition of Dll1 and Dll4 ligands in the peritransplant period was sufficient to confer long-term protection, suggesting a tolerogenic effect. Dll1/4 inhibition blocked the production of multiple T cell inflammatory cytokines and decreased the accumulation of conventional effector T cells in the graft, while leading to a relative increase in the frequency of Foxp3⁺ T cells. These effects of Notch inhibition on the Tconv/Treg balance during graft rejection are reminiscent of our observations in GVHD, suggesting the existence of conserved effects of Notch signaling in alloreactivity (21, 23, 26). In GVHD, we previously showed that the increased relative accumulation of Notch-deficient Tregs correlated with enhanced in vivo expansion of pre-existing Tregs (26). In addition to its effects on T cells, systemic administration of anti-Dll1/4 Abs also blocked the production and deposition of donor-reactive alloantibodies, a major mediator of acute and chronic rejection (2–6). Altogether, these findings suggest a central pathogenic role for signals delivered by Dll1 and Dll4 Notch ligands during early stages after organ transplantation, with long-term immunobiological effects even observed upon transient interruption of these signals. Thus, Dll1 and Dll4 Notch ligands are attractive therapeutic targets to prevent multiple aspects of organ allograft rejection.

In the acute rejection model that we studied, the protective effects of Notch inhibition were most pronounced in CD8-depleted recipients, suggesting that CD4⁺ alloreactive T cells were particularly sensitive to Notch inhibition. Notch inhibition markedly decreased the number of IL-4–producing alloreactive T cells that were previously shown to dominate in the CD8-depleted model (43). This effect of complete Notch blockade in T cells differed from the Th2 bias observed upon concomitant partial Dll1 inhibition and B7/CD28 blockade (31). Moreover, the effects of Notch blockade were not limited to Th2 differentiation, because IFN-γ–producing cells also were affected in this model, as well as in the absence of CD8⁺ T cell depletion. These observations differ from past reports linking Notch signaling preferentially to Th2 differentiation (22, 31, 48), but they are consistent with recent observations that Notch can regulate the differentiation and function of multiple Th cell lineages (21–23, 26, 48–51). In particular, our observations in transplant rejection have multiple features in common with our findings in GVHD (21, 23, 26). In both cases, transient Notch blockade was sufficient to provide long-term protection and was associated with decreased production of multiple inflammatory cytokines, while preserving or enhancing the expansion of Tregs. After bone marrow transplantation, we showed that Notch blockade induces a hyporesponsive state in alloreactive T cells, including features previously reported in models of T cell anergy (26). Thus, Notch signaling may exert a conserved set of effects in T cell alloimmunity, with tolerogenic effects of Notch inhibition in both graft-versus-host and host-versus-graft reactivity.

An interesting consequence of Notch inhibition was a blunting of the accumulation of donor-specific alloantibodies, as well as complement deposition in the allograft. Dll1/4 inhibition had particularly profound effects on B lineage cells differentiating into Ab-producing cells and blocked the alloantibody response that correlated with long-term allograft rejection in the CD8-depleted rejection model. Because T cell–specific Notch inhibition with DNMAML also delayed alloantibody production, decreased T cell cytokine production may be involved in this effect. Alternatively, Notch inhibition may block the differentiation and function of T follicular helper cells that support the germinal center reaction, as shown recently with model Ags (52). However, systemic Dll1/4 inhibition had more profound and durable effects on alloantibody production than did T cell–specific Notch inhibition in the CD8-depleted model, suggesting that direct effects of Dll1/4 blockade on B lineage cells could play a major role. Possible targets include germinal center B cells and plasma cells, consistent with the markedly decreased numbers of these cells in anti-Dll1/4–treated mice (33, 53–55). Alternatively, it is possible that anti-Dll1/4 Abs, but not DNMAML expression, inhibit putative noncanonical effects of Notch signaling in T cells that are independent of transcriptional activation by CSL/RBP-Jk and MAML proteins (47, 56–58). Such noncanonical pathways could control T cell help beyond the effects of canonical Notch signaling. Regardless of the pathways involved, our findings are highly significant, because Ab-mediated mechanisms are increasingly recognized to play an important role in acute and chronic organ rejection (2–6).

We found that Dll1 and Dll4 Notch ligands both contributed to the rejection process, with additive benefits of Dll1 and Dll4 blockade. Thus, the effects of Notch inhibition were underestimated in studies using Dll1 blockade alone, especially because only partial Dll1 inhibition was achieved (31). In mouse models of acute GVHD, we also observed individual effects of Dll family ligands (21), although the relative importance of Dll1 appeared higher in transplant rejection. It remains to be determined whether Dll1 and Dll4 exert their effects on the same immune cells or on distinct aspects of the alloimmune response. Because anti-Dll1/4 Abs were administered systemically, both T cell and non-T cell effects could contribute to the protection. These include inhibition of B cell or plasma cell function, effects on dendritic cells, and thymic effects (7, 33, 53–55, 59, 60). Dll4 inhibition was reported to block early T cell development, but with rebound production of natural Tregs after recovery from Dll4 blockade (60). These effects were shown to dampen autoimmune diabetes and, thus, could also be involved in mitigating transplant rejection, although it cannot explain protection during administration of the Abs during the initial 2 wk after transplantation.

Altogether, our findings have fundamental immunobiological and translational implications. Notch appears to play a central role in the regulation of alloreactivity. Its full impact is best revealed by efficient in vivo loss-of-function strategies that completely block all Notch effects in T cells or all systemic Dll-mediated signals. In view of these profound effects, it will be essential to understand how this pathway is controlled and how it interacts with other critical regulators of T cell differentiation and function. In terms of translational applications, humanized anti-Dll1/4 used in this study cross-react with mouse and human Notch ligands and, thus, could be considered in principle for human interventions, especially because short-term blockade in the peritransplant period has the potential to confer long-term allograft survival.

A.S., M.Y., and C.W.S. are employees of Genentech.