The pertuzumab (anti-HER-2) sequence was obtained from the National Center for Biotechnology Information. The mouse SP34 anti-hCD3 mAb sequence has been published.16 24 The pertuzumab IgG1 mAb, SP34 IgG1_N297Q mAb, the IgG1AAQ BsAb, and tandem Fab heavy chains (HCs) were generated using overlapping PCR and plasmid constructs published previously.16 22 PCR products were cloned into a cytomegalovirus promoter-driven mammalian expression vector (Lonza) using the Clonables Kit (Novagen) and existing HindIII (5’) and EcoRI (3’) expression cassette restriction sites. The entire coding DNA sequence of the HER-2/CD3 tandem Fab HC, was generated using overlapping PCR with a (G₄S)₃ 15 amino acid (aa) linker between the Fab HCs and cloned directly into the same mammalian vector. The pertuzumab and chimeric SP34 light chains (LCs) containing the Fab specificity mutations were synthesized similarly. All constructs utilized a murine kappa leader signal sequence. For generating HC and LC mutants, the QuikChange II Site Directed Mutagenesis Kit (Agilent) was used per manufacturer’s instructions. Plasmid ligations, transformations, DNA preparations were performed using standard molecular biology protocols.

The mAbs and BsAbs were expressed in Chinese Hamster Ovary (CHO) cells as described.25 Briefly, for the HER-2/CD3 IgG BsAbs, two HC plasmids and two LC plasmids were co-transfected at 2 µg DNA/mL culture using a 1:1:1:1 ratio. For the HER-2/CD3 tandem Fab, one HC plasmid and two LC plasmids were co-transfected at 3 µg DNA/mL culture using a 1:1:1 ratio. Transfected CHO cells were grown for 5 days at 37°C in a 5% carbon dioxide (CO₂) incubator while shaking at 125 rpm. Supernatants were harvested by centrifugation at 10 000 rpm for 5 min and filtered through 2 µm filters.

Purification, characterization, and pharmacokinetics evaluation are described in online supplemental methods.

Bio-layer interferometry (BLI) experiments were performed on an Octet RED96 system (FortéBio). All steps of the assay were performed at 30°C with a plate shaking speed of 1000 rpm. Anti-human Fc biosensors were presoaked in Octet kinetic buffer (FortéBio, cat#18–1105) for 5 min prior to measurement. Next, biosensors were collected by the instrument and dipped into the first row of a black 96-well plate containing 200 µL solutions with 10 µg/mL pertuzumab variant (Loading) for 60 s. Biosensors were then moved to a new row containing Octet buffer to generate a baseline for subtraction. Next, biosensors were moved to solutions containing the soluble human HER-2 extracellular domain (Speed Biosystems, cat# YCP1045) for 300 s (Association Phase) followed by transfer to wells containing Octet kinetic buffer for 600 s (Dissociation Phase). The binding data was analyzed using Octet data analysis software and fit globally to a 1:1 binding reaction.

For all experiments, data were analyzed using FlowJo V.10, and, unless specified otherwise, acquisition was performed on a Becton Dickinson LSRFortessa flow cytometer using BD FACSDiva software.

Flow cytometric binding potency on SKOV3 and Jurkat tumor cells. Adherent SKOV3 ovarian cancer cells (ATCC Cat#HTB-77) cultured in RPMI 1640 /10% fetal bovine serum (FBS) Corning/gentamicin (Gibco) at 37°C, 5% CO₂ were resuspended using Accutase (Innovative Technologies Cat#AT104). Jurkat cells (ATCC Cat#TIB-152) grew in suspension. Cells were centrifuged at 170 g for 7 min and washed once with phosphate-buffered saline (PBS). All subsequent steps were performed on ice. ‘Wash buffer’ contained PBS, 12% FBS, 0.05% Sodium Azide (NaN₃) and 10% Normal Goat Serum. Cells were resuspended in Wash buffer and blocked for 15 min. Tumor cells (0.5×10⁶ cells/well) were added to 96-well plates (Corning 3799). BsAbs and control mAbs were added to the wells starting at 100 nM for Jurkat cells and 67 nM for SKOV3 cells (except 100 nM for Tandem Fab), titrated using threefold dilutions, and incubated for 45 min. Cells were washed twice and supernatants were aspirated before adding secondary detection reagents in Wash buffer for 45 min. R-Phycoerythrin (PE)-conjugated Goat anti-Human kappa (Southern Biotechnology Cat#2060-09) was used for SKOV3 cells. Goat anti-Human lambda-PE (Southern Biotechnology Cat#2070–09) was used for Jurkat cells, except for the pertuzumab control where Goat anti-Human kappa-PE was used. Cells were washed twice again and resuspended in Wash buffer with PI (Molecular Probes Cat#P3566) before analysis.

