Flt-3L Expansion of Recipient CD8α⁺ Dendritic Cells Deletes Alloreactive Donor T Cells and Represents an Alternative to Posttransplant Cyclophosphamide for the Prevention of GVHD

Understanding the modifiable determinants of alloreactivity is critical for the design of rational strategies to prevent and treat graft-versus-host disease (GVHD) after allogeneic bone marrow transplantation (BMT). The elimination of GVHD must be balanced against maintenance of protective graft-versus-leukemia (GVL) effects and meaningful separation of these two immunologic phenomena remains the ultimate goal in the field. Furthermore, retaining functionality within the nonalloreactive T-cell pool is paramount for the prevention of severe infection that remains responsible for both short- and long-term mortality after BMT.

Outside of selecting optimally matched donors, the prevention of pathogenic alloreactive T-cell responses (i.e., GVHD) relies on T-cell depletion or pharmacologic immune suppression, usually with calcineurin inhibitors (i.e., cyclosporin or tacrolimus). Posttransplant cyclophosphamide (PT-Cy), which is thought to eliminate activated, proliferating alloreactive donor T cells (while sparing regulatory T-cell populations) is also highly effective in reducing the incidence of chronic GVHD after haploidentical BMT (1–5).

Retrospective data suggest that PT-Cy may be superior to standard immune suppression (with or without anti-thymocyte globulin) with regard to the prevention of chronic GVHD after unrelated donor transplantation (6) although this requires confirmation in prospective randomized studies. Interestingly, the effect of PT-Cy on leukemia relapse is unclear, and importantly, no prospective studies to date have been sufficiently powered to analyze relapse after BMT using PT-Cy compared with standard immune suppression. With the expanding role of alternate donor transplantation, it is critically important to understand the drivers of early T-cell responses that characterize GVL effects and infectious immunity, such that they may be targeted for therapeutic benefit.

It is now clear that recipient antigen-presenting cells (APC), both hematopoietic and nonhematopoietic, initiate alloreactive donor T-cell responses and GVHD (7), while reconstituting donor DCs determine final GVHD severity (8, 9). As yet, no preventive or therapeutic approaches have been developed to specifically target the antigen presentation component of T-cell activation. With regard to GVL responses, recipient APC (putatively dendritic cells, but this is as yet unproven) are thought to determine the magnitude of responses (10), with donor APC playing a limited role (11).

Given the important effects of CD8⁺ T cells in HLA-matched clinical transplantation (12), we have focused on CD8⁺ T cell-driven mouse models of allogeneic BMT. Here, we have defined the contribution of dendritic cells (DC) to GVL effects, and explored the influence of recipient DCs on donor T-cell activation following allogeneic BMT in models that recapitulate both HLA matched and haploidentical transplantation in the clinic. Surprisingly, we find that recipient CD8α⁺ DCs potently induce antigen-specific deletion of donor cytotoxic T cells (CTL). We demonstrate that this effect can be exploited to prevent GVHD by further expanding recipient DCs with recombinant fms-like tyrosine kinase-3 ligand (Flt-3L); notably, this strategy appears superior to current approaches for in vivo T-cell depletion using PT-Cy, as the GVHD outcome is superior and the polyclonal T-cell pool appears maintained.

Female C57BL/6 (B6.WT, H-2^b), Ptprc^a (B6.Ptprc^a, H-2^b, CD45.1; also described as B6.CD45.1 for clarity), and B6D2F1 (H-2^b/d) mice were purchased from the Animal Resources Centre (WA, Australia). The following strains were bred and housed at QIMRB: C3H.Sw (H-2^b, Ly9.1), B6.CD11c.DOG (H-2^b) and B6.CD11c.DOGxDBA/2F1 (H-2D^b/d; diphtheria toxin (DT) receptor, ovalbumin (OVA), and enhanced GFP (eGFP) driven off the CD11c promoter), B6.CD11cGCDL (eGFP, Cre, the DTR, luciferase driven off the CD11c promoter), B6.IL12p40 eYFP (13), OT-I Tg, CD11c.Rac Tg (14, 15), IRF8^−/− mice [>10 x backcrosses to B6 background, lacking in CD8α⁺ DC; ref. 16), Batf3^−/− (lacking in CD8α⁺ DC, (17)], β₂m^−/−, Bm1 (H-2K^bm1), and Bm1.ActmOVA (H-2K^bm1 and ubiquitous ovalbumin expression driven off the β-actin promoter). Female mice at 8 to 12 weeks of age were used throughout the study.

