RG7386, a Novel Tetravalent FAP-DR5 Antibody, Effectively Triggers FAP-Dependent, Avidity-Driven DR5 Hyperclustering and Tumor Cell Apoptosis

This article is featured in Highlights of This Issue, p. 765

Programmed cell death (apoptosis) is a crucial process for normal development, tissue homeostasis, and the immune response in multicellular eukaryotic organisms. Impaired apoptosis, in turn, plays a key role in cancer pathogenesis and contributes to poor chemotherapy responses. Thus, development of agents that restore/activate apoptotic pathways specifically in tumor cells has emerged as an important therapeutic strategy. The extrinsic apoptotic pathway is triggered by external signals, including proapoptotic FAS-ligand and TRAIL, which bind to and activate cell surface-anchored death receptors (DR) such as FAS, DR4, or DR5 through receptor oligomerization. DR activation promotes the formation of the death-inducing signaling complex, leading to subsequent activation of caspases 3, 6, and 7 and cell apoptosis (1, 2). Several proapoptotic receptor agonists (PARA), including mAbs, have been developed to target this extrinsic apoptotic pathway (3, 4). However, despite promising preclinical data, none of these approaches have thus far demonstrated significant clinical efficacy (5–7). Although some antibodies appear to be able to autonomously activate DR4 and DR5, many require additional cross-linking to achieve optimal activity in vitro (8–11). In vivo, and as demonstrated for drozitumab, a fully human DR5 agonistic antibody (8), this cross-linking activity is likely provided by activatory or inhibitory Fcγ receptors (FcγR) on leukocytes in the tumor microenvironment (12, 13). Low or variable infiltration of FcγR-expressing leukocytes, reduced targeting of DR agonistic antibodies due to systemic binding to DRs on nontumor cells or to FcRs on circulating cells, and/or competition for FcRs by high levels of endogenous immunoglobulins, may account for the lack of clinical efficacy observed for these agonistic antibodies.

Fibroblast-activation protein (FAP) is a homodimeric, single-pass type II membrane protein expressed in reactive stromal fibroblasts of more than 90% of epithelial malignancies, including breast, colorectal, and lung cancers and on malignant mesenchymal cells of bone and soft tissue sarcomas (14–16). Here, we exploited FAP expression on the tumor stroma to develop novel bispecific antibodies (BsAb) consisting of a DR5 agonist and anti-FAP antibody fusion to enhance targeted delivery of the DR5 agonist, induce avidity-driven DR5 hyperclustering and activate the extrinsic apoptotic pathway specifically in tumor cells. We demonstrate that FAP-DR5 antibodies significantly enhance tumor cell apoptosis in vitro across several tumor indications compared with the parental drozitumab mAb. Importantly, the antitumor activity of the FAP-DR5 antibodies was dependent on the presence of FAP-expressing fibroblasts. Using an optimized FAP-DR5 BsAb RG7386, we further show that the mode of action of the DR5 agonist can be uncoupled from FcγR-dependent cross-linking when fused to the FAP-binding moiety. RG7386 demonstrates in vivo efficacy in several xenograft models, including patient-derived xenograft (PDX) models with FAP expression either on stroma or on malignant cells. Finally, RG7386 in combination with irinotecan or doxorubicin enhances the antitumor effects of the respective monotherapies in vivo and results in complete tumor remission.

HEK293 EBNA, NIH-3T3, MDA-MB-231, DLD-1, Colo205, and LOX-IMVI were obtained from LGC Standards (European distributor for the ATCC), A549 from ECACC, GM05389 from the Coriell Institute for Medical Research (Camden, NJ), and 1833-PPOP233 from Promega Corp. All cell lines were purchased between 2007 and 2010 and authenticated by morphologic and growth curve analysis. For more details see Supplementary Materials and Methods.

