Hexavalent TRAIL Fusion Protein Eftozanermin Alfa Optimally Clusters Apoptosis-Inducing TRAIL Receptors to Induce On-Target Antitumor Activity in Solid Tumors

TRAIL (Apo2L or TNFSF10) is a member of the TNF superfamily of cytokines that play diverse roles in immune cell modulation and the activation of intracellular signaling pathways that control cell proliferation, survival, and apoptosis. TRAIL is a type II transmembrane protein that can bind as a trimer to membrane-bound or soluble receptors: TRAIL-R1 (DR4, TNFRSF10A), TRAIL-R2 (DR5, TNFRSF10B), TRAIL-R3 (DcR1, TNFRSF10C), TRAIL-R4 (DcR2, TNFRSF10D), and osteoprotegerin (TNFRSF11B). Whereas binding of TRAIL to DcR1 and DcR2 has a limited impact on intracellular cell signaling, binding to DR4 and/or DR5 initiates their trimerization that can trigger the formation of the death-inducing signaling complex (DISC) to drive caspase-dependent apoptotic cell death (the extrinsic apoptosis signaling pathway). In type I cells, caspase-8/10 activation is robust enough to directly activate effector caspase-3/7 signaling that results in apoptosis. However, in type II cells engagement of the intrinsic apoptosis signaling pathway is required to amplify the apoptosis signal. This process requires caspase-8 to induce cleavage of the BH3-only protein BID. The truncated form, tBID, subsequently initiates mitochondrial outer membrane permeabilization (MOMP) following BAX and BAK dimerization to release cytochrome c, which in turn leads to apoptosome formation, effector caspase activation, and ultimately apoptosis (1–3).

The intrinsic apoptosis signaling pathway is regulated by the B-cell lymphoma protein-2 (BCL2) family of proteins that can be divided into three groups each containing at least one BCL2 homology (BH) motif (BH1–4). The proapoptotic BH3-only proteins BIM, BID, PUMA, NOXA, BAD, BIK, BMF, and HRK, and the multidomain effector proteins BAX and BAK, are activated or induced by various cell death stimuli that drive MOMP and subsequently apoptosis. The antiapoptotic members (BCL2, BCL-X_L, MCL-1, BCL-W, and BCL2-A1) possess BH3-binding grooves that function to counter the activity of the BH3-only and multidomain effectors via direct protein–protein interactions. Aberrant expression and/or function of BCL2 family proteins are integral to tumorigenesis and resistance to cancer therapy by enabling malignant cells to evade apoptosis (4, 5), including that induced by TRAIL (6–8). A number of small-molecule inhibitors of the BCL2 family members are now in clinical development in various oncology indications including venetoclax, which received FDA approval in the United States for use in patients with relapsed/refractory chronic lymphocytic leukemia harboring 17p deletions (3, 9). However, the development of anticancer therapeutics that target the extrinsic apoptosis signaling pathway have faced significant challenges in the clinic (1, 2).

Compelling preclinical data demonstrated that TRAIL was capable of selectively inducing the cell death of tumor cells without the adverse effects associated with TNF or CD95 (Fas ligand; refs. 1, 2). Subsequently, a first generation of TRAIL-based therapeutics entered into clinical trials and mainly centered upon the recombinant form of human soluble TRAIL [sTRAIL; exemplified by Apo2L/AMG-951/dulanermin, (10)] or agonistic antibodies that selectively target DR4 [mapatumumab, (11)] or DR5 [drozitumab, (12); conatumumab/AMG-655, (13); lexatumumab, (14); tigatuzumab, (15)]. However, despite being well tolerated in human patients with cancer, these TRAIL-based therapeutics failed to develop compelling objective responses as a monotherapy or in combination with other therapeutic agents (1, 2). Much of these clinical failures have been attributed to the biologic properties of these first generation TRAIL-based molecules; Apo2L possesses a short plasma half-life, is rapidly cleared from the circulation (16, 17) and is inherently weak at inducing higher-order receptor clustering (18, 19), whereas bivalent DR5 agonistic mAbs require additional Fc-FcγR–mediated cross-linking to induce optimal clustering of their respective receptors (20). The requirement of Fc–FcγR interactions has been viewed as a critical activity-limiting step in patients with cancer in part due to the dependence of FcγR-expressing immune cells in the tumor microenvironment, or competition with high circulating levels of endogenous IgG (21). To address these liabilities and improve upon first generation agents, eftozanermin alfa (ABBV-621) was engineered as a hexavalent TRAIL receptor agonistic fusion protein derived from its prototype APG-350 (21). ABBV-621 selectively binds to TRAIL receptors with nanomolar affinity to induce optimal receptor clustering in human solid tumor cancer cells that drives on-target apoptosis and robust antitumorigenic activity that is independent of Fc–FcγR interactions. In addition, ABBV-621 activity is significantly enhanced when combined with sensitizing agents such as taxanes, topoisomerase inhibitors or BCL-X_L small-molecule inhibitors. Together, these data provide a preclinical rationale for evaluating ABBV-621 in multiple different cancer types (NCT03082209).

Cell lines were purchased from DSMZ, ATCC, ECACC, JCRB Cell Bank or from the KCLB and cultivated for 1–15 passages in the recommended culture media supplemented with 20 mmol/L HEPES (Gibco), penicillin/streptomycin and 10% FBS (Invitrogen; Supplementary Tables S1 and S2). Cells were grown at 37°C in a humidified atmosphere with 5% CO₂. Cell lines were tested for authenticity by short tandem repeat profiling and Mycoplasma by the AbbVie Core Cell Line Facility. ABBV-621, anti-DR5 agonistic mAb with or without mutant Fc (DANA or LALA), anti-ABBV-621 mAb, venetoclax, navitoclax, A-1331852, paclitaxel, docetaxel, erlotinib and 5-fluorouracil (5-FU) were all synthesized by AbbVie unless otherwise stated. SN38 was purchase from Sigma-Aldrich.

