hEGFRvIII:CD3 bi-scFv (50.9 kDa) was produced in a similar manner as previously published.6 11 The most significant change arises from the outsourcing of the production to a clinical research organization to produce cGMP and cGMP-representative hEGFRvIII:CD3 bi-scFv. Briefly, stably transduced Chinese hamster ovary (CHO-S) cells, thawed from a cryopreserved cGMP master cell bank vial, were expanded in shaker flasks until of sufficient density to start the production run. Production of hEGFRvIII:CD3 bi-scFv occurred over 12 days, during which time the CHO-S cells were regularly fed to ensure high viability. After cell clarifications, hEGFRvIII:CD3 bi-scFv was harvested from the supernatant using affinity chromatography, specifically using MabSelect Protein A, Sartobind Q, and SP-Sepharose columns. Virus removal was done via a low pH hold and 20 nm ultrafiltration using a Planova BioEX filtration system. Purified antibody was stored in sterile vials at a maximum concentration of 0.5 mg/mL. All experiments in this work were done with cGMP-representative material to ensure the most accurate and clinically relevant dosing calculations.

A flow cytometry assay was utilized to confirm binding of hEGFRvIII:CD3 bi-scFv to CD3-expressing blood cells and to establish species specificity of the antibody. Healthy peripheral blood mononuclear cell (PBMC) samples for the species beagle, human, CD-1 mouse, cynomolgus, Sprague Dawley rat, New Zealand white rabbit, and pig were obtained frozen from BioIVT (New York, USA). Human PBMC samples were also obtained from healthy donors at Duke University. The binding assay consisted of a two-step binding protocol that ensured both arms of the bispecific antibody were functional at the same time. Specifically, 1 µg of hEGFRvIII:CD3 bi-scFv was incubated with 1×10⁶ to 3×10⁶ PBMCs for 30 min, which results in binding of CD3ε. The samples were washed and then incubated for 30 min with a custom-made fluorescent EGFRvIII petide (PEPvIII)/Streptavidin-PE multimer (PEPvIII multimer), which binds the other arm of the bispecific antibody. After another wash step, the cells were analyzed in a BD Fortessa flow cytometer. Fluorescence data were analyzed using Flowjo. A fluorescent signal indicates that hEGFRvIII:CD3 bi-scFv binds to CD3-expressing cells of that species and that both arms of the antibody are functional.

Flow cytometry was used to confirm binding of hEGFRvIII:CD3 bi-scFv to EGFRvIII expressing tumor cells. Cell lines used included the human malignant glioma line U87-MG, both wild type and engineered to express EGFRvIII (U87-MG-EGFRvIII). The binding assay consisted of three steps. In step 1, around 200,000 target tumor cells were incubated with hEGFRvIII:CD3 bi-scFv for 30 min at 4°C. Following that, cells were washed two times in fluorescence-activated cell sorting (FACS) buffer and incubated with Protein-L-biotin (Genscript M00097) for 30 min at 4°C. Finally, after two more washes, cells were incubated with Streptavidin-PE (Biolegend 405204) for 30 min at 4°C. Cells were analyzed in a BD Fortessa flow cytometer. Fluorescence data were analyzed using Flowjo.

We assessed IFN-ɣ, IL-6, IL-1β, MIP-1α, granulocyte-macrophage colony-stimulating factor (GM-CSF), and tumor necrosis factor (TNFα) release using the Luminex multiplex platform (Millipore HSTCMAG-28SK). Target tumor cells were mixed with human PBMCs at a ratio of 1:20 in a 96-well ultra-low adherent U bottom plate. Human PBMCs alone served as control. hEGFRvIII:CD3 bi-scFv was added to the cells at varying concentrations and the cells were incubated at 37°C. Maximal cytokine release was induced using 50 nM phorbol 12-myristate 13-acetate (PMA). After 24 hours, the plate was spun and supernatant was harvested. Cytokines in the supernatant were measured by the Duke core laboratory RBL Immunology using the Luminex kit.

We assessed hEGFRvIII:CD3 bi-scFv-induced T cell activation by measuring CD25 upregulation and proliferation by measuring 5(6)-carboxyfluorescein (CFSE) dilution. Human PBMCs were labeled with 1 µM CFSE (Invitrogen C34554) for 15 min at 37°C, according to the manufacturer’s recommendations. Target tumor cells were mixed with human PBMCs at a ratio of 1:10 in 96-well U bottom plates. After the addition of hEGFRvIII:CD3 bi-scFv, the cells were placed in an incubator for 5 days at 37°C. Subsequently, the plates were spun, the supernatant was discarded, and the cells were stained with anti-CD3-APC (Biolegend 300412) and anti-CD25 antibody-BV421 (Biolegend 302630). CD25 and CFSE signal was measured using a BD Fortessa flow cytometer.

