Combining Antibody-Directed Presentation of IL-15 and 4-1BBL in a Trifunctional Fusion Protein for Cancer Immunotherapy

Immunomodulating cytokines have great potential in cancer immunotherapy (1, 2). Interleukin (IL)-15 and 4-1BB–directed agents have emerged here as promising candidates (3). IL-15 belongs to the cytokines of the common cytokine receptor γ chain (γc) family (4). It is involved in the generation, proliferation, and activation of natural killer (NK) cells, induces proliferation and differentiation of CD8⁺ T cells, and supports the survival of CD8⁺ memory T cells. Unlike IL-2, another clinical relevant member of this cytokine family, IL-15 rather inhibits activation-induced cell death (AICD) and seems not to influence regulatory T (Treg) cells (5). The antitumor potential of IL-15 was shown by using IL-15 gene–modified tumor cells (6), overexpression of IL-15 in transgenic mice (7), or the administration of recombinant IL-15 in several mouse models (8). In addition, the antitumor potential of recombinant IL-15 was shown to be further improved by complex formation with IL-15Rα-Fc (9) or the generation of a fusion protein with the fragment of the IL-15Rα chain involved in ligand binding (extended sushi domain) (RD_IL-15; ref. 10). Also, a fusion protein composed of IL-15, IL-15Rα's sushi domain, and apolipoprotein A-I (Apo A-I) has been reported, inducing therapeutic effects in lung and liver metastases mouse models (11). Furthermore, tumor targeting by antibody fusion proteins with IL-15 (12) or RD_IL-15 (13, 14) resulted in enhanced antitumor effects in different mouse models. On the other hand, improved therapeutic effects could be achieved by the combination of IL-15 with other immunomodulatory approaches, for example, cytokines (IL-21, IL-7, IL-12; refs. 6, 15, 16), agonistic anti-CD40 mAb (17), or inhibitory checkpoint blockers (anti-CTLA-4 mAb, anti-PD-L1 mAb; ref. 18). Thus, IL-15 holds great potential for combination therapies.

4-1BB is a costimulatory member of the TNF receptor superfamily (TNFR-SF) that is upregulated on activated T cells and NK cells (19). Costimulation by 4-1BBL/4-1BB interaction is fundamentally involved in the proliferation, differentiation, and survival of CD8⁺ T cells, therefore thought to play an important role in potentiating cytotoxic T-cell immune responses (20). Antitumor effects were reported in several mouse models with agonistic 4-1BB–specific antibodies (21, 22). Also, vaccination with tumor cells transfected to express a 4-1BB–specific antibody fragment (scFv) on the cell surface or admixture of them to wild-type tumor cells induced strong antitumor responses (23, 24). Furthermore, antibody fusion proteins with the extracellular domain of 4-1BBL showed target-mediated costimulation in vitro (25, 26) and antitumor activity in vivo (27). Improved therapeutic effects were achieved by the combination of 4-1BB costimulation with diverse cytokines [e.g., IL-12, granulocyte macrophage colony-stimulating factor (GM-CSF), IFN-α; refs. 28–30], other costimulatory receptor activation (OX40, HVEM, CD28; refs. 31–33), or inhibitory immune checkpoint blockade (anti-CTLA4 mAb; ref. 34). Thus, 4-1BB costimulation seems particularly suitable for combination therapies, too.

There is evidence that IL-15 and 4-1BBL act cooperatively in the immune response. IL-15 was shown to induce 4-1BB expression on CD8⁺ memory T cells in an antigen-independent manner, increasing their survival (35). Thus, both agents are linked in a model describing memory CD8⁺ T-cell homeostasis (36). On the other hand, successful ex vivo expansion of human NK cells has been achieved by stimulation with genetically modified K562 cells, expressing IL-15 and 4-1BBL. Hence, a potent and long-lived NK cell population could be generated, capable to eradicate leukemia in a xenograft mouse model (37). Therefore, immunotherapy involving the combination of IL-15 and 4-1BB costimulation should promote the induction of a strong antitumor response.