For IRF8^−/− chimeras, bone marrow was harvested from IRF8^−/− mice and 5 × 10⁶ bone marrow cells/mouse transferred by intravenous injection into lethally irradiated (1,000 cGy) B6.WT or B6.CD45.1 (PTprc^a) recipients and allowed to reconstitute for 3 months prior to second transplantation.

B6 recipients of C3HSw grafts received 1 × 10⁶ FACS purified (CD90.2⁺/CD4⁻) CD8⁺ T cells and 5 × 10⁶ bone marrow cells. Where T cell depleted (TCD) bone marrow controls were included, cells were prepared using an antibody incubation followed by complement depletion as described previously (18). In the C3H.Sw model, donor mice were immunized via intraperitoneal injection with B6 splenocytes 2 weeks prior to transplantation. B6D2F1 recipients received 5 × 10⁶ TCD bone marrow cells + either magnetic bead–purified CD8⁺ T cells (MACS-purified according to the manufacturer's instructions, Miltenyi Biotec), CD3⁺ T cells (as previously described; ref. 18), or OT-I Tg CD8⁺ T cells in doses as stated. OT-I were MACS-purified from spleen and lymph nodes. Where OT-I T cells were transferred to read out antigen-specific responses, MHC class I–deficient (β₂m^−/−) bone marrow was used to exclude any contribution of indirect antigen presentation by donor APC.

Total body irradiation (TBI) doses were as follows: B6 background, 900 cGy in the presence of DT treatment and 1,000 cGy otherwise; B6D2F1 mice, 1,100 cGy; Balb/B mice 400 cGy (in conjunction with 2 mg fludarabine d-4 to d-2).

For in vivo depletion of DTR-expressing DCs, diphtheria toxin from Corynebacterium diphtheriae (Sigma Aldrich) was administered intraperitoneally at 160 ng per dose on day −2, −1, 0, 1, and 2 for day 3 analyses and continued on day 5, 7, 9, for day 10 analyses. Primary leukemia cells were generated using the expression of the human oncogenes MLL-AF9 or BCR-ABL + NUP98-HOXA9 to model human AML and myeloid blast-crisis leukemia, as described previously (19) and cryopreserved at disease onset, for subsequent transplantation. Leukemia cells were thawed on the day of injection and included in grafts at 0.5–1 × 10⁶/mouse. Mice were scored according to standard protocols and sacrificed if clinical score reached ≥6, in accordance with animal ethics guidelines and a previously established scoring system (20). For a death to be attributed to leukemia, leukemia burden in peripheral blood at either terminal or last routine bleed had to meet the following criteria to avoid overstatement of leukemic deaths for the rare cases when both GVHD and leukemia were present: greater than 10% GFP⁺ cells in peripheral blood (with any total white cell count) or present in the peripheral blood at any level but with a total WCC ≥ 1 × 10⁷/mL.

For in vivo expansion of DCs, mice received 10 μg of Flt-3 ligand, daily via subcutaneous injection (Flt-3L; Celldex Therapeutics) from d-10 to -1 (21, 22). Posttransplant cyclophosphamide (PT-Cy) was administered on d+3 and d+4 at 100 mg/kg i.p. Cyclosporin was administered at 5 mg/kg i.p. from d0 to d+14.