Because affinity studies with monomeric CD3 has given us misleading information in the past, we have found more relevant to measure affinity by competition with Jurkat cells. Competition flow cytometry with SP34-PE mAb. Jurkat cells were resuspended in Blocking buffer (Wash buffer supplemented with Human BD Fc Block (BD Biosciences Cat#564 220)) for 15 min, pelleted, washed twice, and 50 µL of cells (0.5×10⁶ cells/well in Wash buffer) added to 96-well plates (Corning 3799). Next, cells were centrifuged at 170 g for 7 min and Wash buffer was aspirated. In a separate 96-well plate (Corning 3879), the BsAb and control mAbs were added to wells starting at 1000 nM and titrated using threefold dilutions (working concentrations were 2×). BsAb and mAb dilutions were mixed with equal volume of SP34-PE (final concentration of 3 µg/mL, BD Biosciences Cat#552127). This sample mixture was added to the cells and incubated 45 min. Cells were washed three times before resuspension in Wash buffer with PI (Molecular Probes Cat#P3566) and then analyzed on a BD LSRII flow cytometer.

Flow cytometric evaluation of cell surface proteins on T cells 48 hours after mixing SKOV3 cells and biologic test articles is further described in online supplemental methods. Briefly, human T cells (ALLCELLS frozen normal human peripheral blood CD3⁺ Pan T cells, Cat# PB009-1F, Lot #A5715/Donor 1 and A5647/Donor 2) were removed from redirected lysis assay plates after co-cultured for 48 hours in RPMI 1640 /10% FBS Corning/gentamicin (Gibco) at 37°C, 5% CO₂ with SKOV3 tumor cells, various IgGs and IgG BsAbs (E:T 10:1). Human T cells were stained with Alexa-488 Mouse Anti-Human CD4 (BD Biosciences Cat#557695) and Pacific Blue Mouse Anti-Human CD8 (BD Biosciences Cat#558207). Labeled T cell activation markers: APC-Cy7 Mouse Anti-Human CD69 (BD Biosciences Cat#557756) and APC-Mouse Anti-Human CD25 (BD Biosciences Cat#555434) were used with Donor 1 and incubated for 45 min. Labeled checkpoint inhibitors: PerCP-eFluor 710 Mouse Anti-Human CD279 (PD-1) (Invitrogen cat#46-2799-42), APC-Mouse Anti-Human CD152 (CTLA-4) (Biolegend Cat#349908) and PE-Mouse anti-Human CD137 (4-1BB) (BD Biosciences Cat#561701) were used with Donor 2. Finally, the cells were resuspended in Wash buffer with 1:1000 PI (Molecular Probes Cat#P3566) and analyzed along with OneComp eBeads (eBioscience Cat#01-1111-42).

Quantitation of HER-2 and PD-L1 receptors per cell by flow cytometry is further described in online supplemental methods. The SKOV3 (ATCC Cat#HTB-77) and OVCAR3 (ATCC Cat#HTB-161) adherent tumor cells (RPMI 1640 /10% FBS (Corning), gentamicin (Gibco), 37°C, 5% CO₂) were stained for expression of PD-L1 and resuspended in LIVE/DEAD stain PI (Molecular Probes Cat#P3566) to assess cell viability. Quantibrite PE beads (BD Bioscience Cat# #340495) were used to determine the number of cell surface HER-2 receptors/tumor cell according to the manufacturer’s instructions.