Here, we propose a targeted approach in form of a trifunctional antibody fusion protein to combine and deliver IL-15 and 4-1BBL simultaneously to the tumor. The antibody is directed against the fibroblast activation protein (FAP), a tumor stroma antigen expressed in more than 90% of breast, colorectal, and lung carcinomas (38). Targeting should hold advantages by modulating the activity and accumulate the cytokines at the tumor site, thus reducing the effective dosage and the risk of adverse events associated with a systemic administration. Moreover, application of the 2 cytokines in a single molecule format results in a spatiotemporal coordinated activity, which is expected to focus the respective activity of the two cytokines on the same cell, ensuring cooperativity. Hence, an improvement of the antitumor response could be expected. Therefore, we have generated a trifunctional fusion protein composed of a tumor-directed antibody in the scFv format, IL-15 joint to an IL-15Rα chain fragment and the extracellular domain of 4-1BBL. We investigated the immunostimulatory capacity of this fusion protein format in vitro and evaluated its antitumor potential in a mouse model in vivo.

Antibodies and recombinant proteins were purchased from Biolegends (mouse anti-human 4-1BBL-PE, rat anti-mouse 4-1BBL-PE, mouse anti-human CD3-PerCP), BD Biosciences (mouse anti-human CD3-PE), Immunotools (human recombinant IL-15), KPL [goat anti-mouse IgG (H+L)], Miltenyi Biotec (mouse anti-hexahistidyl-tag-PE), R&D Systems (human 4-1BB-Fc, mouse 4-1BB-Fc, mouse anti-human CD3ϵ mAb), Santa Cruz (mouse anti-human CD107a-FITC), and Sigma (goat anti-human IgG(Fc)-PE). DuoSet ELISA kit for human IFN-γ and CellTrace CFSE cell proliferation kit were obtained from R&D Systems and Life Technologies, respectively. B16-FAP (transfectants with human FAP) and B16wt (K. Pfizenmaier, IZI) were cultured in RPMI-1640, 5% FBS, supplemented with 200 μg/mL zeocin in the case of B16-FAP. CTLL-2 cells (P. Scheurich, IZI) were cultured in RPMI-1640, supplemented with 20% FBS, 10 mmol/L HEPES, 0.05 mmol/L β-mercaptoethanol, 1 mmol/L natriumpyruvate, nonessential amino acids, and 400 IU/mL rhIL-2. Cells were tested for mycoplasms and their morphologic appearance monitored by microscopic means. Antigen (FAP) expression on B16-FAP and cytokine growth dependence of CTLL-2 cells was verified by flow cytometry and proliferation assays, respectively. Human peripheral blood mononuclear cells (PBMC) were isolated from buffy coat of healthy donors [blood bank, Klinikum Stuttgart (Katharinenhospital)] and cultivated in RPMI-1640, 10% FBS. C57BL/6Jrj mice were purchased from Elevage Janvier. Animal care and experiments carried out were in accordance with federal guidelines and had been approved by university and state authorities.

Generation of scFv_RD_IL-15 (13), scFv_4-1BBL (25), and scDbFAPxCD3 (26) had been described previously. RD_IL-15_scFv_4-1BBL was cloned starting from these bifunctional fusion proteins by introducing the human RD_IL-15 N-terminally of the scFv_4-1BBL in the backbone vector pSecTagA (Life Technologies). RD_IL-15_scFv_m4-1BBL and scFv_m4-1BBL were obtained by replacing the extracellular domain of human 4-1BBL (aa 71–254) by the corresponding mouse 4-1BBL (aa 104–309). All recombinant proteins were produced in stably transfected HEK293 cells and were purified by immobilized metal ion affinity chromatography (IMAC) as described elsewhere (13). In brief, producer cells were expanded and grown to 90% confluence in RPMI 5% FBS before switching to serum-free Opti-MEM I medium (Life Technologies). Supernatants were collected and pooled. Proteins were concentrated by ammonium sulfate precipitation (60% saturation), before loading onto a nickel nitrilotriacetic acid column (Qiagen) previously equilibrated with PBS. After a washing step with 50 mmol/L sodium phosphate buffer, pH 7.5, 250 mmol/L NaCl, and 20 mmol/L imidazole, the recombinant fusion proteins were eluted with 50 mmol/L sodium phosphate buffer, pH 7.5, 250 mmol/L NaCl, and 250 mmol/L imidazole. Protein fractions were pooled and dialyzed against PBS. Integrity and purity of the recombinant proteins were determined by SDS-PAGE and the identity was corroborated by Western blotting.

