Glucocorticoid-Induced Tumor Necrosis Factor Receptor–Related Protein Ligand Subverts Immunosurveillance of Acute Myeloid Leukemia in Humans

The development of clinically apparent malignancy after cell intrinsic oncogenic events is largely dependent on the interaction of the transformed cells with the immune system. This reciprocal process substantially influences whether tumor cells are eliminated or progress to a life-threatening malignancy, and induction of tolerance of innate and adaptive immune effector cells seems to be a required factor in tumorigenesis (1, 2). Among others, many members of the tumor necrosis factor (TNF)/TNF receptor (TNFR) family influence the interaction of tumor cells with the immune system (3). Over the recent years, the role of the TNFR family member glucocorticoid-induced TNFR-related protein (GITR) and its ligand (GITRL) in the regulation of immune responses has received considerable interest. In mice, GITR has been implicated in the development of autoimmune diseases, graft versus host disease, and in the immune response against infectious pathogens (reviewed in refs. 4–6). Furthermore, in mouse tumor models, GITR stimulation has been reported to influence animal survival and even lead to the eradication of tumors (7–11). However, available data suggest that GITR may mediate different effects in mice and men (e.g., ref. 12) necessitating the study of GITR functions specifically in human antitumor immunity. Very recently, we and others showed that GITR is expressed on human natural killer (NK) cells, which play an important role in the immunosurveillance of tumors, and GITR triggering may impair NK cell effector functions (13–16). Clinical evidence for the particularly important role of NK cells in leukemia has been provided by studies of haploidentical stem cell transplantation (SCT), wherein recipient's leukemia cells fail to inhibit donor NK cells via KIR receptors, and KIR disparity has been shown to be associated with powerful graft versus leukemia effects and better clinical outcome (17–19). The observation that leukemia cells may down-regulate HLA class I molecules (20, 21), presumably to escape adaptive immunity, suggests that NK cells may also be involved in controlling leukemia in an autologous setting. This is supported by the observation that NK cell counts and activity are reduced in leukemia patients and that activity levels of autologous NK cells are associated with survival (22–24). Because NK cell reactivity is governed by a balance of multiple inhibitory and activating receptors, NK cell–leukemia cell interaction is dependent on various immunoregulatory molecules far beyond HLA class I–specific inhibitory KIR receptors (25, 26). In this study, we report that acute myeloid leukemia (AML) cells can express GITRL and study the role of GITRL in AML immunoediting and the interaction with NK cells.

Patients. Blood samples from patients with AML were obtained at time of diagnosis before therapy. Peripheral blood mononuclear cells (PBMC) were isolated by density gradient centrifugation. All patients gave their written informed consent in accordance with the Helsinki protocol, and the study was performed according to the guidelines of the local ethics committee. Diagnosis was confirmed by study of bone marrow specimens and flow cytometric immunophenotyping. Clinical data were collected the day when blood samples were taken.

Antibodies, cytokines, and fusion proteins. Anti-GITR (clone 110416), anti-GITRL (clone 109101), anti–interleukin-10 (IL-10; clone 25209) monoclonal antibody (mAb) and isotype control were obtained from R&D Systems. The goat anti-mouse-PE conjugate was from Jackson Immunoresearch. The anti-HLA A,B,C-Pe, anti-CD33-PeCy5, and anti-CD34-PeCy5 conjugates and the respective isotype controls were from BD Biosciences Pharmingen. Human IL-2 and IL-15 were from ImmunoTools GmbH. The extracellular signal-regulated kinase (ERK) inhibitor II (no. 328007) was from Merck KGaA. All other reagents were obtained from Sigma.

The GITR and GITRL fusion proteins (human GITR and GITRL with human IgG1 tail) were generated as previously described (13, 16).

Preparation of NK cells. NK cells were isolated from peripheral blood by negative selection using NK cell isolation kit and MACS columns (Miltenyi Biotech). Alternatively, polyclonal NK cells were generated by incubation of non–plastic-adherent PBMC with irradiated RPMI 8866 feeder cells over 10 d, as previously described (13). Experiments were performed when purity was above 90%, as determined by flow cytometry.

