TLR agonists regulate alloresponses and uncover a critical role for donor APCs in allogeneic bone marrow rejection

Cytosine-phosphorothioate-guanine oligodeoxynucleotides (CpG ODNs) are synthetic oligodeoxynucleotides with unmethylated DNA sequences that mimic viral and bacterial DNA and are recognized by Toll-like receptor 9 (TLR9), a pattern-recognition receptor, expressed in certain innate immune cells, including dendritic cell (DC) subsets, macrophages, monocytes, neutrophils, and natural killer (NK) cells, and also in activated CD4⁺ T cells and B cells.^1-18  CpG ODNs are known to trigger innate immune activation, resulting in the expression of costimulatory molecules, resistance to apoptosis, and the secretion of T helper 1 (Th1)–promoting chemokines and cytokines, including macrophage inflammatory protein-1, IFN-inducible protein-10, and other IFN-inducible proteins.^9,18-24  This innate immune activation typically is followed by the generation of potent adaptive immune responses. Immunostimulatory CpG ODNs have been shown to protect against a broad range of viruses, bacteria, intracellular parasites, prions, immune disorders, and tumors in numerous animal and human models of disease (reviewed in Krieg^9,25 ).

Our previous studies showed that CpG ODNs were highly effective in reducing tumor-related mortality in murine recipients of syngeneic bone marrow transplants and acute myeloid leukemia.²⁶  However, although CpG ODNs increased the graft-versus-leukemia effect of delayed lymphocyte infusion (DLI) in allogeneic bone marrow transplantation (BMT) recipients, resulting in enhanced survival, CpG ODNs also increased DLI-induced graft-versus-host disease (GVHD) as indicated by clinical appearance and weight curves.²⁶  The present study was undertaken to directly determine the effects of CpG ODNs on donor antihost alloresponses that can culminate in GVHD and also on host antidonor alloresponses that can culminate in allogeneic BM graft rejection. We found that CpG ODNs, administered at the time of transplantation, accelerated GVHD lethality by TLR9 ligation of host antigen-presenting cells (APCs). In addition, we demonstrated that CpG ODNs promoted allogeneic BM rejection by TLR9 ligation of donor BM APCs. CpG ODN–mediated BM rejection occurred independently of host TLR9 ligation. An important and unexpected result of these studies was the finding that donor BM APCs modulate the strength of the rejection process. These data give new insight into the mechanisms of GVHD and allogeneic BM rejection and suggest that donor BM APCs may be a potential therapeutic target for the enhancement of alloengraftment in BMT.

BALB/c (H2^d) and C57BL/6 (H2^b; termed B6) mice were purchased from The Jackson Laboratory (Bar Harbor, ME) or the National Institutes of Health (NIH; Bethesda, MD). B10.BR (H2^k), B6 IL-12^−/−, B6 IL-6^−/−, B6 IFNγ^−/−, and BALB/c severe combined immune deficient (SCID) mice were purchased from The Jackson Laboratory. B6 Ly5.2 (CD45 allelic) mice were purchased from NIH. B6 MHC class II^−/− mice were purchased from Taconic (Cambridge City, IN). B6 green fluorescent protein (GFP) transgenic (Tg) mice were obtained from J.S.S. TLR9^−/− mice on a BALB/c and B6 background were backcrossed 4 and 6 generations, respectively. N4 BALB/c and N6 B6 TLR9^+/+ mice were used as controls. TLR9^−/− and B6 MyD88^−/− mice were obtained from H.H. and S.A. BALB/c B7–1/B7–2^−/− mice were obtained from A.H.S. B6 IFNγ/IL-12^−/− mice were a kind gift of Dr Zuhair Ballas (University of Iowa, Iowa City). GFP Tg, TLR9^−/−, MyD88^−/−, and B7–1/B7–2^−/− mice were bred at the University of Minnesota. Mice were housed in a specific pathogen–free facility in microisolator cages and were used at 8 to 12 weeks of age. All protocols were approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Minnesota.

Phosphorothioate-modified ODN 2006, used in these studies and previously described, has an ODN sequence of TCGTCGTTTTGTCGTTTTGTCGTT and was provided by the Coley Pharmaceutical Group (Wellesley, MA). CpG ODNs had undetectable endotoxin levels. CpG ODNs were injected at a 100 μg/dose intraperitoneally once weekly for 4 weeks starting at day 0 (d0) for GVHD experiments and on d0 and d7 for engraftment experiments. This dose was determined to be optimal for the induction of antitumor responses.

