Dose-limiting toxicities are thought to temper the efficacy of single-agent 4-1BB agonists. To overcome this hurdle, in this issue of Cancer Discovery, Muik and colleagues report preclinical and clinical studies describing a first-in-class bispecific fusion protein targeting 4-1BB and PD-L1.
See related article by Muik et al., p. 1248 (9).
4-1BB (CD137) is expressed on multiple populations of immune cells and notably on activated but not naïve T cells (1–5). 4-1BB is a member of the TNF receptor superfamily, and its engagement leads to unique signaling dependent on TRAF1/TRAF2 and NF-κB. The downstream consequence of 4-1BB engagement leads to positive effects on T cells including proliferation, functional activation, and possibly reversal of exhaustion. These properties have led to the inclusion of the 4-1BB signaling domain in a number of chimeric antigen receptor constructs, some of which are the basis of FDA-approved cell therapy products. While natural 4-1BB signaling can be induced by the 4-1BB ligand, agonist antibodies have been developed. In preclinical studies, such antibodies can induce potent antitumor immunity, which are further improved by combination with PD-(L)1–targeted therapies. As 4-1BB has been reported to be selectively upregulated on tumor-reactive T cells, the use of 4-1BB agonists is thought to be an attractive means to selectively engage the T cells with most antitumor potential. These discoveries led to the clinical development of agonist antibodies against 4-1BB, including two of which have been tested in patients with cancer. Results for urelumab (BMS-663513, IgG4) were first reported in 2008, and it has now been evaluated in a number of clinical trials. While showing evidence of clinical antitumor activity, dose-dependent, on-target liver toxicities due to activation of liver myeloid cells have limited urelumab's clinical development. Results for a second anti–4-1BB antibody, utomilumab (PF-05082566, IgG2), were first reported in 2014. Although utomilumab appeared to have a more favorable safety profile compared with urelumab, lack of clinical activity has dampened enthusiasm. Collectively, these studies indicate that although a promising pathway for augmenting antitumor activity, agonist anti–4-1BB antibodies suffer from dose-limiting toxicities.
Bispecific antibodies or molecules offer the possibility of achieving superior antitumor activity with a concomitant reduction in toxicity (6–8). In the case of anti–4-1BB antibodies, a bispecific approach may allow for sequestering of the agonist activity to a specific location, such as the tumor, thereby avoiding toxicities associated with systemic localization of traditional anti–4-1BB antibodies. A second advantage of a bispecific antibody may be the ability to direct context-dependent cross-linking or clustering of receptors, which is thought to be important for appropriately activating lymphocytes for antitumor activity. Muik and colleagues take advantage of these principles and create a bispecific antibody, designated GEN1046, with antigen binding regions targeting both PD-L1 and 4-1BB (9). Importantly, the authors demonstrate in vitro that 4-1BB signaling occurs only in the presence of PD-L1+ target cells, suggesting that receptor cross-linking is essential for activity. Furthermore, in contrast to controls, GEN1046 increased the number of T-cell and dendritic cell contacts. GEN1046 also exhibited the 4-1BB–independent ability to block the interaction of PD-1/PD-L1, showing that the PD-L1–specific binding may both direct localization of the molecule and facilitate function as a traditional immune-checkpoint drug. Using in vitro experiments with activated T cells, the authors also found that GEN1046 increased proliferation and cytotoxicity, as well as induced cytokine expression including IFNγ. Thus, GEN1046 has a 4-1BB–dependent ability to directly modulate T-cell activity.
To test the preclinical utility of GEN1046 as an antitumor therapeutic, the authors used a syngeneic mouse model with double knock-in for human PD-L1 and human 4-1BB, and MC38 tumors transfected with human PD-L1. GEN1046 induced potent antitumor immunity, which was accompanied by T-cell infiltration and functional activation. Notably, GEN1046-dependent tumoral T-cell infiltration was dependent on the tumor expressing human PD-L1, thereby again demonstrating the critical importance of PD-L1 binding. In addition to experiments in mice, the authors also evaluated whether GEN1046 modified the expansion of human tumor-infiltrating lymphocytes (TIL) from non–small cell lung cancer tumors and found potentially favorable alterations in the expansion kinetics of TILs, including potential GEN1046-mediated expansion of tumor-reactive T cells identified using T-cell receptor (TCR) sequencing. Finally, in preparation for human studies, the authors also conducted non-human primate studies. In monkeys, GEN1046 was well tolerated. The binding affinities of GEN1046 to recombinant cynomolgus monkey PD-L1 and 4-1BB were similar to the affinity to human PD-L1 and 4-1BB, thus supporting the relevance of these findings to humans.
