Intratumor OX40 stimulation inhibits IRF1 expression and IL-10 production by Treg cells while enhancing CD40L expression by effector memory T cells


  • Alessia Burocchi,

    1. Molecular Immunology Unit, Department of Experimental Oncology and Molecular Medicine, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy
    Search for more papers by this author
  • Paola Pittoni,

    1. Molecular Immunology Unit, Department of Experimental Oncology and Molecular Medicine, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy
    Search for more papers by this author
  • Andrea Gorzanelli,

    1. Molecular Immunology Unit, Department of Experimental Oncology and Molecular Medicine, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy
    Search for more papers by this author
  • Mario P. Colombo,

    Corresponding author
    1. Molecular Immunology Unit, Department of Experimental Oncology and Molecular Medicine, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy
    • Molecular Immunology Unit, Department of Experimental Oncology and Molecular Medicine, Fondazione IRCCS Istituto Nazionale Tumori, via Amadeo 42 20133 Milan, Italy Fax: +39-223902630
    Search for more papers by this author
  • Silvia Piconese

    1. Molecular Immunology Unit, Department of Experimental Oncology and Molecular Medicine, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy
    Current affiliation:
    1. Dipartimento di Medicina Interna e Specialità Mediche, “Sapienza” Università di Roma, viale del Policlinico 155 00161 Rome, Italy
    Search for more papers by this author


Treg cells maintain the tumor microenvironment in an immunosuppressive state preventing an effective anti-tumor immune response. A possible strategy to overcome Treg-cell suppression focuses on OX40, a costimulatory molecule expressed constitutively by Treg cells while being induced in activated effector T cells. OX40 stimulation, by the agonist mAb OX86, inhibits Treg-cell suppression and boosts effector T-cell activation. Here we uncover the mechanisms underlying the therapeutic activity of OX86 treatment dissecting its distinct effects on Treg and on effector memory T (Tem) cells, the most abundant CD4+ populations strongly expressing OX40 at the tumor site. In response to OX86, tumor-infiltrating Treg cells produced significantly less interleukin 10 (IL-10), possibly in relation to a decrease in the transcription factor interferon regulatory factor 1 (IRF1). Tem cells responded to OX86 by upregulating surface CD40L expression, providing a licensing signal to DCs. The CD40L/CD40 axis was required for Tem-cell-mediated in vitro DC maturation and in vivo DC migration. Accordingly, OX86 treatment was no longer therapeutic in CD40 KO mice. In conclusion, following OX40 stimulation, blockade of Treg-cell suppression and enhancement of the Tem-cell adjuvant effect both concurred to free DCs from immunosuppression and activate the immune response against the tumor.


The accumulation of Treg cells at the tumor site is one of the mechanisms developed by tumor cells to elude the immune system 1, through suppression of both innate and adaptive immune responses 2. Their inhibition is thought necessary for the establishment of a successful cancer immunotherapy.

Several pieces of evidence indicate OX40 as a potential mediator of Treg-cell inactivation. OX40 is a costimulatory molecule constitutively expressed by Treg cells and expressed upon activation by T effector (Teff) cells. Triggering of OX40 has opposite effects on these two T-cell populations: Treg cells are inhibited in their suppressive functions 3–6, while Teff cells are stimulated to proliferate, survive and gain memory phenotype 7–11. Treatment of different types of mouse transplantable tumors with the mAb OX86, the agonist of OX40, favors tumor rejection thanks to its double effect on Treg and Teff cells 3, 12.

The tumor microenvironment is characterized by an immunosuppressive cytokine milieu, which promotes immune tolerance and tumor growth. Treg cells secrete interleukin 10 (IL-10), which plays a critical role in suppressing immune responses and in particular the maturation of fully competent DCs 13–15. Among tumor-infiltrating Teff cells, the subpopulation of effector memory T (Tem) cells is the most abundant. Tem cells are CD4+CD44highCD62Llow lymphocytes excluded from resting lymph nodes and mainly localized in peripheral tissues; upon stimulation they rapidly activate and secrete effector cytokines like IFN-γ, but they have limited proliferative capacity 16. In experimental models of immune activation, Tem cells constitutively express CD40L at levels sufficient to induce DC activation in an antigen-independent manner 17. The CD40/CD40L axis is crucial for DC maturation and the subsequent T-cell priming. However in the tumor microenvironment this costimulatory pathway is often dampened, thus impairing the generation of an efficient anti-tumor immune response 18, 19.

In this study we have investigated the mechanisms by which OX86 modulates Treg- and Teff-cell functions and their reciprocal interactions with DCs at the tumor site. We propose a model of the tumor microenvironment in which, after OX86 treatment, DCs receive a lower IL-10-mediated inhibition by Treg cells on the one hand, and a stronger stimulation from Tem cells, via the CD40/CD40L axis, on the other. In this favorable condition, DCs acquire a stronger migratory ability toward the draining LNs (dLNs), thus inducing a specific anti-tumor immune response.


