Preclinical Assessment of AMG 596, a Bispecific T-cell Engager (BiTE) Immunotherapy Targeting the Tumor-specific Antigen EGFRvIII

Glioblastoma multiforme (GBM), or grade IV malignant glioma, is a highly aggressive malignant primary brain tumor and the most common primary brain tumor in adults, with an incidence of two to three new cases per 100,000 people per year in Europe and North America (1–3). The incidence of GBM has been increasing over the last 2 decades (3), and it now accounts for approximately 50% of all primary malignant central nervous system tumors in the United States (1). Patients diagnosed with GBM have extremely poor prognosis despite recent advances in treatment (4). Median survival for patients with GBM receiving standard-of-care therapy, surgery, and adjuvant chemoradiation is 15 to 16 months (5). Consequently, there is an urgent need to develop new therapeutic options that prolong patient survival.

EGFR variant III (EGFRvIII) is a tumor-associated antigen that is a promising therapeutic target because of its specific expression in tumors and its role in tumor cell growth (6). EGFRvIII is a deletion variant of EGFR and is expressed in up to one-third of all cases of GBM, but is not expressed on any normal tissues (6, 7). EGFRvIII has an extracellular domain truncation lacking EGFR exons 2 through 7, which abolishes EGF ligand binding and converts EGFRvIII into a constitutively active protein kinase that upregulates several pathways involved in tumor growth and metastasis (6). We previously developed an EGFRvIII-targeted antibody–drug conjugate, AMG 595, which was evaluated in a phase I study in patients with GBM. Despite evidence for antitumor activity (2/32 patients achieved a partial response and 15/32 patients demonstrated stable disease), further clinical development of AMG 595 was discontinued (8). Chimeric antigen receptor (CAR) T cells binding EGFRvIII (EGFRvIII CAR T cell) have also been evaluated in patients with GBM (9). EGFRvIII CAR T cells demonstrated evidence for on-target activity, including trafficking to and expansion of CAR T cells at tumor sites and direct killing of EGFRvIII-expressing tumor cells, with stable disease achieved by one in 10 patients treated (9). Together, these clinical studies highlight the promise of targeting EGFRvIII and the potential to use immunotherapy for treatment of GBM.

Bispecific T-cell engager (BiTE) molecules (Amgen Inc.) are composed of two single-chain variable fragments (scFv) connected by a short peptide linker. One scFv is derived from an antibody specific to a tumor-associated antigen on tumor cells, and the other scFv binds to CD3 expressed on T cells. Concurrent binding of a BiTE molecule to tumor cells and T cells leads to T-cell activation, T-cell expansion, and redirected T-cell lysis of tumor cells (ref. 10; Supplementary Fig. S1). The first BiTE molecule, blinatumomab, targets CD19 and is the only approved T-cell–bispecific molecule on the global market. Blinatumomab is approved for treatment of B-cell acute lymphoblastic leukemia and has demonstrated clinical activity in non-Hodgkin lymphoma (11, 12). The BiTE molecule, AMG 212, which targets prostate-specific membrane antigen, has shown early evidence for clinical activity in a solid tumor setting (13). These experiences led us to develop AMG 596, a BiTE molecule targeting EGFRvIII, as a novel, targeted immunotherapy for GBM.

AMG 596 is composed of an EGFRvIII scFv derived from AMG 595 connected to an scFv specific for the CD3 molecule expressed on T cells (8, 14). AMG 596 is a fully human BiTE molecule that cross-reacts with cynomolgus monkey CD3. AMG 596 demonstrated potent, specific cell killing activity against EGFRvIII-positive GBM cells in vitro and showed remarkable antitumor activity in an orthotopic xenograft tumor model in mice. In cynomolgus monkeys, AMG 596 was well-tolerated at doses up to 450 μg/kg/day, with no evidence of T-cell activity, consistent with the tumor-specific expression of EGFRvIII. These findings supported the development of AMG 596 as an immuno-oncology therapy for patients with GBM with EGFRvIII-positive tumors.

