The HIF-1α Hypoxia Response in Tumor-Infiltrating T Lymphocytes Induces Functional CD137 (4-1BB) for Immunotherapy

The study of immune cells in culture is routinely conducted under 21% O₂ concentrations. However, in the tissues where immune responses take place, the concentration of O₂ is much lower (1). This is true in secondary lymphoid organs (2) and much more conspicuously so in solid malignancies. Hypoxia not only affects malignant cells but also stromal components including vascular and immune cells such as tumor-infiltrating lymphocytes (TIL; ref. 3). Although there is extensive knowledge on the effects of tissue hypoxia on vascularization and metabolism in tumors, very little is known about the direct or indirect effects of hypoxia on immune system cells (3).

The heterodimeric hypoxia-inducible transcription factors (HIFα/β) 1, 2, and 3 are central elements in the response to hypoxia (4, 5). In normoxic conditions, their HIFα subunits are rapidly degraded as a result of being poly-ubiquitinated by the Von Hippel–Lindau (VHL) E3 ubiquitin ligase complex (6, 7). When O₂ pressure drops under a certain threshold, prolyl-hydroxylases (PHD) that normally act on HIF-1α become nonfunctional. In the absence of hydroxylation, HIFα subunits are not amenable to poly-ubiquitination by VHL and thereby are not degraded by the proteasome. As a result, HIFα subunits translocate to the nucleus and dimerize with the constitutively expressed HIF-1β subunit. The spectrum of genes controlled directly (by binding to hypoxia-responding elements in promoters/enhancers) or indirectly (by intermediary genes induced by HIF-1α/β) is enormous. The goal of the transcriptional hypoxia response is to ensure cell functional survival under hypoxically hostile conditions and to increase vascularization. Once cell normoxia is regained, VHL can act again on HIFα subunits and the response is accordingly repressed.

Our group has recently published that vascular endothelial cells in solid tumors selectively express surface CD137 (4-1BB; ref. 8). Interestingly, the microenvironmental factor driving CD137 expression was found to be hypoxia.

CD137 (4-1BB) is a member of the TNF receptor (TNFR) family (TNFRSF9) originally described on activated T lymphocytes (9). Its transcriptional induction relies mainly on NF-κB and AP-1 (10) as triggered by the TCR-CD3 complex. Once at the cell surface, CD137 provides upon ligation costimulatory signals to T cells (11) and activates natural killer (NK) cell functions (12). Moreover, CD137 is also functionally expressed on dendritic cells (DC; 13) and endothelial cells undergoing inflammation (14) or hypoxia (8). CD137 is stimulated by its only known natural ligand (CD137L, 4-1BBL) whose expression is mainly restricted to the surface of activated antigen-presenting cell (APC; macrophages, DCs, and B cells; ref. 11)

The CD137-CD137L receptor–ligand pair under physiologic conditions is apparently not very important in light of the mild immune phenotypes of CD137^-/- and CD137L^-/- mice, which only suffer from a discrete CTL defect in antiviral responses (15, 16). Nonetheless, artificial stimulation of CD137 under the overwhelming presence of circulating agonist antibodies provides a strong costimulation in vivo (17) that can even revert CTL anergy (18). Importantly, there is plenty of evidence for the efficacy of anti-CD137 monoclonal antibodies (mAb) in the treatment of mice bearing established tumors (17, 19, 20); this is not only achieved by agonist antibodies but also by dimeric RNA aptamers (21) or tumor cells expressing a surface-attached anti-CD137 scFv antibody (22). The main cellular players in those potent antitumoral immune responses seem to be CTLs, but there is also a role for NK cells (12, 18) and DCs (23) at least in some tumor models. This preclinical evidence has led to clinical trials (24) with 2 human mAbs directed against CD137 (BMS-663513 and PF-05082566).

A potential drawback for anti-CD137 mAb is that treatment causes liver inflammation (25, 26). At least in mice, this is dependent on a polyclonal infiltrate dominated by CD3⁺CD8⁺CD11c⁺–activated T cells that mainly occurs in portal spaces. Such an inflammatory response is dose- and target-dependent, and although there are no relevant clinical consequences for mice, it could be otherwise in human patients (24). Therefore, limitation of such liver side effects while keeping therapeutic efficacy would be a remarkable advantage.

In this study, we have documented that CD137 expression is strongly favored in TILs as a result of a HIF-1α–dependent response to hypoxia. In accordance with the selective CD137 expression in TILs, such lymphocytes are responsive to agonist CD137 mAb. This was exploited by injecting small intratumoral doses of the antibody that were effective without systemically causing liver side effects. From the point of view of therapy, intratumorally delivered low doses of anti-CD137 mAb to target TILs achieve systemic therapeutic effects and act in synergy with the systemic blockade of the PD-1/B7-H1 (PD-L1) pathway. Therefore, hypoxia-mediated upregulation of CD137 on TILs offers the opportunity for confining the therapeutic effects of CD137 agonists to tumors as a safety feature.