Intratumoral OX40 triggering significantly reduces IL-10 secretion by Treg cells

Intratumoral OX40 triggering promotes tumor rejection modulating both Treg- and Teff-cell functions 3, through unknown mechanisms. Here, we separately analyzed the consequences of OX40 triggering on Treg and Teff cells. Treg cells infiltrating the transplantable CT26 colon carcinoma expressed OX40 at higher levels than Treg cells in dLNs (Fig. 1A). We evaluated IL-10 secretion as part of the Treg-cell-suppressive activity directly ex vivo. Low levels of IL-10 were produced by Treg cells in dLNs (Fig. 1B and C), whereas about 40% of tumor-infiltrating Treg cells spontaneously produced IL-10 (Fig. 1D and E). Twenty-four hours after OX86 treatment, IL-10 secretion by tumor-infiltrating Treg cells was significantly decreased (Fig. 1D and E). Similar results were obtained also in mice bearing TSA mammary carcinoma (Supporting Information Fig. 1).

Figure 1.

OX40 triggering reduces IL-10 production by tumor-infiltrating Treg cells. (A) Mean Fluorescence Intensity (MFI) of OX40 (open histogram) on Treg cells from tumor or dLNs of CT26 nodule-bearing mice. OX40-negative staining (grey histogram) was done on tumor-infiltrating Treg cells from OX40-null mice. (B) Representative plots and (C) frequency of IL-10-producing Treg cells in the dLNs 24 h after intratumoral injection of OX86 or rat IgG (mock). (D) Representative plots and (E) frequency of IL-10-producing Treg cells in the tumor 24 h after OX86 or isotype control (mock) treatment. (F) Representative plots and (G) frequency of IL-10 GFP expression by tumor-infiltrating CD4+CD25high T cells obtained from tumors grown in IL-10 GFP BM chimeras, after OX86 or isotype control (mock) treatment. Data are presented as means+SEM of n=5 mice per group and are representative of two experiments. **p<0.01, ***p<0.005, two-tailed Student's t-test.

Some authors have reported tumor-infiltrating CD11b+CD11c+ cells expressing OX40 20, while others did not detect OX40 expression on CD11b+ cells, even if OX86 systemic administration could indirectly reduce their frequency in tumors 21. Tumor-infiltrating macrophages (CD45+CD11b+F4/80+), representing the vast majority of immune infiltration in our tumor model, neither expressed OX40 nor was their IL-10 secretion affected by OX40 stimulation (data not shown).

The decreased IL-10 production by Treg cells upon OX40 engagement was confirmed with a different experimental approach. BM chimeras were generated such as to carry an IL-10-GFP reporter transgene 22 in the hemopoietic lineage. IL-10-GFP expression, evaluated in tumor-infiltrating CD4+CD25high Treg cells, was significantly reduced after intratumoral OX86 injection (Fig. 1F and G). Unfortunately, we could not finely locate IL-10-GFP expression into the Foxp3+-gated Treg-cell subset, since the fixation step required for Foxp3 detection led to GFP loss (data not shown). However, the CD4+CD25high compartment was highly enriched for Foxp3+ cells (data not shown), thus representing a suitable surrogate for Treg-cell analysis.

These data indicate that OX86 can directly antagonize IL-10 secretion, thus blocking, in vivo, a relevant Treg-cell-suppressive function.

OX40 triggering on Treg cells leads to interferon regulatory factor 1 (IRF1) downregulation

Analysis of the transcriptome of naïve Treg cells, sorted from spleens of Foxp3-GFP mice and stimulated in vitro with OX86, showed that nine genes were upregulated and 12 downregulated more than 1.3-fold by OX40 stimulation (Fig. 2A). Among the down-modulated targets, we noticed two probes belonging to interferon regulatory factor 1 (IRF1) mRNA, a transcription factor known to promote IL-10 expression in human cells 23. Hence, we evaluated the effects of OX86 on IRF1 modulation in tumor-infiltrating Treg cells by real-time RT-PCR. As shown in Fig. 2B, IRF1 transcription in tumor-infiltrating Treg cells was about four-fold higher than in splenic Treg cells from tumor-free mice. Intra-tumor OX86 treatment produced a 40% reduction in IRF1 mRNA expressed by tumor-infiltrating Treg cells (Fig. 2B). The expression of IRF1 in the different samples mirrored the different amounts of Treg-cell-derived IL-10 as evaluated by FACS analysis (Fig. 1B–E). These data, together with gene expression data, suggest that the effect of OX40 triggering on IRF1 mRNA expression is Treg-cell-intrinsic and that OX40 stimulation may, directly or indirectly, modulate IRF1 mRNA expression in vivo in tumor-infiltrating Treg cells. Future experiments will test IRF1 downregulation by OX40 at the protein level and will address the molecular cascade linking OX40 engagement to IRF1 repression in Treg cells.

Figure 2.

