Host-Derived CD70 Suppresses Murine Graft-versus-Host Disease by Limiting Donor T Cell Expansion and Effector Function

Graft-versus-host disease (GVHD) remains a major obstacle to successful allogeneic hematopoietic cell transplantation (allo-HCT). It has been recognized that alloreactive T cells are the culprits behind this adverse side effect (1). T cells are also beneficial after allo-HCT working to facilitate engraftment (2), provide graft-versus-leukemia effect (3), and ward off infectious diseases (4, 5). Therefore, ideal treatments to reduce GVHD do not completely eliminate T cell function. This idea has led to the study of T cell costimulation in GVHD. T cell costimulation is an essential component to T cell activation and constitutes a multitude of receptor–ligand interactions that play unique roles in activation. This provides a target by which T cell responses can be tuned down, instead of turned off.

CD27/CD70 is a costimulatory receptor ligand pair in the TNFR family that is important for CD4⁺ and CD8⁺ T cell function (6–13). CD27 is present on naive T cells and transiently upregulated after activation (7). CD70 expression is more tightly regulated and is expressed by mature APCs (14), intestinal nonhematopoietic APCs (15), thymic medulla (11), and activated T cells (14). For CD8⁺ T cells, CD27 signaling provides a signal that enhances survival (16) and proliferation (17). CD27 is also important for CD4⁺ T cells, providing survival signals for regulatory T cells (T_regs) in the thymus (11) and periphery (13), increasing Th1 development (18), and decreasing Th17 differentiation (10).

The CD27/CD70 costimulatory pathway has been studied in alloimmune and autoimmune responses. Ab blockade of CD70 improved cardiac allograft survival compared with isotype controls (19). In autoimmunity, blockade and/or genetic deletion of CD27 has shown to be capable of decreasing symptoms of inflammatory bowel disease (20) and rheumatoid arthritis (21). These studies emphasize the important role of CD27/CD70 costimulation in T cell–mediated diseases. In addition, CD70-mediated costimulation has also been implicated in immune regulation. In this regard, CD27 signaling can induce Fas-mediated activation-induced cell death (AICD) in T cells encountering high Ag loads (22). Fas–FasL interactions are essential in controlling T cell AICD and subsequent expansion after allo-HCT (23). Furthermore, CD27^−/− mice control solid tumor growth better than their wild-type (WT) controls (13). This study highlighted an important role for CD27 signaling in T_reg survival (13), and it is well established that T_regs play a prominent role in the control of GVHD (24, 25). Together, these results suggest that CD27/CD70 can function to promote as well as regulate T cell responses.

T cell costimulation has been intensely studied in GVHD (26, 27). Previous work has employed blocking Abs to receptor or ligand (28, 29), knockout donor T cells (30), or hosts that are deficient for costimulatory ligands (31, 32). Although CD27/CD70 is known to be important for both CD4⁺ and CD8⁺ T cell responses in other models, the role for this costimulatory interaction has yet to be evaluated in GVHD. CD70 expression is primarily restricted to hematopoietic cells (14), with the exception of a nonhematopoietic APC population in the intestine (15) and thymic epithelial cells (11). Our work focuses on genetic deletion of recipient CD70, providing an environment in which host hematopoietic and nonhematopoietic APCs, which are paramount for initiation of GVHD (33–35), would lack this costimulatory molecule. In this study, we define a unique suppressive role for host-expressed CD70 after allo-HCT. The presence of host-derived CD70 inhibits expansion of donor effector T cells, leading to concordant decreases in GVHD.

Male and female 8- to 16-wk-old C57BL/6 and BALB/c mice were purchased from the National Cancer Institute and Charles River–Frederick. C57BL/6 CD70^−/− mice were kindly provided by J. Ashwell (National Cancer Institute) (36). CD70^−/− mice were bred in-house and maintained in specific pathogen-free conditions. All experiments were conducted in accordance with protocols approved by the animal studies committee at Roswell Park Cancer Institute.