The construction of expression vectors for standard IgGs (immunoglobulin G) and BsAbs was performed according to standard recombinant DNA technologies. Each antibody chain cDNA was amplified by PCR or generated by conventional gene synthesis at Geneart and transiently expressed in HEK293 EBNA cells under the MPSV promoter (myeloproliferative sarcoma virus). Bispecific FAP-DR5 antibodies were generated by fusion of a FAP-binding domain to the DR5 IgG heavy chain at the C-terminus via a (G₄S)₄ linker. The DR5 portion consisted of the variable-light chain (VL) and variable-heavy chain (VH) of drozitumab (described by Adams in US2007/003141401 A1) or novel DR5 antibodies generated by phage display. To minimize light-chain mispairing side-products, we used CrossMab technology (17). The FAP-binding unit was engineered as a crossed Fab in which the VH was fused to the constant-light chain (CL) and the VL to a CH1 (constant-heavy 1) domain. To prevent FcR binding, three mutations in the Fc region [PGLALA: P329G (patent No. WO 2012130831) and L234A and L235A (18)] were generated by standard site-directed mutagenesis in the indicated BsAbs. All amplified or synthesized genes were confirmed by DNA sequencing.

Antibodies with specificity for human DR5 were selected from a generic phage-display antibody library in the Fab format. Specific binders were identified by ELISA assays and clones exhibiting signals on human DR5 but none on human DCR2 and human IgG1 were short-listed for further analyses. All BsAbs and antigens (if not obtained from a commercial source) were transiently produced in HEK293 EBNA cells and purified using protein A affinity purification and size exclusion chromatography. The purity and molecular weight of the BsAb preparations were determined by Microchip capillary electrophoresis (CE)-SDS on Caliper's LabChip GXII using the Protein Express LabChip Kit. For more details see Supplementary Materials and Methods.

Simultaneous binding of BsAbs to human DR5 as well as hu or muFAP was assessed by surface plasmon resonance (SPR) using Biacore T100. The binding of BsAbs on cells was determined using a time-resolved fluorescence resonance energy transfer (TR-FRET) assay. In the flow cytometry, anti-FAP and anti-DR5 were used as a primary and goat anti-human Fc specific PE-labeled as a secondary antibody. For more details see Supplementary Materials and Methods.

GM05389 fibroblasts (1 × 10⁴ cells/well), harvested with Cell Dissociation Buffer and resuspended in appropriate medium, were plated overnight at 37°C in 96-well plates. Fibroblasts were then incubated with antibodies (anti-DR5 IgGs or FAP-DR5 BsAbs) alone or together in a 1:1 ratio with a goat anti-human IgG cross-linking antibody (Sigma #I2136) at the indicated concentrations for 10 minutes. Tumor cells (1 × 10⁴ cells/well) were harvested in Cell Dissociation Buffer, resuspended in appropriate medium and were added to the fibroblasts. Cells were incubated with the antibodies for 24 hours at 37°C followed by centrifugation for 10 minutes at 200 × g at room temperature. The amount of fragmented DNA was determined using Cell Death detection ELISA^PLUS (Roche Applied Science #11 774 425 001) according to the manufacturer's instructions. Where indicated V_max values were calculated from absorbance readings taken every minute over 20 minutes at 405 nm using a Versamax Microplate Reader (Molecular Devices). Measurements at 490 nm were used as a reference. In other experiments, end-point absorbance readings were taken at 405 nm using a Tecan Spectra Rainbow Reader.

All experiments were conducted according to the guidelines of the German Animal Welfare Act (GV-Solas; Felasa; TierschG).

For the cell line–based xenograft models, tumor cells (2 × 10⁶ cells) were co-injected s.c. onto Balb/c nude mice (Charles River Laboratories) together with murine fibroblasts (NIH3T3, 0.4 × 10⁶ cells). Cell inoculation was supplemented with Matrigel (Basement Membrane Matrix, BD Biosciences). The fragment-based colorectal cancer xenograft Co5896 and sarcoma tumor xenograft Sarc4605 originally obtained from patients were passaged approximately three to five times until establishment of stable growth patterns. For the subsequent in vivo PDX studies, tumors were removed from xenografts in serial passage in donor nude mice, cut into fragments (4–5 mm diameter), and placed in PBS until subcutaneous implantation in recipient mice.

