The Immune Checkpoint Modulator OX40 and Its Ligand OX40L in NK-Cell Immunosurveillance and Acute Myeloid Leukemia

Modulation of immune checkpoints has become a mainstay in oncological treatment. Besides already approved approaches that block inhibitory molecules like CTLA-4 or PD-1, agonistic antibodies that trigger activating receptors on T cells are being developed. One such receptor is OX40 (1–4). This member of the TNF receptor (TNFR) superfamily is upregulated on effector T cells after activation and promotes their differentiation, proliferation, expansion, and long-term survival while inhibiting the suppressive activity of regulatory T cells (5, 6). In patients with cancer, the frequency of tumor-infiltrating OX40⁺ T cells correlates with survival. Moreover, application of OX40 agonists, alone or in combination with other checkpoint modulators, stimulates the cytolytic activity of T cells and causes tumor regression in preclinical models (7–11). First evidence from early clinical trials, most of which are in progress, indicates that OX40 stimulation is also effective in patients with cancer (3, 12).

Besides T cells, which are a component of adaptive immunity, NK cells, as the major cytotoxic lymphocyte subset of the innate immune system, also play an important role in tumor immunosurveillance, particularly in hematologic malignancies. This is supported by observations that NK-cell counts and activity are reduced in patients with leukemia and that activity levels of autologous NK cells are associated with survival of patients with leukemia (13–15). The role of NK cells in acute myeloid leukemia (AML) is highlighted by studies on haploidentical stem cell transplantation (SCT), in which the resulting KIR mismatch seems to be associated with a pronounced graft-versus-leukemia reaction and improved clinical outcome (16). Besides their role in SCT, multiple approaches presently aim to utilize adoptive transfer of allogeneic/KIR-mismatched NK cells for cancer treatment (17).

Beyond KIR, signals mediated by multiple other activating and inhibitory receptors determine whether NK cells do or do not respond to tumor cells. This comprises various members of the TNF/TNFR family which influence NK-cell reactivity upon interaction with their counterparts expressed, e.g., on leukemic cells (18, 19). OX40L is upregulated on NK cells following activation (20), and its counterpart OX40 is expressed by T-cell–derived leukemia cells (21). However, the influence of the OX40/OX40L system on NK function or its role in AML is so far unknown. Here, we show that AML cells express OX40, and exposure to an agonistic OX40 antibody (mAb) promotes leukemia cell proliferation and release of cytokines that influence growth and survival of the malignant cells (22, 23). We found that OX40L is differentially expressed on polyclonal NK cells (pNKC) generated for adoptive transfer depending on the particular protocol utilized, and that OX40L (reverse) signaling alters NK-cell function, including their reactivity to AML cells.

Peripheral blood mononuclear cells (PBMC) and bone marrow (BM) cells of patients and healthy donors were isolated by density gradient centrifugation after informed consent in accordance with the Helsinki protocol. The study was conducted according to the guidelines of the local ethics committee.

Isolation of highly pure (purity >95%) NK cells from pNKC and AML cells from patient PBMC was performed by immunomagnetic separation using the NK-cell isolation kit and negative selection using microbeads CD3, CD14, CD19, and CD56 from Miltenyi Biotec according to the manufacturer's instructions.

pNKC were generated according to standard protocols by incubating nonplastic-adherent PBMC with irradiated RPMI8866 (pNKC-8866) or K562-mb15-41BBL (pNKC-SJ) feeder cells obtained from DSMZ and St. Jude's Children's Research Hospital as previously described (24, 25). Functional experiments were performed when purity of NK cells (CD56⁺CD3⁻) was above 90% as determined by flow cytometry. In addition, K562 cells (received from DSMZ) were transfected using the vector pcDNA3 containing the open reading frame of human 4-1BBL (K562-4-1BBL) or empty vector as control (K562-mock) as described previously (25).

U937 cells (obtained from DSMZ) were transfected using the vector pcDNA3 containing the open-reading frame of human OX40 (U937-OX40) or empty vector as control (mock) and cultured as described previously (25).