We assessed in vitro target cell cytotoxicity using chromium release assays. ⁵¹Cr (200 mCi; Perkin Elmer) was used to label 4×10⁶ target cells. Target cells were incubated with 200 µCi of radioactive ⁵¹Cr for 1 hour at 37°C, and the reaction mixture was agitated every 15 min. Labeled cells were washed three times with phosphate-buffered saline (PBS), resuspended in cell culture media, and allowed to rest for 20 min at room temperature. Target cells were washed an additional time to remove free ⁵¹Cr and then incubated with various combinations of effector PBMCs and/or bi-scFvs at various concentrations. Effector cells were plated at an E:T ratio of 20:1 when present. Maximal lysis was induced using a 2.0% solution (v/v) of Triton X-100 (Sigma Aldrich). Cells were incubated at 37°C for 20 hours after which 50 µL of supernatant from each well was collected and combined with 150 µL of OptiPhase Supermix Scintilation Cocktail (Perkin Elmer). Radioactivity released into the supernatant was measured using a 1450 MicroBeta TriLux Microplate Scintillation and Luminescence Counter (Perkin Elmer).

To obtain a MRSD for our first-in-human study of hEGFRvIII:CD3 bi-scFv, we ascertained the minimum active biological effect level based off of a battery of in vitro assays, the pharmacologically active dose in vivo, and the theoretical receptor occupancy of CD3 in humans.

For all our in vitro assays, we calculated the concentration of hEGFRvIII:CD3 bi-scFv that results in 20% of the maximal effective concentration (EC₂₀). The EC₂₀ value is a conservative measurement of biological activity and can be converted into a HED by equating EC₂₀ to the maximum plasma concentration (C_max) in humans. Thus, the human equivalent starting dose was calculated as follows, based on an average human blood plasma volume of 3 L:

Humanequivalentdose=EC20frominvitroassay×3Lplasmavolume

The FIH starting dose equals the HED from the in vitro assay with the lowest EC₂₀. To place the FIH dose into context, we then compared it to the pharmacologically active dose from our murine studies and to the theoretical occupancy of CD3 receptors in humans.

All animal experiments were performed under the protocol A224-18-09 approved by the Duke University Institutional Animal Care and Use Committee. Studies were performed as previously described.6 Briefly, for orthotopic U87-MG-EGFRvIII glioma models, tumor cells were grown and collected in logarithmic growth phase, washed with PBS, and mixed with an equal volume of 10% methyl cellulose. Cell mixtures were loaded into a 250 µL Hamilton syringe and a 25-gage was attached. Using a stereotactic frame, 5 µL of the cell mixture was injected into the right hemisphere—2 mm to the right of the bregma and 4 mm below the surface of the skull at the caudate nucleus—of anesthetized 8 to 12-week-old female mice. A total of 5×10⁴ U87-MG-EGFRvIII cells mixed with 5×10⁴ human PBMCs were implanted into NSG mice (Jackson Laboratory), with 10 mice per group. In the treatment group, 50 µg hEGFRvIII:CD3 bi-scFv in PBS was given via daily intravenous tail vein injection during the first 5 days after tumor implantation. Mice were euthanized at humane endpoints and survival was recorded.

Flow cytometry was used to test the binding characteristics of hEGFRvIII:CD3 bi-scFv to PBMCs from various species (figure 1A) and from three human donors (figure 1B). Since hEGFRvIII:CD3 bi-scFv was expressed without purification tags and there are no known antibodies to the drug, we developed a PEPvIII-Biotin/SA-PE multimer to detect hEGFRvIII:CD3 bi-scFv on the surface of cells. The multimer was constructed with the EGFRvIII antigenic determining peptide, PEPvIII, allowing for simultaneous detection of hEGFRvIII:CD3 bi-scFv binding to the surface of lymphocytes and tumor antigen. Of the seven species tested, hEGFRvIII:CD3 bi-scFv bound only to human PBMCs. The multidonor binding assay showed that hEGFRvIII:CD3 bi-scFv binds human PBMCs across a wide concentration range (figure 1B). The CD3 receptors were saturated at and above ~1×10⁷ pg/mL hEGFRvIII:CD3 bi-scFv and the EC₂₀ is 150 ng/mL.

To assess and confirm functional binding of hEGFRvIII:CD3 bi-scFv to EGFRvIII, we labeled U87-MG-EGFRvIII and U87-MG control cells with our antibody and secondary fluorescent Protein-L. Protein-L binds variable fragments of antibodies and can be used to detect unlabeled molecules. Using flow cytometry, we determined that hEGFRvIII:CD3 bi-scFv binding to EGFRvIII was orders of magnitude more specific than binding to wild type EGFR (figure 2). The EC₂₀ for hEGFRvIII:CD3 bi-scFv binding U87-MG-EGFRvIII is 10,307 ng/mL. A line of best fit, and thus the EC₂₀, for the U87-MG group could not be accurately calculated.