AMG 596 is a bispecific construct composed of an EGFRvIII-binding scFv at the amino terminus and a CD3-binding scFv at the carboxyl terminus. The binding domains of AMG 596 are cross-reactive to human and cynomolgus monkey targets. However, because EGFRvIII is a tumor-specific antigen, it is not naturally expressed in cynomolgus monkeys. A putative cynomolgus monkey EGFRvIII sequence was derived from comparison of the cynomolgus monkey and human EGFR cDNA sequences and application of the human EGFRvIII mutation pattern to the cynomolgus monkey sequence (Supplementary Fig. S2). AMG 596 was expressed in Chinese hamster ovary (CHO) cells (ATCC) and harvested from the supernatant by a two-step purification process consisting of immobilized metal affinity chromatography followed by size exclusion chromatography, as reported for other BiTE molecules (14, 15). The AMG 596 sequence is disclosed in sequence ID No. 160 in patent No. US 10,519,241 B2; an additional 6× His Tag has been added.

Tumor cell lines, U-251 MG and EGFRvIII-expressing U-251 MG (U-251 MG/EGFRvIII), were described previously (16). Both cell lines were cultured in DMEM (Gibco) containing 10% FBS (Invitrogen) and 100 U/mL penicillin/streptomycin (Gibco). Tumor cell lines, U-87 MG/EGFRvIII and GOS-3/EGFRvIII, were generated by transduction of U-87 MG tumor cells (ATCC) and GOS-3 [German Collection of Microorganisms and Cell Culture (DSMZ)] with human EGFRvIII cDNA, and were cultured in RPMI1640 (Biochrom) and DMEM containing 10% FBS, 100 U/mL penicillin/streptomycin, and 5 μg/mL Blasticidin (Gibco). The tumor cell line, DK-MG (DSMZ), was cultured in RPMI1640 containing 10% FBS, 100 U/mL penicillin/streptomycin, and 1 × nonessential amino acids (NEAA, Biochrom). CHO cells were transduced with wild-type EGFR (HER1), HER2, HER3, HER4, and human and cynomolgus monkey EGFRvIII. CHO cells were cultured in DMEM containing 10% FBS, 10 mmol/L sodium hypoxanthine, 1.6 mmol/L thymidine (1 × HT Supplement, Gibco), and 1 × NEAA (Biochrom). Transduced CHO cells were cultured in DMEM containing 10% FBS, 2 mmol/L l-Glutamine (Biochrom), 100 U/mL penicillin/streptomycin, and 1 × NEAA. T-cell lines, HPB-ALL (DSMZ) and LnPx4119 (described previously; ref. 15), were cultured in RPMI1640 containing 10% FBS and 100 U/mL penicillin/streptomycin. All cells were cultured at 37°C in a 5% CO₂ chamber.

On-cell binding of AMG 596 was assessed on CD3-expressing human and monkey (Macaca mulatta) T-cell lines (HPB-ALL and LnPx4119) and CHO cells expressing human EGFR/HER1, HER2, HER3, HER4, and EGFRvIII, and cynomolgus monkey EGFRvIII by flow cytometry. Cells were incubated with 5 μg/mL of AMG 596 or the Mec14 BiTE molecule specific for the herbicide mecoprop (15), followed by incubations with a biotin-conjugated secondary antibody (Dianova) and APC-conjugated Streptavidin [Becton Dickinson (BD)].

T-cell–dependent cellular cytotoxicity (TDCC) assays were assessed by flow cytometry using human or cynomolgus monkey peripheral blood mononuclear cells (PBMCs) as effector cells and U-251 MG, U-251 MG/EGFRvIII, U-87 MG/EGFRvIII, GOS-3/EGFRvIII, DK-MG, and CHO cells transduced with cynomolgus monkey EGFRvIII (CHO/cynoEGFRvIII) as target cells. PBMC were isolated from human (“Institut für Klinische Transfusionsmedizin und Immungenetik Ulm,” Ulm, Germany) and cynomolgus monkey blood (WIL Research) by Ficoll (Sigma-Aldrich) density gradient centrifugation. Target cells were labeled with Vybrant DiO (Invitrogen) and cocultured with effector cells at effector to target cell (E:T) ratios of 10:1 and increasing concentrations of AMG 596 in RPMI1640, containing 10% FBS and 100 U/mL penicillin/streptomycin, at 37°C in a 5% CO₂ chamber. Target cell lysis was analyzed by flow cytometry as loss of cell membrane integrity (uptake of propidium iodide) after 48 (human assays) and 72 hours (cynomolgus monkey assays). T-cell activation was evaluated by flow cytometry after 48 hours. T cells were identified by antibodies against CD4 and CD8 and their activation status was determined with an anti-CD25 antibody (antibodies from BD). T-cell activation and redirected lysis were analyzed on a FACSCanto II Flow Cytometer (BD) and data were evaluated using FACS Diva Software (BD) and GraphPad Prism (GraphPad Software). Cytokine release was monitored in supernatants of TDCC assays using a Cytometric Bead Array (BD) according to the manufacturer's protocol.