The OX40:Fas reporter cells (Jurkat-JOM2) and their use in cytotoxic assays were previously described (26).

RPMI8866, K562-mb15-41BBL, K562,U937, and NK-92 (DSMZ) were obtained in 2005, 2004, 2003, and 2004, respectively. Authenticity was routinely determined by validating the respective immunophenotype described by the provider using flow cytometry after thawing, and cell lines were cultured for a maximum of 2 months prior to use in experiments. Contamination with mycoplasma was excluded by routine testing of all cultures every 3 months.

OX40 mAb BerAct35, OX40L mAb ANC10G1, and mouse IgG1 isotype control were from Ancell Corporation and BD Biosciences, respectively. OX40L mAb lk-1 and 11C3.1 were from BD Biosciences and Biolegend, respectively. All fluorescence conjugates were from BD Biosciences, secondary goat anti–mouse-PE was from Dako (BIOZOL). Fusion proteins consisting of human OX40 with a murine (OX40–Fc) and human (OX40–huFc) Fc-part were from R&D Systems, and Ancell, respectively. RhIL2 was from ImmunoTools.

In addition, antibodies against human OX40 were raised by immunization of C57BL/6 mice by repeated injection of 20 × 10⁶ OX40-transfected CHO cells. Then, spleen cells were fused with SP2/0-Ag14 cells, and hybridoma cells secreting OX40 mAbs were cloned by limiting dilution. mAbs were purified from hybridoma supernatants using Protein A Agarose columns (GE Healthcare). F(ab′)₂ fragments were generated using previously described standard protocols (27).

Analysis of OX40 and OX40L surface expression was performed using specific mAb or isotype control followed by anti–mouse-PE using a BD FACSCanto II.

Leukemic cells in patient samples were selected by FSC/SSC and employing the surface markers CD33/CD34/CD14/CD117 based on the individual immunophenotype defined upon routine diagnosis. Specific fluorescence indices (SFI) were calculated by dividing median fluorescences obtained with specific mAb by median fluorescences obtained with isotype control. Expression was considered positive in case of SFI ≥ 1.5. Intracellular staining was performed using the Fixation/Permeabilization Solution Kit with BD GolgiStop from BD Biosciences according to manufacturer's instructions.

OX40 primers were 5′-TGTAACCTCAGAAGTGGGAGTG-3′ and 5′-GGTCCCTGTCCTCACAGATTG-3′. 18S rRNA primers were 5′-CGGCTACCACATCCAAGGAA-3′ and 5′-GCTGGAATTACCGCGGCT-3′. OX40L primers were 5′-CTGCTCCTGTGCTTCACCTAC-3′ and 5′-TCCAGGGAGGTATTGTCAGTG-3′. GAPDH primers were 5′-AGCCACATCGCTCAGACAC-3′ and 5′-GCCCAATACGACCAAATCC-3′. Reverse transcriptase–polymerase chain reaction (RT-PCR) was performed as described previously (18).

For quantitative PCR, total RNA was isolated using the High Pure RNA Isolation Kit and transcribed into cDNA using qScript XLT cDNA SuperMix (Quanta Biosciences) according to the manufacturer's instructions. Amplification of OX40 cDNA was performed using PerfeCTa SYBR Green FastMix (Quanta Biosciences) on a LightCycler 480 instrument. Primer assays (QuantiTect Primer Assay) for OX40 and 18S ribosomal RNA were used according to the manufacturer's instructions. Relative mRNA expression was calculated by the ΔΔ cycle-threshold (Ct) method.

Cytotoxicity of NK cells against primary leukemia cells and U937-transfectants was determined by ⁵¹chromium release assays after 4 or 24 hours as previously described (28).

Cytokine determination was performed by ELISA according to the manufacturer's instructions using OptEIA sets from BD Pharmingen or DuoSet ELISA development systems from R&D Systems. Metabolic activity was measured using the cell proliferation reagent WST-1 set (Roche) according to the manufacturer's instructions.