TILs are a prominent component of tumor stroma. These lymphocytes show features of activation and are known to be enriched for lymphocytes specific for tumor antigens. CT26 is an aggressive transplantable colon carcinoma cell line. CT26-established tumors in syngeneic mice harbor CD4⁺ and CD8⁺ T cells. To study whether TILs express the CD137 activation marker, single-cell suspensions were prepared from tumors, lymph nodes, and spleens.

As can be seen in Fig. 1A and B, CD137 was minimally expressed by T cells in spleens and lymph nodes of tumor-bearing animals but it was brightly present on CD4⁺ and CD8⁺ T cells in the tumor compartment. As a control, TILs in CT26-derived tumors grafted onto CD137^-/- mice (8) showed no evidence of CD137 immunostaining, excluding nonspecific binding of the anti-CD137 antibodies. Figure 1C, representative of a large number of microscopic fields, shows in situ immunofluorescence evidence in the tumor microenvironment for CD4 and CD8 T cells clearly co-expressing CD137.

Among CD4⁺ TILs, it was found that CD137 was brightly expressed on the surface of CD4⁺FOXP3⁺ regulatory T cells (Treg; Supplementary Fig. S1A). It is of note that CD137 expression was not found on NK cells (CD3^-DX5⁺) at the tumor, whereas expression on tumor-infiltrating NKT cells (CD3⁺DX5⁺) was found at dim levels (Supplementary Fig. S1B). With regard to other leukocytes infiltrating the tumors, CD137 expression was undetectable on CD19⁺ B cells (Supplementary Fig. S2A), F4/80⁺ macrophages (Supplementary Fig. S2B), CD11b⁺GR-1⁺ myeloid cells (Supplementary Fig. S2C), and CD11c^high DCs (Supplementary Fig. S2D).

CD137 expression on CD3⁺ TILs was not an exclusive feature of CT26-derived tumors as it was also readily observed in transplanted syngenic tumors derived from MC38 colon carcinomas and B16 melanomas (Fig. 2A and B). Indeed, the pattern of selective CD137 expression on T lymphocytes at the tumor, in the absence of CD137 expression on splenocytes or lymph node lymphocytes, was also confirmed in these tumor models. Furthermore, a similar distribution of CD137 was observed in the Her-2/neu transgenic female mice bearing spontaneous adenocarcinomas, albeit CD137 expression was dimmer than in transplanted tumors and the number of TILs was much lower (Fig. 2C). In any case, CD137 is expressed on TILs from spontaneous tumors arising in tumor-prone oncogene-transgenic mice.

The factors upregulating CD137 expression on TILs could be multiple, including the presence of cognate antigen. However, we have reported that hypoxia is a driving factor for ectopic CD137 expression on endothelial cells in tumor blood vessels (8).

We investigated whether hypoxia could be a factor underlying CD137 expression on T cells. First, we found that in tumors in which we detect CD137⁺, TILs are indeed undergoing hypoxia as documented on living mice by sensitive ¹⁸F-MISO positron emission tomographic (PET) imaging based on imidazole chemistry (10). This property applies both to transplanted (CT26, MC38, B16-OVA) and spontaneous tumors (Supplementary Fig. S3A–S3C).

Hypoxia by itself did not upregulate CD137 expression in cultured T lymphocytes in absence of TCR-CD3 triggering (data not shown). However, hypoxia (1% O₂) clearly upregulated CD137 expression upon T-cell stimulation via CD3/CD28. Indeed, as can be seen in Fig. 3A, suboptimal stimulation using latex microbeads coated with anti-CD3 and anti-CD28 mAb rendered little CD137 surface expression on CD4⁺ and CD8⁺ spleen T cells. Culture under hypoxic conditions greatly upregulated CD137 surface expression in comparison with normoxic conditions. Moreover, such effects of hypoxia-elicited CD137 upregulation were confirmed at the protein and mRNA level with T-cell receptor (TCR) transgenic T cells stimulated with their cognate peptide antigens (Fig. 3B and C). In fact, OT-1 and OT-2 TCR-transgenic lymphocytes in culture were susceptible to CD137 upregulation upon exposure to hypoxia.

To exclude indirect effects mediated by other cell subsets present in the splenocyte cultures, we carried out experiments on CD4 and CD8 immunomagnetically sorted T cells isolated by negative selection. Such experiments rendered comparable results with those carried out with total splenocytes (Supplementary Fig. S4A).

In response to hypoxia, several cellular changes occur. The main route responding to hypoxia is under the control of the HIF transcription factors. HIF-1α and HIF-2 are hydroxylated in prolyl residues by specific hydroxylases. The hydroxylated forms are rapidly degraded, whereas the nonhydroxylated forms are stable and exert transcriptional effects following nuclear translocation.

To study whether HIF-1α responses were involved in CD137 upregulation, first we used a PHD inhibitor [dimethyloxalylglycine (DMOG)] and second we carried out experiments with HIF-1α–deficient T cells.

