Twists and turns to translating 4-1BB cancer immunotherapy

See allHide authors and affiliations

Science Translational Medicine  12 Jun 2019:
Vol. 11, Issue 496, eaax4738
DOI: 10.1126/scitranslmed.aax4738


Previous shortcomings of CD137-targeted immunotherapy may be overcome by engineered bispecific agents (Claus et al., this issue).

T cells are a major focus of cancer immunotherapy. These cells integrate signals from various sources to become fully activated and differentiated. This includes antigen recognition, which stimulates the T cell receptor and CD3 signaling. Other signals include various cytokines or costimulatory molecules. CD137 (4-1BB) is an inducible costimulatory molecule in the TNFR superfamily (TNFRSF-9) that is expressed by T cells that have recognized cognate antigen, as well as other leukocytes. CD137 has a single known ligand, CD137L (TNFSF-9), normally only expressed on professional antigen-presenting cells such as activated macrophages, B cells, and dendritic cells. Cell-cell contact causes CD137L trimers to cluster CD137 into signaling trimers upon ligation. CD137 has long been considered a viable immunotherapy target alongside other TNFR family members such as OX40, GITR, and CD40 but has not yet delivered clinical results. In their new study, Claus et al. (1) get one step closer.

A prominent feature of CD137 is that its expression is inducible and therefore works on those T cells that have recently engaged in antigen recognition. CD137 also works on NK cells that had been engaged in antibody-dependent cellular cytotoxicity (ADCC), so they can be further costimulated to more efficiently perform ADCC. Cell-surface CD137 is even more intensely expressed on FoxP3+ regulatory T cells (Tregs) dwelling in the tumor microenvironment (2), but its role in Treg modulation remains unclear.

When CD137 and CD137L genes were cloned and sequenced, they lacked any known intracellular signaling domains; therefore, they ought to rely on adaptors like other members of the TNFR family. Cytoplasmic interactions with TNF receptor–associated factor-1 (TRAF1), TRAF2, and perhaps TRAF3, in part inducible upon ligation, give rise to signals mediated by K63 polyubiquitination reactions that control canonical and non-canonical NF-κB activation, as well as the activation of MAPK and PI3K/AKT pathways. Perhaps, among all the cellular programs, the most prominent functional consequence is inhibition of apoptosis by down-regulating BIM and up-regulating Bcl-xL. CD137 costimulation also results in enhanced IL-2 and IFN-γ cytokine production upon TCR-CD3 ligation and favors memory T cell differentiation.

Ample evidence indicates that CD137L also signals back to the antigen-presenting cell through ill-defined mechanisms that cooperate with Toll-like receptor 4 (TLR-4). Such reverse signaling might be important in several inflammatory conditions and perhaps in antitumor immunity.


Experiments indicated that CD137 monoclonal antibodies (mAbs) could costimulate CD8 T cells in culture. In parallel, it was found that infusion of anti-CD137 mAb to mice bearing transplanted tumors resulted in enhanced tumor-specific CD8 T cell responses and, most interestingly, in rejection of tumors derived from a variety of mouse cancer cell lines (3).

At the time, it was confusing to see that the very same monoclonal antibodies presumed to be agonists ameliorated mouse autoimmune experimental conditions such as systemic lupus erythematosus or autoimmune encephalomyelitis. Of note, such experimental diseases were mediated by autoreactive CD4 T cells that seem to experience activation-induced cell death upon in vivo costimulation with the antibodies but not with the natural ligand. As an exception, mouse models of type 1 diabetes were exacerbated by CD137 stimulation, and a mild degree of CD8 T cell–mediated hepatitis was induced in anti-CD137 mAb–treated mice.

From the beginning, it was considered that forcing CD137L expression on tumor cells could lead to enhanced immunogenicity. This was indeed the case, but the therapeutic potency of the CD137L transfectants did not mirror the potency of antibody therapy. Subsequent studies using membrane-bound, single-chain variable fragment (scFv) antibodies rendered more pronounced effects, suggesting that tumor-targeting of CD137 agonists might result in potent therapeutic effects (4). Intriguingly, CD137 agonists synergized against mouse tumors with chemotherapy, vaccines, adoptive T cell therapy, anti–PD-(L)1 mAbs, and a variety of other immunotherapy modalities. CD137-bright Tregs can be depleted from the tumor microenvironment with anti-CD137 mAb treatment that spares effector T cells with dimmer CD137 expression. This can be followed by subsequent rounds with a purely agonistic antibody (2). To restrict the activity of CD137 mAb to the tumor microenvironment, intratumoral injections have been performed in mice and local injections of anti-CD137 combined with systemic infusion of the anti–PD-1 mAb nivolumab are being tested in the clinic (NCT03792724). In this issue, Claus et al. engineer a recombinant trimeric form of the CD137L to induce potent CD8 T cell costimulation, leading to abundant production of IFN-γ when CD137 signaling occurs in the presence of TCR-CD3 triggering in both mouse models and human samples (1).