As OX40L can be upregulated on NK cells upon activation and ex vivo-preactivated pNKC are being evaluated for cancer treatment (17, 20), we characterized OX40L expression in pNKC generated according to two differing standard protocols [pNKC-8866 (25) and pNKC-SJ (24)]. RT-PCR revealed similar OX40L mRNA levels in both pNKC preparations (Fig. 1A). Next, we used commercially available OX40L mAb to study surface expression using NK-92 cells, which do not express OX40L mRNA (Fig. 1A), as negative control and OX40–Fc fusion protein to ascertain specificity. Comparative FACS analyses revealed that only mAb ANC10G1 can be used to determine OX40L expression. The other two mAbs (but not OX40–Fc) unspecifically bound to NK-92 cells (Fig. 1B). Although resting NK cells of healthy donors were never found positive for surface OX40L and pNKC-8866 displayed only low or no relevant levels, pNKC-SJ displayed substantial and significantly higher OX40L surface expression. This difference between pNKC-8866 and pNKC-SJ cells held true for preparations generated with PBMC of the same or independent donors despite the considerable variation among individual donors (Fig. 1C; P < 0.05, Mann–Whitney test). Expression of OX40L in pNKC-SJ peaked after 4 to 6 days of culture and declined thereafter, but OX40L still was profoundly expressed on day 8 and later, the time these cells usually are used for functional experiments and clinical application by us and others (29). We occasionally observed upregulation of OX40L early during culture of pNKC-8866, but expression was always lower than in pNKC-SJ and OX40L was never detectable after day 9 (Fig. 1D). In line with previous findings (20), activation of NK cells with IL2 alone was not sufficient for induction of OX40L. Upregulation of OX40L expression was dependent on 4-1BB stimulation as revealed by coculture experiments involving K562-41BBL and mock transfectants and transwell settings (Fig. 1E).

In addition to acting as a ligand for OX40, OX40L can, like other TNF family members, itself transduce signals into the ligand-bearing cell (30). To determine whether and how such reverse signaling via OX40L affects NK cells, we cultured pNKC on immobilized OX40–Fc or Fc-control to allow for OX40L crosslinking in the absence of a second, OX40-expressing (target) cell population. FACS analysis of the activation markers CD69 and NKp44 revealed a significant (both P < 0.05; Mann–Whitney test) upregulation on pNKC-SJ following OX40L signaling, whereas the OX40L⁻ pNKC-8866 were not affected (Fig. 2A). In addition, triggering OX40L also led to a significant (P < 0.001; Mann–Whitney test) induction of IFNγ release by OX40L⁺ but not OX40L⁻ pNKC, which again confirmed that signals were mediated via OX40L. The effect of OX40L signaling was observed both in the absence and presence of IL2, the latter serving to mimic an augmented state, which indicates that OX40L signaling may further enhance the activity of activated NK cells (Fig. 2B).

To determine whether OX40L also affects NK lysis, we transfected U937 cells to express high levels of OX40 (U937-OX40) and generated mock-transfectants (U937-mock) as control. When the transfectants were used in cytotoxicity assays, we observed significantly (P < 0.05, Student t test) higher lysis rates for the OX40⁺ targets (Fig. 2C). Next, we disrupted receptor–ligand engagement in this experimental setting to confirm that OX40–OX40L interaction enhances NK lysis of target cells. We used for this purpose a blocking OX40 mAb (instead of using OX40L mAb) to avoid binding to and induction of signaling in NK cells. In addition, to exclude Fc-mediated effects/ADCC after mAb binding to OX40 on target cells, we used F(ab′)₂-fragments. As no OX40 mAb with distinct blocking capacity was commercially available, we generated mouse OX40 mAb as described in the methods section and, after production and definition of specificity, used these in cross-competition experiments using OX40–huFc. Both of our mAb clones M-OX2 and M-OX17 bound to the U937-OX40 transfectants, but only M-OX2 disrupted OX40–OX40L interaction, as revealed by reduced binding of OX40–huFc to OX40L on pNKC-SJ (Fig. 2D). We then produced F(ab′)₂-fragments of M-OX2 according to standard protocols (27) and used these in cytotoxicity assays with pNKC-SJ and U937-transfectants. Although no effect on NK lysis of mock-transfectants was observed, blocking OX40 significantly (P < 0.05, Student t test) decreased the per se higher cytotoxicity observed with the OX40-transfectants, which confirmed the stimulatory effect of OX40–OX40L interaction on NK-cell reactivity (Fig. 2E).