Binding kinetics of ABBV-621 for recombinant sTRAIL receptor extracellular domains (ECD) were determined by surface plasmon resonance (SPR) as described previously (22). Briefly, recombinant ECDs of human and cynomolgus monkey TRAIL receptors (fused with 6His tag at C-terminal end) were purified from HEK-293 cells. Each assay cycle consisted of the following steps: (i) capture of ABBV-621 on anti-human Fc surface only; (ii) TRAIL receptor injection over both reference and test surface, 240 μL at 80 μL/minute, after which, the dissociation was monitored for 300 seconds at 80 μL/minute. All measurements were double referenced against blank surface and running buffer (HBS-EP+). Data were fitted globally to a 1:1 binding model using Biacore T200 Evaluation software to determine the binding kinetic rate constants, k_a (M⁻¹s⁻¹) and k_d (s⁻¹), and the equilibrium dissociation constant K_D (M).

Human tumor cell lines were seeded into 96- or 384-well plates at 5,000 cells or 1,500 cells per well, respectively, and treated with ABBV-621 alone or in combination with venetoclax, navitoclax, A-1331852 paclitaxel, docetaxel, SN38, gemcitabine, 5-FU or erlotinib for the indicated incubation times. Cell viability was subsequently measured using CellTiter-Glo (Promega). Responses were determined as a percentage of the control treated cells and EC₅₀s calculated from the resulting sigmoidal dose–response curves using GraphPad Prism (GraphPad Software). The Bliss independence model (23) was used to evaluate combinatorial activity, positive integers indicating synergy. Bliss scores were calculated for each combination within the dose matrix and totaled to give a “Bliss Sum” value. For determination of apoptosis, DLD-1 cells were treated with ABBV-621 and/or SN38 for 24 hours and the sub-G₀–G₁-phase content of DNA cell-cycle histograms was determined by flow cytometry as described in detail elsewhere (24).

SW620 cells were seeded in 96-well clear-bottom black polystyrene microplates (Corning Inc.) at 5,000 cells per well prior to treatment and caspase-3/7 activity (catalog no. 4440, Sartorius) and cell viability (DRAQ7, catalog no. 7406, Cell Signaling Technology) was determined using an IncuCyte Zoom (Sartorius) programmed to take four images per well at a 1-hour interval for 24 hours. Data were analyzed using the IncuCyte S3 2019 software and plotted as the count of green objects (activated caspase-3/7) and red objects (DRAQ+) divided by area (mm²) per well.

Cells were harvested, washed with ice cold-PBS, and lysed by RIPA buffer (Sigma) containing protease inhibitor cocktail (Roche). The purified lysates were electrophoresed by SDS-PAGE (Invitrogen) and transferred to nitrocellulose membranes utilizing an iBlot 2 (Invitrogen) device. The following antibodies were used; mouse anti-β-Actin (catalog no. ab6276, Abcam). Rabbit anti-BCL-X_L (catalog no. 2764, Cell Signaling Technology), anti-PARP (catalog no. 551025, BD Biosciences), anti-caspase-8 (catalog no. M032-3, MBL, International), anti-caspase-3 (catalog no. ab13585, Abcam), anti-caspase-9 (catalog no. 9502, Cell Signaling Technology) or GAPDH (catalog no. ab8245, Abcam). Blots were incubated with primary antibodies at 4°C overnight followed by Alexa Fluor 680/800 goat anti-mouse IgG (catalog no. A21057/A32730, Invitrogen) or Alexa Fluor 680/800 goat anti-rabbit IgG (catalog no. A21109/A32735, Invitrogen) secondary antibodies for 60 minutes at room temperature. Proteins were visualized using the Odyssey infrared imaging system (LI-COR Biosciences) and were not further manipulated with imaging software.

For the determination of caspase-8 aggregation by high content analysis, NCI-H460 cells were cultured overnight in black collagen I–coated 96-well plates (Corning). Treated cells were fixed in formaldehyde, washed with PBS, permeabilized with 0.1% Triton and blocked with 1% BSA (Invitrogen) before incubating with mouse anti-caspase-8 Ab (1:800, catalog no. M032-3, MBL). Following overnight incubation at 4°C, cells were washed with PBS and incubated with goat anti-mouse secondary Alexa 488 conjugated antibody (1:400, catalog no. A-11029, Thermo Fisher Scientific) + Hoechst 33342 (1 μg/mL, catalog no. H-3570, Life Technologies) for 1 hour. Cells were washed again with PBS and the microplate scanned on an ArrayScan (Thermo Fisher Scientific) using a 20× objective and analyzed with Spot Detector algorithm.

To determine the colocalization of ABBV-621 binding with activated caspase-8, NCI-H460 cells were treated with ABBV-621 (2 μg/mL) and mouse anti-human mAb (1 μg/mL; catalog no. ab99757, Abcam) fixed and permeabilized as described above and incubated with a rabbit anti-cleaved caspase-8 antibody (1:100, catalog no. 9496, Cell Signaling Technology), followed by an anti-mouse Alexa 488 secondary antibody (1:400, catalog no. A-11029, Thermo Fisher Scientific) and an anti-rabbit Alexa 594 secondary antibody (1:400, catalog no. A-32740, Thermo Fisher Scientific). ArrayScan images were analyzed using the colocalization algorithm to measure the overlapping area between independently identified puncta containing ABBV-621 and activated caspase-8.