3M-011 (kindly provided by Dr Anna Masellis, 3M Pharmaceuticals, St Paul, MN), an imidazoquinoline and known TLR7/8 agonist,²⁷  was injected at 1.0 mg/kg subcutaneously every other day starting at d0 for 4 weeks for GVHD experiments and for 2 weeks for engraftment experiments. This dose and schedule were determined to be a potent inducer of antitumor responses in an acute myeloid leukemia murine cell line model (B. J. Weigel and B.R.B., unpublished data, May 2008).

Recipients of the indicated strain and genotype were lethally irradiated with 8.0 Gy total body irradiation (TBI) by x-ray on d−1 and infused intravenously with 10 × 10⁶ T cell–depleted (TCD) donor BM cells and whole splenocytes from the indicated allogeneic donor strain at the indicated cell number on d0. Where indicated, B6 recipients were depleted of NK cells with anti-NK1.1 mAb (clone PK136; National Cell Culture Center, Minneapolis, MN) by intraperitoneal injection of 400 μg on d−1 and d5, a dose and schedule previously determined to be highly depleting for NK cells, while controls received rat IgG (Rockland, Gilbertsville, PA). Nonirradiated BALB/c SCID recipients were depleted of NK cells by injection of anti–ASGM-1 (25 μL on d−4 and d−2 intraperitoneally; Wako, Richmond, VA) and infused intravenously with 10⁶ or 3 × 10⁶ purified T cells obtained from a pool of axillary, inguinal, and mesenteric lymph nodes (LNs) from B6 mice. T cells were purified by negative selection using Miltenyi depletion columns (Miltenyi Biotec, Auburn, CA) and determined to be 98% of the desired phenotype. Mice were monitored daily for survival and weighed twice weekly for the first month, then once weekly thereafter as well as examined for the clinical appearance of GVHD.

Recipients were irradiated on d−1 with a reduced dose of irradiation as indicated. Irradiation dose was chosen to generally result in mixed donor chimerism in most control mice to optimize the likelihood of uncovering a detrimental effect by TLR agonist. On d0, mice received 10 × 10⁶ TCD BM cells of indicated strain by intravenous injection. BM was T-cell depleted to permit the evaluation of engraftment without the complication of GVHD. CD19⁺CD11b⁺ BM cells were purified by incubation with PE-labeled anti-CD19 and anti-CD11b antibodies (eBioscience, San Diego, CA), followed by incubation with anti-PE microbeads (Miltenyi Biotec) and passage over selection columns (Miltenyi Biotec), and were then added to the donor BM inoculum. A subsequent add-back experiment investigating the role of CD19⁺ versus CD11b⁺ BM cells involved the depletion of DX5⁺, CD4⁺, CD8⁺, CD11c⁺, γδ⁺, and either CD11b⁺ or CD19⁺ cells by incubation with PE-labeled antibodies (eBioscience) followed by incubation with anti-PE microbeads (Miltenyi Biotec) and passage over depletion columns (Miltenyi Biotec). This depletion was followed by positive selection of CD19⁺ or CD11b⁺ BM cells by incubation with PE-labeled anti-CD19 or CD11b antibodies (eBioscience), followed by incubation with anti-PE microbeads (Miltenyi Biotec) and passage over selection columns. Cells were determined to be more than 99% of the desired phenotype by flow cytometry. A dosage of 5 × 10⁶ of the positively selected BM subset was chosen for the add-back experiments because this is the approximate absolute number of CD11b⁺ or CD19⁺ cells contained in a 10 × 10⁶ TCD BM inoculum, since BALB/c BM is 47% (± 5%) CD11b⁺ and 44% (± 7%) CD19⁺ (average ± 1 SD; n = 7). In some experiments, packed red blood cell volumes (PCVs) were monitored on d14 as a measure of BM aplasia as previously described.²⁸  Donor chimerism was evaluated by peripheral blood leukocyte (PBL) phenotyping at about d35. PBLs were stained with fluorochrome-conjugated antibodies (class I or Ly allele and isotype controls; BD Biosciences, San Diego, CA) and analyzed using CellQuest software on a FACSCalibur flow cytometer (Becton Dickinson, Mountain View, CA). Mice were monitored for survival and weighed as described for GVHD experiments.