We used FACS analysis to study whether OX40 is expressed on the surface of leukemic cells. We used a total of 111 different AML patient samples and also CD34⁺ progenitor cells obtained from peripheral blood and BM of healthy donors. Leukemic blasts within PBMC were selected as described in the Materials and Methods section. The clinical characteristics of each patient and individual SFI levels are given in Table 1. Although no surface expression was observed on healthy CD34⁺ cells, their malignant counterparts displayed relevant OX40 expression in a substantial proportion of AML cases [SFI ≥ 1.5, n = 60 (54%); SFI ≥ 2.0, n = 41 (37%); Fig. 3A and B]. CD34⁺ cells from patients with chronic myeloid leukemia (CML, n = 10) and myelodysplastic syndrome (MDS, n = 6) showed no relevant OX40 expression (Supplementary Fig. S1). In AML, OX40 expression was significantly associated with the t(15;17) translocation (PML/RARA) and FLT3-ITD mutation (both P < 0.05, Mann–Whitney-U test), whereas no association with other genetic abnormalities, risk according to the ELN classification, FAB classification or disease etiology (i.e., secondary AML from MDS) was observed (Table 2). OX40 expression by AML cells was also confirmed on mRNA level using RT-PCR. Amplicons of OX40 were detected in all 8 investigated samples of patients with at least 80% blast count (Fig. 3C). This also comprised samples of three patients without detectable surface expression on leukemic cells. Very low or no OX40 mRNA was detected in CD34-enriched BM cells of healthy donors (Fig. 3D). Quantitative PCR revealed significantly (P < 0.05, Mann–Whitney test) lower mRNA levels in healthy CD34⁺ BM-cell samples and surface-negative compared with surface-positive AML samples (Fig. 3E). Next, we cultured primary AML cells in the presence of OX40L, G-CSF, GM-CSF, IFNγ, TNF, IL6, IL8, and IL10 and then analyzed OX40 expression by FACS using an (noncompeting) OX40 mAb. We found that TNF significantly induced OX40 expression, whereas other factors had no effect (Fig. 3F).

To elucidate the role of OX40 on AML cells, we characterized our OX40 mAb using reporter cell assays in which Jurkat-JOM2 cells expressing a human OX40–Fas chimeric receptor are killed upon engagement of the OX40 portion of the receptor (26). OX40 mAb clone M-OX17 killed these reporter cells in a dose-dependent manner, indicating that it has agonist activity (Fig. 4A). The same mAb also stimulated a robust IL8 production in U937-OX40 transfectants, but not in mock controls, which further confirmed its stimulatory property (Fig. 4B).

