Selective Tumor Cell Apoptosis and Tumor Regression in CDH17-Positive Colorectal Cancer Models using BI 905711, a Novel Liver-Sparing TRAILR2 Agonist

Tumor necrosis factor–related apoptosis-inducing ligand (TRAIL) is a homotrimeric protein that binds to and activates specific cell surface-anchored death receptors, TRAIL-receptor 1 (TRAILR1) and TRAIL-receptor 2 (TRAILR2). Ligand binding leads to receptor-mediated activation of the initiator caspase-8, which subsequently cleaves and activates the effector caspases-3 and -7 to execute apoptotic cell death (1). TRAIL induces apoptosis in a broad range of cancer cell lines and xenograft models, without exhibiting significant systemic toxicity in mice or nonhuman primates (2, 3), driving interest in exploring the potential of the TRAIL pathway as an anticancer therapy.

Several pro-apoptotic agonists against TRAILR2, including monoclonal antibodies, have been reported. Despite promising preclinical data, none of the TRAILR2 agonistic antibodies demonstrated significant clinical efficacy (4–10). Initial TRAILR2 antibodies required a secondary crosslinking mechanism to enhance their inherently weak agonistic properties (11, 12). Novel strategies were used to induce a more potent and direct activation of TRAILR2. One, a novel TRAILR2 agonist tetramer nanobody, TAS266, was able to bind TRAILR2 across several trimeric receptors leading to TRAILR2 hyper-clustering. This compound demonstrated excellent preclinical in vitro and in vivo activity (13); however, a clinical trial with TAS266 was terminated early due to drug-induced liver toxicity (14), suggesting that the liver is inherently sensitive to potent TRAILR2 activation in humans. These data prompted the development of compounds capable of potently inducing TRAILR2 oligomerization and apoptosis in tumor cells while sparing liver cells.

CDH17 is a cell surface molecule member of the Cadherin superfamily of adhesion molecules. In contrast with classical cadherins, its cytoplasmic domain is very short (20 amino acids) and apparently lacks binding sites necessary for interaction with the cytoskeleton. Under physiological conditions, CDH17 is proposed to regulate the direction and efficiency of epithelial water transport via trans-interaction with cadherins of neighboring cells (15). CDH17 is overexpressed in the plasma membrane of colorectal and gastrointestinal adenocarcinomas (16) and has been reported as both a tumor biomarker and oncogene (17). Notably, CDH17 is not expressed in mouse or human liver (18).

In this report, we characterize BI 905711, an optimized CDH17:TRAILR2 bispecific molecule that binds to CDH17 and uses it as an anchor to trigger potent and tissue selective TRAILR2 activation. Accordingly, we demonstrate that BI 905711 is highly efficacious in several colorectal xenograft models and its efficacy is strictly dependent on CDH17 expression.

Cells were grown in cell culture flasks until 70%–80% confluence using ATCC-recommended media. Cells were cultured at 37°C and 5% CO₂, except when Leibovitz's L-15 medium was used. GP2d CDH17 knockout (KO) cells were generated using the CRISPR/Cas9 system. Briefly, three different Cas9 genome editing constructs encoding gRNAs for CDH17 were obtained from GenScript. CDH17 CRISPR gRNA constructs were transfected to GP2d cells using the X-tremeGENE HP DNA transfection reagent kit (Roche, 0636236001) following the manufacturer's instructions. Three days post-transfection, cells were cloned. OVCAR-3–mKate2 cells were generated by transduction with Lentiviral-based labeling reagents (Incucyte NucLight Red, 4625) and cultured in RPMI medium supplemented with 10% FBS plus 1 μg/mL Puromycin (Sigma, 8833).

Irinotecan was purchased from Accord Healthcare GmbH (20 mg/mL, PZN 12422462). TAS 266′ tetrameric nanobody expression vector was produced as described in WO 2011/098520, SEQ ID No: 032. The sequence source for Lexatumumab' was the World Health Organization, as listed in the Recommended international nonproprietary names, list 57 (19). BI 905711 sequence is described in WO 2018/115231, SEQ ID No: 212 and SEQ ID No: 218. The bispecific negative control was generated as a BI 905711-based bispecific molecule where the CDH17 building block was substituted by an irrelevant sequence. All molecules were produced at Boehringer Ingelheim.