Mice (n = 10/group) were subjected to stratified randomization based on primary tumor size (median tumor volume of approximately 100–150 mm³) before treatment. Unless otherwise indicated, FAP-DR5 BsAb RG7386, drozitumab and drozitumab-PGLALA were administered once weekly i.v. Irinotecan (15 mg/kg) was administered intraperitoneally, doxorubicin (5 mg/kg) was given weekly intravenously, and saline was used as the vehicle control. Tumor volume was measured by caliper twice weekly [(length × width²)/2] and the percentage of tumor growth inhibition (TGI) relative to control animals was calculated. All treatments were well tolerated as indicated by stable body weights throughout the duration of the studies.

1833-PPOP233 cells (4 × 10⁶ cells/mouse) and NIH3T3 cells (1 × 10⁶ cells/mouse) were subcutaneously coinjected into the right flank of 6- to 8-week-old Scid/beige mice (Charles River Laboratories) with body weights of 18 to 24 g. After treatment with BsAbs, mice were injected with a single dose of d-luciferin (Promega; 150 mg/kg in PBS; i.p.) 10 minutes before bioluminescence imaging using IVIS spectrum (Caliper Life Sciences). The acquired bioluminescence signals in a region of interest (ROI) over the tumor site were measured as average radiance (p/sec/cm³/sr) with Living Image software (Caliper Life Sciences). Every ROI signal was divided by the appropriate baseline ROI signal to calculate fold induction (19, 20).

IHC protocols for cleaved caspase-3 and FAP were performed on Discovery or BenchMark XT automated strainers. All reagents except the antibodies were obtained from Ventana Medical Systems. For more details see Supplementary Materials and Methods.

Nonparametric treatment-to-control-ratios (TCR) based on endpoint analysis and the two-sided nonparametric confidence intervals (CI) compared with the vehicle group were calculated to assess statistical significance in in vivo studies. A TCR value of 0 indicates complete tumor regression and CIs of less than 1 represent statistically significant results. Relative TGI was calculated according to the formula (1−[(T−T₀)/(C−C₀)]) × 100.

To overcome the disadvantages of current DR5 antibodies, and to boost antitumor efficacy, we engineered novel BsAbs that simultaneously bind to DR5 on tumor cells and to FAP on the tumor stroma to induce localized DR5 hyper–cross-linking. The BsAbs consisted of a human anti-DR5 agonist (drozitumab-derived huIgG1) with a C-terminal fusion of an anti-FAP binding moiety (mAb007 or affinity-matured mAb082 clone; Fig. 1A). For heterodimerization of this trivalent molecule, the “knobs-into-holes” technology (21) was implemented. All FAP-drozitumab antibodies exhibited similar binding to human DR5 as the parental drozitumab and a nontargeting BsAb (drozitumab fused to non-binding DP47GS Cross Fab as a control), demonstrating that the process of antibody fusion had not impaired antibody binding to the DR5 antigen in cells (Fig. 1B; Supplementary Table S1). Interestingly, we found that the Fab configuration used to fuse FAP to drozitumab was important for preserving BsAb interaction with the FAP antigen. In a VLCH1 format, mAb007_drozitumab demonstrated reduced binding to FAP (>9-fold), compared with parental mAb007 antibody. Binding to FAP was similarly reduced by a BsAb with a monovalent FAP-binding domain (Fig. 1C; Supplementary Table S1), corresponding to a loss in avidity. In contrast, in the VHCL format, mAb007_drozitumab and mAb082_drozitumab demonstrated similar binding to FAP (K_i = 0.32 nmol/L and K_i = 0.53 nmol/L, respectively) that was comparable with the parental antibodies (Fig. 1C; Supplementary Table S1). These data highlight that the VHCL Fab configuration is the most suitable format for generating BsAbs.

Importantly, SPR experiments confirmed simultaneous binding of the FAP-drozitumab antibodies to both DR5 and FAP recombinant proteins, indicative of a fully functioning dual antibody (Supplementary Fig. S1A).

To investigate the antitumor efficacy and specificity of the FAP-drozitumab BsAbs, tumor cell apoptosis (DNA fragmentation) was monitored in the presence or absence of the FAP-positive human GM05389 fibroblast cell line (Supplementary Fig. S1B).