We assessed cytokine release to determine the ability of hEGFRvIII:CD3 bi-scFv to induce target antigen-dependent activation of T cells. Following a 24 hours incubation of hEGFRvIII:CD3 bi-scFv, PBMCs, and human glioma cells U87-MG-EGFRvIII or control U87-MG cells, we harvested the supernatant and measured the concentration of IFN-ɣ, IL-6, IL-1β, MIP-1α, GM-CSF, and TNFα using the Luminex multiplex platform (figure 3). A dose relationship between the amount of hEGFRvIII:CD3 bi-scFv and amount of cytokines released from PBMCs was detected. We determined that for all cytokines, hEGFRvIII:CD3 bi-scFv-induced cytokine release was order of magnitudes higher in the presence of U87-MG-EGFRvIII cells than compared with the control U87-MG cells. In fact, for four of the six cytokines tested, hEGFRvIII:CD3 bi-scFv did not elicit increase cytokine release from the control U87-MG cells above that of background. These data show the specificity of hEGFRvIII:CD3 bi-scFv for its target antigen EGFRvIII. Lines of best fit could be accurately calculated for five out of six cytokines in the experimental PBMC +U87-MG-EGFRvIII group but for none of the cytokines in the two control groups (PBMC +U87 MG and PBMC alone). In the experimental group, the cytokine release EC₂₀ concentrations hEGFRvIII:CD3 bi-scFv for IFN-ɣ, IL-6, MIP-1α, GM-CSF, and TNFα were 42.3, 12.5, 14.7, 27.6, and 16.5 ng/mL, respectively.

We assessed the ability of hEGFRvIII:CD3 bi-scFv to induce T cell activation and proliferation by coincubating hEGFRvIII:CD3 bi-scFv, human PBMCs, and U87-MG-EGFRvIII or U87-MG cells. After 5 days, T cell activation was analyzed by measuring CD25 expression and T cell proliferation was assessed by measuring CFSE incorporation. hEGFRvIII:CD3 bi-scFv activated and induced proliferation of T cells in a target antigen-specific and dose-dependent manner (figure 4). The EC₂₀ concentration of hEGFRvIII:CD3 bi-scFv for T cell activation and proliferation is 1021 and 70.2 ng/mL, respectively. No lines of best fit could be accurately calculated for the U87-MG control group.

We previously showed that hEGFRvIII:CD3 bi-scFv induces tumor cell lysis in a dose-specific and antigen-specific manner.6 In those studies, chromium cytotoxicity assays were performed with two glioma cell lines engineered to express EGFRvIII, U87-MG-EGFRvIII and D54-EGFRvIII, and with a patient-derived cell line that naturally expresses EGFRvIII, D270-MG.

Using cGMP-representative hEGFRvIII:CD3 bi-scFv, we repeated the cellular cytotoxicity assay with U87-MG-EGFRvIII and U87-MG control cells (figure 5). As previously shown, tumor cell killing was specific to the EGFRvIII antigen and dose dependent on hEGFRvIII:CD3 bi-scFv. The EC₂₀ corresponds to 1.34 ng/mL.

As recommended by drug regulatory agencies, we used the EC₂₀ of hEGFRvIII:CD3 bi-scFv from all of our in vitro experiments to determine the human starting dose based on a MABEL approach (table 1). In the following sections, we then compare this dose to a HED of our effective in vivo concentration and ensure that the theoretical human CD3 receptor occupancy is below standard values.

To calculate the human starting dose, we took the concentration where hEGFRvIII:CD3 bi-scFv showed 20% pharmacologic activity (ie, the EC₂₀) for each assay and calculated the amount of drug it would take to reach that concentration in a human, based on an average plasma volume of 3 L and an average patient body weight (BW) of 70 kg (table 1). The MABEL and safe starting dose was thus represented by the most sensitive assay or smallest starting dose.

Our most sensitive assay, based on the lowest EC₂₀ concentration, was the tumor cell cytotoxicity assay (figure 5). With an EC₂₀ of 1.34 ng/mL, this corresponds to a human starting dose of 4020 ng per patient or 57.4 ng drug/patient kg.