B10.BR recipients were lethally irradiated with 8.0 Gy TBI on d−1 and infused intravenously on d0 with 10 × 10⁶ B6 wild-type BM cells and 2 × 10⁶ purified T cells obtained from a pool of axillary, inguinal, and mesenteric LNs from B6 GFP mice. T cells were purified by negative selection using Miltenyi depletion columns (Miltenyi Biotec) and determined to be more than 98% of the desired phenotype. CpG ODN (100 μg) was given intraperitoneally on d0 and d7. To obtain optimal images, mice were killed and dissected for imaging, but no tissue processing was required. A total of 3 to 4 mice per group were examined at d14. Images were taken with a Retiga Exi color camera and QCapture software (Qimaging, Burnaby, BC) mounted onto a Leica MZFLIII stereomicroscope using a GFP2 or a GFP/dsRED-bandpass filter and a 1.0× transfer lens (Leica Microsystems, Bannockburn, IL). Zoom factors from 3.5× to 10× were used for imaging (3.5× for inguinal LN, ileum, peyer patch, and skin; 7.0× for liver, spleen and kidney; and 10.0× for lung). Exposure times were optimized for each organ, and identical times and settings were used for all mice. Mice within a group yielded very similar results at each time point, so a representative image is illustrated.

Survival data were analyzed by life-table methods; actuarial survival rates are shown. Group comparisons were made by log-rank test statistics. To assess chimerism data, group comparisons of percentage donor chimerism were analyzed by the Student t test. Engraftment and survival rates were analyzed by chi-square test. P values less than or equal to .05 were considered significant in all tests.

Numerous studies indicate that immunostimulatory CpG ODNs confer potent anti-infectious and antitumor properties (reviewed in Krieg^9,25 ). Although patients who undergo allogeneic BMT are at significant risk of infectious complications and tumor relapse, and therefore might be expected to benefit from CpG ODNs, we hypothesized that CpG ODNs might increase GVHD via their stimulatory effects on the innate and adaptive immune systems. To study the effect of CpG ODNs in the context of allogeneic BMT, lethally irradiated B10.BR mice were infused with MHC-disparate C57BL/6 (B6) BM and splenocytes, and CpG ODNs were injected weekly beginning on the day of BMT (d0) and continuing for a total of 4 weeks (Figure 1A). CpG ODN reduced the median survival time (MST) from 94 days to 40 days in recipients of 5 × 10⁶ splenocytes (P = .036) and 28.3 days to 16 days in recipients of 25 × 10⁶ splenocytes (P = .015; Figure 1A).

To determine whether CpG ODN–induced GVHD acceleration was strain dependent, CpG ODNs were evaluated in a second strain combination. CpG ODNs accelerated GVHD and increased mortality in lethally irradiated B6 mice receiving fully allogeneic BALB/c BM and splenocytes (Figure 1B). CpG ODNs increased mortality from 29% to 83% in recipients of 5 × 10⁶ splenocytes (P = .004) and decreased MST from 46 days to 23 days in recipients of 15 × 10⁶ splenocytes (P = .009; Figure 1B).

Lethal irradiation creates a proinflammatory milieu that is known to contribute to optimal GVHD generation and lethality.²⁹  CpG ODN–mediated acceleration of GVHD was not dependent upon irradiation because CpG ODNs accelerated mortality in nonirradiated BALB/c SCID recipients of B6 T cells (Figure 1C; P < .05 for both cell doses).

Collectively, these data indicate that CpG ODNs administered early after BMT markedly accelerate GVHD lethality in the presence or absence of radiation-induced injury and in several MHC-disparate donor-recipient strain combinations.

To determine if CpG ODNs had differential effects on effector T-cell expansion and homing, GFP imaging studies were performed. Purified B6 GFP T cells were infused with B6 wild-type BM into lethally irradiated B10.BR mice, and CpG ODNs were given on d0 and d7. Consistent with survival studies indicating acceleration of GVHD mortality, CpG ODNs increased donor T-cell number in lymphoid and nonlymphoid parenchymal and epithelial GVHD target organs 14 days after BMT (Figure 2). CpG ODNs significantly increased effector T-cell number in inguinal and other peripheral LNs, spleen, Peyer patch, and gut-associated lymphoid tissue of the ileum and colon (Figure 2A,B; data not shown). CpG ODNs also increased GFP effector number in parenchymal organs, including liver, lung, and kidney (Figure 2C; data not shown). Donor T-cell number was also dramatically increased in the skin and oral mucosal tissue of CpG ODN–treated mice (Figure 2D).