U-251 MG/EGFRvIII cells were cultured in DMEM with 10% exosome-depleted FBS (Gibco). Medium was centrifuged at 500 × g for 5 minutes to remove dead cells and cell debris, followed by filtration through an 0.8-μm filter (Millipore). Microvesicles were pelleted by ultracentrifugation (Optima L-100K Ultracentrifuge, Beckman Coulter) at 100,000 × g for 80 minutes at 8°C. The pellet was washed with cold PBS and collected by ultracentrifugation at 100,000 × g for 8 minutes at 8°C. Purified microvesicles were suspended in either PBS or fresh cell culture medium. The protein concentration of the microvesicle preparation was determined by BCA Protein Assay (Pierce).

U-251 MG or U-251 MG/EGFRvIII cells were lysed in RIPA Buffer (Thermo Fisher Scientific) with 1× Halt Protease and Phosphatase Inhibitor Cocktail (Thermo Fisher Scientific). Protein (15 μg) from microvesicles or from RIPA lysates of U-251 MG or U-251 MG/EGFRvIII cells was resolved by NuPAGE Bis-Tris Mini Gel 4% to 12% (Thermo Fisher Scientific) and transferred to a nitrocellulose membrane. Immunoblotting was carried out using iBind Western System and iBind FD Solution Kit (Thermo Fisher Scientific) following the manufacturer's instructions. Primary antibodies against EGFRvIII (Amgen) or β-actin (Cell Signaling Technology) were detected with IRDye-conjugated secondary antibodies (LI-COR). Protein bands were visualized using LI-COR Odyssey CLx Imaging System (Li-COR).

U-251 MG cells were incubated with various amounts of microvesicles (0–200 μg) for 24 hours. Cell surface EGFRvIII levels were detected by incubation with AMG 596, followed by a secondary fluorescently labeled antibody (Miltenyi Biotec), and analyzed by flow cytometry using an LSR Fortessa and FlowJo Software (BD).

U-251 MG cells were incubated with various amounts of microvesicles (0–200 μg) for 24 hours. Human pan-T cells (AllCells) were added at an E:T ratio of 10:1 with a dose range of AMG 596. Microvesicles were added simultaneously with T cells and AMG 596 to maintain consistent microvesicle concentrations. After 48 hours, T-cell activation was analyzed by flow cytometry, and tumor cell killing was assessed by the CellTiter-Glo Luminescent Cell Viability Assay (Promega).

Antibodies used for analysis of T-cell activation were directed against CD3, CD4, and CD25 (BD), and CD8 (BioLegend). Cell viability was assessed with a Fixable Viability Dye (eBioscience).

All experiments were performed according to the German Animal Protection Law with permission from the responsible local authorities. Animals were cared for in accordance with the Guide for the Care and Use of Laboratory Animals, 8th edition (17). Antitumor activity was evaluated in xenograft tumor models in 6-week-old female NOD.CB17-Prkdc^scid/J mice (NOD/SCID; Charles River Laboratories). Before tumor cell injection, mice were sublethally irradiated (2 Gy) to facilitate and ensure tumor cell growth and human T-cell engraftment. In addition, a single intravenous bolus injection of anti-asialo GM1 rabbit antibody (Biozol) was administered 1 day before the intraperitoneal injection of human T cells (2 × 10⁷ T cells/mouse) to deplete murine natural killer cells. T cells were originated from a single healthy donor and were activated and expanded using a T-cell Activation/Expansion Kit (Miltenyi Biotec).