On day −1, WT or CD70^−/− C57BL/6 hosts received 965 cGy from a ¹³⁷Cs source (Mark I; JL Shepherd and Associates) at a rate of 114 cGy/min. C57BL/6 hosts were transplanted at day 0 with BALB/c inoculum as indicated. Mice were weighed twice weekly and considered moribund when body weight reached <80% of initial weight. In the C57BL/6 → BALB/c model, host mice received 792 cGy irradiation on day −1 and were transplanted day 0 with indicated doses of bone marrow (BM) + splenocytes or BM + CD25⁻ PanT. T cell depletion was performed using CD90.2 microbeads and LS columns (Miltenyi), resulting in <5% of original T cell composition of BM. PanT and CD25⁻ PanT cells were isolated using Mouse PanT isolation kit II and LS columns (Miltenyi). PanT Ab mixture was supplemented with biotinylated anti-CD25 for depletion of CD25⁺ cells in CD25⁻ PanT sorts. T cell sorts resulted in >97% purity of desired cell types.

Cells were stained with Abs to H-2K^d (SF1-1.1), H-2K^b (AF6-88.5.5.3), CD3 (145-2C11), CD4 (RM4-5), CD8 (53-6.7), CD25 (PC61), Foxp3 (FJK-16s), IFN-γ (XMG1.2), and fixable Live/Dead Aqua (Invitrogen) or Zombie ultraviolet (BioLegend). For intracellular staining, cells were fixed using the Foxp3/Transcription Factor Staining buffer set (eBioscience). All samples were run on either an LSRFortessa (BD Biosciences) or an LSRII (BD Biosciences). All data were analyzed with FlowJo (Tree Star). For IFN-γ staining, mice were injected i.p. with 250 μg of BFA (Sigma) diluted in PBS 6 h before harvest as previously described (37, 38). All steps before fixation were performed in BFA containing PBS. Active caspase-8 staining was performed using CaspGLOW Fluorescein Active Caspase-8 staining kit (eBioscience).

Single-cell suspensions of sorted PanT cells were resuspended in 5 ml of 37°C PBS. Equal volume of 2 μM CFSE in 37°C PBS was added to PanT suspension and incubated for 10 min at 37°C. After 10 min, 5 ml of 10% FBS containing RPMI 1640 was added and cells were washed. Cells were then washed twice in PBS before injection.

Mice were sacrificed at day 65 after allo-HCT, and liver, large intestine, and small intestine were removed, formalin-fixed, sectioned, and stained with H&E. Intestine tissues were examined using a previously established semiquantitative scoring system (39, 40). Blinded assessments were made for the presence of crypt epithelial cell apoptosis, crypt loss, surface colonocyte vacuolization, surface colonocyte attenuation, lamina propria inflammatory cell infiltrate, mucosal ulceration, luminal sloughing of cellular debris, leukocyte infiltration of the lamina propria, and villous blunting. Liver samples were evaluated using the clinical GVHD score system (41). Assessment for the percentage of pathologic small bile ducts designated 0 as normal, 1 as <25%, 2 as 25–49%, 3 as 50–75%, and 4 as >75%. Representative pictures were captured at 100×. Combined score is the sum of the scores from the liver, small intestine, and large intestine from each individual mouse.

Serum was collected by retro-orbital eye bleed on the indicated days after allo-HCT. Blood was immediately placed on ice until all samples were collected. Once the final sample was collected, all samples were incubated at room temperature for 20 min to allow for clotting. After incubation, vials were centrifuged at 4°C for 10 min at 2000 × g. Serum was removed and vials were then frozen at −80°C. Luminex was performed by the Flow and Image Cytometry, Luminex Division at Roswell Park Cancer Institute as per manufacturer’s instructions.

Chimeras were generated between C57BL/6 WT or CD70^−/− hosts with i.v. injection of 5 × 10⁶ BM plus 5 × 10⁶ splenocytes 1 d after the hosts received irradiation with 965 cGy. After 3 mo, the chimeric hosts were irradiated with 800 cGy and transplanted with 4 × 10⁶ BM alone or combined with 3 × 10⁶ PanT cells isolated from BALB/c donors. Hosts are then monitored for GVHD and survival as previously described (42, 43).