A total of 2 × 10⁵ B16-FAP cells were incubated with each fusion protein for 2 hours at 4°C. FAP-bound protein was detected by incubation for 1 hour at 4°C with either phycoerythrin (PE)-conjugated anti-hexahistidyl-tag antibody (scFv_RD_IL-15) or anti-4-1BBL as well as 4-1BB-Fc followed by anti-huIgG(Fc) antibody, respectively (scFv_4-1BBL, RD_IL-15_scFv_4-1BBL). Washing and incubation steps were carried out in PBS, 2% FBS, and 0.02% sodium azide. Cell analysis was conducted in an EPICS FC500 (Beckman Coulter), and data were analyzed using FlowJo (Tree Star).

IL-15 activity was assessed on the cytokine growth-dependent cell line CTLL-2. Therefore, 2 × 10⁴ CTLL-2 cells per well were seeded and starved for 4 hours, before addition of the respective fusion proteins. After 3 days, cell proliferation was measured by MTT assay (13).

Human PBMCs were stained with carboxyfluorescein diacetate succinimidyl ester (CFSE) at a concentration of 625 nmol/L/1 × 10⁶ cells/mL, following the instructions of the manufacturer. Subsequently, 2 × 10⁵ PBMCs per well were applied in each assay and T-cell proliferation (anti-CD3-PE/CFSE) measured by flow cytometry. Proliferation of PBMCs was assessed in response to the fusion proteins in targeted and nontargeted forms. In the former case, 2 × 10⁴ B16-FAP cells per well were seeded and arrested the next day by incubating with mitomycin (10 μg/mL) for 2 hours at 37°C. After washing, cells were incubated for 1 hour at room temperature with the corresponding fusion protein at the indicated concentrations, followed by another washing step and the addition of PBMCs. Also, the effect of nontargeted fusion protein was assessed. Therefore, B16wt cells were seeded and arrested as indicated above. Then, fusion protein was added together with PBMCs. In either case, blocking of the 4-1BBL–mediated effect of the trifunctional fusion protein (10 nmol/L) was achieved by addition of the recombinant receptor 4-1BB-Fc (80 nmol/L). Furthermore, to analyze the effect on activated T cells, cross-linked anti-CD3 mAb (0.01 μg/mL) was added to the targeted and nontargeted setting. Cross-linking of the anti-CD3 mAb was achieved by previous incubation with goat anti-mouse IgG antibodies at a ratio of 1:3.

PBMC stimulation in response to fusion proteins was also analyzed in terms of IFN-γ release. Therefore, the experimental setting with targeted fusion protein as described for the PBMC proliferation assays was conducted. Here, supernatant was removed 5 days after PBMC addition and concentration of IFN-γ determined by sandwich-ELISA following the instructions of the manufacturer's protocol.

The cytotoxic potential of T cells was assessed in terms of degranulation and target cell killing. Fusion proteins at different concentrations were incubated with 2 × 10⁵ PBMCs per well for 5 days, either targeted to arrested B16-FAP cells (see above) or without target cells. Then, PBMCs were transferred to a plate with freshly seeded B16-FAP cells and T cells retargeted and triggered by incubation with 30 pmol/L bispecific antibody scDb(FAP×CD3) for 6 hours in the presence of 1.4 μL/well GolgiStop (monensin, BD Biosciences). Subsequently, PBMCs were harvested and T-cell degranulation measured by flow cytometry (CD107a-FITC/CD3-PerCP). In parallel, T-cell cytotoxicity was determined, analyzing B16-FAP cell viability by MTT assay (13).

Therapeutic efficacy of the recombinant proteins was assessed in a syngeneic B16-FAP lung metastasis mouse model. C57BL/6JRj mice (female, 4 months) were injected i.v. with 8.5 × 10⁵ B16-FAP cells per mouse on day 0. Treatment with (i) PBS, (ii) scFv_RD_IL-15, (iii) scFv_m4-1BBL, and (iv) RD_IL-15_scFv_m4-1BBL (0.02 nmol fusion protein/animal) was administrated intraperitoneally (i.p.) on days 1, 2, and 10. Mice (6 mice/group) were sacrificed on day 21. Lungs were removed, fixed in formaldehyde, and metastases counted.