In the irradiated GVHD models, APCs (host or donor origin) or alloactivated donor CD4⁺ T cells were the most likely target cell candidates for GVHD acceleration after CpG ODN administration. However, donor APCs were not infused in the nonirradiated SCID GVHD model, thereby reducing the putative target cell candidates to host APCs or donor T cells. To determine whether TLR9 signaling of host or donor cells was essential for CpG ODN–mediated GVHD acceleration, studies were performed with TLR9^−/− recipient or donor mice. CpG ODNs did not accelerate GVHD in TLR9^−/− recipients of wild-type BM and splenocytes but did reduce the MST of wild-type recipients given donor BM and TLR9^−/− splenocytes from 50 days to 19 days, indicating that CpG ODN accelerated GVHD by TLR9 ligation of host APCs (P = .012; Figure 3A,B). Because the immune effects of TLR9 (and TLR7) are mediated through the adapter protein MyD88,^30,31  complementary studies were performed examining the effect of CpG ODNs in MyD88^−/− recipients. As expected, and similar to results with TLR9^−/− recipients, CpG ODNs did not accelerate GVHD in MyD88^−/− mice (P = .432; Figure 3C). Of special note, TLR9^−/− and MyD88^−/− mice were not resistant to lethal GVHD.

Previously, we reported that splenocytes exposed to CpG ODN in vitro released proinflammatory cytokines, including IL-6, IL-12p70, and IFNγ.²⁶  To determine whether CpG ODN–induced GVHD was dependent upon specific host proinflammatory cytokine production, a series of experiments was performed using B6 allogeneic recipients deficient in proinflammatory cytokine production. Whereas B6 IL-6^−/− or B6 IL12p40^−/− BM transplant recipients given CpG ODN had a marked acceleration in GVHD lethality (P ≤ .003; Figure 4A,B), mice deficient in IFNγ alone or in conjunction with IL-12p40 did not have a significant acceleration in GVHD mortality (P > .21, control vs CpG; Figure 4C,D). Although CpG ODNs stimulate NK cells that are major producers of IFNγ, host NK production of IFNγ was not required for CpG ODN–mediated GVHD acceleration because CpG ODNs accelerated GVHD mortality in NK-depleted wild-type recipients (P = .041; Figure 4E). At a lower spleen dose, CpG ODNs significantly increased the mortality rate in NK-depleted mice from 0% to 63% (P = .005; Figure 4F). Collectively, these data indicate that CpG ODNs accelerate GVHD and increase GVHD mortality by TLR9 ligation of host APCs in a host IFNγ-dependent, NK cell–independent fashion.

In addition to TLR9 agonists, TLR7 and TLR8 agonists have been developed to target a range of diseases, including cancers, infections, asthma, and allergies, and to serve as potent vaccine adjuvants.^32-40  Like TLR9, TLR7 and TLR8 are located in the intracellular endosomal compartment of innate immune cells with highest expression in plasmacytoid DCs, a DC subset that produces large amounts of type I IFN in response to viral infections (reviewed in Kawai and Akira,³⁶  Krieg and Vollmer,³⁷  and Takeda and Akira⁴¹ ). Whereas TLR9 detects unmethylated CpG motifs in bacterial and viral DNA, TLR7/8 detects viral and synthetic single-stranded RNA.^36,37,41  To determine the effect of other TLR agonists on GVHD, 3M-011, a potent TLR7/8 agonist found to be protective against influenza virus in rats,²⁷  was administered to B6 recipients of BALB/c BM and a low splenocyte dose. Similar to TLR9 ligation by CpG ODNs, 3M-011 increased the GVHD mortality rate from 0% to 50% (P = .006; Figure 4G).