Flow cytometry. Cells were incubated with anti-GITRL mAb or isotype control (all 10 μg/mL) followed by the secondary PE conjugates (1:100) and the anti-HLA class I conjugate or isotype control. Leukemia cells within PBMC of AML patients were selected by staining with anti-CD33-PeCy5 or anti-CD34-PeCy5 conjugates and then analyzed on a FACScan (Becton Dickinson). Specific fluorescence indices (SFI) were calculated by dividing median fluorescences obtained with the respective specific mAb by median fluorescences obtained with the corresponding isotype control.

Detection of soluble GITRL. Levels of soluble GITRL (sGITRL) in sera were determined by ELISA, as previously described, using a polyclonal and a monoclonal GITRL antibody and recombinant GITRL as standard (16).

Reverse transcription–PCR. Reverse transcription–PCR (RT-PCR) for GITRL was performed as described previously (13). GITRL primers were 5-GCTGTGGCTTTTTGCTCA-3 and 5-ACCCCAGTATGTATTATTT-3. Primers for 18S rRNA were 5-CGGCTACCACATCCAAGGAA-3 and 5-GCTGGAATTACCGCGGCT-3.

Cytotoxicity assays. Cytotoxicity of NK cells was analyzed by a 2-h BATDA Europium release assay. Leukemia cells from patients with >80% blast counts were labeled with BATDA (Wallac Oy), washed, and placed in 96-well round-bottomed plates at 5,000 per well before addition of NK cells. After incubation, 20 μL of supernatant per well were removed and mixed with 200 μL DELFIA Europium Solution (Wallac Oy). Cytotoxicity was quantified by measuring the fluorescence of the Europium TDA chelates using a time-resolved fluorometer (VICTOR, Wallac Oy). Alternatively, cytotoxicity of NK cells was analyzed by standard chromium release assays, as previously described (13). Maximum release was determined from target cells lysed in 1% Triton X-100. Percentage of lysis was calculated as follows: 100 × (experimental release − spontaneous release) / (maximum release − spontaneous release).

Determination of IFN-γ, IL-10, and TNF. Cytokine determination was performed by ELISA according to manufacturer's instructions using OptEIA sets from Pharmingen or DuoSet ELISA development systems from R&D Systems.

Patient characteristics. The clinical characteristics of each patient are given in Table 1. There were 39 males and 21 females (male/female ratio, 1:1.9), the median age was 58 years with a range from 18 to 84 years. The median peripheral blood blast count was 78% with a range from 6 to 98%. Bone marrow aspirates were available on all patients and reviewed by an experienced hematologist. Adequate cytogenetic data were available for 49 of 60 patients (82%), of which 27 (55%) had a normal karyotype.

GITRL is expressed by AML cell lines and patient AML cells. As a first step, we determined GITRL and HLA class I surface expression on various AML cell lines by fluorescence-activated cell sorting (FACS). Highly variable levels of HLA class I were observed in the different cell lines. Although NB4 cells displayed no GITRL surface expression, we found substantial levels on the other six cell lines studied (Fig. 1A). Interestingly, we found that all the AML cell lines, including the surface-negative NB4 cells, expressed GITRL mRNA as revealed by RT-PCR analysis, which suggests that GITRL expression may be regulated posttranscriptionally (Fig. 1B). To survey GITRL and HLA class I surface expression on patient AML cells, we performed flow cytometry and selected malignant cells among patient PBMC by staining for CD33 or CD34. The primary AML cells expressed substantial levels of HLA class I. Intensities of GITRL cell surface stainings were, in general, lower compared with HLA class I expression (Table 1). Therefore, we considered leukemia cells positive for GITRL in case of a >1.5-fold increase of median fluorescence above background (SFI >1.5). Of the 60 investigated patients, 34 (57%) expressed substantial surface levels of GITRL whereas no GITRL expression was detected on CD34⁺ cells of six investigated healthy volunteers (Table 1). In Fig. 1C, GITRL and HLA class I surface expression on selected AML cells from patients and CD34⁺ cells of one exemplary healthy donor is shown. To confirm GITRL expression in AML patient cells, we determined GITRL mRNA by RT-PCR using samples of patients with at least 80% peripheral blast count. Analyses of PBMC from five AML patients with substantial GITRL surface expression revealed amplicons in all investigated samples. Among PBMC of three patients that lacked GITRL surface expression on leukemia cells, low amplicons were observed in two cases supporting our notion that GITRL expression may be regulated posttranscriptionally (Fig. 1D). Alternatively, positive results might also be due to GITRL expression on contaminating healthy cells, because, e.g., healthy monocytes and macrophages express substantial levels of GITRL (Supplementary Fig. S1). Taken together, our results indicate that, on the basis of phenotypic analysis, GITRL is expressed on the cell surface of leukemic blasts in a high percentage of human AML cases.