Figure 4A shows that DMOG-treated OT-1 and OT-2 lymphocytes stimulated by their cognate antigens expressed surface CD137 at a higher intensity. Such increases were comparable with those observed upon culture under hypoxic conditions. Results were consistent with robust inductions of CD137 mRNA (Fig. 4A, bottom).

Treg cells expressed bright intensity of surface CD137 in the cultures placed under normoxia and hypoxia, indicating that there was no need for hypoxia to drive intense CD137 expression in this immunosuppressive subset (Supplementary Fig. S4B).

In humans, we observed that CD4 and CD8 T cells stimulated by plate-bound anti-CD3 mAb and cultured under hypoxia upregulated CD137 more intensely in 48 hours than lymphocytes cultured under normoxic conditions (Supplementary Fig. S5A and S5B). These experiments carried out with peripheral blood mononuclear cell (PBMC) derived from buffy coats of healthy blood donors showed indications of individual variability in the degree of CD137 upregulation by hypoxia (low vs. high responders in Supplementary Fig. S5A). Similarly, when the cultures on plate-bound anti-CD3 mAbs were treated with DMOG, CD4 and CD8 T cells upregulated surface CD137 more intensely, indicating the involvement of the PHD controlling HIF-1α (Supplementary Fig. S5C).

In a more direct set of experiments, the involvement of HIF-1α was addressed using T cells from mice induced to become deficient in HIF-1α. For this purpose, we used Hif-1α^floxed-UBC-Cre-ER^T2 mice harboring 2 loxP sites flanking the exon 2 of the murine Hif-1α locus, as described previously (27). These mice also express ubiquitously the Cre recombinase fused to a mutated estrogen receptor ligand–binding domain (Cre-ERT2), which remains confined in the cytoplasm until 4-hydroxytamoxifen is administered (28). Therefore, Hif-1α is inactivated after 4-hydroxytamoxifen treatment in Hif-1α^floxed-UBC-Cre-ER^T2 but remains intact in the corresponding control mice Hif-1α^wt-UBC-Cre-ER^T2 or Hif-1α^floxed (29). In this study, HIF-1α gene inactivation was induced by intraperitoneal injection of 4-hydroxytamoxifen in mice that carry a Hif-1α^floxed allele in homozygosity.

Age-matched Hif-1α^floxed-UBC-Cre-ER^T2 and littermate control mice were treated with 4-hydroxytamoxifen, and 1 week later, splenocytes from these groups were stimulated with anti-CD3 mAb either under normoxic or hypoxic conditions. T cells isolated from 4-hydroxytamoxifen–treated Hif-1α^floxed-UBC-Cre-ER^T2, but not those isolated from 4HT-treated control mice, drastically reduced their Hif-1α gene expression (data not shown) and did not upregulate surface CD137 in 48 hours of culture (Fig. 4B). This occurred both on HIF-1α–deficient CD4 and CD8 T cells present in these cultures, whereas control spleen T cells readily upregulated the CD137 molecule under hypoxic conditions as a positive control on both lymphocyte subsets. These differences were also observed by quantitative reverse transcription (RT) PCR at the mRNA level (Fig. 4B, right). However, HIF-1α was not required for expression of CD137 on Treg cells (Supplementary Fig. S4B).

Furthermore, hypoxia and HIF-1α were required for sustained CD137 expression because if hypoxia was only provided during the first 24 hours of culture, the resulting CD4 and CD8 T cells on days 4 and 6 expressed considerably less surface CD137 than those which remained under hypoxia for the duration of the culture (Supplementary Fig. S6A and S6B).

In Fig. 4C, the effect of DMOG on HIF-1α–deficient T cells undergoing CD3-elicited activation is shown at the protein and mRNA levels. Indeed, DMOG did not cause any upregulation of CD137 on HIF-1α–deficient T cells, whereas CD137 was readily induced in T cells from control mice.

According to these results, TIL expression of CD137 could be at least, in part, an effect of hypoxia as sensed by the HIF-1α system. To further explore this possibility, T cells induced to become HIF-1α–deficient were transferred into mice bearing established MC38-derived subcutaneous tumors. As explained in Fig. 5A, recipient and donor mice were congenic for the CD45.1/CD45.2 allelic polymorphism and thereby donor- and recipient-derived TILs could be recognized by specific antibodies upon fluorescence-activated cell sorting (FACS) analyses. With this experimental design, we could compare whether adoptively transferred T cells (CD45.2⁺) upregulated or not CD137 upon infiltration into the tumors in comparison with endogenous T cells (CD45.1⁺). Figure 5A shows that CD3⁺ TILs from the HIF-1α^-/- CD45.2⁺ donor did not express surface CD137 whereas CD45.1⁺ T cells from the tumor-bearing recipient mice did. As expected, both endogenous and donor T cells were CD137-negative in the spleen (Fig. 5A). The selective lack of CD137 in HIF-1α–deficient TILs applied both to CD4 and CD8 TILs as shown in Fig. 5A.