CpG ODNs accelerated GVHD via TLR9 signaling of host but not donor cells. We hypothesized that TLR9 signaling of host cells, with the resultant proinflammatory cytokine production, would stimulate allogeneic donor BM graft rejection. To study the effect of CpG ODNs on engraftment, recipients were given a lower dose of irradiation to permit the survival of some host T cells that would be able to reject donor BM in the event that CpG ODNs mediated increased rejection, and TCD BM to avoid the complication of GVHD. CpG ODN significantly decreased the d35 survival rate of B6 mice given BALB/c TCD BM from 88% to 55% (Table 1; P < .003). Significantly lower d14 PCVs in CpG ODN–treated mice indicated deaths were due to accelerated BM rejection resulting in BM aplasia (20.2% vs 39.6% PCV; P < .001, n = 25, pool of 2 experiments; data not shown). Consistent with this hypothesis, CpG ODN decreased the engraftment rate of surviving mice from 86% to 50% and reduced the average donor chimerism from 70% to 31% (Table 1; P ≤ .007).

To determine if CpG ODNs increased rejection mediated by both CD4⁺ and CD8⁺ T-cell subsets, B6 BM was infused into sublethally irradiated class II–disparate bm12 or class I–disparate bm1 recipients in which rejection is mediated by CD4⁺ or CD8⁺ T cells, respectively (Table 1). CpG ODNs reduced donor chimerism in bm12 recipients from 77% to 39% and in bm1 recipients from 91% to 37%, indicating that both host CD4⁺ and CD8⁺ T cell–mediated rejection were increased (P < .001).

Despite the important role of IFNγ on Th1-mediated immune responses and CpG ODN–induced GVHD acceleration, CpG ODNs decreased allogeneic BM engraftment in IFNγ^−/− as well as IL-12p40^−/− and IL-6^−/− mice, indicating that CpG ODN–mediated rejection was independent of host IFNγ, IL-12, or IL-6 (Table 1; P ≤ .05).

Given the critical role of host TLR9 signaling and IFNγ production on GVHD acceleration, we considered 2 possibilities. The explanation we favored was that TLR9 signaling of host cells was required for graft rejection, but that these effects were not mediated via IFNγ, IL-12p40, or IL-6 production. The alternative and seemingly less likely explanation was that TLR9 signaling was mediating rejection via effects on donor BM cells rather than host cells. To determine the role of donor versus host TLR9 signaling requirements of CpG ODN–mediated BM rejection, B6 wild-type or TLR9^−/− mice were given allogeneic BALB/c wild-type or TLR9^−/− TCD BM, and engraftment was assessed (Table 2). Contrary to our expectations, CpG ODNs increased rejection in TLR9^−/− recipients, indicating that signaling of host TLR9 was not required for CpG ODN–mediated rejection (P < .001). Moreover, CpG ODNs did not increase rejection of TLR9^−/− TCD BM by wild-type recipients, demonstrating the critical role of donor TLR9 signaling for CpG ODN–mediated rejection of allogeneic donor BM (P = .869). Together, these data indicated that CpG ODN signaling of TLR9 on a donor BM cell was the critical mechanism by which CpG ODNs increased allogeneic donor BM rejection.

To determine whether CpG ODN–stimulated donor APCs were essential for inducing donor BM graft rejection, studies were performed in sublethally irradiated recipients of BM obtained from allogeneic MHC class II^−/− mice that have a functional defect in donor APCs. Importantly, CpG ODNs did not significantly impair engraftment unless CD19⁺CD11b⁺ BM cells from wild-type mice were added to the MHC class II^−/− donor BM inoculum (Table 3; experiment 1). These data implied that either the wild-type MHC class II⁺ donor BM CD19⁺ or the CD11b⁺ subset or both subsets could serve as the target cell for CpG ODN–mediated BM rejection.

To further investigate which donor APC subset was responsible for CpG ODN–mediated BM rejection, sublethally irradiated B6 mice were given BALB/c TLR9^−/− BM. Cohorts of mice received either highly purified BALB/c wild-type DX5⁻CD4⁻CD8⁻CD11c⁻TCRγδ⁻CD19^-CD11b⁺ or DX5⁻CD4⁻CD8⁻CD11c⁻TCRγδ⁻CD11b⁻CD19⁺ BM cells added to the TLR9^−/− donor BM inoculum. Only the add-back of donor wild-type CD11b⁺ BM cells induced CpG ODN–mediated BM rejection of the TLR9^−/− BM, indicating that donor BM TLR9⁺CD11b⁺ but not TLR9⁺CD19⁺ cells were the target cells of CpG ODNs (Table 3; experiment 2).