NSG-His mice engrafted with human T cells were generated in the Stem Cell and Xenograft Core at the University of Pennsylvania. Briefly, immunodeficient NSG mice between 7–9 weeks of age were injected intravenously with 30 mg/kg of busulfan to condition their bone marrow niche. Mice were injected intravenously with ≥100 k cord blood-derived CD34⁺ cells, 22–26 hours later. After 12 weeks, reconstitution levels were assessed and only mice having a sufficient level of T cells (≥5% huCD3+/huCD45+) were retained for in vivo studies.

Xenograft tumors were established by subcutaneous injection of 10⁶ SKOV3 cells expressing Luciferase (SKOV3-Luc) into the flanks of the NSG-His mice. Mice received 50 mg of intravenous immunoglubulin (IVIG, Privigen, Behring) intraperitoneally, 24 hours before tumor cell inoculation. At indicated days, mice received intravenous or intraperitoneal injections of control IgGs or BsAbs. Tumors were measured two times per week with caliper, and volumes were calculated as V=1/2×length×width×height. Tumor implantation was verified by bioluminescence imaging before treatment with BsAbs or control antibodies. Bioluminescence imaging was performed with a Lumina IVIS imaging system and quantified with the Living Image software (PerkinElmer, Waltham, Massachusetts, USA). Mice were injected intraperitoneally with 150 mg/kg D-luciferin (Caliper Life Sciences, Hopkinton, Massachusetts, USA) and imaged under isoflurane anesthesia at the peak of photon emission.

Mice were euthanized when tumor diameter was ≥2 cm or if they exhibited any sign of suffering that could not be relieved.

Statistics are calculated for each of the tumor models’ Max Radiance or Caliper data by transformation to a log scale to equalize variance across time and treatment groups. The log radiance data are analyzed with a two-way repeated measures analysis of variance, with a Spatial Power correlation structure, by time and treatment using the MIXED procedures in SAS software (V.9.3). Treated groups are compared with the control group at each time point. The MIXED procedure is also used separately for each treatment group to calculate least squared means and SEs at each time point. Both analyses account for the autocorrelation within each animal and the loss of data that occurs when animals with large tumors are removed from the study early.

We first investigated potency and pharmacokinetic differences between different formats of HER-2/CD3 BsAbs. We chose pertuzumab (anti-HER-2) and SP34 (anti-CD3) as the parental mAbs for constructing the BsAbs.16 26 Three BsAb formats were generated including a tandem scFv (BiTE format), a tandem Fab, and a human IgG1 rendered effectorless by mutating Leu234 and Leu235 both to alanine and Asn297 to glutamine (backbone denoted hIgG1AAQ, figure 1A).27 The tandem scFv format did not express stably and only fragments were ever recovered (figure 1B). With the more stable Fab components,28 the tandem Fab and the IgG BsAb expressed well and assembled properly (figure 1B). As assessed using an HPLC method with an in-line electrospray mass spectrometer, both the tandem Fab and IgG BsAbs were >90% correctly assembled (figure 1C). Thus, while the BiTE format did not yield properly folded material, the stable tandem Fab and IgG BsAbs were suitable for further functional characterization.

Next, the pharmacokinetics properties of the two HER-2/CD3 BsAbs were tested. Male Balb/c mice were intravenously administered with 10 mg/kg of the BsAbs or SP34 hIgG1_N297Q control mAb. Pertuzumab IgG1 data was published previously.22 Neither arm of the BsAbs have reactivity to the mouse antigens (data not shown). As expected, the presence of the IgG Fc prolonged the serological persistence of the HER-2/CD3 hIgG1AAQ BsAb. The half-life was roughly 8 days, similar to that of the control mAbs (figure 1D). The tandem Fab HER-2/CD3 BsAb lacking an IgG Fc was cleared rapidly. Six hours after administration, >90% of the tandem Fab BsAb had cleared from the serum (figure 1D).