For poly I:C experiments, mice received 100 μg of poly I:C (Invivogen) in saline via intraperitoneal injection 1 hour following the second dose of TBI, and prior to the injection of OT-I T cells.

Female C57BL/6 mice were infected intraperitoneally with 10⁴ PFU of salivary gland propagated MCMV-K181-Perth for >90 days to establish a latent MCMV infection, as determined by the absence of replicating virus. Splenic T cells were isolated from latently infected mice and transplanted using the same methods described below. MCMV-infected mice were housed at the University of Western Australia prior to being used as BMT donors. CMV-specific CD8⁺ T cells were identified using PE-conjugated tetramer for H-2K^b-SSPPMFRV MCMV-m38 (ImmunoID Tetramers, University of Melbourne, Australia).

Total RNA was extracted with the RNeasy Mini Plus kit (Qiagen) from sort-purified (>95% purity) CD8⁺ donor-type T cells and gene expression measured using the Qiagen RT² Profiler PCR Array Mouse Apoptosis kit (Qiagen), and validated using TaqMan GE assays (Applied Biosystems).

Bioluminescent imaging was performed to demonstrate the depletion and expansion of DCs (using B6.CD11c.GCDL mice, and chimeras in which the hematopoietic compartment alone was of B6.CD11cGCDL origin). Recipients were injected subcutaneously (SC) with d-Luciferin (0.5 mg, PerkinElmer) and then anesthetized with isoflurane 5 minutes before imaging using the Xenogen imaging system (Xenogen IVIS 100; Caliper Life Sciences; ref. 23). Bioluminescence (BLI) shown as photons per second (ph/s).

In vivo cytotoxicity assays were performed as follows: 10 days after BMT, recipient mice received 2 × 10⁷ congenic donor-type (PTprc^a, CD45.1⁺) unlabeled splenocytes and 2 × 10⁷ host-type B6D2F1 CD45.2⁺ CFSE-labeled splenocytes IV (24). Eighteen hours later, peripheral blood and spleen were analyzed for remaining donor (CD45.1⁺ PE-stained) and host-type (CFSE-labeled) cells by FACS analysis. The ratio of adoptively transferred donor to recipient cells in spleen after 18 hours reflects degree of cytotoxicity of the resident donor T-cell population and is reported as: n donor-type/n host-type cells recovered.

For in vitro cytotoxicity, allogeneic (B6D2F1 BCR-ABL + NUP98-HOXA9) and syngeneic (B6 MLL-AF9) primary leukemia target cells were labeled with ⁵¹Cr, prior to culture with donor CD8⁺ effector T cells (FACS purified from recipient spleens on day 10 following BMT, CD45.1⁺/CD8⁺) for 5 hours at 37°C, 5% CO₂. ⁵¹Cr release into culture supernatant was determined via gamma counter (TopCount microplate scintillation counter, Packard Instruments). Spontaneous release was determined using wells containing labeled target only, and maximum release from wells containing targets + 1% Triton X−100. Percentage cytotoxicity was determined as follows: percentage cytotoxicity = (experimental release − spontaneous release)/(maximum release − spontaneous release) × 100.

A full list of mAbs utilized is given in Supplementary Table S2. Annexin V was assessed as previously described (18). Intracellular staining with activated caspase-3 was performed according to the manufacturer's instructions (BD Pharmingen). Flow cytometry analysis was performed using an LSR Fortessa II (BD Biosciences) using FACSDiva software (Version 8.0.1). Offline analysis was performed using FlowJo (Version 10, Treestar).

Survival curves were plotted using Kaplan–Meier estimates and compared by log-rank analysis. Unpaired two-tailed Mann–Whitney tests were used throughout. Data are mean ± SEM and P < 0.05 considered significant. GraphPad Prism (Version 6.00 for Windows, GraphPad Software, www.graphpad.com) was used for the generation of graphs and for statistical analysis. Researchers were not blind to the groups at the time of analysis. *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001. Sample sizes for mouse experiments were estimated on the basis of our expected effect size and previous experience with these models. Where all recipient mice were WT, formal randomization was not conducted, but groups were matched for weight at the commencement of experiments.