Next, we asked whether OX40 on leukemic cells of patients with AML was functional. We observed no association of constitutive OX40 expression with the basal release of the cytokines TNF, IL10, IL6, and IL8, metabolic activity/proliferation of leukemic cells or apoptosis/death in vitro (Supplementary Fig. S2). However, M-OX17-induced OX40 signaling resulted in a significant (all P < 0.05, Wilcoxon signed rank test) induction of TNF, IL10, IL6, and IL8 by OX40⁺ AML cells (both in cases with de novo and secondary AML from MDS; Fig. 4C; Supplementary Fig. S3A). We observed inter-individual differences concerning the cytokine release of AML cells upon OX40 signaling: None of the 19 investigated samples released all four cytokines. Release of TNF, IL10, IL6, and IL8 was observed with 15, 10, 15, and 3 of the 19 samples, respectively (Fig. 4D). Analysis of intracellular IL8 and TNF levels by FACS and gating for CD33⁺/CD19⁻/CD3⁻ cells served to ascertain that in fact the leukemic cells among patient-PBMC produced the respective cytokines upon OX40 stimulation (Fig. 4E). WST-1 assays then revealed that OX40 signaling can also significantly (P < 0.05, Wilcoxon signed rank test) enhance AML-cell proliferation/viability: in OX40⁺ AML patient samples, 8 (53%) and 5 (33%) out of 15 investigated patients responded to OX40 stimulation with a 1.5-fold and 2-fold increase, respectively (Fig. 4F; Supplementary Fig. S3B).

Finally, we analyzed the outcome of OX40–OX40L interaction for NK-cell antileukemia reactivity. As a first step, we studied the frequency and activation state of NK cells within patients with AML with different OX40 expression levels on leukemic blasts, which, however, did not reveal a correlation (Supplementary Fig. S4). This can be attributed to the fact that NK-cell reactivity is influenced by a multitude of activating and inhibitory receptors as well as cytokines and other immune cell subsets far beyond the OX40/OX40L molecule system (17). Accordingly, we next aimed to delineate the role of this molecule system by employing our blocking OX40–F(ab′)₂-fragments in long-term (24 hours) chromium release assays with OX40L⁺ pNKC-SJ of 9 different healthy donors and OX40⁺ primary AML cells from 8 different patients (at least 80% blast count). Blocking OX40 significantly (P < 0.0001, Wilcoxon signed rank test) decreased lysis of primary AML cells, whereas the isotype control had no relevant effect (Fig. 5A and B). To further exclude an influence of other immune effector cells remaining in AML patient samples or the pNKC preparations, we conducted lysis assays using primary AML samples after MACS-depletion of CD3⁺/CD14⁺/CD19⁺/CD56⁺ cells as targets with purified NK cells obtained by MACS-isolation from bulk pNKC-SJ as effectors. Again, we found that blocking OX40–OX40L interaction reduced AML cell lysis, which confirmed the stimulatory role of this molecule system in AML-NK-cell interaction (Fig. 5C).

The therapeutic inhibition of immune checkpoints to reinforce antitumor immunity of T cells has become a mainstay of cancer treatment. However, many patients benefit only temporarily or not at all from checkpoint blockers that target CTLA-4 and PD-1/PD-L1 (2). Additional strategies are being developed to better exploit the immune system's potential to combat malignant disease. Such approaches include, among others, therapeutic stimulation of activating immune receptors on T cells as well as ex vivo manipulation/expansion and subsequent transfer of cytotoxic lymphocytes such as chimeric antigen receptor (CAR) T cells (31) or pNKC (29).

The TNFR family member OX40 is an activating receptor that can reinforce T-cell antitumor reactivity in the sense of a “stimulatory” immune checkpoint. Preclinical studies with agonistic mAb revealed that the ability of OX40 to stimulate T cells is comparable with CTLA-4 blockade (12). Due to its ability to sustain T-cell proliferation/survival, OX40 is also used in the costimulatory signaling domain of CAR (32, 33). The latter confines the effects of OX40 activation to the transfected T cells, whereas systemic application of agonistic mAb may also affect other cellular components of the immune system and the many non-immune cells that express OX40 (4–6). Here, we used FACS analysis to study 111 samples from patients with AML. We found that leukemic cells from a large proportion of patients (but not CD34⁺ cells of healthy donors or patients with MDS or CML) express OX40 on the cell surface. OX40 mRNA expression was also observed in AML samples without expression of the protein on the cell surface. Although contamination with OX40-expressing healthy cells may have influenced the respective PCR results, we consider it more likely that posttranscriptional or posttranslational mechanisms blocked expression of transcribed OX40 mRNA. In addition, as for other TNF/TNFR members, protein expressed on the cell surface may be shed and released in soluble form. Indeed, soluble OX40 is found in sera of patients with malignant and autoimmune diseases (34–37). Our finding that OX40 expression can be upregulated by exposure of AML cells to TNF indicates that environmental stimuli in the micromilieu may affect OX40 expression via these or other yet unidentified mechanisms.