Because we recently showed that GITRL is spontaneously released as soluble form from cancer cells and can be detected in sera of patients with solid tumors (16), we determined whether sGITRL was also present in sera of the AML patients included in our study. From 55 of 60 AML patients, sera obtained at time of diagnosis before therapy were available for analysis by ELISA. Elevated levels of sGITRL above the detection limit of 0.1 ng/mL were detected in 18 of the investigated sera (33%) with a mean of 4.48 ng/mL and a range from 0.12 to 10.9 ng/mL (Table 1). Because sera of healthy donors do not contain detectable levels of sGITRL (16), these results clearly suggest that GITRL is released at significant amounts from AML cells in vivo.

GITRL surface expression is associated with monocytic differentiation of AML. Because GITRL expression patterns differed between individual patients, we studied whether GITRL surface expression was associated with certain AML FAB types (Fig. 2A). Among the investigated patients with FAB M4 and M5, leukemia cells were found to be positive for surface GITRL in 12 of 14 and 9 of 10 cases, respectively. The mean of the GITRL SFI levels was found to be 6.4, with SD of 1.4 in the cases with FAB M4 and 9.1 with 1.6 in the cases with FAB M5. In contrast, only a smaller subset of the patients with other FAB types displayed GITRL expression (M0, 2 of 4; M1, 2 of 8; M2, 7 of 18; M3, 2 of 4). The means with SE within the M0, M1, M2, and M3 patient groups were 2.3 with 2.6, 1.7 with 1.8, 1.5 with 1.2, and 2.0 with 2.6, respectively. No substantial GITRL expression was found in the single investigated patients with FAB M6 and FAB M7. Statistical analysis using one-way ANOVA revealed that GITRL expression was significantly (P < 0.01) associated with monocytic differentiation (FAB M4 and M5). No significant association of GITRL expression patterns with particular cytogenetic abnormalities, blast count, or white blood count was observed (data not shown). Of note, no significant correlation of the levels of sGITRL with certain FAB subtypes, GITRL surface expression, or other patient variables was observed (Fig. 2B), which shows that expression of cell surface and sGITRL do not necessarily correlate.