Immune-mediated antitumor effects are increasingly recognized as central to long-term disease control following clinical transplantation. The contribution of recipient DC to CD8⁺ T-cell–mediated GVL effects was examined in a CD8⁺ T-cell–dependent, miHA mismatched C3H.Sw → B6 mouse model of BMT (25). Recipient B6.CD11c.DOG mice (where diphtheria toxin receptor, OVA, and GFP are driven off the CD11c promoter) with either DCs intact (saline treated) or DCs depleted (diphtheria toxin treated) were transplanted with C3H.Sw bone marrow and T cells, as well as B6-derived (recipient-type) MLL-AF9–transduced leukemia, to mimic the clinical potential for relapse in parallel with the immunologic GVL effect exerted by the donor graft (19, 26). Unexpectedly, mice developed higher leukemia burdens and died more rapidly when recipient DCs were present compared with when they had been depleted, suggesting that DCs somehow function to attenuate GVL effects (Fig. 1A and B). This model does not induce GVHD mortality under control conditions, and as such, survival data reflects solely leukemia-related death. Of note, when we performed these experiments using a lymphoma cell line (EL4, which generates hepatosplenic masses rather than a leukemic phase) to model GVL there were no differences between DC-depleted and -replete recipients (data not shown).

As CD8α⁺ DCs have been implicated in tolerance induction (27, 28), we next performed transplants using recipients lacking IFN regulatory factor 8 (Irf8) gene expression in the hematopoietic compartment (Irf8^−/−→ B6 chimeras), and thus missing CD8α⁺ DC (Fig. 1C). The improvement in survival in the isolated absence of CD8α⁺ DC mimicked that observed in the setting of pan-DC depletion (Fig. 1D and E), suggesting CD8α⁺ DCs are the key population required for control of leukemia relapse. Recipient DCs as a whole, and the CD8α⁺ DC subset specifically are lost rapidly after total body irradiation (TBI), and as expected, the CD8α⁺ DCs preferentially produce IL12, both homeostatically and following TBI (Fig. 1F and G).

CD8α⁺ DCs are known to be highly efficient at the uptake of cell-associated antigen from dying cells (29, 30); we therefore investigated the contribution of this pathway (i.e., cross-presentation) to leukemia control after allogeneic BMT. We utilized Rac1 transgenic recipients, where DCs specifically lack the capacity to acquire apoptotic antigen due to deficiency of the Rho GTP-ase Rac1 within CD11c⁺ cells (14). Surprisingly, there was no impairment in GVL (Fig. 1H), suggesting that recipient DCs regulate GVL effects via the presentation of endogenous alloantigen, rather than cross-presentation of exogenous alloantigen.

To understand the profound modulation of CD8⁺ T cell-mediated GVL effects by recipient DCs, we examined the donor T-cell compartment after BMT. Donor CD8⁺ T-cell numbers were equivalent following transplantation into DC-depleted or DC-intact mice at d3, but were significantly reduced by d7 in recipients with DCs intact (Fig. 2A). This was associated with a marked increase in donor cells undergoing apoptosis at both time points examined (Fig. 2B). The same effect was observed in both IFR8^−/− → B6 chimeras and Batf3^−/− recipients (17) that lack the CD8α⁺ DC subset in isolation (Fig. 2C and D).

We next analyzed donor CD8⁺ T cells sort-purified from DC depleted or intact recipients on d7 using a targeted PCR array. We identified four genes that were expressed at higher levels in the T cells from mice that had not been exposed to recipient DCs, all of which are involved in regulation of cellular growth and apoptosis: Insulin-like growth factor 1 receptor (Igfr1; 4.44 fold), transformation-related protein 63 and 73 (Trp63; 2.29 fold and Trp73; 3.00 fold), and death associated protein kinase 1 (DapK1; 2.22 fold; Fig. 2E). This likely reflects the loss of, or lack of requirement for, regulation among the nonalloreactive pool of T cells that remain post-DC-mediated deletion. These changes in gene expression were validated by qPCR in the haploidentical model at d10, with consistent results obtained (Fig. 2F). The full panel of genes analyzed is listed in Supplementary Table S1.