C57BL/6 WT and CD70^−/− BM-derived dendritic cells (BMDCs) were cultured in RPMI 1640 media containing 5% GM-CSF for 7 d. LPS (100 ng/ml) was added to culture on day 6 to fully mature the BMDCs, which were then used as stimulators for MLR. PanT cells as responders were isolated from splenocytes of BALB/c mice. A total of 0.25 × 10⁶ responders and 0.05 × 10⁶ stimulators were then cocultured in a 96-well plate for 4–5 d.

Working under the hypothesis that costimulation is required for full effector function of T cells, we reasoned that either blocking or eliminating CD27/CD70 costimulation would decrease GVHD. To determine whether the CD27/CD70 costimulatory interaction played a role in GVHD, we first used Ab blockade of CD70. Ab blockade of costimulation has been used previously in murine GVHD models (29, 44), and CD70 blockade has been performed in models of autoimmunity and alloimmunity (19, 21, 45). Fig. 1A shows the survival of BALB/c (allogeneic) and C57BL/6 (syngeneic) host mice after transplant of 2 × 10⁶ BM ± 3 × 10⁶ total splenocytes of C57BL/6 origin. Post-HCT administration of a CD70 blocking Ab (FR70) in doses ranged 25–250 μg in separate experiments surprisingly, yet consistently, increased GVHD in allogeneic recipients as evidenced by significantly decreased survival versus IgG-treated controls (Fig. 1A, Supplemental Fig. 1). In contrast, no lethal or otherwise severe GVHD was observed in syngeneic recipients (C57BL/6 → C57BL/6) of equivalent transplants treated with control IgG or anti-CD70 Ab.

Because our Ab blockade could be affecting multiple cell types, we first sought to determine whether CD70 derived from the donor or host was important for suppressing GVHD. To this end, we transplanted mice with WT BM + WT splenocytes or CD70^−/− BM + CD70^−/− splenocytes. In this setting, CD70 is absent from both the donor BM and the splenocytes. We found that CD70^−/− BM + CD70^−/− splenocytes mediated identical GVHD compared with WT controls (Fig. 1B, 1C). Because of the difference in T_reg numbers in the splenic T cell compartment of WT and CD70^−/− mice (11), we examined whether the function of conventional T cells in the absence of T_regs was affected by CD70. We observed no difference in GVHD between recipients of WT or CD70^−/− T_reg-depleted splenic T cells (CD25⁻ PanT) (Fig. 1D, 1E). These data indicate that donor-derived CD70 is not contributing to GVHD.

We next hypothesized that, because of the importance of host APCs in GVHD (33), host-derived CD70 may be responsible for suppressing GVHD. We employed an MHC-disparate HCT system using WT and CD70^−/− C57BL/6 (H-2K^b) host mice transplanted with BALB/c (H-2K^d) BM with or without sorted splenic T cells (PanT). Based on our Ab blockade data, we hypothesized that GVHD would be increased in CD70^−/− hosts compared with WT controls. Indeed, we found that CD70^−/− mice had increased GVHD as evidenced by greater weight loss (Fig. 2A) and increased lethality (Fig. 2B) compared with WT controls. This difference in GVHD was more evident at higher numbers of PanT cells (Fig. 2C, 2D).

Interactions between CD27 and CD70 have been implicated in T_reg survival (13), and it is well established the donor T_regs play a prominent role in the control of GVHD (24, 25). Therefore, we hypothesized that host-derived CD70 could be providing a survival signal to donor-derived T_regs, allowing for better control of GVHD by donor T_regs in WT versus CD70^−/− recipients. To test this hypothesis, we purified CD25⁻ PanT cells and transplanted these in combination with T cell–depleted BM into WT and CD70^−/− recipients. We found that increased GVHD in CD70^−/− recipients was not dependent on donor T_regs, because transplant with T_reg-depleted PanT cells still resulted in increased GVHD in CD70^−/− recipients (Fig. 2E, 2F). Together, these data suggest that host-derived CD70 suppresses GVHD mediated by conventional donor T cells.

To confirm that the observed weight loss and survival differences between WT and CD70^−/− hosts were due to GVHD, we sacrificed mice after allo-HCT and assessed the large intestine, small intestine, and liver. As expected, mice that had more weight loss and death also had increased levels of pathologic GVHD (Fig. 3A). Although some difference in pathology was observed in the large, but not small, intestine (Fig. 3B, 3C), we found a significant increase in damage in the livers of CD70^−/− versus WT hosts (Fig. 3D, 3E). These data confirmed that GVHD was responsible for decreased survival of CD70^−/− hosts after allo-HCT.