GITRL expression and release by AML cells diminish NK cell cytotoxicity. Because GITR can modulate NK cell reactivity, we next investigated how AML-expressed GITRL influenced NK cell cytotoxicity. Using the GITRL-expressing AML cell line THP-1 as target, we observed a substantial and statistically significant increase of NK cell lysis (up to 65% increase; E/T ratio, 20:1) by blocking GITR-GITRL interaction using a GITR-specific mAb whereas isotype control had no effect. In line, NK cell cytotoxicity was also significantly increased (up to 99% increase; E/T ratio, 20:1) after blocking GITR in assays with GITRL-expressing primary AML cells from patients with different AML FAB subtypes (Fig. 3A). Of note, THP-1 cells do not express GITR whereas NK cells do not express GITRL (Supplementary Fig. S1). Furthermore, the GITR mAb had no effect in cocultures of GITRL-negative AML and NK cells or GITRL-positive AML cells and GITR-negative NK92 cells (Fig. 3B). Together with available data on the inhibitory role of GITR for NK cells (13, 15, 16), this indicates that cytotoxicity of NK cells was, in fact, reduced due to triggering GITR by AML-expressed GITRL. To ascertain further that forward signaling via GITR into NK cells mediated inhibition of NK cell cytotoxicity, we cultured NK cells for 24 hours alone or in the presence of the activating cytokines IL-2 and IL-15. Subsequently, we performed cytotoxicity assays with K562 cells, which do not express GITR or GITRL (Supplementary Fig. S1) in the presence or absence of immobilized GITRL-Ig, with human IgG1 as control. The immobilization of the GITRL-Ig to plastic enables multimerization of GITR in the absence of cell surface–expressed GITRL (13), which excludes effects due to transduction of signals into GITRL-expressing target cells. Presence of immobilized GITRL-Ig significantly (all P < 0.05, Mann-Whitney U test) reduced the lysis of K562 cells, whereas presence of IgG1 had no substantial effect (Fig. 3C). Of note, the effect of GITR triggering by GITRL-Ig caused only slightly more pronounced effects in cytokine-activated NK cells, which express higher levels of GITR compared with resting NK cells (13). A potential explanation might be that increased inhibitory signaling via GITR in activated NK cells is balanced by up-regulated expression of activating NK cell receptors, like NKp44 or NKG2D, or the stimulatory effect of the cytokine treatment itself.

Because sGITRL derived from carcinoma cells impairs NK cell reactivity (16), we evaluated whether this also occurred with sGITRL-containing AML patient sera. NK cells were cocultured with the GITR-GITRL double-negative K562 cells in the absence or presence of sGITRL-containing serum. Presence of the serum significantly (P < 0.05, Mann-Whitney U test) diminished NK cell cytotoxicity. Neutralization of sGITRL in patient sera with GITR-Ig before addition to the cocultures significantly increased NK cell cytotoxicity (P < 0.05, Mann-Whitney U test), whereas isotype control had no effect (Fig. 3D). This extends our findings regarding the NK cell suppressive effect of sGITRL derived from solid tumors to AML and also confirms the inhibitory effect of GITR on NK cell cytotoxicity.

AML-derived GITRL inhibits IFN-γ production of NK cells. To investigate whether GITRL on AML cells also diminished NK cell IFN-γ production, we cultured NK cells for 24 hours in the absence or presence of the GITRL-expressing AML cell line THP-1 or GITRL-positive primary AML cells of patients with >80% peripheral blast count. Where indicated, GITR mAb or isotype control was added, and culture supernatants were subsequently analyzed by ELISA (Fig. 4A). In the absence of AML cells, untreated NK cells produced little IFN-γ, and addition of the GITR mAb or isotype control did not alter IFN-γ production. Presence of AML cells significantly increased the detectable levels of IFN-γ. In all assays with GITRL-expressing AML cells, IFN-γ production was significantly (all P < 0.05, Mann-Whitney U test) increased by addition of anti-GITR mAb whereas isotype control had no effect. In contrast, NK cell cytokine production was not altered by anti-GITR in cocultures of GITRL-negative AML and NK cells or GITRL-positive AML cells and GITR-negative NK92 cells, which confirmed the specificity of the effects of the GITR blockade (Fig. 4B). Again, we also used our approach with immobilized GITRL-Ig to ascertain that the reduction of NK cell IFN-γ production by GITRL was, at least in part, due to signals via GITR. NK cells were cultured with or without IL-2 for 24 hours alone or on immobilized GITRL-Ig with IgG1 as control in the presence or absence of the GITR-GITRL double-negative K562 cells. IFN-γ production of NK cells in the presence of K562 cells was significantly (both P < 0.05, Mann-Whitney U test) reduced by immobilized GITRL-Ig, whereas isotype control had no effect (Fig. 4C). In line with the results regarding cytotoxicity, a slightly more pronounced but not markedly differing effect of GITR triggering was observed in cytokine-activated NK cells compared with resting NK cells. Of note, IL-2 stimulation in the absence of target cells was sufficient to induce production of low levels of IFN-γ, and this was already significantly (P < 0.05, Mann-Whitney U test) reduced by GITR triggering due to the presence of immobilized GITRL-Ig whereas only a weak effect of GITRL-Ig on IFN-γ production of resting NK cells was observed.