In the absence of FAP-expressing fibroblasts, low concentrations of drozitumab or BsAb resulted in minimal MDA-MB-231 breast cancer cell apoptosis. However, DNA fragmentation was observed at the highest antibody concentrations used (7 and 35 nmol/L; Fig. 1D–I). Induction of drozitumab cross-linking by addition of a secondary anti-human Fc antibody promoted dose-dependent tumor cell apoptosis with a significant 100-fold reduction in EC₅₀ values from 15.2 nmol/L (drozitumab alone) to 0.14 nmol/L (cross-linked drozitumab; Fig. 1D and E; Supplementary Table S2), demonstrating that cross-linking is critical for antibody efficacy.

In the presence of FAP-expressing fibroblasts, all BsAbs substantially increased tumor cell apoptosis (Fig. 1F–I). Under these conditions, BsAb treatment resulted in dose-dependent cell death at concentrations as low as 0.0112 nmol/L with maximum activity at 7 nmol/L. In addition, the EC₅₀ value of the bispecific molecules was approximately 20-fold lower as compared with the anti-Fc cross-linked drozitumab (0.014 vs. 0.33 nmol/L; Supplementary Table S2). Importantly, the parental drozitumab antibody did not trigger apoptosis in the same MDA-MB-231 and fibroblast coculture assay without cross-linking. These data clearly implicate the requirement for FAP expression in the tumor stroma for bispecific FAP-drozitumab antibody function. Interestingly, the antibody configuration did not appear to hinder the in vitro potency of the bispecific molecules (Fig. 1G and H; Supplementary Table S2) despite differential binding to the FAP antigen (Fig. 1B and C). However, it appeared that antibody valence for the FAP antigen was crucial for BsAb potency (EC₅₀ value increased from 0.01 in the bivalent molecule to 0.57 nmol/L in the monovalent format; Supplementary Table S2). Altogether, the BsAb containing the mAb007 anti–FAP-binding moiety appeared to be less specific and induced tumor cell apoptosis in the absence of fibroblasts at lower concentrations. Therefore, the mAb082 anti–FAP-binding moiety was adopted for further BsAb development and in vivo experiments.

The FAP dependency of mAb082_drozitumab to trigger DR5 clustering and apoptosis was further evaluated in the colorectal cancer DLD-1 cell line. In the presence of NIH3T3 fibroblasts that lack FAP expression in vitro (Supplementary Fig. S1B), the antibody failed to trigger tumor cell apoptosis (Supplementary Fig. S1C). In contrast, stable expression of murine FAP in NIH3T3 cells (muFAP NIH3T3; Supplementary Fig. S1B) resulted in significant antibody-mediated tumor cell death (Supplementary Fig. S1D). Moreover, the untargeted control molecule DP47GS_drozitumab, which does not bind FAP, failed to induce apoptosis even in the presence of FAP-expressing cells (Supplementary Fig. S1C–S1D). Together, these data demonstrate the fundamental requirement for FAP-positive cells in cross-linking of the FAP-drozitumab BsAbs and in the activation of the DR5 apoptotic pathway.

To assess the generalizability of the FAP-drozitumab antibody, five tumor cell lines were evaluated for DR5 cell surface expression and apoptosis sensitivity in response to BsAb treatment. Flow-cytometry analyses revealed two cell lines with high (ACHN; renal carcinoma and DLD-1 colorectal cancer), two with moderate (Colo205; colorectal cancer and MDA-MB-231 breast cancer), and one cell line with low DR5 expression (A549; lung carcinoma; Fig. 2A). All of these tumor cell lines demonstrated susceptibility to FAP-drozitumab antibody-mediated apoptosis (Fig. 2B) in a dose-dependent manner. Moreover, BsAb action required FAP targeting as the untargeted BsAb control (DP47GS_drozitumab) failed to induce apoptosis at low doses (Fig. 2B). As previously discussed, the apoptotic activity of DP47GS_drozitumab at higher concentrations could be attributed to the action of drozitumab alone (Fig. 1D). Intriguingly, and in agreement with a previous publication (22), no obvious correlation was apparent between DR5 cell surface expression and FAP-drozitumab BsAb-induced apoptosis (Fig. 2; Supplementary Table S3).

Altogether, these data indicate the potential for the broad application of FAP-drozitumab BsAbs in a range of tumor indications and highlight the importance of FAP targeting for BsAb function.