We next measured antigen-specific T-cell function using CD8⁺ OT-I transgenic donor T cells and B6.CD11c.DOGxDBA/2 F1 recipients, with or without DT treatment to deplete DCs. Ovalbumin (OVA), which is expressed by recipient DC (driven by the CD11c promoter), serves as a model alloantigen in this system. OT-I T cells exclusively recognize the OVA-derived peptide SIINFEKL in the context of MHC class I, and therefore OT-I responses quantify DC-specific, endogenous presentation of OVA-derived peptide. Once again, we observed DC-associated compartment size contraction and enhanced apoptosis (Fig. 3A and B), as well as hyperactivation of OT-I transgenic T cells (CD25 and CD69 expression, Fig. 3C). OT-I acquired an effector phenotype (CD62L^neg/CD44^high) more rapidly in the presence of DCs (78.0 ± 0.9% in DC replete vs. 33.2 ± 1.8% in DC deplete at 3 days after BMT; Fig. 3D). Strikingly, recipient DCs induced high levels of exhaustion markers PD-1 and Lag3 on T cells early after BMT, and Tim3 was present at high levels, but was not differentially expressed (Fig. 3E). Exhaustion marker expression was markedly decreased in both T-cell groups by d10, with enhanced expression seen in the T cells from DC-depleted recipients likely reflecting the delayed, more measured activation that occurs when initial interaction with DC is denied.

To assess the impact of DCs on antigen-specific T cells within a polyclonal pool (containing both CD4⁺ and CD8⁺ T cells), we used the parent-into-F1 (B6 → B6D2F1) model of haploidentical transplantation and waited until donor T-cell numbers were equal between DC-depleted and -replete recipients (d10, Fig. 3F). We noted reduced cytotoxic function in vivo (Fig. 3G), and a striking absence of in vitro cytotoxic function against recipient-type leukemia (Fig. 3H), consistent with the specific deletion of alloreactive CTL, even when a full complement of polyclonal T cells are present in the initial graft.

Having established recipient CD8α⁺ DCs as key drivers of alloantigen-specific donor T-cell deletion, we sought to further harness this effect by administering Flt-3L pretransplantation (Pre-T Flt-3L). Expansion of CD8α⁺ DC was confirmed by bioluminescence and flow cytometry (Fig. 4A and B). When OT-I T cells were transferred into Pre-T Flt-3L conditioned B6.CD11c.DOGxDBA/2F1 recipients, OT-I were present in increased numbers in spleen and lymph nodes 12 hours after transfer when compared with saline-treated controls (Fig. 4C and D). Strikingly, by day 3, the reverse was true, with OT-I undetectable in Flt3-L–treated animals, consistent with the induction of widespread deletion following initial stimulation and early expansion (Fig. 4E). By day 10, splenic OT-I numbers were equivalent in the Flt-3L and saline groups, suggesting a reexpansion of surviving T cells in the Flt-3L pretreated mice. These were, however, markedly different to the OT-I present in saline pretreated mice, skewed toward a T_EM/T_EFF phenotype and a high proportion were undergoing active apoptosis, as measured by activated caspase-3 intracellular staining (Fig. 4F).