BI 905711 was expressed transiently in the CHO-E/pTT5 system. Transient expression titer ranged from 30 to 60 mg/L in 10 days. Final sample was 98.9 % monomer ± 1.0% by analytical SEC. The endotoxin level in the purified sample was determined by LAL cartridges (Charles River Laboratories); a standard release threshold of <1 EU/mg of protein was applied. Molecular weight of intact BI 905711 was observed to be 201 KDa, analysis using LC-TOF-MS.

Cells were seeded at 2–3 × 10⁶ cells in T80 (ThermoFisher, 178905) and cultured overnight at 37°C and 5% CO₂. The next day, cells were detached with Versene solution (Gibco, 15040033), centrifuged at 300 × g for 5 minutes and resuspended in FACS buffer containing phosphate-buffered saline (PBS; Gibco, 20012–019), 1% FBS, and 0.01% sodium azide (Sigma, 71289). Approximately 2 × 10⁵ cells in 50 μL were aliquoted per well into round-bottom 96-well plates (Costar, 3799), followed by adding anti-CDH17 or anti-TRAILR2 (produced at Boehringer Ingelheim according to WO 2010/123874, SEQ ID No: 038 and SEQ ID No: 049 and WO 2002/094880, SEQ ID No: 029 and SEQ ID No: 031), or human IgG (Sigma, I5154) at final concentration of 1 μg/mL. Staining was performed on ice for 30 minutes. Cells were washed once with cold FACS buffer, and further incubated with 100 μL per well of allophycocyanin (APC)-conjugated anti-human IgG Fc Antibody (1:40 pre-diluted in FACS buffer; Biolegend, 409306) on ice for another 30 minutes. Finally, cells were washed twice with FACS buffer and analyzed with a BD FACSCanto II system (BD Biosciences). A minimum of 10,000 events per well were collected using FACSDiva software and analyzed using FlowJo version 10.1. Wells containing buffer only were included as unstained controls.

The cells were plated at 1,000–7,000 cells in 50 μL per well in white opaque 96-well microtiter plates (PerkinElmer, 6005680) and incubated overnight at 37°C and 5% CO₂ (except for cells grown in Leibovitz's Medium). On the following day, serial dilutions of the corresponding agonists were added and cells were then transferred back to incubator and incubated for an additional 24 or 72 hours as indicated. Cell viability was measured using the CellTiter-Glo luminescent cell viability kit (Promega, G7571) according to the manufacturer's instructions. Luminescent signal was recorded with EnSight multimode plate reader (PerkinElmer). Average background values were subtracted from each measurement.

COLO 205 cells were plated at 5,000 in 50 μL per well in white opaque 96-well microtiter plates (PerkinElmer, 6005680) and incubated overnight at 37°C and 5% CO₂. On the following day, agonists were added and cells were incubated for an additional 2, 4, 6, or 16 hours as indicated. Caspase 8 or caspase 3/7 was determined using the Caspase-Glo 8 kit (Promega, G8200) or Caspase-Glo 3/7 (Promega, G8090), respectively. Luminescent signal was recorded with EnSight multimode plate reader (PerkinElmer). Average background values were subtracted from each measurement.

DLD-1 and OVCAR-3-mKate2 cells were pre-mixed in a 3:1 ratio (6,000:2,000 cells), 1:1 ratio (4,000:4,000 cells) or 1:3 ratio (2,000:6,000 cells) in suspension and plated in 100 μL per well in clear-bottom 96-well microtiter plates. In additional plates, OVCAR-3–mKate2 cells alone were seeded at 4,000 cells in 100 μL per well and used as control. After overnight incubation at 37°C, agonists at a 3-fold of the final concentration were added in 50 μL into each well. The plates were then placed in the IncuCyte S3 Live-Cell Analysis System (Sartorius) for image acquisition. Images in experimental wells were captured every 2 hours in both phase contrast and the red fluorescent channels. By the use of the integrated software, the red fluorescent nuclei counts were measured to represent the viable OVCAR-3–mKate2 cell number in each well.

GP2d, COLO 205, and DLD-1 tumors were established by subcutaneous injection of tumor cells in 6-week-old athymic female BomTac:NMRI-Foxn1nu purchased from Taconic, Denmark. Standardized irradiated diet (PROVIMI KLIBA) and autoclaved tap water were provided ad libitum. Subcutaneous microchips implanted under isoflurane anesthesia were used to identify each mouse. Mice were randomly distributed between the treatment and the vehicle control group when tumors were well established (150–200 mm³) and treatment was initiated. A vehicle control group (citrate buffer) was included and applied at the same volume/route. Seven to 10 animals were included per treatment group. Tumor volume was determined three times a week using a caliper and body weight was measured daily as an indicator of tolerability. Animals were sacrificed at the end of the study or when the tumors reached a size of 1,500 mm³. Animals with necrotic tumors or a body weight loss <18% were euthanized for ethical reasons. All xenograft experiments have been approved by the relevant Institutional Review Board.