OX86 modulates IRF1 expression in Treg cells. (A) Mean expression plot representing genes differentially expressed by Treg cells activated with OX86 versus rat IgG. Externalblue lines in the graph represent a fold change (FC) of 1.3. Probes for genes of interest are shown. (B) Foxp3-GFP Treg cells were sorted from tumors according to GFP expression, 24 h after OX86 or rat IgG (mock) injection. Control Treg cells were sorted from spleens of tumor-free Foxp3-GFP mice. IRF1 expression was evaluated by real-time RT-PCR. Data are shown as mean of n=3 pooled spleens or n=6 pooled tumors per group and are representative of two independent experiments. (C) IRF1 BS in IL-10, VCAM-1 and Viperin promoter sequences. Analysis was performed using the web tool TESS.

The binding of IRF1 to IL-10 promoter was previously demonstrated in human cells 23; to confirm this interaction in the mouse system, we performed a computational analysis of the mouse IL-10 promoter with the web tool Transcription Element Search System (TESS). We found a putative IRF1 binding site (BS) of six nucleotides (AAGTGA) between −1470 and −1476 nucleotides. To reinforce this data, we investigated if the same IRF1 BS was in the promoter sequence of two other genes known to be regulated by IRF1: VCAM-1 and Viperin 24, 25. TESS analysis confirmed the presence of the IRF1 BS also in the promoter of these additional target genes (Fig. 2C). Even if additional experiments are needed to confirm IRF1 recognizing and regulating IL-10 promoter in murine Treg cells, our data point to a possible role for IRF1 in sustaining IL-10 expression in tumor-infiltrating Treg cells.

OX86-induced tumor rejection requires the CD40/CD40L axis

To investigate the Teff-cell subpopulation relevant for OX86 anti-tumor effect, we classified CT26 tumor-infiltrating CD4+Foxp3 lymphocytes into four main subsets according to their expression of CD44 and CD62L. We found that in tumor microenvironment the prevalent subset was composed of CD4+Foxp3CD44highCD62Llow Tem cells, conversely they were poorly represented in dLNs (Fig. 3A and B). The increased accumulation of Tem cells in tumor mass was confirmed also in TSA and MCA203 tumor models (Supporting Information Fig. 2). We found Tem cells highly expressing OX40, even if at lower level than Treg cells (Fig. 3C); thus intratumoral OX86 injection could directly target Tem cells at this site. Conversely, Tem cells obtained after immunization of naïve BALB/c mice with two consecutive injections of BM-derived dendritic cells (BMDCs) activated with LPS, as previously described 17, expressed low or absent OX40, even after in vitro activation (Supporting Information Fig. 3). In BMDC-injected animals, Tem cells were shown to constitutively express CD40L at sufficient levels to induce DC activation 17. The CD40/CD40L interaction is crucial for DC activation, survival and proliferation 26. Many data suggest that this axis is involved in inducing protective anti-tumor immune response 18, 19, 27, 28 and that activation of this pathway may represent a strategy for tumor treatment. To investigate whether OX86-induced tumor rejection was dependent on the CD40/CD40L axis, WT and CD40−/− mice were inoculated subcutaneously with CT26 cells and treated intratumorally with OX86. As shown in Fig. 4A, OX86 treatment was ineffective in CD40-deficient mice, while inducing tumor rejection or impaired tumor growth in CD40-sufficient mice. Such failure to induce anti-tumor response in the absence of CD40 could be either due to defective licensing of DC reactivation and migration from the tumor or due to impaired T-cell priming at the dLNs by DCs licensed, but not competent, for optimal T-cell costimulation. To discriminate between these two possibilities, an in vivo DC migration assay was performed. Tumors growing in WT and CD40−/− mice were treated with OX86 or rat IgG co-injected with green fluorescent microbeads. Only DCs that have up-taken the beads at the tumor site can be detected as fluorescent in dLNs. Although OX86 rescued DC migration from tumor to dLNs in WT mice, the same treatment was ineffective in CD40−/− mice (Fig. 4B), a finding implying that in absence of the CD40/CD40L axis tumor-associated DCs cannot be reactivated. We also confirmed that, in CD40−/− mice, the OX40 expression by tumor-infiltrating Tem cells was not defective, indicating that the CD40/CD40L system did not affect the T-cell activation status at the tumor site or Tem-cell responsiveness to OX86 treatment (Supporting Information Fig. 4).

Figure 3.

T-cell subsets in the tumor microenvironment. (A) Frequency and (B) representative plots of Teff-cell subsets, classified according to CD44 versus CD62L in gated CD4+Foxp3 cells from tumors and dLNs. (C) MFI of OX40 on tumor-infiltrating Tem and Treg cells. Filled grey histogram, isotype control. Tcm: central memory T cells; Tn: naïve T cells; Tact: recently activated T cells. Data are presented as mean+SEM of n=6 mice per group and are representative of three independent experiments.

Figure 4.