Conventional anti-DR5 antibodies such as drozitumab rely on FcγR-dependent mechanisms to achieve DR5 clustering (13). Therefore, efficient DR5 activation may be hindered by several factors, including high concentrations of endogenous IgG, present in patient sera, and by the activity of FcγR-expressing immune cells within normal tissue leading to limited antitumor efficacy and unwanted side-effects. In addition, FcγR-dependent cross-linking in the liver has been shown to mediate hepatotoxicity for murine FAS agonistic antibodies (23), and a similar mechanism may play a role in the hepatic side-effects reported for previous DR5 agonists in the clinic (24). Moreover, our data suggested that the parental drozitumab antibody triggers cross-linker–independent cytotoxicity at concentrations above 1 nmol/L (Fig. 1D). To eliminate these confounding variables and to improve future clinical safety profiles, we generated a BsAb composed of another DR5 antibody mAb096, which does not trigger apoptosis in the absence of cross-linking (Supplementary Fig. S2A). Furthermore, to prevent interaction with host FcγR on host immune cells, we introduced previously reported IgG1 point mutations [PGLALA: P329G (patent No. WO 2012130831) and L234A and L235A (18)] in mAb096 that completely abolish FcγR and C1q binding while still retaining cross-linking activity by anti-Fc antibodies.

In the absence of fibroblasts, this optimized FAP-DR5 BsAb (mAb096-mAb082 in VHCL configuration) named RG7386 (characterized in Supplementary Table S4 and Supplementary Fig. S2B), promoted MDA-MB-231 tumor cell apoptosis only with the addition of a secondary anti-Fc antibody similarly to the parent mAb096–anti-Fc combination (Fig. 3A). Importantly, the BsAb displayed more potent tumor cell apoptosis in the presence of FAP-expressing GM05389 fibroblasts (EC₅₀ = 0.3 nmol/L) compared with the cross-linked parent molecule (EC₅₀ = 5.3 nmol/L) and the addition of the anti-Fc secondary antibody had no additive effects on RG7386-mediated MDA-MB-231 tumor cell death in cocultures (EC₅₀ = 0.5 nmol/L; Fig. 3B). In addition, RG7386 had minimal effects on fibroblast cell survival in monocultures (Fig. 3C). These data demonstrate the strong and specific antitumor effects of RG7386 in response to FAP-mediated cross-linking.

The antitumor efficacy of RG7386 was further investigated in vivo in xenograft mouse models. To obtain comparable FAP expression levels with those observed in the human tumor stroma, FAP-expressing NIH3T3 cells were s.c. coinjected with DLD-1 tumor cells onto nude mice and FAP expression in the tumor microenvironment was confirmed with IHC (Fig. 4A). Dose-optimization studies revealed effective TGI following once weekly administration of RG7386 at 3 and 10 mg/kg relative to vehicle control (Supplementary Fig. S3A). In terms of pharmacokinetics, the exposure to RG7386 (AUC_0–72h) appeared to increase in a dose-dependent manner with exposure of the largest dose (10 mg/kg) being somewhat higher than expected (Supplementary Fig. S3B).

Importantly, treatment with RG7386 (10 mg/kg) significantly inhibited tumor growth (TCR, 0.31; 95% CI, 0.17–0.66; TGI = 89%) whereas drozitumab-PGLALA (mutant drozitumab unable to be cross-linked without addition of a secondary anti-Fc antibody), failed to have any effect on tumor size as compared with vehicle-treated mice (TCR, 1.075; 95% CI, 0.65–2.43; TGI = 22%; Fig. 4A). As RG7386 also lacks these FcγR-binding sites, these data strengthen the notion that FAP-dependent cross-linking drives the antitumor effects of this FAP-DR5 BsAb. Importantly, the in vivo efficacy of the FAP-DR5 antibody complex to inhibit tumor growth was recapitulated using the previous mAb082_drozitumab molecule in the same DLD-1 colorectal cancer model (TCR, 0.12; 95% CI, 0.00–0.33; TGI > 100% compared with vehicle control) and in the MDA-MB-231 breast cancer model (TCR, 0.34; 95% CI, 0.24–0.47; TGI: 77% compared with vehicle control; Supplementary Fig. S4), demonstrating the potential broad application of these bispecific molecules. Moreover, drozitumab coupled to the nontargeting DP47GS molecule did not affect tumor growth, further confirming the FAP-dependent targeting action of the FAP-DR5 antibodies to inhibit tumor growth in vivo (Supplementary Fig. S4).