Functional analyses using newly generated mAb with defined specificity and agonistic properties revealed that OX40 signaling can induce the release of cytokines that act as autocrine/paracrine growth and survival factors in AML and are associated with development and progression of the disease (22, 23). OX40 signaling did not always induce release of the same cytokines. Rather, we found distinct patterns of cytokine release upon OX40 signaling: TNF and IL6 were released in more than 70% of the investigated cases, IL10 released in about 50% of cases, and IL8 released in <20% of cases. In none of the AML samples were all four cytokines released, although all investigated OX40⁺ AML patient samples responded to OX40 signaling by release of at least one cytokine. OX40 may thus (variably) contribute to the cytokine milieu associated with AML. As with T cells, OX40 signaling enhanced viability/metabolic activity of the leukemic cells in a substantial proportion of the AML cases. It seems thus possible that OX40 confers a survival benefit for leukemic cells that is enacted through interaction with OX40L-bearing immune or bystander cells. This is in line with evidence regarding the role of the immune and stromal microenvironment in malignancies in general, which also holds true for AML (38). Moreover, these findings concur with our hypothesis regarding therapeutic application of “untargeted” agonistic OX40 mAb. Other investigators have reported on OX40 expression on cancer cells of various origins other than AML, although without analyzing functionality (39). Another layer of complexity is added by the issue of whether mAb-binding to OX40 affects interaction with cells that express its cognate ligand. OX40–OX40L interaction can lead to transduction of bidirectional signals, this is into the receptor- and the ligand-bearing cell, a characteristic feature of many ligands of the TNF family (30, 40). Besides healthy tissues like endothelial cells, antigen-presenting cells—including B cells and monocytes/dendritic cells—express OX40L, and various cellular functions of these cells are affected by OX40L "reverse signaling" (30, 41–43), that may also occur upon their interaction with OX40-expressing AML cells.

Zingoni and colleagues reported that NK cells, which play an important role in immune surveillance, particularly of leukemia (44), also express OX40L following activation and stimulate OX40-expressing T cells via this receptor–ligand system (20). Despite the multiple approaches presently used to evaluate the clinical efficacy of ex vivo expanded/activated NK cells upon adoptive transfer (17), the expression and function of OX40L by such cell preparations has so far not been analyzed. When we studied OX40L expression on pNKC generated as described in the work of Zingoni and colleagues (; pNKC-8866; ref. 20) or generated according to a protocol used for large-scale clinical grade expansion (pNKC-SJ; refs. 24, 29, 45), we observed no or very low OX40L expression on the first which are generated by culture with RPMI8866 feeder cells in the presence of IL2. In contrast, OX40L was highly expressed on pNKC-SJ that are expanded in the presence of K562 cells transfected with 4-1BBL and membrane-bound IL15 as well as soluble IL2. In line with the findings of Zingoni and colleagues (20), cytokine activation of NK cells alone was not sufficient to induce OX40L expression, and additional expression of 4-1BBL as revealed by analyses involving transfectants and transwell settings was required. In a clinical study in which patients with pediatric solid tumors received IL15/4-1BBL-activated NK cells after allogeneic SCT, acute GVHD was observed despite the fact that the T-cell dose in the grafts was below the threshold usually required for GVHD in this setting (46). Although the underlying mechanism is unknown, we speculate that the ability of NK cells to stimulate T cells via OX40–OX40L interaction may have contributed to the result. This hypothesis is supported by observations in mouse models of GVHD, where OX40 signals in effector T cells play a crucial pathophysiological role (47–49).