OX86-induced tumor rejection requires CD40/CD40L axis at the tumor site. (A) WT BALB/c (WT) or CD40−/− mice were inoculated with CT26 cells. Tumor growth was monitored twice per wk and recorded as longest diameter×(shortest diameter)2 (in cubic millimeters). When tumor size was around 3×3 mm, mice were treated intratumorally with 50 μg of anti-OX40 mAb (OX86) or rat IgG (mock). The frequency of mice rejecting (R) the tumor, among treated mice, is indicated in each plot. (B) Tumor-bearing BALB/c (WT) and CD40−/− mice received OX86 or rat IgG (mock) intratumorally combined with green fluorescent microbeads. After 24 h, the absolute count of DCs migrated to the dLNs was estimated by flow cytometry evaluation of FITC+ CD11c-PE-Cy7+ cells. (C) Percentage of tumor-infiltrating Tem cells 24 h after OX86 or rat IgG injection. (D) MFI of CD40L expression on tumor-infiltrating Tem cells 24 h after OX86 or rat IgG injection. Representative histogram plots of CD40L expression on tumor-infiltrating Tem cells are shown. Filled grey line, isotype control; dotted line, Tem cells from mock-treated tumors; black line, Tem cells from OX86-treated tumors. (E) MFI of CD40L on tumor-infiltrating Tact cells 24 h after OX86 or rat IgG injection. Data are presented as means+SEM of n=6 mice per group and are representative of three independent experiments *p<0.05, two-tailed Student's t-test.

OX86 increases CD40L expression on Tem cells in the tumor

We hypothesized that, in the immunosuppressive tumor microenvironment, Tem cells were inhibited in their ability to license DCs via CD40L, and that OX40 triggering might provide the right signal for Tem cells to supply an effective CD40/CD40L mediated co-stimulation. Although the intratumoral injection of OX86 did not change the percentage of Tem cells (Fig. 4C), it significantly upregulated CD40L expression on Tem cells (Fig. 4D). Such CD40L up-modulation was specific for Tem cells, as no other T-cell subsets and especially CD44lowCD62Llow recently activated T (Tact) cells responded accordingly (Fig. 4E). Pd1, a marker of T-cell exhaustion, which is expressed by tumor infiltrating T cells 29, was not modulated by OX40 stimulation (Supporting Information Fig. 5). We explain the lack of tumor rejection and DC migration by OX86 treatment in CD40−/− as a consequence of insufficient CD40L upregulation by Tem cells and therefore insufficient DC reactivation in the tumor microenvironment.

Tem cells directly activate DCs through CD40 engagement

To demonstrate that OX40 stimulation promoted in vivo the direct adjuvanticity of Tem cells toward DCs via CD40/CD40L, Tem cells were sorted from tumors 24 h after treatment with OX86 or rat IgG and were co-cultured with WT or CD40−/− BMDCs. After 24 h, BMDC maturation was estimated by the expression of CD80 and CD86 (Fig. 5A). We found that WT BMDCs received a stronger stimulation by Tem cells pre-treated in vivo with OX86, than with isotype matched control Ab. However, CD40-deficient BMDCs could not increase the expression of maturation markers after co-culture with Tem cells obtained from either OX86 or mock-treated tumors (Fig. 5B and C). We cannot exclude that a reverse CD4/CD40L-mediated interplay may occur between Tem cells and DCs, thus explaining the superior capacity of OX40-triggered Tem cells to costimulate WT DCs. Indeed, OX40-stimulated Tem cells, expressing higher CD40L levels, could be more receptive to CD40-mediated signals provided by WT but not CD40-null DCs, thus in turn boosting WT DCs via signals other than the CD40/CD40L axis, for instance through enhanced cytokine secretion. However, we failed to detect an increased production of IFN-γ, TNF-α, IL-17 or IL-6 ex vivo by tumor-infiltrating lymphocytes (TILs) upon OX86 intratumoral administration (Supporting Information Fig. 6). These data demonstrate that tumor-infiltrating Tem cells, stimulated in vivo with OX86, directly provided the adequate stimuli for DC ex vivo reactivation in a CD40/CD40L-dependent manner.

Figure 5.

Tem cells activate DCs more efficiently after OX86 treatment via CD40L. (A) Experimental scheme. BALB/c mice were inoculated with CT26, and treated with OX86 or rat IgG. After 24 h, Tem cells were sorted from TILs and co-cultured for 24 h with WT or CD40−/− BMDCs at 1:1 ratio. (B) BMDC activation was evaluated as the percentage of the relative increase (compared with non-stimulated BMDCs) of the percentage of CD11c+CD80+ and CD11c+CD86+ cells. (C) Representative plots of CD86 expression by WT or CD40−/− BMDCs in each culture condition. (B and C) Data are presented as means+SEM of n=7 mice per group and are representative of three independent experiments *p<0.05, **p<0.01, two-tailed Student's t-test.


The effects of OX40 triggering on Treg and Teff cells in tumor rejection were separately investigated. In different contexts, Treg cells may adopt preferential suppression mechanisms among a variety of possibilities 2. IL-10 is one of the best-known cytokines endowed with immune-suppressive functions. Il10 gene expression characterizes Treg-cell signature 30, even though a significant IL-10 expression at the protein level can be detected in naïve mice only in the intestine 15, 31. Treg-cell-derived IL-10 is redundant for the control of systemic autoimmunity but becomes crucial for the control of inflammation at the mucosal interfaces with the external environment, such as in lungs and colon 32. In chronic inflammation-related tumorigenesis, Treg cells may turn from anti- to pro-inflammatory and pro-tumorigenic. Indeed, along the development of colon polyposis, Treg cells lose the ability to secrete the anti-inflammatory IL-10 and switch to the pro-inflammatory and pro-tumorigenic IL-17 33. In other tumor models not related to chronic inflammation, such as transplanted cell lines, IL-10 is pro-tumoral, inhibiting DC activation and the generation of an effective anti-tumor response 13. Here we show for the first time, using two experimental approaches, that abundant IL-10 is spontaneously produced by Treg cells in tumors subcutaneously injected in mice. Of note, IL-10 was not detectable anymore after FACS-sorting and culture of Treg cells (data not shown), an observation suggesting that IL-10 induction may be a transient and reversible feature of tumor-infiltrating Treg cells, closely dependent on microenvironmental cues at the tumor site.