The human patient derived colorectal cancer xenografts were established and characterized at EPO (Experimental Pharmacology and Oncology, Berlin-Buch GmbH, Germany; ref. 20). Main clinical characteristic of the tumors and a description of the efficacy experimental procedures are described in the Supplementary Material and Methods.

mRNA-expression (TPM, RNAseq) in non-diseased samples and various cancer entities was obtained from the Genotype-Tissue Expression (GTEx) project and the Cancer Genome Atlas database, respectively.

FFPE-blocks of liver samples from different donors with transition to colorectal cancer metastases were acquired from Indivumed GmbH (Falkenried 88, D-20251 Hamburg, Germany). Commercially available FFPE human normal tissue microarrays and FFPE-blocks of normal human organs (Supplementary Materials And Methods) from various biobanks, meeting all ethical standards and their use for oncological research has been authorized by the donors via informed written consent (templates approved by the relevant Institutional Review Boards), were used.

Two-μm thick serial sections of each FFPE-block were prepared on a microtome, put on glass slides and dewaxed. Unmasking solution (Vector Laboratories #H3300) was applied at 121°C/1 bar, followed by blocking steps using 3% H₂O₂ and subsequently normal goat serum (Vector Laboratories #S-1000) in PBS/2% BSA. The serial sections were incubated with the primary antibody [anti-cleaved caspase-3 (Asp175) Cell Signaling Technology #9661; anti-cleaved Caspase-8 (Asp391; 18C8) Cell Signaling Technology #9496; anti-CDH17 Clone 141713 from R&D Systems #MAB1032 and anti-TRAILR2, Cell Signaling Technology, #8074] for 1 hour at room temperature, followed by detection reagent (EnVision+, HRP, Rabbit, DAKO #4002) and incubation in DAB-solution. Counterstain was performed using hematoxylin. Washing steps in PBS were performed where appropriate. Corresponding isotype controls were performed in parallel.

EDTA-plasma samples were taken during final sacrifice and immediately frozen in liquid nitrogen. They were stored at −20C until evaluation with the Caspase-Glo 3/7 assay from Promega # G8091. According to the manufacturer's instructions.

We aimed to identify cell surface proteins absent in non-neoplastic liver samples that were expressed at significant levels in one or more major cancer types.

We analyzed 188 non-diseased liver tissue samples using mRNA expression as a classifier. We selected CDH17, as none of the samples analyzed had a value higher than 0.24 TPM, with 27% of the samples (51 of 188) displaying less than 0.01 TPM (Fig. 1A). We then examined the distribution of CDH17 in normal human organs by IHC using a panel of FFPE-blocks. CDH17 expression was observed in the epithelium of the small intestine, colon, and pancreatic ducts. In agreement with the RNA analysis, no CDH17 expression was detected in the liver (Fig. 1B).

High expression levels of CDH17 and TRAILR2 were observed in colon (455 samples) and rectum (167 samples) adenocarcinomas (97% of the samples have > 10TPM for both targets). Additional malignancies of the gastrointestinal tract, stomach adenocarcinoma (373 samples), pancreatic adenocarcinoma (178 samples), and esophageal carcinoma (162 samples) also included relevant populations of CDH17-high/TRAILR2-high cases (66, 42 and 38% of the samples >10TPM for both targets respectively; Fig. 1A). A significantly lower number of related normal tissues (colon, stomach, pancreas and esophagus) displayed high levels of CDH17 and TRAILR2 RNA expression (Supplementary Fig. S1). IHC analysis of a series of colorectal cancer hepatic metastases (n = 10) confirmed widespread expression of CDH17. Heterogeneous to diffuse TRAILR2 expression was detected in 6 of 10 cases (Fig. 1C; Supplementary Table S1). No expression of TRAILR2 by IHC was observed in any normal human tissue analyzed.