To examine the functional consequences of Flt-3L expansion on alloreactivity, we first enumerated polyclonal B6 T cells (both CD4⁺ and CD8⁺) following transplantation into saline or Flt-3L–treated B6D2F1 recipients, a model of haploidentical transplantation in the clinic. Interestingly, polyclonal T cells were preserved in the setting of Flt3L treatment (Fig. 5A). To explore the role of antigen-specific T-cell deletion in leukemia relapse, Pre-T Flt-3L recipient mice were transplanted with BCR-ABL + NUP98-HOXA9 cotransduced primary leukemia with bone marrow ± polyclonal CD3 T cells (31, 32). Median leukemia survival was just 11 days in the Pre-T Flt-3L BM + T group, equivalent to recipients of T-cell–depleted (TCD) control grafts. The saline pretreated mice receiving bone marrow + T grafts survived long-term (Fig. 5B and C).

We confirmed that this was a DC-specific effect by using Flt-3L pretreatment in combination with the DT-depletion approach previously described (Fig. 5D and E). The mice treated with Flt-3L and DT (such that DCs were expanded and subsequently deleted) had equivalent leukemia control to that seen in the WT bone marrow + T recipients pretreated with saline (Fig. 5E), confirming the primacy of recipient DCs in this effect.

To assess whether this deletion was due to the potentially tolerogenic nature of CD8α⁺ DCs, we treated mice with Flt-3L or saline as described, and then activated recipient DCs with toll-like receptor 3 (TLR3) ligand polyinosinic:polycytidylic acid (poly I:C) prior to the infusion of antigen-specific OT-I T cells. As expected, DCs were highly activated following poly I:C treatment (Fig. 5F; ref. 33). Importantly, complete deletion of antigen-specific T cells in Flt-3L pretreated recipients was maintained in the setting of this DC activation (Fig. 5G and H). Interestingly, while deletion was most effective in the Flt-3L pretreated mice, poly I:C activation of residual DCs also lead to marked donor T-cell depletion in saline pretreated recipients. Thus, the ability of recipient DCs to attenuate GVHD reflects direct alloantigen presentation and their capacity as highly potent APC rather than any intrinsic regulatory property.

We next assessed the impact of posttransplant cyclophosphamide (PT-Cy) on antigen-specific T-cell depletion, maintenance of the polyclonal T-cell pool and ultimately, GVL effects. As expected, the high-dose chemotherapy effectively eliminated the alloreactive OT-I T cells (Fig. 5I), but also had significant impact on the polyclonal T-cell pool when WT B6 grafts were transplanted (Fig. 5J). PT-Cy had a significant impact on leukemia relapse. While the PT-Cy TCD group experienced delayed relapse compared with saline-treated animals (median survival 22 days compared with 14.5), likely due to direct effects of the cyclophosphamide on the leukemia cells. PT-Cy resulted in rapid relapse in the bone marrow + T recipients (median survival 26 days, compared with unreached in the saline-treated controls, as no mice died of leukemia; Fig. 5K).

Following on from this striking and rapid relapse in the setting of PT-Cy, we next performed experiments using clinically relevant calcineurin inhibition (CsA, 5 mg/kg, day 0–14, achieving trough levels between 200 and 300 mg/dL; Fig. 5L). CsA-treated mice relapsed in an equivalent fashion to the saline treated controls, and the PT-Cy mice once again relapsed at similar rates whether T cells were present in the graft, or grafts were TCD marrow alone.

Until there are clear therapeutic strategies which separate GVHD and GVL, it is likely that any strategy chosen to prevent and treat GVHD will have some influence on relapse risk. In that setting, we sought to assess the relative benefit from a GVHD point of view from PT-Cy and Flt-3L pretreatment, given that both have a detrimental (and equivalent) impact on leukemia relapse. We found that Flt-3L pretreatment was superior to PT-Cy (median survival 45 vs. 36 days) for GVHD prevention (Fig. 6A) in a haploidentical model. Given that Flt-3L is a myeloid growth factor, efforts to translate this strategy for GVHD prevention would need to begin in the lymphoid malignancies, in which setting patients are often conditioning with reduced intensity regimens. We therefore performed transplants in an additional miHA mismatch model (B6 to Balb/B) with fludarabine and low-dose TBI conditioning. The benefit of Flt3L pretreatment was maintained in this setting, with no GVHD deaths in the Flt-3L pretreated recipients of bone marrow + T grafts (Fig. 6B), and equivalent engraftment.