For the orthotopic tumor model, mice were anesthetized with a Ketamine/Xylazine Mixture (100/8 mg/kg, i.p. injection, CP-Pharma) and received Rimadyl (5 mg/kg, s.c., Pfizer) for analgesia. Mice were fixed in a Stereotactic Device (Kopf Instruments) and the skull was trepanned with a 22 G needle at the injection site (3.5 mm anterior of the interaural line and 2 mm lateral of the midline). A Microliter Syringe (Hamilton) adjusted to the stereotactic device was used to inject U-87 MG/EGFRvIII tumor cells (1 × 10⁵ cells/mouse) into the right cerebral hemisphere. T cells were injected at day 4, and AMG 596, the vehicle control, and the Mec14 BiTE molecule were administered by intravenous bolus into the lateral tail vein once daily for 16 consecutive days starting on day 8. Statistical analysis of survival was performed using the log-rank (Mantel–Cox) test.

For the subcutaneous tumor model, U-251 MG/EGFRvIII cells (1 × 10⁷ cells/mouse mixed with Matrigel, R&D Systems) were injected subcutaneously into the right dorsal flank. T cells were injected at day 9 when tumors had reached a volume of approximately 200 mm³. At day 11, AMG 596 and vehicle were administered by intravenous bolus into the lateral tail vein once daily for 11 consecutive days. Tumor growth was monitored by external caliper measurements, and tumor volumes were calculated using a standard hemi ellipsoid formula: [length (mm) × width (mm)²]/2.

Whole mouse brains from the orthotopic tumor model were embedded in Tissue-Tek OCT Compound (Hartenstein) and cut to 3- to 5-μm-thick sections. Sections were transferred onto slides (Hartenstein), air dried, and fixed in cold Acetone (Sigma-Aldrich). For histologic evaluation, brain sections were stained with hematoxylin (Vector Laboratories) and eosin (Sigma-Aldrich). Evaluation of EGFRvIII and CD3 expression included blocking of endogenous peroxidase activity by a Peroxidase Blocking Reagent (Dako North America), and incubation with 10 μg/mL of murine-antihuman EGFRvIII (NewEast Biosciences) and murine-antihuman CD3 (Sigma-Aldrich) or a murine IgG isotype control antibody (Bio-Rad). Antibody binding was detected by incubation with a peroxidase-labeled polymer conjugated to goat-anti mouse IgG followed by the incubation with a DAB + Peroxidase Substrate (both Dako). Slides were counterstained with hematoxylin, dehydrated in ethanol, cleared in xylol, and covered with Roti HistoKitt II (Carl Roth GmbH). Tissue slides were recorded with a LM2500 light microscope using a DFC295 Camera (Leica Microsystem).

Male cynomolgus monkeys were cared for in accordance with the Guide for the Care and Use of Laboratory Animals, 8th edition (17). Animals were housed individually at an indoor, Association for Assessment and Accreditation of Laboratory Animal Care International–accredited facility, and all research protocols were approved by the Institutional Animal Care and Use Committee. Before continuous intravenous infusion of AMG 596, animals were implanted with a vascular access port/catheter into the femoral vein. The catheter was attached to a battery-powered Infusion Pump (SIMS Deltec) located in the pocket of a jacket placed on each animal, and dosing solutions were administered via continuous intravenous infusion for 7 days. Throughout the study, animals were monitored for general health/mortality and moribundity twice daily. Both before and during dosing, body temperature, body weight, and food consumption were monitored, and hematology, coagulation, serum chemistry, lymphocyte subtypes, and serum cytokine levels were analyzed. At study end, animals were necropsied and macroscopically assessed for noticeable changes. Tissues from selected organs were prepared for subsequent histopathologic examinations.

Key resources of “Materials and Methods,” their sources, and identifiers can be found in Supplementary Table S1.