A fully human IgG4 anti-CD137 mAb termed urelumab entered the clinic following a very good preclinical safety profile, including nonhuman primates. It showed single-agent activity in dose escalation phase 1, but the maximal tolerated dose was overestimated; a phase 2 trial in patients with advanced cancer had to be stopped because of liver inflammation severely affecting about one-third of the patients (5). As a result, antibody doses were reduced to 0.1 and 0.3 mg/kg, conferring a safe liver toxicity profile, but losing single-agent activity with the exception of few responses in patients with non-Hodgkin’s lymphoma (6). Ensuing phase 1b single arm combination of such low urelumab doses with the anti–PD-1 mAb nivolumab resulted in good tolerability and a promising overall response rate in patients with melanoma, even more evident when considering cases with fewer than 1% PD-L1+ tumor cells (NCT01471210 and NCT02253992).

Urelumab has shown agonist activity on human CD137, but reduced affinity to the cynomolgus homolog protein, perhaps explaining why liver safety problems were not observed in experimentation with nonhuman primates. An IgG2 anti-hCD137 mAb termed utomilumab also entered the clinic in spite of very weak intrinsic CD137 agonistic activity. Utomilumab is safe in patients with tolerability up to 10 mg/kg, even though it has very modest therapeutic activity only seen against Merkel cell carcinoma. Combination regimens with the anti–PD-1 mAb pembrolizumab were also found to be safe in phase 1 clinical trials (7).

Some of the most remarkable effects in clinic have been observed with two FDA-approved chimeric antigen receptor (CAR) T cells that encompass the CD137 cytoplasmic sequence in the signaling domain of the chimeric receptors. The most prominent effect of CD137 in CAR T cells, an ongoing revolution in the therapy of hematological malignancies, seems to be T cell persistence. Aside from antiapoptotic functions, recent evidence suggests that prominent invigoration of mitochondrial mass and functions (8), as well as chromatin remodeling (9), may underlie such functions that are also observed to be elicited with CD137 agonists.


The concern is to secure maximal efficacy of CD137 monotherapy while avoiding liver toxicity. The solutions pursued, and under clinical development, include tumor-targeting, intratumor injections (NCT03792724), bispecific/multispecific constructs, and inert probodies that become selectively unmasked in the tumor microenvironment (Fig. 1). All these strategies are effective in mouse models and show synergy with PD-1 blockade.

Fig. 1 Protein engineering strategies to target CD137 (4-1BB) costimulation to the tumor microenvironment.

Different approaches can be undertaken to target the effect of CD137 costimulation to the tumor microenvironment. This is considered important to avoid side effects in the form of hepatitis (observed in humans) or myelosuppression (observed in mice). Protein engineering approaches include targeting CD137 agonists to tumor-associated antigens, to proteins selectively expressed in the stroma such as fibroblast activated protein (FAP), or using antibodies targeting CD137 that become unmasked and active in the tissue microenvironment (probodies). In other instances, dual targeting of costimulatory and coinhibitory receptors enriched in the tumor microenvironment by a single multispecific protein is being attempted with potential for synergistic effects in a single moiety.


One of the first bispecific constructs targeting CD137 and the tumor-associated antigen EGFR showed efficacy and enhanced immune functions without noticeable liver side effects in a mouse model (10). This approach has the limitation of efficacy being expected only on tumors that express such tumor-associated antigens.

Claus et al. (1) designed a bispecific construct, including a trimeric form of CD137L and an antigen-targeting portion against fibroblast activation protein (FAP). FAP is a common moiety expressed on cancer-associated fibroblasts and is present in most tumors. The construct, FAP-4-1BBL, should bring T cells in contact with FAP-expressing cancer-associated fibroblasts and activate CD137 on the T cells. The authors confirmed that 4-1BB–related liver inflammation induced by anti-CD137 mAbs is highly dependent on FcR-mediated cross-linking and does not occur with the engineered compounds lacking an intact Fc. The construct was able to stimulate T cells from cancer patient biopsies, indicating that it could work on tumor-infiltrating lymphocytes. Biodistribution and liver safety were demonstrated in mice along with strong therapeutic activity with a bispecific antibody engaging another tumor antigen, CEA, and CD3. Most interestingly, the biodistribution of FAP-4-1BBL was studied in a monkey with spontaneous colon cancer, confirming that the construct actually targets tumor tissue. The concept is further proven with a similar construct that combines the trimeric form of CD137L with CD19 targeting mAbs in a model of B cell lymphoma. Provided that these compounds are not hampered by anti-drug antibodies or lack of penetration into malignant tissue in patients, the compound has excellent potential for clinical development. Importantly, there was also preclinical efficacy combining this construct and an anti–PD-L1 mAb in a colon cancer model.

FAP-41BB-L (RG7826) has entered the clinic in combination with anti–hPD-L1 mAb atezolizumab in a first-in-human trial that is also addressing biodistribution by PET tracing (EUDRACT CT#: 2017–003961-83). Preclinical data strongly encourage clinical testing of combinations with targeted CD3-engagers in a way that would be mimicking the physiology of CAR T cells. However, caution is required since such compounds may result in cytokine release syndrome and other toxicities that might be exacerbated by CD137 costimulation.

Creative biotechnology approaches to achieve bispecificity or selective localization/activation in the malignant tissue microenvironment seem to be the engines to move CD137-based immunotherapy toward clinical efficacy in the near future.


Acknowledgments: M.F.S. was supported by a Miguel Servet contract from Instituto de Salud Carlos III, Fondo de Investigación Sanitaria (Spain).
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article