Furthermore, we determined whether NK cell IFN-γ production was also diminished by sGITRL contained in AML patient sera. Again, we used our approach of adding GITR-Ig or isotype control to the serum before addition to cocultures of NK cells and K562 leukemia cells. As observed for cytotoxicity, presence of sGITRL-containing serum significantly reduced NK cell cytokine production. The presence of the GITR-Ig fusion protein significantly neutralized the inhibitory effect of sGITRL (P < 0.05, Mann-Whitney U test), whereas addition of isotype control had no effect (Fig. 4D). Thus, AML-derived sGITRL and cell surface–expressed GITRL, in addition to reducing cellular cytotoxicity, also diminish IFN-γ production of NK cells.

GITRL signaling stimulates the release of immunomodulatory cytokines by AML cells. Because GITRL itself can transduce signals in various healthy cell types and carcinoma cell lines (e.g., refs. 13, 27), we determined whether this also occurred in GITRL-expressing primary AML cells. We stimulated GITRL on primary AML cells by immobilized GITR-Ig with human IgG1 as control for the indicated times and analyzed the culture supernatants by ELISA. GITRL signaling led to a significant (both P < 0.05, Wilcoxon-signed rank test) induction of IL-10 and TNF by AML cells, whereas isotype control had no effect. Of note, enhanced cytokine production after GITRL stimulation was not only observed with AML cells of monocytic differentiation but also with cells of undifferentiated FAB types (Fig. 5A and B). Secretion of IL-10 peaked after 9 hours of GITRL stimulation, whereas maximal TNF levels were already observed after 3 hours (Fig. 5B). No effect of GITR-Ig was observed when cells of GITRL-negative AML patients were used, confirming that the effects of the GITR-Ig fusion protein were due to signaling via GITRL (data not shown). Because GITRL signaling in macrophages is dependent on ERK1/2 activity (27), we next studied whether this was also the case for GITRL signaling in AML. We stimulated GITRL by immobilized GITR-Ig with human IgG1 as control in the presence or absence of a specific inhibitor of ERK1/2 (ERK inhibitor II) and analyzed the culture supernatants by ELISA. ERK inhibition nearly completely abolished the GITRL-induced cytokine production of AML cells, whereas DMSO, as vehicle control, had no effect (Fig. 5C). Next, we wanted to determine whether and how the cytokine production of AML cells affected NK cell reactivity in our experimental setting. TNF is substantially produced by both AML and NK cells. Furthermore, TNF surface expression and release constitute one of the mechanisms by which NK cells exert antitumor reactivity, and NK cell functions are vice versa influenced by reverse signaling via membrane-expressed TNF (28, 29). Therefore, we focused on the analysis of effects of IL-10 and performed cocultures of GITRL-expressing AML and NK cells in the presence or absence of blocking mAb against GITR and IL-10 alone or in combination. Regarding cytotoxicity, no relevant effect of the IL-10 antibody was observed, neither alone nor together with the anti-GITR, which enhanced NK cell cytotoxicity, thus validating that GITR-GITRL interaction occurred in the experimental setting (Fig. 5D,, left). This is in line with our above-described finding that substantial IL-10 release by AML cells requires a longer time than the 2 hours during the cytotoxicity assay, wherein GITR-GITRL interaction can occur. In contrast, IL-10 blockade significantly enhanced NK cell IFN-γ production after 24 hours of coculture with GITRL-expressing AML cells. As described before, the detectable levels of IFN-γ were significantly (P < 0.05, Mann-Whitney U test) enhanced by the presence of GITR mAb. Even higher IFN-γ levels were observed in the presence of anti-IL-10, and the combination with anti-GITR resulted in an additive effect (Fig. 5D , right). These data show that AML-expressed GITRL, in addition to its direct inhibitory effect on NK cell cytotoxicity and cytokine production, may also impair NK cell reactivity due to its capacity to stimulate the release of immunomodulatory cytokines.