AMG 596 is a targeted immunotherapy that binds to human EGFRvIII on tumor cells and human CD3 on T cells. It is composed of an EGFRvIII-binding scFv located at the amino terminus that is connected via a short peptide linker to the carboxy terminal CD3-binding scFv (Fig. 1A). AMG 596 is a fully human BiTE molecule that was designed to cross-react with cynomolgus monkey CD3 and with a putative cynomolgus monkey EGFRvIII. Flow cytometry analysis confirmed AMG 596 binding to CHO cells that stably expressed either human or cynomolgus monkey EGFRvIII, but not to CHO cells that did not express the target (Fig. 1B). AMG 596 binding was specific for EGFRvIII as no binding was observed to CHO cells engineered to express either wild-type EGFR (HER1) or the EGFR family members HER2, HER3, and HER4 (Fig. 1B). An anti-EGFR BiTE molecule, based on the sequence of an anti-EGFR antibody binding to both wild-type EGFR and EGFRvIII (15), was used as a positive control. No binding to EGFRvIII was observed with the nontargeting Mec14 BiTE molecule. AMG 596 also demonstrated specific binding to the CD3-expressing human and monkey T-cell lines, HPB-ALL and LnPx4119, by flow cytometry (Fig. 1C).

BiTE immunotherapy acts by directly connecting target cells to T cells, leading to T-cell activation, cytokine release, and redirected lysis. These pharmacologic effects of AMG 596 were evaluated in human and cynomolgus monkey cell–based assay systems to investigate whether cynomolgus monkey could serve as a relevant species for pharmacologic and toxicologic assessments.

AMG 596–induced activation of human and cynomolgus monkey T cells was demonstrated in cocultures containing U-251 MG/EGFRvIII cells and human or cynomolgus monkey PBMC. Flow cytometric analysis revealed that AMG 596 treatment led to dose-dependent upregulation of the cell surface T-cell activation marker, CD25, in both species. No T-cell activation was detected when U-251 MG cells that did not express EGFRvIII were treated with AMG 596, demonstrating the requirement of EGFRvIII expression for AMG 596–mediated T-cell activation (Fig. 2A and B).

The BiTE molecule mechanism of action is associated with dose-dependent release of cytokines (14, 18). Therefore, cytokines were analyzed in supernatants from AMG 596 cytotoxicity assays containing U-251 MG/EGFRvIII cells and human or cynomolgus monkey PBMC. AMG 596 induced a dose-dependent release of cytokines by human and cynomolgus monkey effector cells as exemplified by IFNγ (Fig. 2C and D). With effector cells from both species, IFNγ secretion was observed at similar AMG 596 concentrations and was strictly dependent on the presence of EGFRvIII-positive target cells. No IFNγ was observed in assays with U-251 MG cells that did not express EGFRvIII, confirming that AMG 596 activity requires engagement of both EGFRvIII and CD3 on T cells.

Specificity of redirected lysis by human and cynomolgus monkey effector cells was confirmed in AMG 596 cytotoxicity assays. While EGFRvIII-positive U-251 MG cells were efficiently lysed, AMG 596 treatment did not affect the viability of target-negative U-251 MG cells (Fig. 2E and F). Moreover, cytotoxicity data obtained with different EGFRvIII-expressing GBM cell lines (U-251 MG/EGFRvIII, GOS-3/EGFRvIII, U-87 MG/EGFRvIII, and the EGFRvIII endogenously expressing cell line DK-MG) revealed that T-cell–mediated lysis was comparable between the two species (Fig. 3), although EC₅₀ values calculated from dose–response curves with human T cells were somewhat lower (∼5-fold; Supplementary Table S2), suggesting that AMG 596 is slightly more potent in human assay systems. Finally, AMG 596 cross-reactivity to cynomolgus monkey was tested in cytotoxicity assays with cynomolgus monkey PBMC and CHO/cynoEGFRvIII cells. These assays confirmed the capability of AMG 596 to effectively engage cynomolgus monkey EGFRvIII and CD3 to promote target cell lysis (Fig. 2G), and further supported the use of the cynomolgus monkey as a relevant species for safety testing.