IL-10 is a crucial cytokine for immune suppression in tumors. Tumor-associated macrophages constitutively express IL-10 34, thus maintaining an impaired immune status. We and others 35, 36 have reported that IL-10 receptor blockade, when combined with TLR agonists and/or other immunostimulatory agents, rescue the functional paralysis of tumor-infiltrating DCs and macrophages toward an efficient cancer therapy. However, macrophages are not the sole IL-10 source in tumors. Studies in human cancer have shown that Treg cells recruited at tumor sites produce abundant IL-10 37, 38, which may work as the main mediator of Treg-cell functional suppression 37. Conversely, in a murine tumor model, others have shown that CD25+-cell depletion and IL-10 receptor blockade exert distinct, though partially overlapping, effects in suppressing DC activation and anti-tumor CD8+ response 13. Even if a Foxp3-directed, rather than CD25-directed, Treg-cell depletion may provide more reliable results about the functional redundancy of Treg cells and IL-10, it is likely that Treg cells are not the only source of IL-10 at the tumor site 13 and that sole IL-10 receptor blockade cannot recapitulate the efficient anti-tumor activity of combination therapies 35, 36, of the sole OX40 triggering 3, 21 or of Foxp3-targeted Treg-cell depletion, when combined to vaccination 39 or even as single treatment 40.

A link between OX40 stimulation and IL-10 production has been already highlighted in human Tr1 cells 6. OX40L exposure not only prevented the generation of IL-10-producing Tr1 cells from both naïve and memory T cells under different differentiating stimuli, but also repressed IL-10 production and suppressive functions of pre-established Tr1 cells 6. Completely distant regulatory pathways may operate in thymus-derived and tumor-expanded murine Treg cells, expressing Foxp3, as in our system, compared with in vitro generated human Tr1 cells, likely not expressing Foxp3 41. However, OX40 signal may influence conserved pathways regulating IL-10 secretion in divergent lineages. For instance, OX40 engagement inhibits IL-10 production along Th2 differentiation 42 and during anti-viral immune responses 43. Moreover, we show here that OX40 signal may regulate IL-10 secretion through the modulation of IRF1, a Th1-related transcription factor 44.

We found IRF1 expressed in tumor-infiltrating but not peripheral Treg cells producing or not IL-10, respectively. Even if murine IL-10 promoter contains in silico-predicted sites for IRF1 binding, additional experiments are required to confirm the direct IL-10 induction by IRF1 in murine Treg cells, as demonstrated in human cells 23. Whether IRF1 is the major or the sole activator of IL-10 transcription in tumor-infiltrating Treg cells versus other cell populations is unknown. However, we noticed with great interest that Irf1 expression marks the signature of Treg cells obtained from the lamina propria of the intestine, a Treg-cell compartment endowed with a well-known competence for IL-10 production 45. Very little information exists about a role for IRF1 in Treg-cell suppression. The Foxp3 promoter contains IRF1-responsive elements, negatively regulating its transcription 46. However, we could not detect any Foxp3 downregulation in tumor-infiltrating compared with peripheral Treg cells, or in IL-10-producing versus IL-10-negative Treg cells. IRF1 is a transcription factor playing essential roles in Th1 differentiation, inducing IL-12Rβ1 in CD4+ T cells 44. Germane is the expression of IL-12Rβ1 in lamina propria Treg cells 45. The expression by Treg cells of a T helper-specific gene is not surprising. Indeed, recent reports demonstrate that Treg-cell subsets, expressing distinct Th-associated factors, selectively suppress the respective Th classes 47. Treg-cell-specific expression of the Th1 factor T-bet 48, or of miR146a restraining Stat1 activation 49, are required for the optimal suppression of Th1 response. Similarly, IRF1 may represent a Th1-associated factor that, when expressed in Treg cells, dictates a program specifically directed to Th1 suppression, for instance through the IL-10 induction. We are tempted to speculate that IRF1 may represent a transcriptional regulator of the Treg-cell subset functionally oriented toward the suppression of Th1-cell responses in tumors. Through a still unknown signaling pathway, OX40 stimulation may block Treg-cell suppression at the tumor site by directly affecting the IRF1-driven program. Therefore, the effects of OX40 triggering in vivo may differ in peripheral compared with tumor-infiltrating Treg cells, which express different levels of IRF1 and are likely governed by different transcriptional programs. This observation may explain the higher anti-tumor efficacy of the intra-tumor compared with the systemic treatment with OX86 3. More importantly, our data support the notion that distinct Treg-cell subtypes, molecularly and functionally defined, can populate different body districts of healthy individuals as well as pathological tissues such as tumors 50. Future experiments will explore the role of IRF1 in Treg cells' physiological and pathological role and will address whether and how the OX40 signaling pathway affects IRF1 expression at the protein level, thus compromising in Treg cells the IRF-1-driven program.