BI 905711 is a novel bispecific molecule with bivalent CDH17 and TRAILR2 binding. The CDH17-building block was derived from a mouse hybridoma humanized lead and comprises the conventional monoclonal part of the bispecific molecule. The TRAILR2 building block is attached as a scFv to the C-terminus of the molecule. The two building blocks are coupled via a six amino acid Gly-Ser linker. The L234A/L235A mutation, ablating binding to FcγR and complement (21), was incorporated to avoid CDH17-independent crosslinking (Fig. 2A).

When evaluated using surface plasmon resonance, BI 905711 binds to human TRAILR2 and human CDH17 with K_D values of 176 and 33.9 nmol/L, respectively, resulting in an almost 50-fold difference in affinity. Kinetic and steady-state measurements are summarized in Supplementary Table S2.

BI 905711 does not bind to human TRAILR1, human CDH16, or unrelated charged proteins. Human and cynomolgus TRAILR2 and CDH17 extracellular domain proteins display between 90% and 95% sequence identity. BI 905711 is cross-reactive to the corresponding TRAILR2 and CDH17 cynomolgus orthologues, with similar K_D values (Supplementary Table S2). Mouse and rat show weak sequence identities in their extracellular domain compared with the human proteins [10% and 80% (mouse), 29% and 80% (rat) for TRAILR2 and CDH17, respectively].

The COLO 205 cell line was selected as a representative CDH17/TRAILR2 double-positive colorectal cancer cell line. Membrane expression of both proteins was confirmed by FACS (Fig. 2B). Two clinical TRAILR2 agonists, the monoclonal antibody lexatumumab (refs. 4, 5; Lexatumumab'), and the more potent tetrameric nanobody agonist, TAS266 (ref. 14; TAS266′) were included in the assay as controls. BI 905711 was 1,000-fold more potent (IC₅₀) than Lexatumumab' (0.0012 vs. 1 nmol/L), being comparable in potency (IC₅₀) and inducing equivalent decreased cell viability (%) when compared with TAS266′ (0.0012 nmol/L/91% vs. 0.0024 nmol/L/94%; Fig. 2C). The CDH17 building block of BI 905711, engineered as a conventional monoclonal antibody, did not affect the viability of COLO 205 cells (Fig. 2C).

Intracellular caspase activity was assessed to confirm activation of the extrinsic apoptotic pathway. Caspase-8 and caspase-3/7 activity was significantly increased upon treatment with 0.01 nmol/L of BI 905711, the concentration inducing maximal effect on cell viability, at the earliest time points analyzed (2 and 6 hour, respectively; Fig. 2D). In a co-culture system of immune activation, in which dendritic cell activation is measured upon tumor cell death, no effects were observed with BI 905711, in contrast with culture in the presence of LPS (Supplementary Fig. S2).

Correlation between BI 905711 sensitivity and protein expression of CDH17 and TRAILR2 was evaluated across a panel of 24 CDH17/TRAILR2 double-positive colorectal cancer cell lines. Minimal percentage of control (minPOC), corresponding to the minimal number of viable cells detected at any given compound concentration compared with non-treated controls, was used. BI 905711 minPOC inversely correlated with both TRAILR2 and CDH17 expression (Supplementary Fig. S3 and Supplementary Table S3). 54% of cell lines were defined as sensitive to BI 905711 (minPOC <35%), 25% of cells displayed intermediate sensitivity (minPOC between 35% and 60%), and 21% of cells were insensitive (minPOC >60%; Fig. 2E, y-axis). The minPOC of BI 905711 and the reference molecule TAS266′ significantly correlated (Fig. 2E).

CDH17-negative GP2d clones (CDH17-KO GP2d) were generated to assess the selectivity requirement for BI 905711. Loss of CDH17 expression, identified by sequencing, was confirmed by western blot and assessed by FACS analysis (Fig. 3A and C). In agreement with the differences in affinity toward CDH17 and TRAILR2, BI 905711 FACS binding was strongly reduced in CDH17-KO GP2d, signal detected at higher concentrations corresponding to TRAILR2 binding (Fig. 3E and F). Deletion of CDH17 led to an approximately 1,000-fold decrease in cell killing by BI 905711 (IC₅₀ GP2D and CDH17-KO GP2D cells 0.0002 versus 0.250 nmol/L, respectively; Fig. 3B and D). No differences in TRAILR2 membrane expression or sensitivity to TAS266 was observed (IC₅₀ GP2D and CDH17-KO GP2D cells 0.0027 vs. 0.0022 nmol/L, respectively; Fig. 3A–D).