Briefly, Colo205 cells were transfected with pCMV-Cas9-RFP containing the CRISPR sequence 5′-GTACATGTAGGTGCGTTCCTGG targeting TNFRSF10A (Target ID HS0000284207, DR4, Sigma-Aldrich) or sequence 5′-AGAACGCCCCGGCCGCTTCGGG targeting TNFRSF10B (Target ID HS0000284184, DR5, Sigma-Aldrich). SW620 cells were transfected with pCMV-CAS9-GFP containing the CRISPR sequence 5′AGGAACACCAGCCGGTCGGAGG targeting BID (Target ID HS0000187506, Sigma-Aldrich). H460 cells were transfected with CASP8 CRISPR/Cas9 KO plasmid and CASP8 HDR plasmid (sc-400147 and sc-400147-HDR, Santa Cruz Biotechnology,). Transfection was performed using Lipofectamine 3000 according to manufacturer's recommendation (Life Technologies). Cells were assessed for loss in DR5 and/or DR4 expression by flow cytometry, loss of BID or caspase-8 expression by Western blot analysis and sensitivity to ABBV-621–induced cell death.

Stable overexpression of DcR1 and DcR2 was conducted by lentiviral transduction using the lentiviral expression vector PLVX-IRES-PURO containing either full-length human DCR1 or DCR2 cDNA, respectively. Overexpression of BCL-X_L was conducted by transduction of the expression vector pLOC-BCL-X_L (Thermo Fisher Scientific). Colo205 cells expressing PLVX or pLOC vectors were selected for using 1 μg/mL puromycin (Gibco) or 5 μg/mL blasticidin (Gibco), respectively. Expression of DcR1 and DcR2 was confirmed by flow cytometry and BCL-X_L by Western blot analysis.

Harvested cells were washed twice with ice-cold staining wash buffer (Dulbecco PBS without calcium and magnesium containing 0.5% BSA and 0.1% sodium azide). Cell samples were subsequently treated with anti-DR4-PE (catalog no. 12-6644-42, eBiosciences), anti-DR5-PE (catalog no. 12-4714-42, eBiosciences), anti-DcR1-PE (catalog no. FAB6302P, R&D Systems), anti-DcR2-Pe (catalog no. FAB633P, R&D Systems) or mouse IgG1-PE isotype control (catalog no. 12-4714-42, eBiosciences) and incubated for 45 minutes on ice in the dark. Cells were washed with staining wash buffer and resuspended in 350 μL of staining wash buffer containing DRAQ7 (catalog no. 7406S, Cell Signaling Technology) prior to analysis by flow cytometry. TRAIL receptor number was quantified using Quantum Simply Cellular ABC beads (Bang Labs) per the manufacturer's instructions. The isotype control results are subtracted from the resulting TRAIL receptor number, weighted by the percent positive population of cells.

To determine the DR4 and DR5 protein expression in vivo, ABBV-621 was used as a probe to immunoprecipitate DR4 and DR5. Briefly, ABBV-621 naïve tumors derived from xenograft models were homogenized in PBS using a Precellys (Bertin Instruments) and ABBV-621 was added to a final concentration of 200 nmol/L. Tumor samples were incubated on ice for 4 hours and then centrifuged at 15,000 × g for 10 minutes at 4°C. Supernatants were discarded, the remaining cell pellet washed in ice-cold PBS and then resuspended in IP lysis buffer (Pierce Biotechnology Inc.) containing protease inhibitors (Sigma-Aldrich) and phosphatase inhibitors (Roche). Streptavidin multiarray 96-well plates [meso scale discovery (MSD)] with immobilized biotin-labeled anti–ABBV-621 were incubated overnight at 4°C with 5 μg of protein sample/well in duplicate. After three washes with PBS-Tween, sulfo-tagged mouse anti-human DR4 (2 nmol/L) and sulfo-tagged mouse anti-human DR5 antibody (2 nmol/L) were added to each well and incubated for 1 hour at room temperature with rotation. Samples were then washed three times with PBS-Tween and 150 μL of read buffer (MSD) was added per well. Fluorescence was subsequently measured with an MSD Sector Imager 6000.

Tumors harvested from xenograft models treated with ABBV-621 were homogenized in IP lysis buffer (Thermo Fisher Scientific, catalog no. 87787) using a Precellys (Bertin Instruments) under dry ice for 45 seconds twice. Cell supernatants were prepared and 20 μg of protein per sample were used to determine caspase-3/7 activation (Caspase-3/7 Glo; Promega, catalog no. G8090) according to manufacturer's instructions.

Tumors derived from xenograft models treated with ABBV-621 were homogenized in IP lysis buffer using a Precellys (Bertin Instruments). The protein concentration from whole cell or tumor lysates was subsequently determined (Pierce Biotechnology Inc.). Preblocked streptavidin coated multiarray 96-well plates (MSD) were then incubated with 10 nmol/L biotinylated anti-ABBV-621 mAb for 2 hours at room temperature with shaking. Plates were then washed three times with PBS-Tween, 5 μg of protein lysate added in duplicate to each well and samples incubated overnight at 4°C with shaking. Sulfo-tagged anti-human DR4, anti-human DR5 or anti-human caspase-8 antibodies were then added to each sample for 1 hour with rotation to detect ABBV-621 binding to DR4 or DR5, or formation of the ABBV-621–induced DISC, respectively. Finally, plates were washed three times with PBS-tween, 150 μL of MSD reading buffer added and the signal determined using an MSD Sector Imager 6000.