Tumor cell lines expressing varied levels of the HER-2 antigen were used to evaluate the ability of the HER-2/CD3 BsAbs to redirect primary human T cells to kill tumor cells. First, the BsAbs were assessed for their ability to bind both HER-2^HI SKOV3 tumor cells and CD3⁺ Jurkat cells by flow cytometry (figure 2A and online supplemental figure 1A,B). Both BsAbs bound both cell types in a dose-dependent manner. As expected, pertuzumab only bound SKOV3 cells while SP34 only bound Jurkat cells (figure 2A and online supplemental figure 1A,B). Three tumor cell lines, SKOV3, OVCAR3, and OVCAR5, expressing high to low levels of human HER-2, respectively (quantified using BD Quantibrite Beads, online supplemental figure 2A,B) were subjected to human T cells (E:T 10 T cells/) and titrated amounts of either of the BsAbs or their parental mAbs. Adherent mouse 3T3 cells (human HER-2 negative) were used as a negative control. After 48 hours, both BsAbs potently (EC₅₀ <10 pM) induced lysis of SKOV3 and OVCAR3 tumor cells (figure 2B). The difference in potency was much smaller than differences observed between the two formats with epidermal growth factor receptor (EGFR)/CD3 BsAbs,16 indicating that BsAb format does not generally lead to substantial differences in potency. Interestingly, the primary T cells displayed an allogeneic response to OVCAR3 when stimulated by SP34 without TAA-mediated redirection (figure 2B). It is unclear why, but a possibility is that with the donors used, there may be more allogeneic reactivity with this cells line. SP34 demonstrated the typical bell-shaped activation curve known for CD3 agonist mAbs.29 30 The IgG BsAb had little activity towards HER-2^LO OVCAR5 cells and no activity towards the HER-2^neg 3T3 cell line (figure 2B). After 48 hours, the SP34 control mAb led to weak activation of the T cells with an slight increased of CD69 expression (but not CD25) regardless of the expression of HER-2 antigen, while the BsAbs strongly activated T cells only in the presence of HER-2^HI cells with strong upregulation of both CD69 and CD25 (online supplemental figure 3).

The two BsAbs were next evaluated for their in vivo T cell redirected lysis activity against SKOV3-Luc tumor cells implanted in human CD34⁺ stem cell engrafted NSG mice (NGS-His) with evidence of developed populations of circulating human T cells. A transduced firefly luciferase gene in SKOV3-Luc cells enables their tracking using bioluminescent imaging. Given the short half-life of the tandem Fab BsAb, we first compared the BsAbs using daily dosing starting 1 day after tumor cell inoculation. Under daily dosing conditions, both BsAbs significantly inhibited tumor growth (figure 2C). Tumor outgrowth was observed for all control mice (13/13) and all mice treated with the combination of pertuzumab hIgG1 and SP34_IgG1_N297Q (5/5) (figure 2C). Based on caliper measurements, the IgG BsAbs eliminated tumor cells with late tumor outgrowth in only a single mouse each group (1/5) (online supplemental figure 4A).

Next, the BsAbs’ ability to eliminate tumor outgrowth was evaluated following a single injection on day 1. Sera were collected at different time points after BsAb administration. The IgG BsAb persisted in serum as observed in Balb/c mice (online supplemental table 1 and figure 1D). BsAbs eliminated the initial outgrowth of tumor, suggesting that the majority of T cell redirected lysis activity occurs early in the murine model (figure 2C). Eventual tumor outgrowth was evaluated by caliper measurement. After 50 days only two out of five mice treated with the IgG BsAb demonstrated outgrowth, suggesting a benefit to the persistence of the IgG BsAb (online supplemental figure 4B). Elimination was dependent on the expression on the target cells as the IgG BsAb had no effect on the outgrowth of the HER-2^LO OVCAR5 tumor (online supplemental figure 5).