After allo-HCT, proinflammatory cytokines play an essential role in mediating tissue damage and providing signals to enable allogeneic T cells to mediate GVHD (46). Because of the differences in GVHD between WT and CD70^−/− hosts, we speculated that proinflammatory cytokines would be increased in CD70^−/− hosts compared with WT controls. We evaluated TNF-α, IL-2, IFN-γ, and IL-17 levels 5 d after allo-HCT. In hosts that received BM alone, cytokine levels were low and no difference was observed between WT and CD70^−/− hosts (Fig. 4). In contrast, CD70^−/− mice that received allogeneic donor PanT cells had significantly higher levels of TNF-α, IL-2, and IFN-γ, and higher levels of IL-17 compared with WT controls.

Because donor T cells are mediators of GVHD, our data suggest that host CD70^−/− may be increasing donor T cell expansion, function, or both after allo-HCT. Because the spleen is a primary site of T cell activation after allo-HCT (47, 48), we evaluated the spleen of WT and CD70^−/− hosts after allo-HCT. In host mice receiving BM only, we observed very few donor-derived T cells and no significant differences in total cell number or the number of donor CD4⁺ and CD8⁺ T cells between WT and CD70^−/− hosts (data not shown). In contrast, in hosts that received PanT cells, we found that the total number of splenocytes, as well as the percentage and number of donor CD4⁺ and CD8⁺ T cells, was significantly increased in CD70^−/− hosts compared with WT hosts from day 5 after allo-HCT (Fig. 5), a time when substantial lethality occurred to CD70^−/− hosts. In fact, absolute numbers of donor CD4⁺ and CD8⁺ were 26-fold and 20-fold higher, respectively, in CD70^−/− hosts at day 5 after allo-HCT. These data indicate that the absence of CD70 provides an environment that is conducive to marked donor T cell expansion after allo-HCT.

Even though the hosts received lethal irradiation conditioning, we consistently observed residual host T cells present after allo-HCT. To test whether CD70 may affect the residual host T cell pool, we first analyzed the number and accumulation of host T cells in WT and CD70^−/− mice. Compared with T cell numbers pre-HCT, both CD4⁺ and CD8⁺ host T cells were reduced to <1% between days 3 and 7 post-HCT, yet there was no significant difference between WT and CD70^−/− hosts either pre- or post-HCT (Fig. 6A). Furthermore, because host NK cells have been shown to mediate antidonor function and inhibit GVHD (49), we examined whether CD70 affects host NK cell number and function. Even though WT and CD70^−/− mice have equivalent NK cell numbers (Fig. 6B), we further performed NK depletion to determine the role of NK cells in this CD70-dependent phenotype. We injected NK1.1 Ab to deplete host NK cells before irradiation and allo-HCT. Depletion of host NK cells significantly expedited GVHD for both WT and CD70^−/− hosts, which presented similar death curves as a result of acute GVHD (Fig. 6C). These data show that the antidonor response mediated by residual host NK cells plays an important role in slowing down GVHD progression. However, this response does not appear to be the dominant mechanism by which host-derived CD70 suppresses GVHD because NK cell depletion in CD70^−/− hosts also significantly expedited GVHD.

Because previous work indicated that CD27 signaling can promote T cell proliferation and survival (8, 17), we sought to determine why donor T cell expansion was increased in CD70^−/− hosts after allo-HCT. We first evaluated the proliferation of donor T cells. Using CFSE-stained donor T cells, we identified day 3 after allo-HCT as a time point when we could observe robust, but not complete, dilution of CFSE (Fig. 7A). At this time point, we found identical CFSE dilution in both CD4⁺ and CD8⁺ donor T cells in WT and CD70^−/− hosts (Fig. 7B), ruling out an increase in T cell proliferation accounting for higher expansion in CD70^−/− hosts.