The EGFRvIII antigen is reported to be transferred between tumor cells through microvesicles derived from the plasma membrane (19). To test whether this occurs in our GBM cell line models, we isolated microvesicles from U-251 MG/EGFRvIII tumor cells and used Western blot analysis to confirm the presence of EGFRvIII protein in the microvesicle preparation (Fig. 4A). We then incubated U-251 MG tumor cells with increasing amounts of isolated EGFRvIII-positive microvesicles and evaluated EGFRvIII transfer to the cell surface of the U-251 MG cells by flow cytometry. Increasing levels of cell surface EGFRvIII were detected after incubation with increasing amounts of EGFRvIII-positive microvesicles (Fig. 4B), thus confirming transfer of EGFRvIII to the U-251 MG cells and expression on the cell membrane. To test whether U-251 MG cells could be engaged by AMG 596 after EGFRvIII microvesicle uptake to drive T-cell activation and redirected lysis, we preincubated U-251 MG cells with increasing amounts of EGFRvIII microvesicles and, thereafter, cocultured them with human T cells and AMG 596 for 48 hours. AMG 596 effectively induced upregulation of the T-cell activation marker, CD25, on both CD4⁺ and CD8⁺ T cells (Fig. 4C) and mediated redirected lysis of microvesicle EGFRvIII-transferred U-251 MG cells (Fig. 4D). Both T-cell activation and redirected lysis increased in a dose-dependent manner when higher concentrations of AMG 596 were applied or when U-251 MG cells with higher EGFRvIII levels were used in the assay. Together, these data demonstrate that EGFRvIII can be transferred by microvesicle from EGFRvIII-positive cancer cells to EGFRvIII-negative cancer cells, thereby converting target-negative cells to target-positive cells that can be effectively engaged by AMG 596 immunotherapy.

The antitumor activity of AMG 596 was evaluated in an orthotopic xenograft brain tumor model in NOD/SCID mice. U-87 MG/EGFRvIII cells (AMG 596–induced redirected lysis, T-cell activation, and cytokine production were demonstrated in assays with U-87 MG/EGFRvIII cells; Supplementary Fig. S3) were implanted intracerebrally and tumors were allowed to grow. Human T cells were injected intraperitoneally once before initiation of BiTE molecule treatment. Animals were treated with AMG 596 (0.5 or 0.1 mg/kg), the Mec14 BiTE molecule (0.5 mg/kg, negative control), or vehicle only by daily intravenous bolus injection for 16 consecutive days starting on day 8 after the tumor cells were implanted. Mice treated with AMG 596 showed significantly prolonged survival compared with mice treated with either the Mec14 BiTE molecule or vehicle only (P < 0.001; Fig. 5A). At the end of the study (day 44), eight of 11 and 4 of 10 animals treated with 0.5 and 0.1 mg/kg AMG 596, respectively, were alive without macroscopic brain tumors at necropsy. In contrast, all mice treated with either the Mec14 BiTE molecule or control vehicle did not survive longer than 24 days after study start. These findings were confirmed by IHC and hematoxylin and eosin (H&E) staining of brain sections (Fig. 5B). No visible tumors were observed in AMG 596–treated animals, while large EGFRvIII-positive tumors were observed in vehicle-treated mice. IHC analysis of human T cells in brains of vehicle- and AMG 596–treated animals indicated T-cell infiltration, and possibly a trend toward increased numbers of T cells after AMG 596 treatment (Supplementary Fig. S4).

Furthermore, AMG 596 antitumor activity was also assessed in NOD/SCID mice bearing established subcutaneous GBM tumors. U-251 MG/EGFRvIII tumor cells were subcutaneously injected into NOD/SCID mice and grown to tumor volumes of approximately 200 mm³. Then, human T cells were injected intraperitoneally once, followed by daily intravenous bolus injection of vehicle and AMG 596 (0.5 mg/kg) for 11 consecutive days. AMG 596 induced significant antitumor activity of U-251 MG/EGFRvIII tumors, while a steady increase in tumor volume was observed for vehicle-treated mice until day 22 (averaged tumor volume of 41 vs. 568 mm³ for treatment and vehicle group; Fig. 6). These data again demonstrate the high antitumor activity of AMG 596 as monotherapy against established GBM tumors in mouse models.