Activity of both tumor cells and immune effector cells, including NK cells, is substantially influenced by various members of the TNF/TNFR family (3). The TNFR family member GITR, also known as TNFRSF18, and its ligand have, in humans, independently been identified by two groups in 1999 (30, 31). Two years earlier, GITR was identified in mice (32). GITR is constitutively expressed at low levels on CD4⁺ and CD8⁺ responder T cells and is up-regulated after activation and on regulatory T cells (Treg). In addition, GITR has been detected in B cells, mast cells, granulocytes, monocytes, macrophages, dendritic cells (DC), and, more recently, NK cells (4, 13–15). GITRL is expressed on DC, monocytes, macrophages, B cells, activated T cells, endothelial cells, osteoclasts, and various healthy nonlymphoid tissues (reviewed in ref. 6). We reported recently that GITRL is constitutively expressed and released as soluble form by solid tumors of different histologic origin (13, 16). Here, we show that a high proportion of AML cell lines and primary AML cells from patients, but not healthy CD34⁺ cells, displays GITRL surface expression, which suggests that GITRL may be a marker for a malignant phenotype of hematopoietic precursor cells. Interestingly, presence of GITRL mRNA in leukemia cells did not always translate into detectable GITRL surface levels. A potential explanation for this observation could be a regulatory or mutational blockade of GITRL surface expression. Alternatively, surface expression of many TNF family members, like CD137 ligand (CD137L), which is closely related to GITRL, can be reduced by shedding (e.g., ref. 33). Whereas elevated levels of sGITRL were detectable in sera of about a third of the investigated AML patients, which is in agreement with our recent data on sGITRL in epithelial tumors (16), it is not yet clear whether the soluble form of GITRL is produced by shedding. Our observation that GITRL surface expression, but not the presence of sGITRL, is significantly associated with monocytic differentiation (FAB types M4/5) may be indicative of other posttranscriptional/posttranslational mechanisms regulating GITRL expression and release. Thus, the molecular mechanisms mediating GITRL expression and release require further elucidation. The association of surface expression with monocytic differentiation is in line with our observation that healthy cells of myeloid lineage, like monocytes, and immunostimulatory type I and immunosuppressive type II macrophages express substantial levels of GITRL.

Bidirectional signaling is a characteristic feature of many ligands of the TNF family (34), and GITRL signaling, among others, can alter cytokine production of healthy myeloid cells, but also of carcinoma cells (13, 27, 35). We report that GITRL signaling significantly induced the production of IL-10 and TNF by AML patient blasts. The GITRL-induced cytokine production was abolished by an inhibitor of the mitogen-activated protein kinases (MAPK) ERK1/2, which confirms previous results that GITRL activates MAPK signaling in healthy macrophages and extends these findings to primary malignant cells (27). Whereas IL-10 has been shown to mediate immunosuppression, e.g., by inhibition of T-cell cytotoxicity and DC functions or by favoring the induction and differentiation of Treg (36), TNF is a molecule with potent immunomodulatory properties that largely depend on cellular context and cell type. Besides profound proinflammatory effects, TNF can mediate equally impressive immunosuppressive effects, like induction of lymphopenia or inhibition of T-cell signaling and DC functions (37). Because both forward and reverse signalings via membrane-expressed TNF modulate NK cell functions (28, 29), we focused our functional studies on the role of AML-derived IL-10. In our experimental setting, neutralization of IL-10 enhanced NK cell reactivity, which confirms the functional relevance of GITRL-induced cytokine production by AML cells. In this context, it is of importance that certain TNF family ligands may transduce signals, even in the absence of their cognate counterpart. For example, CD137L can mediate signals that promote cytokine production by macrophages in the absence of the CD137 receptor (38). Our recently reported observations with C1R-GITRL and mock transfectants provide evidence that this may also occur with GITRL (13). By inducing the release of immunomodulatory cytokines, GITRL may, thus, substantially affect the reciprocal interaction of leukemia cells with the immune system and influence whether malignant cells are eliminated or develop into a lethal malignancy (1). This is even more because GITRL also directly diminishes the reactivity of NK cells by mediating inhibitory signals via GITR. Blocking GITR and neutralization of sGITRL in cocultures of leukemia cells with NK cells significantly increased their cytotoxicity and IFN-γ production. IFN-γ plays a crucial role in tumor immunoediting and participates in cancer elimination by inhibiting cellular proliferation and angiogenesis, promoting apoptosis, and stimulating the adaptive immune system and is instrumental for enhancing antigen processing and presentation for optimal recognition of tumor cells by T cells (2).