CD27/CD70 costimulation can promote survival (8) but can also drive T cell AICD in the presence of abundant Ag (22, 50). We hypothesized that host-derived CD70 caused an increase in AICD, thus accounting for decreased T cell expansion in WT hosts. Therefore, we monitored T cell apoptosis 4 d after allo-HCT. Because CD27/CD70 costimulation has been shown to upregulate Fas-mediated AICD (22), we measured active levels of an apoptotic effector downstream of Fas signaling termed caspase-8 (51, 52). We found that 4 d after allo-HCT, donor CD4⁺ and CD8⁺ T cells in WT hosts had more active caspase-8 than donor T cells in CD70^−/− hosts (Fig. 7C). When adding together all donor T cells with activated caspase-8 (active caspase-8⁺ dead⁻ + active caspase-8⁺ dead⁺), we found that donor T cells within WT hosts had significantly more caspase-8–dependent AICD than donor T cells in CD70^−/− hosts (Fig. 7D). Although only a moderate difference (5–10%) is consistently observed, it may be because of the fast turnover of dead cells. These data suggest that host CD70 increases caspase-8–dependent AICD in donor T cells in WT hosts. Therefore, donor T cells in CD70^−/− hosts gain a survival advantage that leads to marked accumulation (Fig. 5).

Because CD70 is expressed by hematopoietic cells including mature APCs, B cells, and T cells (14), as well as a population of nonhematopoietic intestinal APCs (15), we generated BM chimeras to determine whether hematopoietic CD70 or nonhematopoietic CD70 dominantly contributes to the suppression of GVHD. We first performed syngeneic transplants generating BM chimeras that have normal CD70 in the hematopoietic compartment but are deficient for CD70 in nonhematopoietic cells (WT → CD70^−/−) and chimeras in which CD70 deficiency is confined to the hematopoietic compartment (CD70^−/− → WT). After allo-HCT to these chimeras, we observed significantly increased lethal GVHD in hosts that lack CD70 in the hematopoietic compartment (CD70^−/− → WT) compared with hosts that lack CD70 in the nonhematopoietic compartment (WT → CD70^−/−) (Fig. 8A). This result suggests that it is hematopoietic CD70 that makes the dominant contribution to the suppression of GVHD. To further define whether hematopoietic APCs are responsible for this CD70-dependent mechanism, we first compared dendritic cell number and subset composition in WT versus CD70^−/− mice. We found that CD70 deficiency does not affect these parameters (Supplemental Fig. 2). Next we generated BMDCs from C57BL/6 WT and CD70^−/− mice, and used them as stimulators in an MLR to mimic the alloreactive T cell response. Indeed, >40% of mature BMDCs expressed substantial levels of CD70 (Fig. 8B). When BALB/c T cells were used as responders in the MLR, WT BMDCs induced significantly higher levels of caspase-8 activation than CD70^−/− BMDCs (Fig. 8C). These data indicate that CD70 expressed by host hematopoietic APCs stimulates caspase-8–dependent AICD in alloreactive donor T cells.

Our data indicate that T cell expansion is increased in CD70^−/− hosts after allo-HCT (Fig. 5). We wondered whether the decrease in AICD resulted in an increase in effector T cells. To evaluate effector T cells after allo-HCT, we used an in vivo assay to determine the amount of IFN-γ–producing T cells (37). The percentage of donor CD4⁺IFN-γ⁺ and CD8⁺IFN-γ⁺ T cells was significantly increased in CD70^−/− versus WT hosts (Fig. 9). When accounting for the increased T cell numbers in CD70^−/− recipients, the absolute numbers of IFN-γ–producing CD4⁺ and CD8⁺ effector T cells were increased 83-fold and 34-fold, respectively, in CD70^−/− versus WT hosts (Fig. 9B). Together, these results suggest that host-derived CD70 limits the number of donor effector T cells after allo-HCT.

Our study provides evidence showing that, in the absence of host-derived hematopoietic CD70, expansion of donor effector T cells is increased, leading to concordant increases in GVHD. This increase in GVHD was also observed in a second system in which we used a CD70 blocking Ab. Our results are in contrast with the studies that have shown decreased GVHD when hosts are deficient for ligands to other TNF family receptors such as OX40L (44), 4-1BBL (53), and CD30L (54). Therefore, this work suggests that host-derived CD70 plays a unique role by suppressing GVHD.