Hep G2 cells, derived from a hepatocellular carcinoma, were used as a surrogate for TRAIL-sensitive hepatocytes. This cell line was selected as TRAILR2 expression was not observed in cultured systems derived from primary human hepatocytes, similar observations were described in ref. 22. TRAILR2 membrane expression and the lack of CDH17 were confirmed by FACS (Fig. 3G). In contrast with the reference TAS266′, no effect was observed upon treatment of HepG2 with any concentrations of BI 905711 concentration tested (up to 100 nmol/L; Fig. 3H).

Both CDH17 and TRAILR2 are homogeneously expressed in GP2d xenografts (Fig. 4A). In this model, a single dose of BI 905711 led to sustained tumor regressions up to 30 days following treatment. Subsequent administration at day 36, when only 2/8 tumors were still in regression, resulted in tumor growth delay that lasted until the end of the experiment on day 52 (Fig. 4B). Increased cleaved caspase-3 and cleaved caspase-8 (IHC) 24 hours after initial administration indicated induction of tumor cell apoptosis (Fig. 4C). Cleaved caspases 3/7 activity in the plasma correlated well with intra-tumoral apoptosis. A statistically significant increase of cleaved caspases 3/7 plasma activity was also detected 24 hours after first treatment (Fig. 4D). Increased caspase activation was again detected following the second administration of BI 905711 (Supplementary Fig. S4).

DLD-1 cells are less sensitive to TRAILR2 agonism in vitro (Fig. 2E), but display similar homogeneous CDH17 and TRAILR2 expression (Fig. 4E). Following initial regression, tumor regrowth was observed 10 days after administration of BI 905711. Subsequent administration of BI 905711 (day 14), caused additional tumor growth delay, TGI of 67% by day 22 when the control group had to be terminated (Fig. 4F). Combination of BI 905711 and the standard-of-care, irinotecan resulted in a statistically significant tumor growth control, as compared with vehicle and either compound as monotherapy (Fig. 4F).

A bispecific modality, such as BI 905711, can potentially induce apoptosis in both a cis (same cell) as well as trans (adjacent cell) fashion. Evidence for cis-activation was initially supported by the >99% decrease in cell viability observed in cell lines such as COLO 205 (Fig. 2C) and GP2d (Fig. 3B) in vitro. Confidence in this hypothesis was strengthened by tracking individual GP2d cells after BI 905711 administration. Analysis of sequential images following treatment with BI 905711 administration demonstrated induction of apoptosis in individual cells (Fig. 5A).

The ability of BI 905711 to activate TRAILR2 in adjacent tumor cells (trans-activation) was assessed in Red Fluorescent Protein (RFP)-expressing, CDH17-negative, human ovarian carcinoma cell line OVCAR-3 (OVCAR-39–mKate2). Similar to parental cells, OVCAR-3–mKate2 cells, were sensitive to TAS266′ but unresponsive to BI 905711 (Fig. 5B). An additional negative control, a bispecific molecule where the CDH17-binding domain of BI 905711 was substituted by an irrelevant sequence, had no effects on OVCAR-3–mKate2 cell number (Fig. 5B).

Consistent with the hypothesis that BI 905711 is able to mediate TRAILR2 trans-activation, a reduction in the number of detected OVCAR-3–mKate2 red cells was observed when co-cultured with CDH17-positive DLD-1 cells, in the presence of antibody. This reduction was detected in all DLD-1:OVCAR-3–mKate2 seeding ratios tested and was comparable with the reduction observed after TAS266′ administration. Importantly, treatment with the negative control bispecific molecule had no detectable effect (Fig. 5B). Similar results were observed when images captured over time following administration of the different compounds were analyzed (Fig. 5C).

The implication that BI 905711-induced trans-activation can enhance antitumor activity was evaluated in xenograft mouse models with heterogeneous CDH17 expression. COLO 205 cells were proven CDH17-positive when cultured in vitro, and represented one of the most sensitive cell lines to BI 905711 administration amongst the 24 colorectal cancer cell lines tested (Fig. 2B and E). However, when grown as a xenograft, heterogeneous CDH17 expression was observed (5%–10% of viable tumor; Fig. 5D). Despite this, BI 905711 administration induced initial tumor regressions and strong growth inhibition by the end of the experiment (day 14; Fig. 5E). Accordingly, increased cleaved caspase-8 and cleaved caspase-3 was observed 24 hours after BI 905711 administration. Using a double staining for cleaved caspase-3 and CDH17, we showed that the cleaved caspase 3–positive area in the treated tumors extended beyond the boundary of the CDH17-positive area (Fig. 5F). To demonstrate that BI 905711 effect was dependent on CDH17 targeting, mice were also treated with the bispecific molecule where the CDH17 binding domain was substituted by an irrelevant sequence. This negative control failed to demonstrate in vivo efficacy in COLO 205 xenografts, even at dose levels providing over 35-fold plasma concentration when compared with BI 905711 efficacious dose (Supplementary Fig. S5).