For comparison of multiple groups, the one-way ANOVA followed by the Tukey posttest was applied, using the GraphPad Prism software (GraphPad Software). P < 0.05 was considered to be significant.

Before the generation of the trifunctional fusion protein with human cytokines, the feasibility to improve immune cell response by combining IL-15 stimulation and 4-1BBL costimulation had to be shown in vitro. Therefore, B16-FAP cells were cocultured with human PBMCs in the presence of cross-linked anti-human CD3 mAb, scFv_4-1BBL, and recombinant IL-15. Indeed, the combined application of human IL-15 and targeted human 4-1BBL led to an enhancement in PBMC proliferation and IFN-γ release (Supplementary Fig. S1), providing a rational base to assess the immune-stimulating capacity of a corresponding trifunctional fusion protein. The trifunctional recombinant protein RD_IL-15_scFv_4-1BBL was generated by the genetic fusion of human IL-15 (IL15) linked to an IL-15Rα (IL15RA) fragment (aa 31–107, containing the IL-15 binding sushi domain), an antibody in the single chain Fv (scFv) format directed against the FAP, a hexahistidyl tag and the extracellular domain of human 4-1BBL (TNFSF9; aa 71–254; Fig. 1A). As 4-1BBL is a member of the TNF superfamily, noncovalent homotrimer formation of the ligand and therefore the fusion protein was expected, resulting in a molecule with 3 subunits of RD_IL-15 and scFv, respectively. The bifunctional antibody fusion proteins used for reference were scFv_RD_IL-15 and scFv_4-1BBL. They were also directed against FAP (same antibody) and had been described previously (refs. 13, 25; Fig. 1A). Analysis by SDS-PAGE under reducing conditions revealed a band of 76 kDa for RD_IL-15_scFv_4-1BBL (Fig. 1B). Considering predicted N-glycosylation for IL-15 at position 127, this correlates with the calculated molecular mass of 71 kDa of the monomer. Furthermore, size exclusion chromatography (SEC) of RD_IL-15_scFv_4-1BBL showed a main peak at approximately 240 kDa, corroborating TNFSF ligand–mediated trimer formation (Fig. 1C). The latter was also shown for scFv_4-1BBL, whereas scFv_RD_IL-15 was present as a monomer and a small dimer fraction.

Antibody-mediated binding of the fusion protein was analyzed by flow cytometry (Fig. 2A and B). Here, the trifunctional and the bifunctional fusion proteins showed similar binding capacity to FAP-expressing target cells (B16-FAP; Fig. 2B). No binding was detected on B16 wild-type cells (Fig. 2A). In functional assays using the IL-15 (mouse/human) growth-dependent mouse cell line CTLL-2, RD_IL-15_scFv_4-1BBL was observed to be approximately 4-fold less active than scFv_RD_IL-15 (Fig. 2C). As expected, no activity was measured for scFv_4-1BBL (human 4-1BBL is not cross-reactive with mouse 4-1BB; Fig. 2C). Finally, ligand–receptor interaction of the human 4-1BBL compound with the corresponding human 4-1BB receptor was confirmed by flow cytometry (Fig. 2D). Therefore, binding of recombinant 4-1BB-Fc to target-bound RD_IL-15_scFv_4-1BBL and scFv_4-1BBL was detected via PE-conjugated anti-IgG(Fc) antibody. Thus, all 3 components of the fusion protein RD_IL-15_scFv_4-1BBL proved to be functional.