To determine whether the RG7386-mediated suppression of tumor growth was specifically a result of DR5-activated tumor cell apoptosis and not through an unidentified mode of antibody action, cell apoptosis was monitored noninvasively in a human tumor xenograft mouse model by bioluminescence imaging. The 1833-PPOP233 cells (human bone-metastasizing MDA-MB-231 subclone) expressing a split-luciferase construct with a caspase 3/7 cleavage site (20), together with FAP-expressing NIH3T3 cells were s.c. inoculated in mice. Cleavage of the luciferase construct, resulting in a detectable luminescence signal, was assessed over time and used as a measure of apoptosis. RG7386 treatment (1.0 and 10 mg/kg) of tumor-bearing mice resulted in a rapid and significant increase in the luminescence signal indicative of apoptosis induction downstream of activated caspase-3/7 (Fig. 4B). Importantly, the enhanced luminescence signal denoting increased apoptosis is in agreement with inhibition of tumor growth, in the same mouse model, following RG3786 treatment (Fig. 4C). Increased caspase-3 activation (brown staining) was also confirmed in the vicinity of stromal bands by IHC staining of RG7386-treated DLD-1 / 3T3 coinjected tumor sections (Fig. 4D). These data clearly demonstrate that RG7386 induces tumor cell apoptosis in vivo to restrict tumor growth.

To further validate the in vivo efficacy of RG7386 in a more relevant disease background, a patient-derived fragment-based colorectal cancer tumor (Co5896) was s.c. transplanted onto NMRI nude mice and FAP expression was confirmed with IHC analysis (Fig. 5A). RG7386 treatment (30 mg/kg, twice weekly) of these mice led to a significant decrease in tumor burden (TCR, 0.412; 95% CI, 0.24–0.84; TGI = 76%; Fig. 5B). Interestingly, comparison of treatment with control groups revealed that a combination of RG7386 and the standard-of-care irinotecan (chemotherapeutic agent) translated into superior efficacy (TCR, 0.00; 95% CI, 0.00–0.004; TGI > 100%) relative to either RG7386 (TCR, 0.19; 95% CI, 0.05–0.37) or irinotecan monotherapy (TCR, 0.036; 95% CI, 0.02–0.09) alone. Direct comparison between each treatment group further confirmed that the enhanced efficacy of RG7386 and irinotecan combination was more than additive (TCR, 0.00; 95% CI, 0.00–0.019, and TCR, 0.00; 95% CI, 0.00–0.017, compared with RG7386 and irinotecan monotherapies, respectively) and resulted in prolonged antitumor activity with complete eradication of tumors in all mice (Fig. 5C). Together, these data demonstrate the potential cumulative effects of RG7386 when used with the standard-of-care for the treatment of human colorectal cancers.

Thus far, we have focused on the role of RG7386 in initiating tumor cell apoptosis with a complete dependence on the presence of FAP-positive fibroblasts. However, FAP expression is not restricted to the tumor stroma and can be found as a minor or major component of malignant cells in mesenchymal-derived tumors. To investigate whether RG7386 could be used to directly target FAP-positive tumors, LOX-IMVI desmoplastic melanoma cells were s.c. injected onto mice and tumors were allowed to develop. Treatment with 10 mg/kg of RG7386 triggered a significant inhibition of tumor growth compared with vehicle control (TCR, 0.073; 95% CI, 0.03–0.32) and compared with drozitumab-PGLALA (TCR, 0.27; 95% CI, 0.17–0.48), indicating that FAP-dependent cross-linking contributes to enhanced efficacy (Fig. 6A). Moreover, RG7386, administered as a single agent, substantially restricted tumor development (TCR, 0.16; 95% CI, 0.06–0.36; TGI, 99%) in an osteosarcoma Sarc4605 fragment-based PDX model with 100% FAP expression (Fig. 6B). In the same model, complete tumor remission (TGI > 100%) was achieved when RG7386 was combined with the standard-of-care doxorubicin (5 mg/kg; TCR, 0.03; 95% CI, 0.007–0.09; TCR, 0.15; 95% CI, 0.04–0.54 and TCR, 0.06; 95% CI, 0.02–0.14 compared with vehicle, RG7386, and doxorubicin as single agents, respectively; Fig. 6C). Altogether, these data support the use of RG7386 not only in the treatment of multiple carcinomas associated with a FAP-positive stroma but also as a potential viable therapy for mesenchymal tumors directly expressing FAP.