An exploratory toxicology study was performed to evaluate EGFRvIII-independent effects that might be attributable to AMG 596 binding to only CD3 on T cells. A 7-day exploratory dose-range–finding study was conducted in which three male cynomolgus monkeys per group were administered vehicle or AMG 596 at doses of 15, 150, or 450 μg/kg/day via continuous intravenous infusion at dose volumes of 10 mL/kg/day. AMG 596 administration was very well-tolerated. Even at the highest dose with a steady-state serum concentration of 355 ng/mL, no clinical signs of toxicity or changes in body temperature or body weight were observed. Strikingly, the serum level of 355 ng/mL was about 15- and 197-fold higher than the effective concentration for 90% response (EC₉₀) values required for redirected lysis in AMG 596 TDCC assays with cynomolgus monkey (EC₉₀ value, 24 ng/mL) and human PBMC (EC₉₀ value, 1.8 ng/mL). In addition, levels of serum cytokines (MCP-1, IL6, IFNγ, TNFα, and IL2), usually an indicator for BiTE molecule activity in vivo, were not detected during the study. Moreover, clinical pathology parameters indicative of an acute phase response, such as C-reactive protein, did not differ between AMG 596- and vehicle-treated animals. Lack of AMG 596–mediated toxicity or activity was accompanied by the absence of histopathologic findings. This study clearly highlights the need for expression of both the target antigen and CD3 to obtain BiTE activity in vivo.

BiTE molecules are a clinically validated modality. The anti-CD19/CD3 BiTE molecule, blinatumomab, was approved by the FDA more than 5 years ago, and since that time several other BiTE molecules have demonstrated clinical activity, including in solid tumors (20). EGFRvIII is a promising immunotherapeutic target for the treatment of GBM owing to its specific expression on tumor cells. As with other BiTE molecules (14, 18, 21), nonclinical characterization of AMG 596 included analysis of in vitro and in vivo properties, as well as toxicologic evaluation in cynomolgus monkeys. Because EGFRvIII expression has not been reported in the cynomolgus monkey, AMG 596 was designed to bind to both human and a putative cynomolgus monkey EGFRvIII protein created by applying the human EGFRvIII mutation pattern to the highly homologous cynomolgus monkey EGFR sequence. AMG 596 possesses almost identical equilibrium KDs for human and cynomolgus monkey antigens (single-digit nmol/L range) with exclusive binding to EGFRvIII and no cross-reactivity to wild-type EGFR or other members of the EGFR family. Most importantly, in vitro human and cynomolgus monkey T-cell responses were comparable in terms of AMG 596–induced T-cell activation, cytokine secretion, and the magnitude of redirected lysis, which was shown to be similarly induced by CD4⁺ and CD8⁺ T cells (Supplementary Fig. S5). Slight differences were observed in EC₅₀ values for redirected lysis (∼5-fold lower in assays with human T cells), which could reflect variability in T cells from different donors, as well as differences between the two species. Similar findings were reported for other BiTE molecules and were considered when modeling human exposures based on toxicology and toxicokinetic data from cynomolgus monkey studies (14). In summary, in vitro data support the use of cynomolgus monkeys as a pharmacologically relevant species for toxicology testing. However, the absence of EGFRvIII expression on healthy cynomolgus monkey tissues indicated that toxicology studies with AMG 596 would investigate only EGFRvIII-independent effects attributable to engagement of CD3 on T cells alone. This mimics the clinical situation in which EGFRvIII is reported to be expressed exclusively on tumors, but not on normal tissues (22–25). In fact, the EGFRvIII exploratory toxicology study showed that, even at high doses, AMG 596 did not induce any signs of activity or toxicity, demonstrating that BiTE molecule activity is not attributable solely to CD3 binding on T cells, but requires the additional concurrent binding of a tumor-associated antigen, in this case EGFRvIII. In addition, the absence of toxicity at high AMG 596 serum concentrations in this study suggested that there was no potential cross-reactivity to EGFR, which has widespread expression on healthy tissues that caused massive cytokine release leading to intolerability in cynomolgus monkeys treated with a cross-reactive EGFR BiTE molecule (15).

In vivo activity of AMG 596 was shown in two xenograft tumor models in NOD/SCID mice. In an orthotopic intracranial tumor model, AMG 596 prolonged survival, with up to 73% of mice surviving to study end. In a second tumor model, AMG 596 treatment induced regression of established subcutaneous GBM tumors. Both studies indicated that AMG 596 can be a promising therapy for GBM.