Similar to the other TNF family receptor interactions, early studies evaluating the role of CD27/CD70 costimulation showed this pairing to be essential for optimal T cell responses (6, 8). However, after further dissecting this pair in multiple disease models, it became evident that it could also serve as a negative regulator of T cell responses. Perhaps most similar to our study is the fact that CD70^−/− mice have increased experimental autoimmune encephalomyelitis compared with WT controls (10). Experimental autoimmune encephalomyelitis is highly dependent on Th17 responses, and CD27 signaling works to dampen Th17 differentiation (10). In line with these data, we found that IL-17 is increased in CD70^−/− hosts (Fig. 4D), which show increased GVHD (Fig. 2). Therefore, the ability of CD27/CD70 costimulation to decrease GVHD may be in part caused by the reduction in Th17 response because Th17 cells have been shown to contribute to GVHD severity (55–57).

Also consistent with the literature is the ability of CD27/CD70 costimulation to drive AICD in T cells (22). AICD is of marked importance in GVHD, because it provides a regulatory checkpoint to prevent T cell expansion (23). Our data suggest that CD27/CD70 costimulation is essential for governing T cell expansion after allo-HCT. Work from others has shown that CD27/CD70 costimulation increases AICD in the presence of abundant Ag (22). It appears that the abundant Ags present in our GVHD model highlight the role of CD27/CD70 costimulation in AICD. Similar to our study, GITR (another member of the TNFR family) stimulation increases AICD of donor CD4⁺ T cells, which results in a decrease in GVHD (58). Both CD27 and GITR have been shown to increase Fas-mediated AICD (22, 58), but have also been shown to interact with the proapoptotic protein Siva-1 (59, 60). Interestingly, both Fas and Siva-1 can increase levels of active caspase-8 (51, 52, 60). Although OX40, CD40, and 4-1BB have similar cytoplasmic tails to CD27 and GITR (59), it appears that GITR and CD27 may be fundamentally different from other members of the TNFR family because of this association with Siva-1. This may help explain the seemingly paradoxical roles of GITR/GITRL and CD27/CD70 costimulation in GVHD.

Previous work from others has shown that CD27/CD70 costimulation promotes IFN-γ production (12, 18, 61). In contrast, we find a decrease in IFN-γ production in the presence of CD27/CD70 costimulation. We believe that the ability of host-derived CD70 to drive AICD may help to explain the increase in IFN-γ–producing T cells in our CD70^−/− hosts. WT APCs may be better at activating T cells than CD70^−/− APCs because of the fact that they can provide CD27/CD70 costimulation. However, in the context of allo-HCT with abundant allogeneic Ags, T cell activation may become overabundant, meaning that the full repertoire of costimulatory molecules may drive T cells past the point of adequate activation to overactivation (i.e., AICD). Similar to viral models when CD27/CD70 costimulation works to eliminate dominant T cell clones (50), CD27/CD70 costimulation may be eliminating highly alloreactive T cells. These may be the same effector T cells that are expressing IFN-γ and other inflammatory cytokines, resulting in the selective loss of IFN-γ⁺ T cells in the presence of CD27/CD70 costimulation. This mechanism therefore spares alloreactive IFN-γ–producing T cells in CD70^−/− hosts, resulting in increased donor T cells and IFN-γ production.

In summary, this study shows a suppressive role for CD27/CD70 costimulation in GVHD, suggesting that targeting this interaction may be beneficial for ameliorating GVHD in the clinic. This work also highlights the importance of studying CD27/CD70 costimulation in different contexts, because this costimulatory pairing has been found to have both stimulatory and inhibitory potential. In addition, this study suggests that Ab blockade of T cell costimulation does not always reduce T cell responses and can have unexpected deleterious effects. In fact, strategies eliminating host-derived CD70 may lead to selection for alloreactive T cells that are pathogenic. Thus, preclinical work must ensure that costimulatory blockade strategies, in particular combined blockade strategies, are truly decreasing T cell responses and are not selecting for highly alloreactive T cells.

We thank Ree Dolnick and Dr. Jason Muhitch for providing helpful advice and technical assistance. We also thank the Roswell Park Cancer Institute departments of laboratory animal resources and flow and image cytometry.

The authors have no financial conflicts of interest.