Having observed profound effects on broad alloreactivity (i.e., both GVL and GVHD) in response to both Pre-T Flt-3L and PT-Cy, we sought to examine maintenance of virus-specific immunity using murine cytomegalovirus (MCMV) immune B6 donor mice in the haploidentical B6 → B6D2F1 model. We hypothesized that in the absence of CMV-specific antigen in the peritransplant period, the CMV-specific memory T-cell pool would be intact following transplant into Flt-3L-pretreated recipients. At the time of maximal clonal deletion of alloreactive T cells, CMV-specific (m38 tetramer⁺ cells) were indeed present in equivalent numbers in Flt-3L pretreated mice compared with saline pretreated controls (Fig. 6C). In contrast, there was profound reduction in all T lymphocytes in the PT-Cy mice, with almost complete loss of CMV-specific CD8 T cells (Fig. 6D). Thus unlike PT-Cy, pre-T Flt-3L preserves virus-specific T cells.

In this study, we have examined the role of recipient DCs in controlling CD8⁺ T-cell–mediated alloreactivity and propose that pretransplant Flt-3L therapy may provide a new strategy for the prevention of GVHD in the clinic. It is clear that this comes at the cost of decreasing T-cell–mediated GVL effects, but in the era of posttransplantation cellular therapies (e.g., with engineered CAR T cells or suicide-gene transfected polyclonal donor T cells) it is likely that this can be overcome. Importantly, pretransplant Flt-3L therapy appears to spare the polyclonal T-cell pool and in light of the importance of opportunistic infection after transplant, this represents a significant advantage over nonselective chemotherapy-based approaches for the deletion of alloreactive T-cell populations that are currently used. In addition to the polyclonal T-cell pool, Flt-3L appears to spare CMV-specific immunity where PT-Cy does not. This is likely explained by the relative absence of viral antigen in the immediate post-transplant period.

Mechanistically, we demonstrate that alloantigen presentation within MHC class I by recipient CD8α⁺ DC results in activation induced cell death and exhaustion of allospecific CD8⁺ cytolytic T cells. We thus demonstrate the ability of DCs to delete MHC class I–dependent GVL effects, as well as CD8⁺ T-cell–mediated GVHD. A previous study using Batf3-deficient recipients and a T-cell lymphoma cell line to model relapse after BMT reported that CD8α⁺ DCs were required for optimal GVL effects, which is in contrast to our findings using primary myeloid leukemias (34). Indeed, we could see no effect of recipient DC depletion on GVL against the EL4 cell line (data not shown). This likely reflects the high mutational burdens in the multiply passaged cell lines used in those studies relative to primary leukemia (35) and the fact that these models are poor discriminators of quantitative defects in GVL (12). The results presented here also serve to confirm and extend data from ourselves and others (7, 36, 37) demonstrating that recipient DCs are not required for the induction of GVHD. Indeed, we have previously shown that recipient DCs also potently delete antigen-specific donor CD4 T cells to attenuate GVHD (7).

Our work demonstrates that DCs act to control alloreactivity via their profound stimulatory capacity, which results in clonal T-cell deletion. This effect can be augmented by expanding recipient CD8α⁺ DCs (e.g., with Flt-3L), or enhancing residual DC activation (e.g., with poly I:C). Interestingly, the administration of exogenous IL12 within 12 hours of BMT has been shown to prevent GVHD by inducing donor T-cell apoptosis and is again consistent with the CD8α⁺ DC–mediated effect demonstrated here (38–40). While the administration of IL12 has significant potential for toxicity in clinical BMT and has not progressed into the clinic, Flt-3L has been widely administered to patients (41, 42) and does represent a feasible clinical approach.