Human PBMC response was assessed in the presence of the fusion protein in targeted and nontargeted forms on B16-FAP cells and B16wt cells, respectively. PBMCs were applied either unstimulated or prestimulated with cross-linked anti-CD3 mAb. In the presence of CD3-stimulated PBMCs, proliferation could be efficiently enhanced by all 3 fusion proteins in targeted form (Fig. 3A). Here, RD_IL-15_scFv_4-1BBL showed similar activity than scFv_RD_IL-15 and stronger activity than scFv_4-1BBL. In nontargeted form, the RD_IL-15_scFv_4-1BBL fusion protein was slightly less active than scFv_RD_IL-15 (Fig. 3B). As expected, nontargeted scFv_4-1BBL showed no activity, consistent with previous reports (25). When experiments were carried out with unstimulated PBMCs, the IL-15 containing fusion proteins were able to initiate T-cell stimulation, whereas scFv_4-1BBL alone did not, according to its costimulatory nature (i.e., requirement of first signal; Fig. 4). In nontargeted form, RD_IL-15_scFv_4-1BBL and scFv_RD_IL-15 exerted at 10 nmol/L similar, but rather limited proliferation inducing activity (Fig. 4A). The situation changed completely, when the fusion proteins were presented in targeted form. Here, RD_IL-15_scFv_4-1BBL was clearly more active than scFv_RD_IL-15, inducing stronger T-cell proliferation (up to 2.6-fold; Fig. 4B) and IFN-γ release (up to 5.9-fold; Fig. 4C), thus effectively promoting T-cell activation. The participation of 4-1BBL in the signal enhancement was shown by a blocking assay with recombinant 4-1BB-Fc receptor, reducing the proliferation induced by RD_IL-15_scFv_4-1BBL to the level of that retrieved by scFv_RD_IL-15 only (Fig. 4D). Interestingly, the proliferation induced by RD_IL-15_scFv_4-1BBL could not be reached by the combination of the respective bifunctional fusion proteins (scFv_RD_IL-15 and scFv_4-1BBL; Fig. 4D). In summary, on unstimulated PBMCs, the activity of the trifunctional fusion protein in nontargeted form seems restricted to the IL-15 module, whereas 4-1BBL remains inactive. The situation changes in the targeted form, where 4-1BBL becomes active. Under these conditions the 4-1BBL module enhances the IL-15 activity, thus conferring higher immune stimulating activity to the trifunctional fusion protein in comparison to the corresponding bifunctional IL-15 fusion protein.

Furthermore, the impact of the fusion proteins on the cytotoxic potential of T cells was analyzed (Fig. 5). Therefore, unstimulated PBMCs were incubated for 5 days with the fusion proteins either targeted to B16-FAP cells or without target cells. Subsequently, PBMCs were transferred to a fresh plate and T cells retargeted to B16-FAP cells via a bispecific antibody (scDbFAP×CD3). Thus, T cells were triggered and their cytotoxic potential determined by measuring T-cell degranulation and tumor cell killing. In targeted form, RD_IL-15_scFv_4-1BBL and scFv_RD_IL-15 showed comparable capacity in enhancing the cytotoxic potential of T cells (Fig. 5A), whereas in nontargeted form, at low concentrations, the effect of RD_IL-15_scFv_4-1BBL was slightly reduced in comparison to scFv_RD_IL-15 (Fig. 5B). In addition, the enhancement of the cytotoxic potential of the T-cell population showed to correlate with an increase in concomitant tumor cell killing, corroborating the antitumor potential of RD_IL-15_scFv_4-1BBL and scFv_RD-IL-15 in vitro (Fig. 5C, D).

To evaluate the antitumor potential of the RD_IL-15_scFv_4-1BBL in a syngeneic tumor mouse model, the fusion protein had to be adjusted for mouse compatibility. Therefore, the extracellular domain of human 4-1BBL was replaced by the corresponding mouse domain (m4-1BBL). As RD_IL-15 is species cross-reactive in human and mice, no adjustment was necessary for this module of the fusion protein. The resulting trifunctional fusion protein RD_IL-15_scFv_m4-1BBL retained binding and cytokine activity (Supplementary Fig. S2). The syngeneic B16 lung metastasis model has been used previously by us (13) and other groups (9, 10, 39) to evaluate immunotherapeutic treatments with IL-15–based reagents. B16-FAP cells were injected i.v. into C57BL/6 mice, and animals were treated with RD_IL-15_scFv_m4-1BBL, scFv_RD_IL-15, or scFv_m4-1BBL applied i.p. on days 1, 2, and 10. After 3 weeks, lungs were removed and tumor metastases counted. Significant antitumor effects were obtained with both, RD_IL-15_scFv_m4-1BBL and scFv_RD_IL-15 (Fig. 6). However, the best therapeutic result was obtained by treatment with the trifunctional molecule RD_IL-15_scFv_m4-1BBL. Here, the number of metastases was reduced approximately to half the number of metastases grown after treatment with scFv_RD_IL-15. Thus, superior antitumor activity could be shown for the trifunctional fusion protein RD_IL-15_scFv_m4-1BBL in vivo.