The in vivo efficacy of BI 905711 in a more disease-relevant context was assessed across a panel of three colorectal cancer PDX models. Expression of TRAILR2 and CDH17 was analyzed by IHC after tumor establishment. In alignment with our prevalence data in human samples (Fig. 1), all three models showed a strong and homogeneous CDH17 expression, with more variation in TRAILR2 intensity and expression pattern (Fig. 6A, C, and E). Single-agent administration of BI 905711 on a q14d schedule, resulted in significant antitumor efficacy in all three models (Fig. 6B, D, and F), with >50% of the tumors remaining in regression on termination of the experiment (days 18 and 28, respectively) in two of the models analyzed (Fig. 6B and F).

Here, we report the preclinical antitumor efficacy of BI 905711, a novel bispecific molecule with bivalent CDH17 and TRAILR2 binding. BI 905711 binds to CDH17 and uses it as an anchor to trigger potent TRAILR2 activation, resulting in apoptosis and tumor regression in a CDH17-expression dependent manner.

In agreement with published data, CDH17 expression was observed in 100% of the colorectal cancer samples analyzed. In addition, published studies described a high level of concordance in CDH17 expression between primary colorectal adenocarcinomas and their corresponding lung and lymph node metastases (23), suggesting that CDH17 expression is preserved during the metastatic process. TRAILR2 expression in several solid cancers, including colorectal tumors, has also been described previously (24–26). In agreement with these published studies, we were able to confirm expression of TRAILR2 in colorectal cancer hepatic metastases. Altogether, these data point to colorectal cancer as a relevant cancer type to evaluate the antitumor activity of BI 905711. Evidence of co-expression of both targets in a large subset of gastric and pancreatic cancers suggest that BI 905711 may have additional activity in these indications (16, 27–31). Preclinical studies to confirm this hypothesis are currently ongoing.

Using CDH17 as an anchor, BI 905711 is approximately 1,000-fold more potent than conventional TRAILR2 targeting antibodies, examples of which displayed weak and inconsistent signs of efficacy in human clinical trials (4, 5). The activity of these anti-TRAILR2 antibodies can be enhanced by secondary crosslinking (32) and their antitumor activity relies on interaction with Fcγ receptors (11). The superior TRAILR2 agonistic activity of BI 905711, independent of any secondary mechanism, can overcome this deficiency increasing its probability to translate into a robust therapeutic activity in patients.

BI 905711 displayed comparable potency with TRAILR2 second-generation agonists, developed to work independently of a secondary cross-linking mechanism, across a panel of CDH17-positive colorectal cancer cells. This potent in vitro activity translated into single-agent antitumor efficacy in different CDH17-positive colorectal cancer xenograft models, including patient-derived xenografts. Importantly, BI 905711 activity is strictly dependent on CDH17 expression, as deletion of CDH17 resulted in a approximately 1,000-fold loss of potency. In addition, BI 905711 had no significant effect on the viability of Hep G2 cells, a liver derived CDH17-negative cell line used as surrogate for TRAIL-sensitive hepatocytes. Collectively, and considering the lack of detectable CDH17 protein in non-neoplastic liver tissue, our data suggest that BI 905711 has the potential to trigger potent tumor cell apoptosis while improving the therapeutic window, by sparing hepatotoxicity associated with the activation of TRAILR2 in the liver.

Currently, there are three biologics (ABBV-621, GEN1029, and INBRX-109) designed to induce superior clustering of TRAILR2 in phase-I clinical trials (NCT03082209, NCT03576131, and NCT03715933). As none of them includes a domain or motif that specifically enables sparing the targeting of normal liver cells, it is unclear how well these new molecules will overcome the reported clinical hepatotoxicity associated with TRAILR2 agonism in patients.