When Flt-3L therapy was previously studied in GVHD, it appeared to have mixed effects, dependent on the timing of administration (21, 43). Flt-3L administration post-BMT expanded donor DCs and resulted in the expected acceleration of GVHD (43). Conversely, treatment of recipients prior to BMT appeared to reduce GVHD, and the authors observed lower numbers of donor T cells early after transplantation. At that time, this was interpreted as a failure of Flt-3L–expanded DCs to stimulate T-cell expansion, as DCs were thought to have an exclusively stimulatory function (21).

Expansion of CD8α⁺ DC using Pre-T Flt-3L represents a potent immunomodulatory strategy that appears to be at least equivalent to high-dose PT-Cy for the deletion of alloreactive T cells using our haploidentical transplant model. While PT-Cy has been shown to spare regulatory T cells (4), it is still highly active in systems where CD4⁺ T cells are absent (5), suggesting that the augmentation of recipient DC-induced clonal deletion is highly relevant to the protection seen with this strategy. DCs are specialized cells that are highly efficient at presenting limiting quantities of antigen and their paradoxical ability to decrease alloreactive T-cell responses reflects the nonphysiologic nature of transplantation. Importantly, the stimulatory capability of recipient DCs, including their secretion of IL12, is markedly augmented by TBI (12). Furthermore, unlike pathogen-derived antigen, alloantigen is ubiquitous, present in excess, and persists indefinitely.

It is clear from our data that both PT-Cy and Pre-T Flt-3L have important implications for GVL effects, and their use for GVHD prevention must be balanced against the inevitable impact on GVL activity. Thus, only well-designed and large prospective clinical studies can analyze the relative impact of these comparative GVHD prophylaxis strategies on GVL effects and relapse after BMT. Given that Flt-3L is a myeloid growth factor, its receptor (Flt3) is ubiquitously expressed by AML blasts (44), and that Flt3 mutations portray an adverse prognosis in AML (45), we would not advocate its use in patients undergoing BMT for myeloid malignancies. Instead, we propose that initial trials be undertaken in patients with lymphoid malignancies. In addition, we hypothesize that Pre-T Flt-3L, like PT-Cy, may be most appropriate for use in haploidentical BMT, where HLA mismatches are permissive of strong GVL effects mediated by low numbers of donor T cells escaping deletion. These strategies would also seem particularly suited to nonmalignant conditions, indolent hematopoietic malignancies, or settings where an engineered graft is employed (e.g., CAR or suicide-gene T cells; ref. 46) such that there is little reliance on donor T cells within the graft for antitumor effects. Our data demonstrating maintenance of CMV-specific memory T cells with the Flt-3L approach may mean that this strategy results in lower infection risk in the clinic, when compared with other T-cell depletion approaches that inevitably target both allo-specific and bystander T cells.

No potential conflicts of interest were disclosed.

Conception and design: K.A. Markey, M.A. Degli-Esposti, G.R. Hill

Development of methodology: K.A. Markey, S.W. Lane, G.R. Hill

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K.A. Markey, R.D. Kuns, D.J. Browne, K.H. Gartlan, R.J. Robb, J.P. Martins, S.A. Minnie, M. Cheong, M. Koyama, R.J. Steptoe, G. Belz, M.A. Degli-Esposti

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.A. Markey, K.H. Gartlan, A.S. Henden, M. Koyama, G. Belz, M.A. Degli-Esposti

Writing, review, and/or revision of the manuscript: K.A. Markey, R.D. Kuns, K.H. Gartlan, R.J. Robb, M.J. Smyth, G. Belz, M.A. Degli-Esposti, S.W. Lane, G.R. Hill

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.J. Browne, M.J. Smyth

Study supervision: G.R. Hill

Other (provision of mice): T. Brocker

The authors would like to acknowledge the assistance of Grace Chojnowski, Paula Hall, and Michael Rist. T. Brocker is supported by DFG SFB 1054 B03. The project was supported by grants from the NHMRC, Australia.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.