Our observation that AML-derived GITRL inhibits NK cell cytotoxicity and IFN-γ production is in line with previous finding by us and others that GITR diminishes reactivity and nuclear factor-κB activity of human NK cells (13, 15, 16). The fact that GITRL is expressed on monocytes of healthy donors and AML patients and on different macrophage subsets suggests that, in light of available data on the reciprocal interaction of NK cells with these cell types (e.g., ref. 39, 40), NK cell functions may also be inhibited after interaction with GITRL-expressing healthy cells of myeloid origin.

Whereas being seemingly contradictory to the stimulatory role of GITR described in mouse antitumor immunity (7–11), our findings regarding the inhibitory effect of GITR on human NK cells have recently been confirmed and extended by another group (15). Furthermore, species-dependent differences regarding the effect of GITR have also been reported in T cells. In contrast to the murine system, GITR does not inhibit suppression by human Treg (12). In addition, both increased and reduced proliferations as well as proapoptotic and antiapoptotic effects have been reported in T cells after GITR stimulation due to activation of different TNF receptor–associated factors (TRAF), the proapoptotic protein Siva, and/or the CD95/Fas pathway, suggesting that GITR effects may depend on the biological environment (reviewed in refs. 5, 41). We did not observe consistent results regarding the effect of GITR blockade in cocultures of human allogenic CD3/CD28-stimulated or IL-2–stimulated CD8 T cells with GITRL-expressing patient AML cells (data not shown). More detailed studies using, e.g., antigen-specific T-cell clones, are required to elucidate the consequences of GITR-GITRL interaction for T-cell antileukemia immunity in humans. It will also be interesting to study whether, e.g., association with different TRAF molecules (42) determines whether GITR mediates activating or inhibitory signals. Opposing effects of seemingly analogue receptors in mice and men have also been described with other molecules, like, e.g., the NK cell receptor 2B4/CD244 (reviewed in ref. 25).

Our results underline the necessity to study the role of GITR specifically in human tumor immunology. This is even more because the results obtained in mouse tumor models prompted some authors to suggest the evaluation of therapeutic GITR stimulation, e.g., with agonistic antibodies in tumor patients (reviewed in ref. 4). The use of primary AML cells and allogenic NK cells mimics the situation in patients that undergo allogenic SCT and is thus suitable to study the role of the GITR-GITRL system in NK cell-AML interaction in humans. Our findings may help to define the susceptibility of a given AML to allogenic NK cell responses, which is particularly important in view of the relevance of a sufficient NK cell response for preventing leukemic relapse in patients after haploidentical SCT (18, 19). Despite the fact that GITR triggering can costimulate T cells and, in mice, can induce rejection of tumors, our data implicate that, in the interaction of AML and NK cells in humans, GITRL is a mediator of immunosubversion. Further studies are required to elucidate clearly whether GITR mediates different effects in mice versus men and/or in T cells versus NK cells. This is even more because numerous attempts are made to engraft NK cells in the treatment of leukemia by optimizing autologous or allogeneic NK cell activity using approaches or interventions that prevent NK cell suppression or stimulate NK cell reactivity for immunotherapy of cancer (43).

No potential conflicts of interest were disclosed.

Grant support: Deutsche Krebshilfe grant 10-2004-Sa2, German Research Foundation grants SA 1360/2-2 and SA 1360/4-1, and Wilhelm Sander-Stiftung grant 2007.115.1.

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.