Many cancer therapeutic approaches target the specific molecules/signaling pathways that promote the hyperproliferative/antiapoptotic states in tumor cells. The continuing challenge in these methodologies, as opposed to conventional chemotherapeutics that target all rapidly dividing cells, is to be able to specifically discriminate between tumor cells and their healthy counterparts without loss in efficacy. For conventional DR5 agonistic antibodies, such as drozitumab and conatumumab, efficacy is strongly dependent on cross-linking that is mediated by FcγRs in in vivo mouse models (11, 13, 25). However, despite very convincing and potent in vitro and in vivo preclinical data, this mode of agonistic antibody action has severe drawbacks that may explain the limited clinical efficacy observed for first-generation DR agonists (5, 26–28).

In mouse xenograft models, significant tumoricidal activity of DR5 agonists is dependent on sufficient infiltration of FcγR-expressing immune cells. However, this situation does not appear to be relevant in humans, where FcγR-mediated antibody cross-linking may be reduced by physiologic IgG levels. On the other hand, potential adverse events such as elevated hepatotoxicity are related to nontumor specific, systemic crosslinking of agonistic antibodies via FcγRIIb molecules present in the liver (23). Consequently, second-generation DR agonistic antibodies are currently under development with a focus on both DR cross-linking to enhance efficacy, and on reducing the risk of toxicity due to non–tumor-targeted effects. Conatumumab-coated nanoparticles encasing a cytotoxic drug that allow a sufficient density of antibody paratopes at the tumor site to activate DR5 are currently under investigation (29). APG350 is a hexavalent scTRAIL–RBD (single-chain TRAIL receptor–binding domain), which in contrast with previous agonistic antibodies triggers multimer formation independently of FcγRs (30).

In addition to full-length antibodies, different scaffolds such as atrimers (31) or multivalent monobodies (32, 33) are under evaluation to target DRs. However, the lack of clinical efficacy observed with untargeted DR5 agonistic antibodies illustrates that the activity of these molecules in vitro may not necessarily translate into clinical efficacy. Multivalent DR5 binders might provide a more potent effect as compared with agonistic antibodies, but these also bear the risk of toxic side effects. TAS266, a novel tetravalent and highly potent DR5 agonistic Nanobody, was recently entered into a phase I clinical trial, but was deemed unsafe for human administration due to acute hepatotoxicity (34) that may have been exacerbated by an absence of a tumor-targeting entity. The DR ligand, TRAIL, untargeted, targeted, or in multimeric formats (35–38) also has some disadvantages such as a short half-life and potential toxicity, if oligomerized. In addition, as TRAIL does not only bind to DR4 and DR5, but also to decoy receptors DCR1, DCR2, and OPG (osteoprotegerin; ref.39) in vivo tumor activity may be limited.

In our study, we eliminated FcγRs-dependent formation of antibody multimers by introducing three mutations (PGLALA) in the Fc portion of a novel DR5 agonist (mAb096) that abolish FcγR binding. In addition, we fused the DR5 agonistic antibody to an anti–FAP-binding moiety to take advantage of FAP expression in the tumor stroma as an alternative mechanism to promote hyperclustering of DR5 in a targeted manner. The tumor-associated stroma contributes to a significant proportion of the mass of many malignancies (14) and the stromal fibroblasts play an important role in the development, growth and metastasis of carcinomas (40–42). FAP, normally restricted in healthy adult tissue (except in granulation tissue, e.g., during wound healing), is expressed in tumor stromal fibroblasts of more than 90% of epithelial cancers (15, 16, 43), and therefore is an ideal candidate for targeted delivery of cancer therapeutics. The FAP-DR5 BsAbs we described here, including RG7386, exhibited high specificity and a clear dependency on FAP-based cross-linking to promote tumor cell death. In vitro, apoptosis induction of tumor target cells only occurred in the presence of FAP-expressing fibroblasts and replacement of the FAP CrossFab unit with a non-binding IgG (DP47GS) was sufficient to prevent tumor cell apoptosis. In addition, the FAP-dependent mode of BsAb action successfully translated into in vivo efficacy in several xenograft models, including a colorectal cancer PDX and a sarcoma PDX model. The antitumor effects observed in the sarcoma PDX model also indicate that RG7386 function is not restricted to an environment consisting of a FAP-positive stroma, but can also be applied to mesenchymal-derived tumors that express FAP directly on malignant cells.