We have reported here a trifunctional antibody fusion protein that combines a tumor-directed recombinant antibody, RD_IL-15, and 4-1BBL into a single molecule. Each component retained its functionality, thus simultaneous antibody-mediated presentation of the IL and the costimulatory ligand of the TNF superfamily on target cells, accompanied by their respective activity, was achieved. Previous studies with trifunctional antibody fusion proteins had focused mainly on combinations involving the cytokines IL-12, IL-2 and GM-CSF, using different tumor targets (EpCAM, Her2/neu, CD30), antibody formats (whole IgG, heterominibody, scFv-Fc), and cytokine combinations (IL-2/IL-12, IL-12/GM-CSF, IL-2/GM-CSF; refs. 40–44). Although activity of the respective antibody and cytokine components was shown in each case, the antitumor effect of trifunctional and the respective bifunctional antibody cytokine fusion proteins (alone or in combination) was strongly dependent on the molecular design, cytokine combination, and tumor mouse model. Thus, not only the feasibility of the concept but also the particular challenge of implementation became evident. For members of the TNFSF, only a trifunctional antibody fusion protein with TNFα and IL-12 has been described so far (45). In this case, IL-12 was fused to the N-terminus and TNFα to the C-terminus of a scFv targeting the extradomain B of fibronectin (ED-B). According to the structural properties of TNFα, assembling of a homotrimer was expected. Antibody binding and bioactivity of the fusion protein in untargeted form was shown in vitro. Unfortunately, biodistribution studies with the fusion protein in an immunocompetent 129SV mice bearing s.c. grafted F9 teratocarcinoma failed to show tumor accumulation and therefore further therapeutic studies were not attempted. Thus, although an active fusion protein was engineered, the functional requirements for the in vivo model were not fulfilled.

Here, we report for the first time a trifunctional antibody fusion protein combining human 4-1BBL, another member of the TNFSF, and human IL-15, a member of the common γc receptor cytokine family, crucial for proliferation of effector T cells. Also in this case, TNFSF ligand–mediated homotrimer formation was expected and confirmed. Antibody-mediated targeting properties were not impaired by the molecular design of this trifunctional fusion protein. Thus, the same functional affinity was shown for the trifunctional and the bifunctional, also homotrimeric scFv_4-1BBL fusion protein, which was also similar to that of the mainly monomeric scFv_RD_IL-15. For the latter, improved binding capacity of the antibody fusion protein in comparison to the scFv alone had been described previously (13). Higher avidity in terms of RD_IL-15 units seemed not to confer further advantages to the trifunctional molecule. Thus, in untargeted form, RD_IL-15_scFv_4-1BBL in comparison to scFv_RD_IL-15 showed similar or even lower activity on PBMC proliferation. This might be partially influenced by the position of the cytokine in the trifunctional molecule, as RD_IL-15_scFv in comparison to scFv_RD_IL-15 was recently shown to have slightly lower activity on PBMC and CTLL-2 proliferation (Supplementary Fig. S3). Remarkably, in the trifunctional fusion protein, 4-1BBL retained its characteristic bioactivity feature. Thus, in the nontargeted form of the fusion protein, the 4-1BBL remained inactive and only in targeted form, that is, after scFv-mediated cell surface presentation, activity was observed in terms of an enhancement in IL-15–induced T-cell proliferation and IFN-γ release. This property has been described for several other members of the TNFSF (25, 46, 47) and was confirmed here for scFv_4-1BBL. Targeting-dependent acquisition of bioactivity might confer an advantage to the trifunctional fusion protein, considering potential unwanted side effects of systemically active cytokines.