Female SCID and SCID-Beige mice were obtained from Charles River Laboratories, NSG mice were from The Jackson Laboratory. All experiments were approved by AbbVie's Institutional Animal Care and Use Committee (IACUC) and the NIH Guide for Care and Use of Laboratory Animals Guidelines in a facility accredited by the Association for the Assessment and Accreditation of Laboratory Animal Care. Cell line–derived xenograft (CDX; Supplementary Table S2) and PDX mouse models were weighed at the onset of therapy. Each experimental group included 8–10 animals. Tumors were measured two to three times weekly. Measurements of the length (L) and width (W) of the tumor were obtained via electronic calipers and the volume was calculated according to the following equation: tumor volume (TV) = L × W²/2. Mice were euthanized when TV reached a maximum of 2,000 mm³ or upon presentation of skin ulcerations or other morbidities, whichever occurred first.

Studies were conducted at Champions Oncology and approved by the IACUC of Champions Oncology. Athymic nude female mice obtained from Envigo between 6 and 8 weeks of age were used in the studies. Tumor fragments (∼5 × 5 × 5 mm³) from Champions TumorGraft model were implanted subcutaneously in the left flank of the female prestudy mice. When the TV reached approximately 150–300 mm³, animals were matched by tumor size and assigned into control (n = 1) and treatment groups (n = 3) and dosing was initiated on day 0. TV was subsequently measured twice per week and on the final day of study or on the day moribund animals were euthanized. The study was terminated when the mean TV in the vehicle control reached 1,500 mm³ or up to day 60, whichever occurred first. Analysis was based on a modified RECIST 1.1 (25). Each tumor response was determined by comparing TV change at time t with its baseline [% TV change = 100% × ((Vt-Vi)/Vi)]. The criteria for response were defined as follows: overall response (OR) = CR + PR; complete response (CR) < −90%; partial response (PR) between 0% and −90%; stable disease (SD) > 0% and ≤ 100%; progressive disease (PD) > 100%. PDX models in which the control tumor did not reach 3 x initial TV were removed from the dataset.

Whole blood exposure of ABBV-621, DR5 mAb (conatumumab/AMG-655) or DR5 mAb possessing a LALA or a DANA mutation (20) was measured in Colo205 xenograft or nontumor-bearing CD-1 mice treated at the indicated concentrations and route of administration. Whole blood was collected at various timepoints and measured for test article concentrations (N = 12 samples/group) by electrochemiluminescence (ECL) immunoassay (QuickPlex 120 Reader, MSD). The pharmacokinetic parameters assessed included: Terminal half-life (t_1/2), maximum observed serum concentrations (C_max), C_max divided by dose (C_max/D), serum concentration at first timepoint post dose (C_t), C_t divided by dose (C_t/D) volume of distribution at steady state (V_ss), clearance (CL), AUC from zero to infinity, (AUC_o-inf), AUC up to the last measurable concentration (AUC_0-t) and AUC divided by dose (AUC/D). All pharmacokinetic studies were approved by AbbVie's IACUC.

The toxicokinetic properties of ABBV-621 were assessed in a Good Laboratory Practice repeat dose toxicology study and approved by the IACUC of MPI Research. Cynomolgus monkeys were treated with weekly intravenous doses of 4, 20, and 100 mg/kg for a total of five doses (4–6 animals per sex per dose level) followed by a 4-week recovery period (high dose only). Toxicokinetic profiles for all animals were analyzed after the first and fourth doses by ECL immunoassay (QuickPlex 120 Reader, MSD). All animals were euthanized 24 hours after last dose for histologic evaluation.

In all cases, n refers to the number of independent experiments. Linear regression and Spearman correlation were used to determine the correlation between two variables, and the Mann–Whitney U test or one-way ANOVA with Dunnett multiple comparison test was used to determine statistical significance between datasets, all using Prism 7.03 (GraphPad Software).

ABBV-621 is a hexavalent TRAIL receptor agonistic Fc fusion protein derived from APG-350 (21) that contains a single IgG1-Fc point mutation (asparagine to serine 297; Fig. 1A) that effectively removes a glycosylation site to eliminate binding to all Fcγ receptors and complement component C1q. ABBV-621 binds to human recombinant DR4 receptor or DR5 receptor with affinities of approximately 780 and 635 nmol/L, respectively, as measured by SPR (Supplementary Table S3). ABBV-621 is highly active across a panel of solid tumor cell lines in vitro, inducing cell death with an EC₅₀ of less than 1 nmol/L in approximately 36% (45/126) of cell lines assessed (Fig. 1B; Supplementary Table S1), and possesses similar activity to APG-350 in Colo205 cells in vitro and in vivo (Fig. 1C). ABBV-621 cellular activity is mechanism based as demonstrated by the rapid dose-dependent activation of downstream apoptotic signaling in vitro (cleavage of caspase-8, caspase-3, and PARP) and in vivo (caspase-3/7 activation; Fig. 1D). DR4 plasma membrane expression was consistently lower than DR5 across all solid tumor cell lines in vitro (Supplementary Fig. S1A). Reflecting this disparity, DR5 protein or gene (TNFRSF10B) expression correlated with ABBV-621 EC₅₀ in vitro across solid tumor cell lines, whereas DR4 or TNFRSF10A did not (Fig. 1E). However, the tumor type may influence these associations (Supplementary Fig. S1B–S1D). ABBV-621 activity requires DR4 and DR5 tumor cell expression since Colo205 cells deficient in the genes encoding these proteins completely abrogated ABBV-621–induced cell death in vitro and in vivo (Fig. 1F). Reflecting the lower plasma membrane expression of DR4 compared with DR5 in Colo205 cells, deletion of TNFRSF10B reduced ABBV-621 activity to a greater extent than TNFRSF10A deletion in vitro (Supplementary Fig. S2). Although ABBV-621 binds to human recombinant DcR1 and DcR2 (Supplementary Table S3), endogenous expression of DcR1 or DcR2 across human tumor cell lines did not impact the cell killing activity of ABBV-621 (Supplementary Fig. S3A). Similarly, overexpression of DcR1 or DcR2 in Colo205 cells to levels greater than 2-fold of DR4 and DR5 combined did not dramatically influence ABBV-621–induced cell death (Fig. 1G; Supplementary Fig. S3B).