Unlike the implanted tumors in the xenograft study, GBM tumors in humans have been reported to show heterogeneous EGFRvIII expression (26, 27). In vitro observations from this study showed that AMG 596 only redirected T cells in the presence of EGFRvIII-expressing cells, suggesting that only glioblastoma cells carrying the EGFRvIII mutation are suitable targets for AMG 596–mediated cell lysis, leaving the surrounding EGFRvIII-negative tumor cells unaffected. This view is supported by publications demonstrating that the BiTE molecule–triggered formation of a cytotoxic synapse between tumor and T cells and the subsequent release of cytotoxic granules exclusively induces the death of tumor cells involved in synapse formation (28, 29). However, another EGFRvIII-targeted CD3-bispecific molecule demonstrated potent killing of patient-derived explants ex vivo, even with heterogeneity of EGFRvIII expression (30). We propose an additional potential mechanism of AMG 596–induced redirected lysis of tumor cells that do not endogenously express EGFRvIII. We showed that EGFRvIII-negative tumor cells acquired EGFRvIII on their cell surface after the uptake of EGFRvIII-positive microvesicles isolated from EGFRvIII-expressing tumor cells. AMG 596 was not only able to bind to vesicle-transferred EGFRvIII, but also mediated lysis of EGFRvIII-transferred tumor cells. Comparable results have not been observed with EGFRvIII-negative GBM cells previously cocultured with EGFRvIII-positive GBM cells separated by a cell impermeable membrane (Supplementary Fig. S6), indicating that the microvesicle transfer requires close proximity of the two cell types. Future studies are needed to clarify whether the above-described mechanism is specific to the EGFRvIII target and is exclusively implicated in GBM or whether it is a general feature of solid tumors.

Altogether, our studies demonstrate that AMG 596 is a very potent and specific BiTE molecule with impressive in vivo antitumor activity and high tolerability in cynomolgus monkeys. These properties make AMG 596 a very promising immunotherapy for patients with GBM, a devastating and, thus far, incurable disease. A phase I study with AMG 596 is open (NCT03296696) and includes combination with an anti-programmed cell death (PD)-1 antibody to circumvent immunosuppression potentially caused by the upregulation of PD-1 on BiTE molecule–activated T cells (31).

A. Sternjak reports employment with and stock ownership in Amgen. F. Lee reports employment with and stock ownership in Amgen. O. Thomas reports employment with and stock ownership in Amgen. M. Balazs reports stock ownership in and prior employment with Amgen. J. Wahl reports employment with and stock ownership in Amgen. G. Lorenczewski reports employment with and stock ownership in Amgen. I. Ullrich reports a patent for US20170029512A1 issued and employment with and stock ownership in Amgen. M. Muenz reports a patent for antibody constructs for EGFRVIII and CD3 antibody constructs for EGFRVIII and CD3 issued and employment with and stock ownership in Amgen. B. Rattel reports employment with and stock holder of Amgen. J.M. Ballis reports employment with and holds stock in Amgen Inc. M. Friedrich reports employment with and stock ownership in Amgen.

A. Sternjak: Conceptualization, investigation, methodology, writing–original draft. F. Lee: Investigation, methodology, writing–review and editing. O. Thomas: Investigation, methodology, writing–review and editing. M. Balazs: Investigation, methodology, writing–review and editing. J. Wahl: Investigation, methodology, writing–review and editing. G. Lorenczewski: Investigation, methodology, writing–review and editing. I. Ullrich: Investigation, methodology, writing–review and editing. M. Muenz: Investigation, methodology, writing–review and editing. B. Rattel: Investigation, methodology, writing–review and editing. J.M. Ballis: Investigation, methodology, writing–review and editing. M. Friedrich: Investigation, methodology, writing–review and editing.

We thank Sid Roychoudhury, Sabine Stienen, and Angela Coxon for critical review of the article. Assistance with writing and editing the article was provided by Allison R. Gillies, PhD (ICON), Lee B. Hohaia, PharmD (ICON), and Christopher Nosala, PhD (Amgen, Inc.) and was funded by Amgen Inc. This study was funded by Amgen Inc.

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.