The authors also reported first-in-human results in 61 patients with advanced solid tumors treated with GEN1046 enrolled in the dose-escalation portion of their phase I/IIa trial (NCT03917381). The most common tumor types were colorectal (n = 12; 19.7%), ovarian (n = 9; 14.8%), pancreatic (n = 6; 9.8%), and non–small cell lung cancer (n = 6; 9.8%), and patients were heavily pretreated with a median of three prior treatments. Although treatment-related adverse events (AE) occurred in 43 (70.5%) patients, only 17 patients (27.95%) had at least one treatment-related grade 3–4 AE. While there was some evidence of liver toxicities, the authors demonstrated a manageable safety profile that compared favorably with urelmuab. While six patients (9.8%) had dose-limiting toxicities, the maximum tolerated dose was not reached. Although an early-stage trial, there was disease control in 40 (65.6%) of the 61 trial patients and overall a favorable response rate. Interestingly, pharmacodynamic analysis showed greater immune responses at lower dose levels, which is consistent with the complicated nature of a bispecific antibody that depends on binding to both 4-1BB and PD-L1. In summary, this first-in-human trial suggests that GEN1046 is a safer agent than urelumab, and furthermore, there was evidence of antitumor activity.
The results reported here are important for several reasons. Foremost, this represents the first publication describing the use of a bispecific antibody targeting 4-1BB and PD-L1 in humans. Although other investigators have conceptually shown that 4-1BB/PD-L1 is an attractive target based on preclinical data, the advancement of this concept into the clinic is significant and reflective of a trend toward the development of drugs not simply targeting one pathway. These results are also notable as the authors provide data suggesting that it is possible to dissociate the toxicities associated with single-agent anti–4-1BB antibody with retention of the beneficial properties of engaging the 4-1BB pathway. Conceptually, there are several aspects of GEN1046 worth considering from a mechanistic perspective (Fig. 1). Foremost, GEN1046 may facilitate the conversion of a PD-L1+ tumor to a 4-1BB agonist+ tumor. Thus, the PD-L1 inhibitory signal is not only removed but converted into a stimulatory signal. Furthermore, as GEN1046 is dependent on PD-L1 expression, 4-1BB agonistic activity may be restricted to PD-L1hi locations such as the tumor or lymph node. Another important mechanistic component of GEN1046 is its ability to induce inflammatory cytokines such as IFNγ that upregulate PD-L1, thereby allowing GEN1046 to generate a positive feedback loop. Although there is a lot of promise for GEN1046, even if it does not meet future clinical endpoints, this study illustrates the possibilities of therapeutically converting any inhibitory molecule to an immune agonist with an appropriately designed bispecific molecule.
A number of areas warrant attention for future studies. Foremost, although there is an encouraging efficacy signal, further studies will be needed to demonstrate that clinical activity is not simply a consequence of blockade of the PD-1/PD-L1 pathway and rather 4-1BB–dependent. A second area of importance relates to the identification of biomarkers such as tumoral PD-L1 that might predict response or toxicities. Such biomarker studies may be especially important given the complex pharmacokinetics whereby a bispecific antibody needs to be at a concentration that appropriately engages both target moieties, which creates complexities not apparent with traditional antibodies. Another area yet to be resolved is whether GEN1046 might induce antidrug antibodies (ADA) and if so whether such ADAs might limit efficacy. Also important for future studies is assessment of liver toxicity in real-world settings. Preclinical mechanistic studies on the potential impact of GEN1046 on liver myeloid cells may warrant further study. It will also be informative to broaden the study to more preclinical cancer models especially those that do not express high levels of PD-L1, or are “cold” tumors with no or low levels of tumor-infiltrating T lymphocytes, to provide better guidance for the future clinical development of this interesting agent. Finally, some cells, including regulatory T cells, might express both 4-1BB and PD-L1, and the impact of GEN1046 binding on such cells would be of interest to know (10).
Although this GEN1046 study is important, it is also reflective of a significant ongoing effort to optimally target 4-1BB agonistic activity for therapeutic intervention. Other 4-1BB and PD-(L)1 bispecific and related molecules are in clinical development (Supplementary Table S1). These drugs have differences in their construction including in the choice of antibody clones and how such clones are linked. Unlike conventional antibodies, these molecules may be quite construct-specific in their optimal activity. It will thus be of interest to compare results from those studies with those of GEN1046. It is also of note that there are clinical studies with other bispecific and related molecules seeking to combine 4-1BB agonistic activity with different targeting molecules beyond PD-L1 (Supplementary Table S2). Results from all these studies will be important for making informed decisions to understand existing 4-1BB bispecific molecules and to facilitate the design of future drugs.
Although the specific path forward may need to be defined, it is clear that there is now a momentous shift in the development of new antitumor immunotherapeutic agents. Although utilization of single target agents either alone or in combination is clearly important, it is apparent that the biological activity achievable with the creation of bispecific molecules will uncover novel opportunities. In the case of GEN1046, the potential ability to avoid systemic toxicities by localizing 4-1BB to a PD-L1–rich environment and also to convert PD-L1–positive tumors to 4-1BB agonist–positive tumors offers great potential for improving outcomes for patients with cancer.
Z. Li reports other support from Heat Biologics, Alphamab, Henlius, Ikonisys, and HanchorBio outside the submitted work. M.P. Rubinstein reports a patent for WO2018/064255 pending and licensed to XOMA and a patent for US10377988B2 issued. No disclosures were reported by the other author.