As further evidence of the importance of donor APCs in modulating host antidonor rejection, B6 mice were irradiated with 5.0 Gy TBI and infused with 10 × 10⁶ TCD BM from BALB/c wild-type or B7–1/B7–2^−/− mice. CpG ODNs were not administered to any mice in this experiment. None of the 10 mice receiving wild-type BM engrafted. In contrast, 8 of 10 mice receiving B7–1/B7–2^−/− BM had donor chimerism levels ranging from 86% to 98% as measured by PBL phenotyping 3 months after BMT (data not shown; P < .001). Collectively, these data indicate that TLR9 signaling of donor BM CD11b⁺ cells, but not CD19⁺ cells and not TLR9 signaling of host cells, was critical for CpG ODN–mediated BM rejection. These data further identify B7 ligands expressed on donor BM cells as an important factor in host rejection of donor BM.

Since the TLR7/8 agonist 3M-011 increased GVHD, it was also evaluated for its effect on engraftment. B6 mice were irradiated with 6.0 Gy TBI on d−1 and given 10 × 10⁶ TCD BALB/c BM on d0; 3M-011 or drug vehicle was administered every other day for 2 weeks (1.0 mg/kg subcutaneously). Like CpG ODN, 3M-011 promoted BM rejection. Day 14 PCVs were significantly lower in 3M-011–treated mice, indicating accelerated BM rejection resulting in BM aplasia (12.9% vs 38.1% PCV; P < .001, n = 12-15/group; data not shown). Moreover, all 3M-011–treated recipients died by d34 compared with an 85% d60 survival rate in control mice (P < .001; n = 15/group; data not shown). These data indicated that acceleration of graft rejection by TLR ligation could be mediated by either a TLR9 or a TLR7/8 agonist.

CpG ODNs have demonstrated efficacy in numerous animal models of infectious agents and tumors (reviewed in Krieg^9,25 ). We showed that although beneficial for increased antitumor effects in syngeneic BMT, CpG ODNs increased the graft-versus-leukemia (GVL) effect of DLI at the risk of increased GVHD in allogeneic BMT.²⁶  Data presented here directly demonstrate that a TLR9-agonistic CpG ODN administered at the time of allogeneic BMT increased donor antihost and host antidonor T-cell responses, resulting in increased GVHD and increased rates of donor BM rejection, respectively. Similarly, a TLR7/8 agonist also increased GVHD mortality and promoted donor BM rejection, indicating that these findings were not unique to TLR9 stimulation. These data counsel caution in the use of TLR agonists in the early period following allogeneic BMT for the intended purpose of increasing GVL effects and/or promoting immune responses against infectious agents.

CpG ODN–mediated acceleration of GVHD was dependent on TLR9 ligation of host APCs and host IFNγ production, but was independent of host NK cells, host IL-12 or IL-6 production, or TLR9 ligation of donor BM APCs. These data are consistent with reports implicating host APCs as being critical in optimal GVHD generation.^42-47  CpG ODNs given on d0 and d7 or a single injection on d7 were sufficient to accelerate GVHD mortality, the latter potentially due to a second burst of proinflammatory cytokine production, resulting in sustained expansion of effector T cells (data not shown). The effect of CpG ODNs on driving effector T-cell expansion was supported by d14 GFP⁺ effector imaging indicating markedly increased donor T-cell number in lymphoid and nonlymphoid organs. Moreover, the GVHD acceleration was potent since CpG ODNs significantly increased the d14 histologic GVHD scores in colon, liver, skin, and spleen as well as reduced the long-term survival of lethally irradiated mice that received only allogeneic donor BM with no supplemental T cells (data not shown). Although T-cell number in donor BM is low and donor inoculum was depleted of T cells, sufficient numbers of donor BM T cells may have escaped depletion to mediate GVHD tissue destruction sufficient to impair survival of CpG ODN–treated recipients. Alternatively, the release of proinflammatory cytokines induced by CpG ODNs may have resulted in tissue destruction independent of the donor T cells.⁴⁵  Although CpG ODNs significantly impaired engraftment of mice conditioned with lower doses of irradiation, PBL phenotyping revealed that the reduced long-term survival of the lethally irradiated mice given allogeneic TCD donor BM was not due to poor engraftment (data not shown).