Under physiologic conditions, 4-1BBL and IL-15 are presented on the cell surface, modulating the immune response upon cell–cell contact. Therefore, 4-1BBL is expressed as a transmembrane protein (48), whereas IL-15 is presented in trans by the IL-15Rα chain on the cell surface (49). It was shown that the activity of IL-15 is strongly enhanced in the context of IL-15Rα chain presentation (50). Thus, IL-15×IL15RαFc complexes or fusion proteins of IL-15 with the extended sushi domain of IL-15Rα (RD) showed to be very effective in enhancing NK cell and T-cell proliferation and increasing antitumor effects in diverse mouse models (9, 10, 39, 51). Furthermore, tumor-directed antibody fusion proteins with RD_IL-15 showed improved antitumor effects in comparison to untargeted RD_IL-15 in different mouse models (13, 14). The strong potential of RD_IL-15 was also demonstrated and shown to be predominant in the trifunctional fusion protein of the present study. Thus, the activity of RD_IL-15_scFv_4-1BBL and scFv_RD_IL-15 were comparable in enhancing proliferation of activated T cells and promoting the expansion of cytotoxic T cells in vitro. The benefit of targeted 4-1BBL–mediated costimulation became apparent under suboptimal activation conditions, that is, when the stimulatory activity of IL-15 was limited. Of note, simultaneous presentation of RD_IL-15 and 4-1BBL in one molecule was required, as the combination of equimolar amounts of the bifunctional antibody fusion proteins did not retrieve the cooperative effect observed for the targeted trifunctional antibody fusion protein. Thus, the targeted 2-in-1 cytokine fusion protein constellation apparently disposes of structural characteristics that favor cooperative activity. Encouraged by the antitumor effect observed in the animal experiment after treatment with the mouse compatible equivalent of the trifunctional fusion protein, following in vivo studies will have to focus now on the evaluation of the antitumor response of immune effector cell subpopulations as well as pharmacokinetic and pharmacodynamic properties of the fusion protein to determine the eligibility of such a trifunctional molecule for clinical translation. So far, clinical development focuses on recombinant IL-15 in phase I/II trials [acute myelogenous leukemia (AML), malignant melanoma, and solid tumors; ref. 2] and 2 agonistic human 4-1BB–specific monoclonal antibodies (PF-05082566 and BMS-663513) in phase I (non–Hodgkin lymphoma) and phase I/II (melanoma/solid tumors), respectively (52).

Modulations of the immunologic conditions at the tumor site are gaining increasing interest, as it has been recognized that solid tumors constitute an important place for the initiation and development of an antitumor immune response. For example, it was shown in mouse models that naïve CD8⁺ T cells can infiltrate tumors and become activated and differentiated into functional effector cells in situ (53). Studies in mice provided evidence that localized activity of IL-15 and 4-1BB in the tumor microenvironment was relevant for the respective antitumor response (54, 55). Thus, in a Rag1^−/− mouse model, IL-15–secreting tumors were grown in the presence of neutralizing IL-15 antibody and then after withdrawal of the antibody, eradicated by an NK cell–mediated immune response. Co-expression of IL-15Rα by the tumor cells was needed for the efficient induction of the densely granulated NK effector cells in the tumor microenvironment. Moreover, the regression of the IL-15–secreting tumors did not stop the growth of contralateral non–IL-15-secreting control tumors, suggesting that the effect of IL-15 was largely restricted to the local microenvironment of the IL-15–secreting tumor (54). On the other hand, in another mouse model, it was shown that the hypoxic conditions in the tumor induced the upregulation of 4-1BB on tumor-infiltrating lymphocytes (TIL), which could be selectively activated by local, that is, intratumoral application of 4-1BB–specific monoclonal antibodies, promoting the development of an effective antitumor response without causing liver side effects. (55). Thus, local copresentation of IL-15 and 4-1BBL in the tumor microenvironment might be of especial value to support an antitumor immune response.

In summary, the feasibility and benefit of the trifunctional antibody fusion protein concept could be shown for the targeted combination of RD_IL-15 and 4-1BBL. These 2 cytokines display remarkably different structural and functional properties that could be successfully bundled, enhancing immune cell stimulation in vitro and antitumor response in vivo.

R.E. Kontermann is a consultant/advisory board member of BioNTech. No potential conflicts of interest were disclosed by the other authors.

Conception and design: D. Muller

Development of methodology: V. Kermer, D. Muller

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): V. Kermer, M. Harder, A. Bondarieva

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): V. Kermer, M. Harder, D. Muller

Writing, review, and/or revision of the manuscript: V. Kermer, N. Hornig, R.E. Kontermann, D. Muller

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): V. Kermer

Study supervision: D. Muller

The authors thank Robert Lindner for technical support on the size exclusion chromatography analysis.

This work was supported by a grant from the Deutsche Forschungsgemeinschaft (MU 2956/2-1; D. Müller).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.