In these experiments, comparisons were made between adoptively transferred T cells and endogenous T cells. To rule out possible artefacts due to the lack of adoptive transfer of the HIF-1α–sufficient lymphocytes in Fig. 5A, experiments were carried out upon adoptive co-transfer of HIF-1α–sufficient and -deficient T splenocytes to Rag^–/– mice bearing MC38 tumors. As can be seen in Fig. 5B, expression of CD137 on TILs was much more intense on CD45.1⁺ HIF-1α–sufficient T lymphocytes than on CD45.1^- HIF-1α–deficient co-infiltrating lymphocytes.

In conclusion, HIF-1α in response to hypoxia plays a critical and necessary role in the induction of CD137 expression on TILs.

The fact that CD137 expression is largely confined to the tumors could be exploited by intratumoral delivery of the agonist immunotherapeutic mAb. Intratumoral injections of as little as 5 μg/dose of 1D8 mAb induced regressions in nearly half of CT26-bearing BALB/c mice treated when the tumors had already reached 5 mm in diameter (day 7).

Systemic treatment with anti-CD137 mAb and anti-B7-H1 (PD-L1) has been described to exert powerful synergistic effects (30). In this CT26 tumor model, repeated 100 μg doses of anti-B7-H1 given intraperitoneally (i.p.) attained 5 of 12 complete regressions (Fig. 6A). Importantly, combined treatments of minute doses of anti-CD137 mAb given intratumorally and anti-B7-H1 mAb given systemically achieved complete regressions in 10 of 12 mice (Fig. 6A). These cured mice were immune to a rechallenge with CT26 cells 3 months later (Supplementary Fig. S7) as a result of systemic immune memory.

The effect of B7-H1 blockade is probably due to the fact that grafted CT26 cells forming the tumor mass express B7-H1 on the surface as described in other tumors (31). This is in contrast to the lack of expression on cultured CT26 cells, which can be induced to express B7-H1 by 48-hour culture in the presence of IFN-γ (Fig. 6B). In conjunction with the expression of B7-H1 (PD-L1) on the tumor cells, CD4 and CD8 TILs express the checkpoint co-inhibitory molecule PD-1. Hence PD-1⁺ TILs should be sensitive to de-repression by B7-H1 blockade (Fig. 6C). It was also found that PD-1^bright TILs coincided with CD137⁺ TILs. Taken together, these observations probably provide a reason for the synergistic effects of artificial costimulation via CD137 and simultaneous relief from PD-1 inhibition upon blockade of its ligand as previously described by Hirano and colleagues (30).

Importantly, low doses of anti-CD137 mAb achieved remarkable systemic therapeutic effects against concomitant established tumors. This was shown in a bilateral model of CT26-derived tumors transplanted on day 0 to one of the flanks and on day 4 to the opposite side to resemble metastatic disease. On days 7, 9, and 11, the primary lesion was intratumorally treated, whereas the smaller tumor was left untreated. As can be seen in Fig. 6D, 5 of these 7 distant tumor nodules underwent complete rejection, thus indicating the elicitation of concomitant antitumor immunity against tumor nodules that had not been directly treated with the 1D8 anti-CD137 mAb.

Absence of CD137 immunostaining on T cells in tumor-draining lymph nodes was confirmed (Supplementary Fig. S8), further suggesting that the therapeutic costimulatory events were taking place on the TILs. However, upon intratumoral treatment with low doses of 1D8 mAb, there was a clear hyperplasia of tumor-draining lymph nodes that became enlarged 3- to 4-fold over control IgG (Supplementary Fig. S9A).

In addition, the number of TILs increased upon intratumoral treatment with low-dose anti-CD137 mAb (Supplementary Fig. S9A). CD8⁺ T cells increased both in terms of percentage and absolute numbers, whereas CD4⁺ T cells decreased in absolute and relative numbers (Supplementary Fig. S9B and S9C). As a consequence of a decrease in CD4⁺FOXP3⁺ T cells among TILs, a clear increase in the ratio of CD8 T cells/Treg was observed upon 1D8 mAb treatment (Supplementary Fig. S9D). It is of note that following intratumoral treatment, a tendency in the CD8⁺ TILs to express higher levels of CD137, Granzyme B, and KLRG-1 was observed (Supplementary Fig. S9E). These results are in line with the observations reported by Curran and colleagues (32) upon systemic treatment with an anti-CD137 antibody. Also, consistent with a more active effector phenotype of such CD8 TILs, these T lymphocytes if restimulated in vitro with the AH1 tumor antigen peptide, expressed more IFN-γ and TNF-α (Supplementary Fig. S10A–S10C).