ABBV-621 binds to human and cynomolgus DR4/DR5 receptors with similar affinities (Supplementary Table S3), exhibiting a greater affinity for cynomolgus monkey DcR2 than human DcR2 (∼25-fold increase by SPR). No binding of ABBV-621 to cynomolgus DcR1 could be detected (Supplementary Table S3). Considering these binding parameters, and published data generated across species [rat, mouse, monkey (16, 26)], cynomolgus monkey was therefore considered the only relevant species for which to assess the toxicology profile of ABBV-621. ABBV-621 was well tolerated when administered to cynomolgus monkeys by weekly intravenous infusion (30 minutes) at doses of 0, 4, 20, or 100 mg/kg/week for 4 consecutive weeks (five total doses). No ABBV-621–related clinical signs of toxicity, changes in body weight, clinical pathology parameters, gross necropsy, or microscopic findings were observed during the study with no injection site reactions. Importantly, in this cynomolgus monkey study, ABBV-621 demonstrated approximately dose-proportional increases in maximum observed serum concentrations post dosing with C_max/D as 21.8–25.4 μg/mL/mg/kg and AUC/D as 501–567 μg*hour/mL/mg/kg across 4–100 mg/kg i.v. doses tested. No accumulation was noted in Cmax or AUC between the first and fourth dose of ABBV-621. Serum concentrations and toxicokinetic parameters for ABBV-621 did not exhibit any sex-specific differences. Approximate dose proportionality was also observed in nontumor-bearing CD1 mice treated with a single dose of ABBV-621 with a C_t/D of 6.6–9.06 μg/mL/mg/kg and a terminal half-life of between 43.8 and 51.9 hours (Supplementary Table S4).

Several TRAIL-related therapeutics have been advanced into clinical trials in patients with cancer (10–15). One of these, conatumumab (anti-DR5 mAb; ref. 13) possesses an extended systemic exposure profile when compared with ABBV-621 in tumor-bearing SCID beige mice (Fig. 2A; i.p. routes of administration). Despite this, ABBV-621 induced superior antitumorigenic activity than the anti-DR5 agonistic mAb in the Colo205 CDX model, the latter requiring Fcγ-Rs to drive DR5 receptor clustering because Fc-mutant anti-DR5 mAbs (DANA/LALA mutants; ref. 20) do not impact tumor growth (Fig. 2B). Unlike anti-DR5-agonistic Abs, cross-linking does not significantly enhance the activity of ABBV-621 (Fig. 2C). Utilizing a fluorescent-based approach (12, 18), exposure of the ABBV-621-sensitive human tumor cell line H460 in vitro to ABBV-621 causes active caspase-8 to be rapidly redistributed to regions of ABBV-621 binding (Fig. 2D; Supplementary Fig. S4A). Procaspase-8 was also redistributed from a diffuse staining pattern in untreated H460 cells to defined puncta in ABBV-621 treated cells. This effect was absent with agonistic anti-DR5 mAb treatment without cross-linking (Fig. 2E). To evaluate binding of ABBV-621 to DR4 and DR5, and the formation of the DISC in vivo, we developed an ELISA-based assay to measure ABBV-621 binding to DR4, DR5 or caspase-8 (a component of the TRAIL-induced DISC; refs. 1, 2) that is required for ABBV-621 activity (Supplementary Fig. S4B). DR5 (and/or DR4) engagement and subsequent DISC formation could be observed within 3 hours of ABBV-621 administration to mice bearing Colo205 tumors and occurred in a dose-dependent fashion (Fig. 2F). Collectively, these data demonstrate a requirement for enhanced receptor clustering to induce potent tumor cell death in vitro and in vivo, providing further mechanistic insights as to why other death receptor agonists may have failed in the clinic (18, 20, 21) and thereby distinguishing ABBV-621 from first generation agents.

By capturing tumor heterogeneity, human PDX models can be more reflective of human disease (27). To further assess the spectrum of ABBV-621 activity, a screen of 162 different human PDX models covering 15 solid tumor types was performed applying modified RECIST 1.1 to define PD, SD, and OR (complete response and partial response, CR and PR, respectively) in individual PDXs. Across all PDXs, an OR rate of 29% (47/162) was observed in PDX models treated with ABBV-621 monotherapy, with SD and PD observed in an additional 11% (18/162) and 60% (97/162) of PDX models tested, respectively (Table 1). Importantly, this activity was independent of treatment history or tumor sample harvest site; naïve versus pretreated and primary versus metastatic site (Supplementary Tables S5A and S5B). Of note, ABBV-621 demonstrated better single-agent activity within specific tumor types that include colorectal cancer, pancreatic cancer, gastric cancer, and Ewing sarcoma, with OR rates ranging from 35% to 60% (Table 1). To understand the impact of DR4 and DR5 protein expression on the response of PDX models to ABBV-621 monotherapy, we developed a quantitative ELISA-based assay that leveraged ABBV-621 as a probe to immunoprecipitate DR4 or DR5 from tumors derived from colorectal cancer, gastric, and pancreatic cancer PDX models naïve to ABBV-621 treatment. Although higher DR5 plasma membrane expression positively correlated with ABBV-621 potency in solid tumor cell lines in vitro (Fig. 1E), the measured levels of DR5 protein or gene expression (or DR4/TNFRSF10B expression) did not predict the level of response to ABBV-621 monotherapy in PDX models (Fig. 3A; Supplementary Figs. S5A and S5B). Of note however, elevated DR4 (TNFRSF10A) and DR5 (TNFRSF10B) gene expression was associated with gain-of-function mutations in KRAS (Fig. 3B). Furthermore, colorectal cancer PDX models with strong responses to ABBV-621 treatment (CR and PR) were enriched with KRAS mutations (Supplementary Table S6). These data indicate that while DR4 and DR5 are required for ABBV-621 binding and subsequent antitumorigenic activity in vitro and in vivo, predictive response can be multi-factorial and regulated by additional genetic or downstream cell signaling resistance mechanisms.