The broad application of our FAP-DR5 bispecific molecules was evident in five different tumor cell lines (lung, renal, colorectal, and breast cancer cells). However, the degree of apoptosis was different for each cell line and did not necessarily correlate with DR5 expression levels, indicating that there are additional factors influencing this mechanism. In some cells the DR-initiated signal requires amplification to induce maximum apoptosis. Correspondingly, we demonstrate increased inhibition of tumor growth in patient-derived colorectal cancer and sarcoma models in the presence of both RG7386 with irinotecan or with doxorubicin, respectively, which may be indicative of synergistic extrinsic and intrinsic apoptosis signaling (7, 44). It is, however, important to note that p53, an important tumor suppressor and a key regulator of the intrinsic pathway, is inactivated in many human cancers and the high prevalence of p53 mutants (dominant negative forms) contributes significantly to chemotherapeutic resistance (44–46). As a consequence, activation of the extrinsic pathway offers an attractive target for cancer therapy as the apoptotic cascade is p53 independent, and hence not susceptible to development of apoptosis resistance (47).

Taken together, our preclinical data strongly suggest that RG7386, a novel FAP-DR5 tetravalent BsAb, is superior to existing DR5 agonists as it offers a targeted strategy to specifically activate the extrinsic apoptosis pathway in tumor cells by avidity-driven DR5 hyperclustering. Future studies will be essential for clinical development of RG7386 particularly for the successful application of combination therapies.

C. Klein has ownership interest (including patents) in Roche. No potential conflicts of interest were disclosed by the other authors.

Conception and design: P. Brünker, K. Wartha, T. Friess, S. Grau-Richards, B. Weiser, H. Niu, K. Packman, C. Klein, P. Umaña

Development of methodology: P. Brünker, T. Friess, S. Grau-Richards, M. Majety, V. Runza, H. Niu, K. Packman, N. Feng, S. Daouti, E. Mössner, T.G. Weber, W. Scheuer, H. Sade, C. Shao, B. Liu, P. Wang, C. Klein, P. Umaña

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Friess, S. Grau-Richards, I. Waldhauer, C.F. Koller, M. Majety, H. Niu, N. Feng, S. Daouti, R.J. Hosse, E. Mössner, T.G. Weber, F. Herting, H. Sade, C. Shao, B. Liu, P. Wang, G. Xu, S. Vega-Harring, K. Bosslet, P. Umaña

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): P. Brünker, K. Wartha, T. Friess, S. Grau-Richards, I. Waldhauer, C.F. Koller, B. Weiser, M. Majety, V. Runza, H. Niu, S. Daouti, R.J. Hosse, E. Mössner, T.G. Weber, F. Herting, W. Scheuer, H. Sade, C. Shao, G. Xu, S. Vega-Harring, P. Umaña

Writing, review, and/or revision of the manuscript: P. Brünker, T. Friess, S. Grau-Richards, I. Waldhauer, C.F. Koller, M. Majety, V. Runza, H. Niu, F. Herting, S. Vega-Harring, C. Klein, K. Bosslet, P. Umaña

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): P. Brünker, S. Grau-Richards, B. Weiser, E. Mössner, F. Herting, C. Shao, S. Vega-Harring, K. Bosslet, P. Umaña

Study supervision: T. Friess, M. Majety, H. Niu, F. Herting, K. Bosslet, P. Umaña

Other (conducted preclinical studies and histologic analysis of explanted tumor tissue): W. Scheuer

All authors were employees of Roche at the time of this study. The authors thank Aurexel Ltd. (www.aurexel.com) for editorial support funded by Roche.

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.