The results implicating host APCs as a critical target for CpG ODN–mediated GVHD acceleration were not so surprising given the data incriminating host APCs as the major target of GVHD effector T cells.^42-47  However, the results implicating donor APCs as the target for CpG ODN–mediated BM graft rejection were unexpected. Because CpG ODN is referred to as the universal vaccine due to the potent adjuvant-like, nonspecific immune stimulus delivered to the recipient's innate immune system,^{13,14,18,48-52}  we predicted that CpG ODN signaling of host TLR9 would be sufficient to increase rejection. We hypothesized that CpG ODN would ligate host TLR9, resulting in the increased production of the proinflammatory cytokines IL-12, IFNγ, TNFα, and IL-6 that would amplify a host antidonor T-cell response sufficient to increase BM rejection regardless of the precise target cell or whether T cell–mediated rejection was the result of direct or indirect alloantigen recognition. Contrary to our expectations, CpG ODN promoted rejection in TLR9^−/− recipients and even more surprising, CpG ODN did not promote rejection in wild-type recipients given TLR9^−/− donor BM. Nor did CpG ODN significant impair engraftment of MHC class II^−/− BM unless wild-type CD19⁺CD11b⁺ cells were added to the donor BM inoculum. These data implicated a donor APC as the target cell of CpG ODN–mediated rejection, although the precise APC subset was not identified. As TLR9 is expressed in many BM cell types, including myeloid cells and B cells, we hypothesized that either the CD19⁺ or CD11b⁺ BM subset could serve as the target cell for CpG ODN–mediated donor BM rejection. Contrary to expectations, only wild-type CD11b⁺ donor BM cells added to the TLR9^−/− donor BM restored CpG ODN–mediated BM rejection. Imaging data indicate that donor BM homes to lymphoid organs very early after intravenous infusion.²⁸  CpG ODN stimulation of TLR9⁺CD11b⁺ myeloid donor BM cells may result in an early burst of proinflammatory cytokines that amplifies and accelerates the host antidonor immune response locally in the microenvironment of secondary lymphoid organs where host T cells directly encounter donor BM. Perhaps it is the absence of proinflammatory cytokine production that makes wild-type TLR9⁺CD19⁺ donor BM cells unable to restore CpG ODN–mediated BM rejection. Alternatively, the array of costimulatory molecules induced and up-regulated on donor myeloid cells by TLR agonists may provide a more potent signal to rejecting host T cells compared with that induced and up-regulated on donor B-lineage cells.

Although these studies were undertaken to discover the mechanism behind CpG ODN–mediated graft rejection, the data identified an important and heretofore overlooked role for donor BM APCs in modulating the strength of the rejection process. The finding that costimulation-deficient B7–1/B7–2^−/− allogeneic BM was far less likely than wild-type BM to be rejected further underscores the critical role of donor BM APCs and in particular APC B7 interactions in the BM rejection process. A critical role for donor APC B7 expression has also been demonstrated for the elicitation of resistance following transplantation of MHC-matched, minor antigen–disparate BM in reduced intensity–conditioned recipients (R.B.L., manuscript in preparation). It remains to be determined what other cell-surface molecules on donor BM cells play a role in BM rejection.

Finally, although not directly addressed, these data, implicating donor APCs as the cellular target of rejection, suggest that BM rejection is more likely to be the result of direct host T-cell recognition of alloantigen rather than indirect presentation of donor alloantigen by host APCs to host T cells.

Clinical intervention for the enhancement of BM engraftment has focused primarily on reducing the number and function of host T and NK cells. Our data suggest that donor BM APCs merit further investigation as potential therapeutic targets for the promotion of engraftment in BMT.

This work was supported by NIH grants R01 AI034495, R37 HL56067, R01 HL63452 (B.R.B.) and P01 AI056299 (B.R.B. and A.H.S.).

National Institutes of Health

Contribution: P.A.T. designed and executed experiments, analyzed data, and wrote the paper; M.J.E. and C.J.L. executed experiments; A.P.-M. interpreted and scored histopathologic sections; A.M.K., A.H.S., W.J.M., J.S.S., H.H., S.A., and R.B.L. contributed unique reagents and insights; and B.R.B. designed experiments, analyzed data, and edited the paper.

Conflict-of-interest disclosure: A.M.K. is an employee of Pfizer. The other authors declare no competing financial interests.

Correspondence: Bruce Blazar, MMC 109, University of Minnesota, 420 Delaware Street SE, Minneapolis, MN 55455; e-mail: blaza001@umn.edu.