Agonist anti-CD137 mAbs cause liver inflammation (25). One advantage of giving smaller and more localized doses of anti-CD137 mAb is that the side effects of CD137 known to be dose-dependent would be mitigated. In this regard, Fig. 7 shows that although three 100-μg doses of anti-CD137 mAb given i.p. induced mild liver transaminase increases, 5 μg given intratumorally 3 times did not cause this side effect (Fig. 7A). This is because CD8 T mononuclear cells do not accumulate forming inflammatory infiltrates in the liver of mice treated with the intratumoral 5-μg doses (Fig. 7B–D). Importantly, combinatorial treatments with local anti-CD137 mAb and systemic anti-B7-H1 mAb do not cause liver inflammation (Fig. 7B–D). The idea of conducting intratumoral injections of lower doses of immunostimulatory mAbs had been previously applied successfully by Fransen and colleagues (33) with agonist anti-CD40 mAb. Their strategy also limits systemic toxicity while preserving antitumoral efficacy.

In our case with anti-CD137 mAb, we are exploiting the selective presence of the target for the immunostimulatory antibody at the hypoxic tumor microenvironment to spatially focus the immunotherapeutic effects while avoiding liver toxicity.

Stromal components and malignant cells dialog and mutually change their functional responses. TILs are a prominent feature of malignant tissue whose presence has been correlated with a better survival outcome in various malignant diseases (34, 35). However, TILs are severely restrained in their ability to tackle tumor cells as a result of multifactorial immunosuppressive mechanisms, which are especially intense in the tumor microenvironment (36). Nonetheless, ex vivo cultures of TILs to expand and reinvigorate these cells have been used in adoptive cell therapy with reports of impressive objective responses in patients with melanoma (37).

Solid tumors are typically compromised in vascularization; as a result, they become hypoxic and respond by producing vascular growth factors to promote angiogenesis. Tumor hypoxia affects both malignant cells and stromal compartments presumably including infiltrating lymphocytes, although there is a conspicuous lack of information in this regard.

Hypoxia has been recently found to induce tumor cell production of a chemokine (CCL28) that attracts Tregs that would locally depress effector cellular immunity (38). However, very little is known about what tumor hypoxia does to tumor-infiltrating T cells directly. Experiments on T-cell cultures set up under hypoxia suggest a decrease in T-cell activation (39, 40), for instance, upon CD3/CD28 stimulation (39). The reasons could involve the HIF response to hypoxia and the effects of released adenosine that acts on inhibitory lymphocyte receptors for this moiety (1).

Moreover, a recent report involves HIF-1 transcription factors in the control of T-helper (T_H) 17/Treg balance acting directly on T cells (41). Experiments on conditional HIF-1α^-/- T cells indicate that the HIF-1 response downregulates acquisition of cytokine production and cytolytic activity in CD4 and CD8 T cells (42). Lack of detailed information about T-cell activation under low O₂ tensions is worrying as immune responses in tissues occur at ≥5% O₂ (2) and tumors are even more hypoxic, at least in tissue regions located far from blood supply.

We had previously observed that the T-cell activation marker CD137 was upregulated on cultured endothelial cells as a result of hypoxia (8). Anti-CD137 mAb on these CD137⁺ endothelial cells elicited proinflammatory changes believed to be involved in the prominent immunotherapeutic effects of anti-CD137 agonist antibodies against tumors because of promoting homing of T lymphocytes to the tumor lesions.

In this study, we found that CD137 is upregulated on TILs but not in secondary lymphoid organs of tumor-bearing mice. Hypoxia was a candidate factor to explain CD137 expression on TILs. Indeed, grafted and spontaneous tumors are clearly hypoxic. However, hypoxia by itself does not induce CD137 on cultured T cells and TCR-CD3 stimulation is required. The striking finding is that CD137 is much more intensely upregulated on T cells undergoing activation when hypoxia is provided. Therefore, we can speculate that both antigen stimulation and hypoxia concurrently coexist in the tumor microenvironment to cooperatively give rise to CD137 expression on TILs. It is intriguing that NK lymphocytes in the tumor that ought to experience the same level of hypoxia do not upregulate CD137. Further studies will address the differential behavior of this lymphocyte subset which also expresses HIF-1α mRNA.

Hypoxia causes various cellular changes such as induction of functionality of HIF-1α and HIF-2 transcription factors, generation of free radicals, and adenosine release to the surrounding milieu (1). To ascertain which hypoxia response was involved in CD137 upregulation, we carried out compelling experiments using T cells from HIF-1α–inducible knockout mice. Such T cells upregulate CD137 much less efficiently when undergoing activation under hypoxia. Accordingly, experiments in which HIF-1α–deficient and –sufficient T cells co-infiltrate tumors show that CD137 is selectively expressed on the hypoxia-responsive lymphocytes.

We considered the possibility that CD137 was a direct transcriptional target of HIF-1α/1β, but detailed bioinformatic analyses of the cd137 locus did not reveal any interspecies-conserved hypoxia response elements in the noncoding regions 10 kb upstream and downstream of the cd137 locus including introns (43). Hence, the alternative hypothesis of indirect CD137 enhancing effects mediated by genes upregulated by the transcriptional activity of HIF-1α is under investigation. In fact, previous reports have revealed a functional interplay of HIF-1α and NF-κB members (44), whereas canonical NF-κB sites are leading inducers of the CD137 promoter (10). In fact, CD137 upregulation in our experiments seems to take place mainly at the transcriptional level. An antecedent for indirect effects exerted by HIF-1α in T cells has been recently published by Dang and colleagues (41), who unraveled unexpected functions of HIF-1α that are critical for the immune system such as protein-to-protein regulation of FOXP3 degradation in CD4 T cells and the control of the expression of secondary transcription factors.