A current topic is how the cytokine milieu influences Treg cells' response to different stimuli. Recent papers have shown that Treg cells could be reprogrammed into effector cells, produce inflammatory cytokines (IFN-γ, IL-17) 51, 52 and induce anti-tumor CD8+ T-cell activation 53. Interestingly, several pieces of evidence support the idea that the cytokine milieu greatly affects Treg-cell response to OX40 triggering. We have previously shown that OX86 reverses Treg-cell suppression in graft versus host disease (GVHD) 54 and in tumors 3, while others have reported that OX86 administration to naïve mice promotes Treg-cell expansion, thus reinforcing suppression 55. Therefore, the outcome of OX40 stimulation may vary depending on microenvironmental cues. Conversely, OX40 may affect Treg-cell response to cytokine stimulation. Indeed, OX40 signal supports Treg-cell susceptibility to IL-2 by sustaining miR155 expression and restraining SOCS1 availability 56. These data highlight the importance of understanding how different microenvironments influence Treg-cell behavior and how to take advantage of Treg-cell plasticity for the development of efficient cancer immunotherapies.

The strictly Treg-cell-intrinsic modifications detected in the transcriptome of sorted Treg cells, treated or not with OX40 agonist Ab, were relatively few and of limited extent (all modulations were below 1.8-fold). However, according to the above considerations about OX40 tuning cytokine susceptibility, far wider effects may be elicited by OX40 stimulation in Treg cells embedded in a complex microenvironment and exposed to a panoply of signals.

Among downregulated genes, beside Irf1, attention should be paid to Igtp and Iigp2 (also called Irgm2), belonging to p47-GTPase family that, like Irf1, are downstream IFN-γ during the immune responses to pathogens 57. Again, the expression levels of both Igtp and Irgm2 were particularly high in Treg cells derived from lamina propria 45. Other modifications induced in Treg-cell transcriptome by OX40 triggering seemed to affect Treg-cell homing or Treg-cell ability to recruit other cells: Ccr8 and Itgae (encoding for CD103) were increased, Ccl4 and Xcl1 were decreased. A general interpretation of these changes is complex. CD103 is an integrin dictating gut homing, and OX40 is required for Treg-cell accumulation in the colon 58. However, in a model of T-cell transfer-induced colitis, OX40-deficient Treg cells expressed normal levels of CD103 and properly accumulated in the lamina propria 56.

Contrary to Treg cells, effector T cells express OX40 only upon activation 11, 59. We found that Tem cells, representing the most abundant TIL subset, highly expressed OX40. This class of memory lymphocytes was reported to constitutively express CD40L at sufficient levels to induce DC activation 17. We hypothesized that, at the tumor site, the presence of immune-suppressive elements could render the basal CD40/CD40L-mediated interaction insufficient for optimal DC stimulation by Tem cells, and that OX40 triggering may supply to Tem cells the adequate boost. Accordingly, OX40 stimulation significantly upregulated CD40L expression by effector memory lymphocytes, thus licensing DCs for migration to the dLNs 3. We do not know at the moment whether OX40 signaling induces directly or indirectly CD40L upregulation in Tem cells. Along T-cell activation, CD40L expression is induced by TCR ligation, and further enhanced by CD28 costimulation 60. Less clear are the signals sustaining constitutive CD40L expression in memory T cells. Of note, OX40 ligation can assemble a TCR-related signalosome also in the absence of an antigen, providing a sustained level of NF-κB activity necessary for effector memory responses 61. However, CD40L modulation may be also an indirect consequence of OX40 stimulation in Tem cells. For instance, OX40 may induce a complete molecular reprogramming in Tem cells, resulting in an enhanced responsiveness to activatory stimuli or an increased expression of costimulatory molecules and cytokines fostering CD40L expression in an autocrine/paracrine fashion, thus amplifying the initial trigger. We could not detect any change in IFN-γ, TNF-α, IL-17 or IL-6 secretion by Tem cells; however, we cannot exclude that other cytokines or surface molecules may mediate the OX40–CD40L link.

In an experimental model of immune activation, Tem cells licensed DCs in vivo via CD40L when recruited into reactive LNs 17. In that setting, Tem-cell induction and recruitment bypassed the need for any immunization adjuvant 17. Conversely, in our tumor model, Tem cells were abundant at the tumor site but seemed unable to license DCs unless stimulated via OX40. Moreover, Tem-cell adjuvanticity likely occurred at the tumor site, rather than at the dLNs, since OX86 administration increased first of all DC migration from the tumor to the dLNs in a CD40-dependent fashion. Apparently, tumor-infiltrating Tem cells are held in a dysfunctional state, recalling T-cell exhaustion. This condition of poor T-cell responsiveness may be generated by chronic immune stimulation and may also contribute to immune tolerance in cancer 29. In our tumor model, Tem cells highly expressed Pd1, a feature revealing their exhausted phenotype. Even if Pd1 expression was not affected by OX40 stimulation, the CD40L-dependent adjuvanticity was clearly restored in Tem cells. This may suggest that Pd1 blockade might work additively to OX40 triggering toward a full reactivation of tumor-associated Tem cells. Of note, tumor-infiltrating, but not immunization-elicited 17, Tem cells expressed OX40, possibly as a consequence of chronic stimulation.