Because this in vivo PDX screen is limited by the number of mice within each treatment group (n = 3), we evaluated the impact of dose on the growth of two colorectal cancer PDX tumors with an increased number of mice in each treatment arm (n = 8). ABBV-621 dose-dependently inhibited growth of colorectal cancer tumors CTG-0064 and CTG-0069, the 3 mg/kg response group reflecting those obtained under screening conditions (Fig. 3C). To validate the high monotherapy response rate observed in some tumor types, we further assessed activity of ABBV-621 in an independent set of pancreatic cancer PDX models. ABBV-621 treatment induced a dose-dependent reduction in tumor growth characterized by robust tumor regression that was maintained for more than 20 days at the maximum dose tested (Fig. 3D).

Although single-agent ABBV-621 activity was demonstrated in 36% of human tumor cell lines in vitro and an OR of 29% of solid tumor PDXs across tumor indications, these data also illustrate tumors that are resistant to ABBV-621 monotherapy. To address tumor resistance, we initially evaluated the combinatorial activity of ABBV-621 with therapeutic agents utilized in the management of patients with pancreatic cancer (28) in a panel of pancreatic cancer cell lines in vitro. Broad and robust synergy was observed between ABBV-621 and the microtubule inhibitor paclitaxel or the topoisomerase inhibitor SN38, the active metabolite of irinotecan (Fig. 4A and B; Supplementary Fig. S6A and S6B; Supplementary Table S7). To further interrogate these combinations in vivo, we evaluated their activity in the pancreatic cancer CDX model BxPC3. Cotreatment of mice bearing BxPC3 tumors with ABBV-621 and either docetaxel or irinotecan each enhanced the ABBV-621 monotherapy response (Fig. 4C; Supplementary Table S8). We subsequently evaluated the docetaxel/ABBV-621 combination in several pancreatic cancer PDX models characterized by differential sensitivities to ABBV-621 monotherapy. In all models tested, the combination of ABBV-621 with docetaxel exhibited more pronounced antitumorigenic activity than either agent alone (Fig. 4D and E; Supplementary Table S8). Expanding the combination activity analyses to colorectal cancer, an indication where irinotecan is an established treatment option (29), we also observed robust synergistic cell death between ABBV-621 and SN38 in colorectal cancer cell lines (Fig. 5A; Supplementary Fig. S7; Supplementary Table S9), with SN38 enhancing the sub-G₀–G₁ DNA content of DLD-1 cells treated with ABBV-621 (Fig. 5B). In vivo, the combinatorial activity between ABBV-621 and irinotecan was superior to either agent alone in colorectal cancer CDX or PDX models that exhibit differential ABBV-621 monotherapy response profiles (Fig. 5C and D; Supplementary Table S10).

BCL2L1 (gene encoding BCL-X_L protein) is amplified in colorectal cancer (30) and is associated with resistance to standard of care (SoC) chemotherapeutics in this and other tumor indications (31, 32) thereby restricting apoptosis mediated by other TRAIL-related therapeutics (7, 33–35). In agreement with these observations, overexpression of BCL-X_L in Colo205 cells inhibited ABBV-621 activity (Fig. 6A). Subsequently, we evaluated the combinatorial activity between ABBV-621 and the BCL-X_L selective small-molecule inhibitor A-1331852, the BCL2–selective inhibitor venetoclax, or the BCL2/BCL-X_L dual inhibitor navitoclax (31, 36, 37) in numerous preclinical colorectal cancer models. ABBV-621 in combination with A-1331852 induced broad synergistic cell death in a panel of colorectal cancer cell lines in vitro that was superior to that observed between ABBV-621 and venetoclax or navitoclax (Fig. 6B; Supplementary Fig. S8A; Supplementary Table S11). To further probe the tumor sensitization mechanism of BCL-X_L inhibition combined with TRAIL receptor agonism, we utilized the colorectal cancer cell line SW620, which is resistant to both ABBV-621 or A-1331852 monotherapy. Cotreating SW620 cells with these two apoptosis-targeting agents induced robust caspase-dependent apoptosis in vitro (Fig. 6C; Supplementary Fig. S8B) and robust tumor regression in vivo (Fig. 6D; Supplementary Table S12). As expected, ABBV-621 was shown to bind to DR4 and DR5 and induce DISC derived from treated SW620 tumors (Fig. 6E), indicating that resistance mechanisms to ABBV-621 monotherapy can be regulated by downstream cell signaling. Of note, BID deletion completely abrogated ABBV-621/A-1331852 synergy (Fig. 6F) and caspase-9 activation (Fig. 6G), indicating that the BH3-only protein BID is essentially required to link the extrinsic and intrinsic apoptosis signaling pathways to drive combinatorial activity. The antitumorigenic activity of ABBV-621 in combination with A-1331852 was also superior to either agent alone in pancreatic cancer PDX models (PA117 and PA20; Supplementary Fig. S9; Supplementary Table S12) emphasizing the breadth of activity that can potentially be achieved by combining this TRAIL agonistic fusion protein with BCL-X_L inhibitors in solid tumors.