We cannot be sure that cognate antigen and HIF-1α response are the only factors behind CD137 expression on TILs, but our adoptive transfer experiments with HIF-1α–deficient T cells clearly indicate that HIF-1α is a critical factor for CD137 upregulation on TILs. The contribution of hypoxia to CD137 expression on T lymphocytes in vivo might extend to other hypoxia situations such as tissues destroyed by infection or atherosclerotic lesions.

The fact that CD137 is a target for a powerful strategy of immunostimulation against cancer makes the observation that TILs are selectively expressing CD137 very appealing. However, the tumor microenvironment is a difficult place for T-cell stimulation because of the abundance of immunoinhibitory factors. Among these suppressive factors, the PD-1/B7-H1(PD-L1) pathway stands out as hierarchically very important (45, 46). The importance is highlighted by the fact that its blockade is immunotherapeutic in mice and humans (47).

Tregs are perceived also as an important repressor of therapeutic immunity against tumors. We found that CD4⁺FOXP3⁺ lymphocytes among TILs express CD137 and would be able to interact with injected anti-CD137 mAb. The effect of CD137 stimulation on Tregs has been reported as dual because despite promoting Treg proliferation, Treg exposure to agonist anti-CD137 mAb weakens the suppressor activities and also desensitizes effector T lymphocytes to the suppressor effects (48). In our study, we confirm bright expression on these suppressive lymphocytes; but in our hands, sensing hypoxia via HIF-1α is not a prominent driving factor for their expression of CD137.

Agonist anti-CD137 mAbs as monotherapy attain impressive responses against many transplanted tumors, although some of these are resistant including spontaneous breast carcinomas in MMTV-Neu mice (20, 49). Although treated mice survive curative anti-CD137 mAb treatments without signs of distress, transient liver inflammation and myelosuppression have been reported in repeatedly treated animals (26). Fransen and colleagues (33) showed that anti-CD40 mAb at low doses inside tumors can be efficacious without systemic toxicity. We have found that because of the selective presence of CD137 on TILs, the effect of the antibody can be delivered locally without liver inflammation. Importantly, CD137 delivered intratumorally is synergistic with B7-H1 blockade; this synergy had been previously described by Hirano and colleagues (30). Studies in mice chronically infected with LCMV also show synergistic therapeutic effects of anti-CD137 and anti-B7-H1 mAb that correlate with an enhancement of CD8-mediated antiviral immunity (50). Combinatorial treatments with systemic B7-H1 blockade and local agonist anti-CD137 mAbs are highly efficacious. This probably reflects that the very same TILs released from PD-1 inhibition are being provided costimulation via CD137. Importantly, local injection of anti-CD137 mAb attained systemic therapeutic effects against concomitant tumors that are postulated to be mediated by recirculation of the properly stimulated TILs.

Recently, an elegant report from Alizadeh and colleagues (51) has shown that TILs in human lymphomas express surface CD137 and that the amount of CD137 on human pathology specimens of this malignancy correlated favorably with prognosis. Further research should extend these observations to TILs in other human tumors and study a potential correlation with hypoxia.

In summary, CD137 is expressed on TILs because of a response to hypoxia mediated by HIF-1α. The physiologic reason behind these facts is unknown, but it can be exploited to target CD137 stimulation to where it is most needed, thus limiting systemic side effects. Combinatorial PD-L/B7-H1(PD-1) pathway blockade and anti-CD137 local stimulation offer many potential clinical advantages, and human immunostimulatory mAbs to all these targets are being tested (24).

BALB/c and C57/BL6 wild-type mice (5 to 6 weeks old) were purchased from Harlan Laboratories. OT-1, OT-2, CD45.1, Rag^-/-, and MMTV-NeuT mice (52) were purchased from the Jackson Laboratory and bred in our animal facility under specific pathogen-free conditions. CD137^-/- mice in pure BALB/c background have been previously described (8).

Hif-1α^floxed-UBC-Cre-ER^T2 mice were generated from B6.129-Hif-1α^tm3Rsjo/J mice (Jackson Laboratories, stock no. 007561), which harbor 2 loxP sites flanking exon 2 of the murine Hif-1α locus. These mice were crossed with Tg(UBC-Cre/ER^T2)1Ejb/J mice (Jackson Laboratories, stock no. 008085) as described above to generate Hif-1α^floxed-UBC-Cre-ER^T2 mice and their corresponding controls, Hif-1α^wt-UBC-Cre-ER^T2 and Hif-1α^floxed mice. For Hif-1α gene inactivation, Hif-1α^floxed-UBC-Cre-ER^T2 and the corresponding control mice (10 to 15 weeks old) were treated with 3 every other day i.p. doses (total of 3 doses) of 1 mg of 4-hydroxytamoxifen (Merck); 7 days after the last dose, the animals were used for experiments. All animal procedures were conducted under Institutional Guidelines that comply with National Laws and Policies (study approval 070-10).