A huge body of data supports the notion that CD40 signal releases DCs from paralysis in the tumor microenvironment. DC-restricted CD40 proficiency is necessary and sufficient to induce protective Th1 immunity, through IL-12 production, in a tumor vaccination setting 18. Three seminal papers demonstrated that CD40 triggering in vivo by an agonist mAb can overcome CD4+ and CD8+ T-cell tolerance to tumor-associated antigens and induce tumor eradication 62–64. We have shown that CD40 engagement by CD40L expressed by a tumor-cell vaccine can increase immunity against tumor antigens cross-presented by DCs 27 and that the CD40/CD40L axis is required for CTL induction by vaccination with GM-CSF/OX40L-transduced tumor cells 65. T cells expressing high levels, but not low or null levels, of CD40L can adoptively transfer an efficient anti-tumor immunity 19. We propose here that OX40 triggering can indirectly enhance CD40 stimulation to tumor-infiltrating DCs by increasing CD40L expression by tumor-infiltrating Tem cells, otherwise kept in a quiescent state.

In conclusion, in the present study we provide a mechanistic insight into the effects of OX40 stimulation, separately in Treg and in Teff cells, and specifically in the tumor microenvironment. Indeed, tumor-infiltrating Treg and Teff cells express peculiar molecular programs and functions compared with their peripheral counterparts, and consequently OX40 stimulation elicits tumor microenvironment-specific modifications and allows the “local” correction of “local” defects in both cell types, thus finally leading to the restoration of a functional anti-tumor immunity.

Materials and methods

Mice and treatments

BALB/c mice were from Charles River Laboratory (Calco, Italy); CD40−/−, OX40−/− and Foxp3-GFP mice were provided by L. Adorini (Intercept Pharma, Perugia, Italy), N. Killeen (UCSF), respectively and R. Furlan (San Raffaele Scientific Institute, Milan, Italy) upon agreement with A. Rudensky (New York, USA). All these strains were backcrossed for ten generations to BALB/c. Mice were maintained under pathogen-free conditions in our animal facility and used at 8 wk of age. CT26 cell line (ATCC) was cultured in DMEM (Invitrogen) supplemented with 10% FBS; 5×104 CT26 cells were inoculated subcutaneously in the left flank of mice. When tumor was about 8×8 mm in size, mice were injected intra-tumor with 50 μg of purified anti-OX40 mAb (clone OX86, European Collection of Cell Cultures) or rat IgG (mock) and were sacrificed after 24 h for analysis. Animal experiments were authorized by the Fondazione IRCCS Istituto Nazionale dei Tumori Ethical Committee for animal use and were performed in accordance to the national law (DL116/92).

Abs and flow cytometry analysis

FITC and PerCPCy5.5 anti-CD44 (IM7), PE anti-OX40 (OX86), PE and PerCPCy5.5 anti-IL-10 (JES5-16E3), PE and allophycocyanin anti-Foxp3 (FJK-16S), PE-Cy7 anti-CD4 (L3T4), PE anti-Kd (SF1-1.1.1), PE-Cy7 anti-CD11c (N418), allophycocyanin anti-CD62L (Mel14), PE anti-CD80 (16-10A1), allophycocyanin anti-CD86 (GL1), PE anti-CD8 (53–6.7), PE anti-B220 (RA3-6B2), PE anti-CD11b (MI/70) and streptavidin-PE were from eBioscience. Biotin anti-CD40L (MR1) was from BD Pharmingen. Abs were used at 5 μg/mL. Surface staining was performed in PBS supplemented with 2% FBS for 30 min on ice, except for CD40L (1 h on ice). Intracellular staining of Foxp3 was performed according to manufacturer's instruction (eBioscience). Before IL-10 intracellular staining, cells were in vitro re-stimulated for 4 h at 37°C with phorbol myristate acetate (50 ng/mL, SIGMA), Ionomycin Calcium Salt (500 ng/mL, SIGMA), with Monensin (eBioscience). Flow cytometry data were acquired on a LSRFortessa (Becton Dickinson) and analyzed with FlowJo software (version 8.8.6, Tree Star).