Targeted strategies coupling TRAILR2 activation to distinct tumor restricted antigens have been previously proposed (33–35). This has resulted in initiation of clinical trials with RG-7386 (33), a bispecific molecule that co-targets fibroblast activation protein alpha (FAP) to increase tumor specificity of TRAILR2 activation. The phase-I trial of RG-7386 has been completed and initial data presented showed a single partial response among 32 heavily pretreated patients with NSCLC (36). FAP is not expressed directly in the cancer cells in epithelial tumors, such as colorectal cancer, but rather in mesenchymal tumors and on activated fibroblasts, which are prevalent in the tumor stroma (37, 38). Thus, with the exception of mesenchymal tumors, killing can be only achieved at sites of contact between tumor and stromal cells, where RG-7386 bound to FAP on stromal cells can activate TRAILR2 on adjacent tumor cells (trans-activation). Tumor cells that are not in direct contact with activated fibroblasts will not be affected by this treatment. In contrast, BI 905711 is directed by CDH17, an anchor target expressed on tumor cells, increasing the total tumor area that can be targeted (cis- and trans-activation). Altogether, these data suggest that the mode of action of BI 905711 is a promising approach to target colorectal cancer and overcome the limitations of current TRAILR2 agonists.

High expression of both targets appears beneficial for BI 905711 activity and the use of CDH17 and/or TRAILR2 expression as putative biomarkers will be further explored in the ongoing clinical trial. We anticipate that not only the level of CDH17 expression on individual tumor cells, but also, the proportion and distribution of CDH17-positive cells across the entire tumor mass will determine BI 905711 activity. However, we believe that BI 905711 can still offer benefit to patients with rather low fractions of CDH17-positive cells in their tumors, as demonstrated by our data in COLO 205 xenograft. Although high levels of TRAILR2 are expected to be beneficial, the level of TRAILR2 expression alone has been shown to incompletely correlate with TRAIL sensitivity (39–43). Investigation of other genes/proteins within the extrinsic apoptotic pathway that may modulate TRAIL sensitivity are under active investigation.

Colorectal cancers, with the exception of metastatic DNA mismatch repair/microsatellite instability-high colorectal cancers, have been largely refractory to cancer immunotherapies (44). For this reason, we also explored whether BI 905711 as a monotherapy can evoke an immune response. Apoptotic cell death induced by BI 905711 did not activate dendritic cells by itself. However, it cannot be excluded that BI 905711 might be combined with immunotherapy, where the antigen needed for presentation is provided by cells dying upon BI 905711 treatment. Nevertheless, for a productive T-cell activation, the adjuvant signal would need to be provided by the combination partner. TRAIL agonists are also reported to eliminate myeloid-derived suppressor cells (45, 46). The selective elimination of those immunosuppressive cells, potentially via BI 905711-mediated TRAILR2 trans-activation, could remodel the tumor microenvironment and thereby enhance the activity of an immunomodulatory combination partner.

Expression of CDH17 is not tumor-specific and we confirmed CDH17 expression in a subset of normal tissues (small intestine, colon and pancreatic ducts). CDH17-positive non-neoplastic tissues may be spared from BI 905711-mediated apoptosis due to their lower sensitivity to TRAILR2 activation, which also contributes to a potentially wider therapeutic window. In vitro studies demonstrated that, unlike colorectal cancer cells, freshly isolated normal crypt cells are insensitive to TRAIL stimulation (47). This hypothesis is further supported by the results of preclinical safety studies in cynomolgus monkeys, where administration of BI 905711 was well tolerated and no gastrointestinal toxicity was detected by clinical observations or histopathological evaluation (Table S4).

TRAILR2 transactivation may raise concerns regarding unwanted toxicity in the liver, especially in TRAILR2-positive normal cells surrounding colon cancer metastases. In the clinical setting, a rim of desmoplastic stroma or a layer of reticulin fibers may prevent direct contact between tumor cells and hepatocytes in the large majority of cases (48). Similar growth patterns were observed in the colorectal cancer liver metastases samples analyzed in this study (Supplementary Table S1). BI 905711 is designed to spare normal liver tissue; however, some liver damage cannot be absolutely excluded and the ongoing phase I clinical trial in patients with advanced/metastatic colorectal cancer will allow us to confirm the liver-sparing concept.