Herein we characterize the antitumorigenic activity of ABBV-621, a hexavalent TRAIL agonistic fusion protein that induces apoptotic cell death in preclinical models of human tumors through optimal clustering of the apoptosis-inducing TRAIL receptors DR4 and DR5. In preclinical toxicology studies in monkeys, ABBV-621 is well tolerated, with no-test-item–related effects recorded at doses of up to 100 mg/kg i.v. weekly for 4 weeks. When activation of the extrinsic apoptosis signaling pathway is insufficient to drive robust cell death despite binding to DR4/DR5 and subsequent DISC formation, deeper antitumor activity is observed by combining with chemotherapeutics (taxanes or topoisomerase-1 inhibitors) or a tool selective inhibitor of BCL-X_L. A phase I clinical trial to assess the safety, tolerability, and activity of ABBV-621 in patients with previously treated solid tumors and hematologic malignancies is underway (NCT03082209).

ABBV-621 binds with high affinity to all four human TRAIL receptors to rapidly drive caspase-dependent apoptosis that is dependent upon DR5 and DR4 expression. In vivo, binding to DR4 or DR5 can be observed within 3 hours of ABBV-621 treatment and is dose proportionally associated with the concomitant formation of the DISC. Agonistic anti-DR5 mAb requires binding to Fcγ receptors because Fc mutations ablated anti-DR5 Ab activity in vivo from our studies and consistent with previous observations (20). Moreover, cellular imaging data indicate that caspase-8 (a component of the DR4/DR5 DISC; refs. 1, 2) redistributes into defined cellular regions post-ABBV-621 treatment, a phenotype that is absent from agonistic anti-DR5 Ab treated cancer cells unless cross-linked artificially. This ability of ABBV-621 to induce enhanced receptor clustering translates into superior activity in vivo despite having a lower AUC and shorter plasma half-life in mice than an agonistic anti-DR5 Ab. ABBV-621 exhibits a dose-proportional pharmacokinetic profile in cynomolgus monkeys and a serum exposure that is superior to that observed with Apo2L (16). A mutation in the Fc domain (Asn297Ser) of ABBV-621 removes a glycosylation site to eliminate Fcγ-R binding and prevents additional cross-linking that may mediate against hyper-clustering of the therapeutic; a process hypothesized to drive the hepatotoxicity (26) and observed following excessive multimerization with other TRAIL receptor–based therapeutics (38). Collectively, these studies substantiate the ability of ABBV-621 to induce optimized receptor clustering independent of Fcγ-R–mediated receptor cross-linking that translates into on-target antitumor efficacy that is distinguishable from other TRAIL-based therapeutics that have entered human clinical trials.

ABBV-621 is active as a single agent in vitro (EC₅₀ < 1 nmol/L) across a panel of human cell lines representative of different solid tumor indications; minimal activity observed in small cell lung cancer cell lines is attributed to the previously reported low levels of caspase-8 expression (39), and we have demonstrated that caspase-8 is required for ABBV-621 activity. Screening of ABBV-621 against 162 PDX models covering 15 different tumor types revealed an OR (CR + PR) of 29% with SD observed in a further 11% of all models tested. ABBV-621 activity was independent of patient treatment history (naïve vs. previously treated; Supplementary Table S5A) or tumor sample site (i.e., primary vs. metastatic origins; Supplementary Table S5B). Of note, OR or SD was observed in 77% of colon cancer PDX models, 60% of gastric cancer PDX models and 45% of pancreatic cancer PDX models following ABBV-621 monotherapy. These high single-agent response rates under screening conditions were also reflected in PDX studies with an increased number of test subjects in each treatment arm either in models identified as sensitive (OR or SD) under screening conditions (CTG-0064 and CTG-0069) or from an independent PDX model source (PA20, PA55, and PA-117). Although ABBV-621 activity in vitro significantly correlated with DR5 plasma membrane or gene (TNFRSF10B) expression across all solid tumor cell lines (TNFRSF10A expression also correlated, but not the plasma membrane expression of the corresponding protein), this correlation was insufficient to predict ABBV-621 response in PDXs in vivo. However, elevated DR4 (TNFRSF10A) and DR5 (TNFRSF10B) gene expression was associated with PDX models with KRAS mutations, aligning with previously reported data (40). KRAS mutations are frequently observed in patients with colorectal cancer (41) and pancreatic cancer (28), and these mutations are enriched in colorectal cancer PDX models responsive to ABBV-621 monotherapy.

While the breadth of ABBV-621 monotherapy activity observed in preclinical models of solid tumors in vitro and in vivo is encouraging, resistance to ABBV-621 is evident to varying degrees across solid tumor indications. To address this, we demonstrated that ABBV-621 synergized with taxanes or SN38, the active metabolite of irinotecan, to induce dell death across a broad panel of pancreatic cancer and colorectal cancer cell lines in vitro. Subsequent assessment of these combinations in CDX or PDX models of colorectal cancer or pancreatic cancer demonstrated superior antitumorigenic responses to either agent alone; capturing models with differential sensitivities to ABBV-621 monotherapy including those identified in the PDX screen as progressing under ABBV-621 single-agent treatment.