When we analyzed how OX40L reverse signaling affected NK cells, we found that OX40L-induced activation and production of IFNγ, a cytokine which participates in cancer elimination by inhibiting cellular proliferation and angiogenesis, promoting apoptosis, and stimulating the adaptive immune system (50). Analyses with OX40 transfectants revealed that OX40–OX40L interaction also enhanced NK lysis of target cells. This was further confirmed in experiments using F(ab′)₂-fragments of a blocking OX40-antibody, which reduced NK-cell killing. Independent experiments (n = 17) with primary OX40-expressing AML cells and OX40L⁺ pNKC of different patients/donors also revealed that disruption of OX40–OX40L interaction reduces target cell lysis despite donor variability. The latter might be due to different OX40 and OX40L expression (or differences in other immunoregulatory molecules that influence NK function) on AML cells and pNKC, respectively. Thus, even though the pathways and mechanisms that mediate OX40L signaling in NK cells remain unclear, our results demonstrate that OX40L enhances the reactivity of NK cells.

Here, we have studied the expression of OX40 in AML and its involvement in disease pathophysiology. In addition, we report on the consequences of reverse signaling via its counterpart OX40L for NK cells including their antileukemia reactivity. We were not able to demonstrate expression of OX40L on NK cells of patients with AML directly ex vivo, possibly because all OX40L antibodies available for FACS analysis compete with OX40 for OX40L binding. NK-cell–expressed OX40L could thus be masked by soluble OX40 (36, 37). OX40L can be released by shedding, which might be enhanced after binding to its AML-expressed counterpart. So far, the reason for our failure to detect OX40L on NK cells in our ex vivo analyses remains unclear. In addition, it remains to be determined whether and how OX40/OX40L interactions affect AML cells or NK cells in an autocrine/paracrine manner, as the latter themselves can express OX40.

The potential consequences of bidirectional signaling through OX40 and OX40L should be considered when using adoptive transfer of ex vivo expanded pNKC. This bidirectional signaling may also affect therapeutic strategies targeted at checkpoint stimulation by agonistic OX40 mAbs, as OX40 can affect multiple cell types beyond T cells including malignant cells. The effects of OX40 agonists may be more complex than anticipated: the treatment may also influence (i) cellular properties of the malignant cells, (ii) their interaction with the microenvironment, (iii) other OX40-expressing immune cells (including NK cells), and, in case of blocking properties of the applied mAb, (iv) reactivity of other OX40L-expressing immune cells, including NK-cell immunosurveillance. Thus, additional work is warranted to unravel the role of the OX40/OX40L molecule system and to exploit the potential of OX40 stimulation for cancer immunotherapy.

No potential conflicts of interest were disclosed.

Conception and design: T. Nuebling, C.E. Schumacher, H.R. Salih

Development of methodology: T. Nuebling, C.E. Schumacher, M. Hofmann, B.J. Schmiedel, P. Schneider

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Nuebling, C.E. Schumacher, M. Hofmann, I. Hagelstein, S. Maurer, B. Federmann, K. Rothfelder, M. Roerden, D. Dörfel

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Nuebling, C.E. Schumacher, I. Hagelstein, B.J. Schmiedel, S. Maurer, K. Rothfelder, M. Roerden, D. Dörfel, H.R. Salih

Writing, review, and/or revision of the manuscript: T. Nuebling, C.E. Schumacher, I. Hagelstein, B.J. Schmiedel, D. Dörfel, P. Schneider, H.R. Salih

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.E. Schumacher, I. Hagelstein, S. Maurer, M. Roerden, D. Dörfel, G. Jung

Study supervision: T. Nuebling, H.R. Salih

This study was supported by grants from the Deutsche Forschungsgemeinschaft (NU341/1-1, SA1360/7-3, SFB685, project A07) and Deutsche Krebshilfe (projects 111828 and 111134). P. Schneider is supported by grants from the Swiss National Science Foundation.

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.