In summary, we demonstrate here that BI 905711 potently triggers the extrinsic apoptosis pathway, specifically in CDH17-positive tumor cells, which translates into strong antitumor activities in different colorectal cancer models in vivo, including patient-derived xenograft tumors. BI 905711 is a novel bispecific molecule that shows superiority to existing TRAILR2 agonists, representing a targeted strategy for the treatment of colorectal cancer and additional CDH17-positive cancer types. Together with its potential for a favorable safety profile, these data support the ongoing phase I trial of BI 905711 in these patient populations (NCT04137289).

J.M. García-Martínez reports grants from Austrian Research Promotion Agency (grant reference numbers: 839361 and 844335) during the conduct of the study, as well as a patent for (binding molecules for the treatment of cancer; US20180179287A1) pending to Boehringer Ingelheim, and reports employment with Boehringer Ingelheim. S. Wang reports grants from Austrian Research Promotion Agency (grant reference numbers: 839361 and 844335) during the conduct of the study, as well as reports employment with Boehringer Ingelheim. C. Weishaeupl reports grants from Austrian Research Promotion Agency (grant reference numbers: 839361 and 844335) during the conduct of the study, as well as reports employment with Boehringer Ingelheim. A. Wernitznig reports grants from Austrian Research Promotion Agency grants (grant reference numbers: 839361 and 844335) during the conduct of the study, as well as a patent for (binding molecules for the treatment of cancer; US20180179287A1) pending to Boehringer Ingelheim, and reports employment with Boehringer Ingelheim. C. Pinto reports grants from Austrian Research Promotion Agency (grant reference numbers: 839361 and 844335) during the conduct of the study, as well as reports employment with Boehringer Ingelheim. D. Dutcher reports employment with Boehringer Ingelheim Pharma. M.A. Impagnatiello reports grants from Austrian Research Promotion Agency (839361 and 844335) during the conduct of the study, as well as reports employment with Boehringer Ingelheim. I. Tirapu reports grants from Austrian Research Promotion Agency (grant reference numbers: 839361 and 844335) during the conduct of the study, as well as reports employment with Boehringer Ingelheim. F. Hilberg reports personal fees and other from Boehringer Ingelheim RCV (employee) during the conduct of the study, as well as personal fees from Boehringer Ingelheim RCV (employee) outside the submitted work. N. Kraut reports grants from Austrian Research Promotion Agency (grant reference numbers: 839361 and 844335) during the conduct of the study, as well as reports employment with Boehringer Ingelheim. M. Pearson reports grants from Austrian Research Promotion Agency (grant reference numbers: 839361 and 844335) during the conduct of the study, as well as reports employment with Boehringer Ingelheim. K.P. Kuenkele reports a patent for (binding molecules for the treatment of cancer; US20180179287A1) pending. No disclosures were reported by the other authors.

J.M. García-Martínez: Conceptualization, formal analysis, supervision, validation, investigation, visualization, methodology, writing-original draft, writing-review and editing. S. Wang: Formal analysis, validation, investigation, visualization, methodology, writing-review and editing. C. Weishaeupl: Formal analysis, validation, investigation, methodology. A. Wernitznig: Data curation, software, formal analysis, validation, visualization. P. Chetta: Formal analysis, supervision, validation, investigation, visualization, methodology, writing-review and editing. C. Pinto: Formal analysis, validation, investigation, visualization, writing-review and editing. J. Ho: Formal analysis, supervision, validation, investigation, writing-review and editing. D. Dutcher: Formal analysis, validation, investigation. P.N. Gorman: Formal analysis, validation, investigation. R. Kroe-Barrett: Supervision, validation, writing-review and editing. J. Rinnenthal: Formal analysis, supervision, validation, investigation. C. Giragossian: Formal analysis, supervision, validation, investigation, writing-review and editing. M.A. Impagnatiello: Formal analysis, supervision, validation, investigation. I. Tirapu: Formal analysis, supervision, validation, investigation, visualization, writing-review and editing. F. Hilberg: Formal analysis, supervision, validation, investigation. N. Kraut: Supervision, writing-review and editing. M. Pearson: Supervision, writing-review and editing. K.P. Kuenkele: Conceptualization, supervision, writing-review and editing.

We thank Dr. Verena Supper for kindly providing the OVCAR-3-mKate2 cells, and colleagues at Boehringer Ingelheim RCV and Boehringer Ingelheim Biotherapeutics Discovery for scientific and technical support. We also thank Mahmoud Ould Kaci and Holly Dursema for insightful discussions. This work has been supported by the Austrian Research Promotion Agency (grant reference numbers: 839361 and 844335).

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.