Genetic experiments in cell lines indicate that while DR4 and DR5 are required for ABBV-621 activity, additional genomic aberrations within the apoptosis signaling pathway downstream of the TRAIL receptor may regulate apoptosis induction. CASP8 for example is a well-established gene that regulates the activity of various TRAIL-based therapeutics at the DISC (1, 2) and is required for ABBV-621 activity. Further downstream, mitochondria are required to amplify apoptosis signaling originating from death receptors in type II cells and is a process regulated by the BCL2 family of proteins (1, 2). BCL-X_L is highly expressed in colorectal cancer (30) and pancreatic cancer (33–35); its elevated expression limiting the activity of ABBV-621 and other TRAIL-related therapeutics (7, 33–35). Small-molecule inhibition of BCL-X_L with A-1331852 but not BCL2 with venetoclax induced substantial synergy with ABBV-621 across a breadth of colorectal cancer cell lines via caspase-dependent apoptosis. This strong combinatorial activity between ABBV-621 and A-1331852 observed in vitro was mirrored in colorectal cancer and pancreatic cancer in vivo models, even in the SW620 model of colorectal cancer, which minimally responds to either agent alone. Of note, BID was absolutely required for combinatorial activity between ABBV-621 and A-1331852, reflecting the role of this BH3-only protein in linking the extrinsic to the intrinsic apoptosis signaling pathway. BID is variably expressed in human neoplasms (42) and this differential expression may dictate the combinatorial activity of ABBV-621 with BCL-X_L inhibitors.

In summary, we demonstrate a well-tolerated safety profile coupled with the ability of ABBV-621 to optimally induce TRAIL receptor clustering independent of Fcγ-R–mediated receptor cross-linking that translates into on-target antitumor activity in preclinical models of solid tumors. These properties distinguish ABBV-621 from other TRAIL-based therapeutics that have entered human clinical trials. Moreover, the depth and breadth of ABBV-621 activity is enhanced by combining with small-molecule inhibitors of BCL-X_L or SoC therapeutics such as taxanes or topoisomerase-I inhibitors. Collectively, these data support the ongoing phase I clinical trial evaluating the activity of ABBV-621 in patients of multiple different cancer types (NCT03082209).

D.C. Phillips reports personal fees from AbbVie Inc. and being a stock holder in AbbVie Inc. during the conduct of the study; personal fees from and employment with AbbVie Inc. and being a stock holder in AbbVie Inc. outside the submitted work; in addition, D.C. Phillips has a patent for US9908927 issued to AbbVie Inc., a patent for WO19178433 issued to AbbVie Inc., and a patent for WO19178438 issued to AbbVie Inc. F.G. Buchanan reports personal fees from AbbVie during the conduct of the study; personal fees from AbbVie outside the submitted work; in addition, F.G. Buchanan has a patent for US9908927 issued to AbbVie, a patent for WO19178433 issued to AbbVie, and a patent for WO19178438 issued to AbbVie. S.K. Tahir is an employee of AbbVie. H. Zhang reports personal fees from Abbvie during the conduct of the study and personal fees from Abbvie outside the submitted work. V.C. Abraham reports receiving salary from AbbVie. V.C. Abraham also owns stock in AbbVie as part of compensation. The work described was funded by AbbVie. Z. Liu reports personal fees from Abbvie Inc. and being a stock holder in AbbVie, Inc. during the conduct of the study and personal fees from and employment with AbbVie, Inc. outside the submitted work. L. Zhou is an employee of Abbvie and may hold its stocks. E. DiGiammarino is an employee of AbbVie and may hold stock in AbbVie. S.E. Morgan-Lappe reports personal fees from AbbVie, Inc. and being a stock holder in AbbVie, Inc. during the conduct of the study; personal fees from and employment with AbbVie, Inc. and being a stock holder in AbbVie, Inc. outside the submitted work; in addition, S.E. Morgan-Lappe has a patent for US9908927 issued to AbbVie, Inc., a patent for WO19178433 issued to AbbVie, Inc., and a patent for WO19178438 issued to AbbVie, Inc. The design, study conduct, and financial support for this research were provided by AbbVie. AbbVie participated in the interpretation of data, review, and approval of the publication. No disclosures were reported by the other authors.

D.C. Phillips: Conceptualization, resources, data curation, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. F.G. Buchanan: Conceptualization, resources, data curation, formal analysis, supervision, validation, investigation, writing–original draft, project administration, writing–review and editing. D. Cheng: Resources, data curation, formal analysis, validation, investigation, methodology, writing–review and editing. L.R. Solomon: Conceptualization, resources, data curation, formal analysis, supervision, validation, investigation, writing–review and editing. Y. Xiao: Resources, data curation, formal analysis, validation, investigation, methodology, writing–review and editing. J. Xue: Resources, data curation, formal analysis, validation, methodology, writing–review and editing. S.K. Tahir: Conceptualization, resources, data curation, formal analysis, supervision, validation, investigation, writing–review and editing. M.L. Smith: Data curation, formal analysis, validation, methodology, writing–review and editing. H. Zhang: Resources, data curation, formal analysis, validation, investigation, visualization, writing–review and editing. D. Widomski: Data curation, validation, investigation, writing–review and editing. V.C. Abraham: Data curation, formal analysis, investigation, visualization, methodology, writing–review and editing. N. Xu: Data curation, formal analysis, validation, investigation, methodology, writing–review and editing. Z. Liu: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing–review and editing. L. Zhou: Resources, data curation, formal analysis, validation, investigation, methodology, writing–review and editing. E. DiGiammarino: Conceptualization, resources, data curation, validation, investigation, methodology, writing–review and editing. X. Lu: Data curation, formal analysis, validation, investigation, methodology, writing–review and editing. N. Rudra-Ganguly: Formal analysis, investigation, methodology, writing–review and editing. B. Trela: Conceptualization, data curation, formal analysis, supervision, investigation, writing–review and editing. S.E. Morgan-Lappe: Conceptualization, formal analysis, supervision, investigation, methodology, writing–original draft, project administration, writing–review and editing.

The authors thank Keith Hickman of AbbVie and Wei Lian of AbbVie for establishing ABBV-621 production protocols and Leyu Wang of AbbVie for plasmid and recombinant TRAIL receptor generation.

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.