We previously showed that hEGFRvIII:CD3 bi-scFv effectively treats patient-derived and murine EGFRvIII-positive malignant glioma in both subcutaneous and orthotopic models.6 To show equivalency of cGMP-representative hEGFRvIII:CD3 bi-scFv in vivo, we performed a xenograft experiment in immunodeficient NSG mice (figure 6). Mice were coimplanted orthotopically with human malignant glioma U87-MG-EGFRvIII and human PBMCs. Subsequently, mice in the treatment group were injected intravenously daily for 5 days with 2.5 mg/kg hEGFRvIII:CD3 bi-scFv. Treatment with clinical grade drug significantly extended survival compared with the control group (p=0.0079). Furthermore, a single-dose toxicity study of hhEGFRvIII:CD3 bi-scFv in human CD3 transgenic mice showed that the drug was safe up to at least a dose of 10 mg/kg (maximum tested dose; data unpublished). Overall, based on the data presented here and our other published studies in various glioma and mice models, we see robust treatment efficacy when mice are administered daily doses of either 2.5 or 5 mg/kg hEGFRvIII:CD3 bi-scFv.6 19

Based on this collective data, a daily dose of 2.5 or 5 mg/kg hEGFRvIII:CD3 bi-scFv is efficacious in mouse models. Using the average concentration at steady state (C_ave,ss) of this dosing regimen in immunocompetent C57BL/6 mice, a human equivalent dose was calculated from the following equation, using predicted human clearance values:

Dose=CLhu×Ceff,ss×Tau,

where CL_hu is the predicted human clearance, C_eff,ss the predicted efficacious concentration at steady state, and Tau the dosing interval.

The predicted human clearance value was based on the clearance of the clinically approved bi-scFv blinatumomab, which has similar architecture and pharmacokinetic properties. Blinatumomab’s clearance is 43 mL/h/kg in adults.20 The C_eff,ss was assumed to be equal to the average concentration at steady state (C_ave,ss) based on our pharmacokinetic study using the 5 mg/kg dose in human CD3 transgenic mice (0.583 µg/mL).11 A Tau of 24 hours was used.

Using these parameters, a human equivalent dose of 0.6 mg drug/kg of BW was calculated. In comparison, our calculated MABEL dose of 57.4 ng drug/kg of BW is 10,400-fold lower than the murine in vivo efficacious dose, showing a clear safety margin of the MABEL approach.

Theoretical receptor occupancy of target antigens constitutes an important consideration for dose selection. We assessed the theoretical receptor occupancy of hEGFRvIII:CD3 bi-scFv based on our MABEL-based starting dose. As outlined by Lowe et al, there are two commonly used algorithms for calculating drug-target occupancy.21 Method 1 is based on the Michaelis-Menten equilibrium.

Method 1: %RO=[Ab][Ab]+KD×100%

where RO is the receptor occupancy, Ab is the concentration of antibody (which corresponds to our MABEL-based starting drug concentration of 1.34 ng/mL (26.3 pM)) and K_D is the equilibrium dissociation constant.

However, if there are known to be significant quantities of target receptors in the system, then the Michaelis-Menten equation is no longer valid. In these situations, Method 2, which corrects for the high target expression levels, can be applied.

Method 2:

%RO=(KD+TD+TT)−((KD+TD+TT)2−4×TD×TT)1/22×TT×100%

where RO is receptor occupancy, K_D is the equilibrium dissociation constant, TD is total drug concentration (which corresponds to our MABEL-based starting drug concentration of 1.34 ng/mL (26.3 pM)), and TT is the total target concentration.

The TT concentration was calculated as follows:

TT=n1∗n2NA

where n₁ is the number of CD3 receptors per cell (6.11×10⁴), n₂ is the number of T cells per volume (1.3×10⁹ L⁻¹), and N_A is the Avogadro constant.

Using surface plasmon resonance, we previously determined the target binding kinetics of hEGFRvIII:CD3 bi-scFv.6 The association rate constant (k_a), dissociation rate constant (k_d), and equilibrium dissociation constant (K_D) for EGFRvIII were 4.45×10⁴ M⁻¹s⁻¹, 1.24×10⁻³ s⁻¹, and 2.78×10⁻⁸ M, respectively. The association rate constant (k_a), dissociation rate constant (k_d), and equilibrium dissociation constant (K_D) for CD3 epsilon were 5.35×10⁴ M⁻¹s⁻¹, 8.36×10⁻⁴ s⁻¹, and 1.56×10⁻⁸ M, respectively. No binding to EGFR wild type was detected. However, since hEGFRvIII:CD3 bi-scFv is a bispecific antibody that requires simultaneous engagement of both target antigens expressed on different cell types (EGFRvIII on tumor cells and CD3 on T cells), one cannot calculate the receptor occupancy of both targets at the same time. Furthermore, since EGFRvIII expression is restricted to tumor cells within the brain (and thus behind the blood-brain barrier), the concentration of hEGFRvIII:CD3 bi-scFv available to bind EGFRvIII is still unknown. Thus, we restricted our receptor occupancy calculations to the ubiquitous CD3 receptor.

According to both methods, the theoretical CD3 receptor occupancy is 0.17% for hEGFRvIII:CD3 bi-scFv based on our MABEL-based starting dose.