CT26, B16-OVA, and MC38 tumor cell lines were from American Type Culture Collection and the master cell banks authenticated in 2011 by RADIL (Case Number: 6592-2012). Cells were cultured in complete RPMI medium [RPMI-1640 with GlutaMAX (Gibco) containing 10% heat-inactivated FBS (Sigma-Aldrich), 100 IU/mL penicillin, and 100 μg/mL streptomycin (BioWhittaker) and 5 × 10^-5 mol/L 2-mercaptoethanol (Gibco)]. For hypoxic culture conditions, splenocytes were incubated for the indicated times under 1% O₂ atmosphere in a modular incubator chamber (Billups-Rothenberg Inc.) or the H35 Hypoxystation (Don Whitley) for those experiments in which retrieval of part of the cultures from hypoxia was required.

Anonymous buffy coat by-products of blood donations were procured under informed consent of the donors and under approval of the regional committee of ethics of the local government of Navarra. PBMCs were prepared by Ficoll gradients and T cells were stimulated in 6-well plates precoated with 1 μg/mL of OKT3 mAb in the same culture medium as in the mouse experiments but without 2-mercaptoethanol. Forty-eight cultures were placed under hypoxia (1% O₂) or in the presence of 0.2 mmol/L of DMOG when indicated. Retrieved cells were analyzed by flow cytometry upon multicolor immunostaining with fluorochrome anti-CD3, CD4, CD8 (Becton Dickinson) and CD137 (e-bioscience)-specific mAb.

A total of 0.5 × 10⁶ CT26, MC38, or B16-OVA cells were injected subcutaneously into the flank in 100 μL PBS. Mice and tumor size were monitored and mice were sacrificed when tumor size reached 300 mm² (53). For some experiments, mice were injected bilaterally in opposite flanks with 0.5 × 10⁶ CT26 tumor cells with a delay of 4 days between sites. In those experiments, only the larger primary tumor was injected with the antibody.

For immunofluorescence and flow cytometric analyses of TILs, cleanly excised tumor nodules were placed in Petri dishes, minced finely with a scalpel blade, and incubated for 25 minutes at 37°C in a solution containing Collagenase-D and DNAse-I (Roche) in RPMI. The entire material was passed through a 70-μm cell strainer (BD Falcon, BD Bioscience) and pressed with a plunger to obtain unicellular cell suspensions. Single-cell suspensions were pretreated with FcR-Block (anti-CD16/32 clone 2.4G2; BD Biosciences-Pharmingen). Afterward, cells were stained with the following antibodies or reagents obtained from BD Pharmingen: CD3, CD4, CD8, CD137, CD49b, CD45.1, PD-1, PD-L1, KLRG-1, F4/80, Gr-1, CD11b, CD11c, Granzyme B, FOXP3, TNF-α, IFN-γ, and/or the respective conjugated isotype controls. For TNF-α and IFN-γ intracellular staining, tumor mononuclear cells were restimulated with AH1 peptide (SPSYVYHQF 10 μg/mL, NeoMPS) for 1 hour and incubated overnight in the presence of brefeldin-A (10 μg/mL; Sigma) and then fixed and permeabilized with Fix/perm buffer (eBiosciences) before FACS staining. OT-1 and OT-2 peptides were also from NeoMPS.

FACSCanto II and FACSCalibur (BD Pharmingen) as indicated were used for cell acquisition and data analysis was carried out using FlowJo (TreeStar software).

The hybridomas producing agonistic mouse anti-CD137 (clone 2A) and anti-B7-H1 (PD-L1; clone 10B5) were kindly provided by Dr. Lieping Chen (Yale University, New Haven, CT). Mean fluorescence intensity (MFI) values represent the subtraction of specific versus control MFI. Polyclonal rat IgG with undetectable lipopolysaccharide (LPS; Sigma) was used as control. Mouse anti-CD3/CD28 microbeads were from Dynabeads (Invitrogen). DMOG was purchased from Enzo Life Sciences.

Tissue immunofluorescence staining was conducted on 10-μm thick cryosections with the following antibodies: CD4, CD8 (eBiosciences) and goat anti-CD137 (R&D Systems) followed by anti-goat AF488 (Invitrogen) as a secondary antibody. For confocal microscopy, LSM 510 META (Carl Zeiss) equipment was used. The images were analyzed using Zeiss LSM Image Browser software.

Tumor hypoxia was measured by PET with the radiotracer fluorine-18-fluoromisonidazole (¹⁸F-FMISO). On the day of the study, mice were anesthetized with 2% isoflurane in 100% O₂ gas for ¹⁸F-FMISO injection (14.9 ± 4.9 MBq in 100 μL) in the tail vein. Four hours after the tracer injection, mice were placed prone on the PET scanner. A static 30-minute study (sinogram) was acquired using a Mosaic (Philips) small-animal dedicated imaging tomography with 11.9 cm axial field of view (FOV) and 12.8 cm transaxial FOV and a 2-mm resolution. Images were reconstructed using the 3-dimensional (3D) Ramla algorithm (a true 3D reconstruction) with 2 iterations and a relaxation parameter of 0.024 into a 128 × 128 matrix with a 1-mm voxel size applying dead time, decay, random, and scattering corrections.