BM transplantation

Female (BALB/c×C57BL/6) F1 mice were irradiated at 600+600 Rad with an interval of 3 h and received 107 BM cells from IL-10-GFP C57BL/6 female mice 22, provided by Giorgio Trinchieri (NCI, Frederick). After 8 wk correct reconstitution was checked by flow cytometry, after staining peripheral blood cells with PE anti-Kd, PE-Cy7 anti-CD4 and allophycocyanin anti-Foxp3. Transplanted mice were inoculated with CT26 subcutaneously and treated with OX86 or PBS. After 24 h, tumors were collected and GFP fluorescence was evaluated in CD4+CD25high cells without any restimulation. In this experiment we could not identify Treg cells by Foxp3 staining because the fixation/permeabilization step induced the loss of GFP expression.

DC in vivo migration

BALB/c and CD40−/− tumor-bearing mice were intratumorally injected with OX86 or rat IgG plus 4×107 FITC-conjugated latex micro-spheres of 1 μm diameter (Polysciences). After 24 h, dLNs were mechanically and enzymatically disaggregated (by incubation for 30 min at 37°C with 400 U/mL of collagenase D). The absolute counts of FITC+ CD11c PE-Cy7+ cells were done for each sample.

Differentiation of BMDC

BMs were collected from femurs and tibias of BALB/c and CD40−/− mice. Cells were cultured for 10 days in IMDM with 10% FBS supplemented with conditioned medium from a murine fibroblast cell line engineered to express mouse GM-CSF (corresponding to 20 ng/mL of recombinant GM-CSF). The differentiation state of cells was checked by flow cytometry.

BMDC–Tem cell co-culture

TILs were enriched by ficoll gradient from single-cell suspensions of mechanically disaggregated tumors 24 h after OX86 or rat IgG treatment. CD4+CD44highCD62Llow Tem cells were sorted using a FACSAria (Becton Dickinson) from TILs pooled from different mice and cultured with BMDCs at 1:1 ratio. After 24 h, BMDC activation was analyzed by flow cytometry.

Treg-cell sorting and gene expression profiling

Treg cells pooled from splenocytes from different Foxp3-GFP mice were sorted using a FACSaria (Becton Dickinson) as CD4+GFP+CD8B220CD11b cells. Purity after sorting was assessed around 98%. Sorted Treg cells were activated overnight with coated anti-CD3 (1 μg/mL) plus OX86 or rat IgG (10 μg/mL). RNA was purified using mirVana Kit (Ambion), and checked for integrity and purity by Agilent Bioanalyzer. Each sample was analyzed in duplicate.

RNA (0.2 μg) was reverse transcribed, labeled with biotin and amplified using the Illumina RNA TotalPrep amplification kit (Ambion). Biotinylated sample (1 μg) was hybridized at 58°C overnight to an expression Bead Chip MouseRef_8_v2.0 array (Illumina). Array chips were washed, stained with 1 μg/mL Cy3-streptavidin (GE Healthcare Europe GmbH) and scanned with an Illumina BeadArray Reader (Illumina). Data were analyzed using the BeadStudio Gene Expression Module v3 (Illumina). Intensity values were quality checked, and the data set was normalized using a cubic spline algorithm. A detection p value <0.05 was set as a cut-off to filter reliable genes. All array data have been deposited in NCBI's Gene Expression Omnibus (GEO) and are accessible through GEO Series accession number GSE32373.

Class comparison analysis to identify differentially expressed genes between Treg cells activated with OX86 or isotype control was performed using the GenePattern Software (Broad Institute-MIT).

Real-time RT-PCR

Foxp3-GFP mice were subcutaneously inoculated with CT26 and intratumorally injected with OX86 or rat IgG. After 24 h, Treg cells were sorted from TILs according to GFP expression. Control Treg cells were sorted from spleens of Foxp3-GFP tumor-free mice. RNA was extracted according to the manufacturer's instructions (RNeasy MICROKIT, Qiagen) and reverse transcribed using High-Capacity® cDNA Reverse Transcription Kits (Applied Biosystem). Real-time RT-PCR was performed on 7900 HT (Applied Biosystem), using TaqMan® Fast Universal PCR masterMix (Applied Biosystem). Assays (Applied Biosystem) and samples were normalized over HPRT1 expression. Data were analyzed using the comparative Ct method.

Promoter analysis

To predict the IRF1 binding site in IL-10, VCAM-1 and Viperin promoters, we identified the genomic sequences using the web tool Gene ( Analysis of promoters was performed with the software TESS, developed by the Computational Biology and Informatics Laboratory of the University of Pennsylvania (

Statistical analysis

Statistical analysis was performed using Prism software (GraphPad Software). Results are expressed as mean±SEM. Statistical analysis was performed using a two-tailed Student's t-test. Data were considered significantly different at p<0.05 (*p<0.05, **p<0.01, ***p<0.005 by Student's t test).


This study was supported by grants from the Italian Ministry of Health and Associazione Italiana Ricerca sul Cancro (AIRC). S.P. is supported by My First AIRC grant (8726). P.P. is supported by a fellowship from FIRC (Fondazione Italiana Ricerca sul Cancro). We thank Arioli Ivano for technical assistance, Gabriella Abolafio and Andrea Vecchi for cell sorting, and Loris De Cecco for gene expression analysis. We are grateful to Christopher Karp and Giorgio Trinchieri for providing BM from IL-10 GFP mice.

Conflict of interest: The authors declare no financial or commercial conflict of interest.