For the assessment of tumor ¹⁸F-FMISO uptake, all studies were exported and analyzed using the PMOD software (PMOD Technologies Ltd.). Regions of interest were drawn on coronal 1-mm thick small-animal PET images on consecutive slices including the entire tumor. Finally, maximum standardized uptake value (SUV) was calculated for each tumor using the formula SUV = [tissue activity concentration (Bq/cm³)/injected dose (Bq)] × body weight (g).

Total RNA was extracted from lymphocytes using the RNeasy Mini Kit (Qiagen). We treated RNA with DNaseI (Gibco-BRL) before reverse transcription with M-MLV reverse transcriptase (Gibco-BRL) in the presence of RNaseOUT (Gibco-BRL). Real-time PCR was carried out with iQ SYBR green supermix in an iQ5 real-time PCR detection system (Bio-Rad). Copy numbers of CD137 cDNA were quantified by quantitative PCR with primers annealing the mouse CD137 cDNA (forward primer, 5′-AACATCTGCAGAGTGTGTGC-3′; reverse primer, 5′-AGACCTTCCGTCTAGAGAGC-3′; product length, 252 bp). Samples were analyzed in triplicate, and data were normalized by comparison with β-actin as an internal control (forward primer, 5′-CGCGTCCACCCGCGAG-3′; reverse primer, 5′-CCTGGTGCCTAGGGCG-3′; product length, 194 bp). The amount of each transcript was expressed according to the formula 2^{C_t(b-actin)-DC_t(CD137)}, where C_t is the cycle at which the fluorescence increases appreciably above background fluorescence.

Hif-1α^floxed-UBC-Cre-ER^T2 or CD45.1 donor mice were treated with tamoxifen. One week later, total splenocytes were inoculated intravenously (10⁷ spleen cells per mouse) into CD45.1 or Rag^–/– recipient mice carrying subcutaneous MC38 tumors. Ten days following T-cell transfer, tumors were harvested and analyzed for CD137 expression on CD45.1 versus CD45.2 T cells in a FACSCanto II flow cytometer.

Prism software (GraphPad Software) was used to determine statistical significance of the differences between groups by applying the unpaired Student t tests or 2-way ANOVA tests. P values of <0.05 were considered significant.

M. Jure-Kunkel is a full-time employee of Bristol-Myers Squibb, I. Melero is a consultant for Bristol-Myers Squibb, Merck-Serono, Pfizer Inc., Miltenyi Biotec, and DIGNA-Biotech and has Ownership Interest (including patents) in Bristol-Myers Squibb. No potential conflicts of interests were disclosed by the other authors.

Conception and design: A. Palazón, I. Martínez-Forero, J.L. Perez-Gracia, S. Hervás-Stubbs, I. Melero

Development of methodology: A. Palazon, A. Morales-Kastresana, C. Alfaro, J.L. Perez-Gracia, S. Hervás-Stubbs, A. Rouzaut, J. Aragonés

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Palazón, I. Martínez-Forero, A. Teijeira, I. Peñuelas, A. Rouzaut, M.O. de Landázuri, M. Jure-Kunkel, J. Aragonés

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Palazón, I. Martínez-Forero, C. Alfaro, J.L. Perez-Gracia, I. Melero

Writing, review, and/or revision of the manuscript: A. Palazon, M.F. de Sanmamed, J.L. Perez-Gracia, M. Jure-Kunkel, I. Melero

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Palazón, C. Alfaro

Study supervision: S. Hervás-Stubbs, I. Melero

Carried out all experiments: A. Palazón

The authors acknowledge long-term collaborations on CD137 with Dr. Lieping Chen (Johns Hopkins, Baltimore, MD) and scientific discussion with Dr. Luis del Peso (Universidad Autónoma de Madrid, Madrid, Spain) and Dr. Jesus Prieto (Universidad de Navarra, Navarra, Spain); excellent animal care by Eneko Elizalde and Elena Ciordia; flow cytometric assistance by Dr. Diego Alignani; and English editing by Dr. Paul Miller.

The financial support was obtained from MEC/MICINN (SAF2005-03131 and SAF2008-03294), Departamento de Educación del Gobierno de Navarra, Departamento de Salud del Gobierno de Navarra; Redes temáticas de investigación cooperativa RETIC (RD06/0020/0065), European commission VII framework program (ENCITE), SUDOE-IMMUNONET, and “UTE for project FIMA”. S. Hervás-Stubbs has a Ramon y Cajal contract from MICINN and A. Palazón a scholarship from FIS. A. Morales-Kastresana receives an FPI scholarship from MICINN.