Next, the IgG BsAb’s affinity to HER-2, CD3, or both arms was attenuated to determine the impact on T cell redirection. By evaluating the structure of pertuzumab bound to HER-2,31 a small library of approximately 20 point mutations was designed to lower pertuzumab’s affinity. The mutations were generated in the pertuzumab IgG format, expressed at small scale in CHO cells and tested for their affinity towards soluble, monomeric HER-2 using Bio-layer interferometry (B-LI). Three individual point mutations reduced the affinity of pertuzumab to HER-2 across two orders of magnitude. D31S and S54G reduced the affinity from 8 nM (wild-type) to 56 and 70 nM, respectively (Kabat numbering, figure 3A),32 while N53D reduced affinity ~100-fold with very poor binding detected by B-LI. These three mutations were chosen for further study.

SP34 affinity variants were chosen rationally by making single mutations in heavy chain complementarity determining region 3 (HCDR3) since antibody HCDR3 residues are typically at the center of the epitope.33 We evaluated the relative affinities of SP34 variants in the IgG BsAb format by competition flow cytometry with wild-type, labeled SP34 on Jurkat cells (figure 3B). Four mutations reducing CD3 affinity, N97S (>21-fold), S100aA (fourfold), V100cT (sixfold), and F100fH (13-fold) were chosen for further study.

T cell redirection potency of affinity-attenuated IgG BsAbs was evaluated on both SKOV3 cells, expressing approximately 2×10⁵ HER-2 receptors per cell, and OVCAR3 cells expressing approximately 2×10⁴ HER-2 receptors per cell, in vitro. The two cells lines demonstrated an initial difference in susceptibility to redirected cells lysis that was about 100-fold, much more importance that their difference in level of HER-2 expression. On SKOV3 cells, reducing CD3 affinity by as much as 13-fold (IC₅₀ ~600 nM), had little impact on T cell lysis potency at the 10:1 E:T ratio, suggesting that HER-2 engagement enabled significant avidity towards CD3 on T cells. On HER-2^LO OVCAR3 cells, decreasing CD3 affinity led to reduced potency (figure 4A). As expected, reducing HER-2 affinity, led to decreases in BsAb potency on both tumor cell lines (figure 4B) likely due to an intrinsically fast off-rate for the wild-type SP34 arm.16 The redirected lysis potency of the dual HER-2 and CD3 affinity-attenuated IgG BsAbs were nearly indistinguishable from the IgG BsAbs with only HER-2 affinity attenuation when tested on SKOV3 HER-2^HI cells (figure 4). On HER-2^LO OVCAR3 cells; however, the affinity attenuation of both HER-2 and CD3 combined to reduce the potency of the IgG BsAbs (figure 4C). Thus, while the potency of the IgG BsAb is driven primarily by HER-2 potency on HER-2^HI tumor cells, reducing CD3 affinity can reduce the potency of these BsAbs towards HER-2^LO cells (as expected on normal cells), potentially improving their overall safety profile.

Next, a subset of affinity-attenuated IgG BsAbs was evaluated for their ability to kill HER-2^HI SKOV3-Luc tumor cells in vivo. BsAbs with attenuated affinity for HER-2, CD3, or both were compared with the wild-type, high-affinity IgG BsAb. Because of the limited number of NSG-His mice that can be engrafted with the CD34⁺ cells obtained from one cord blood, not all constructs could be tested in vivo. As observed previously, the wild-type IgG BsAb significantly reduced SKOV3-Luc cell growth in vivo (figure 5A). Strong attenuation of HER-2 binding affinity (~100-fold) eliminated BsAb activity in vivo (figure 5A), while it only impacted the potency in vitro (figure 4B). There are many possible reasons for this, including differences in E:T ratios, inadequate tissue diffusion of the BsAbs, or poor tumor mass penetration by T cells, needs for higher doses in vivo or different dose frequency. Moderate HER-2 affinity attenuation (~10-fold) led to a significant reduction in tumor growth versus the vehicle for all imaging measurements after day 13 except for day 39 (p=0.093) and on all caliper measurements past day 13 (figure 5A). The high affinity HER-2 (8 nM)/low affinity CD3 (600 nM) IgG BsAb significantly reduced tumor growth versus vehicle for all measurements past day 6 except for day 39 (p=0.082) based on imaging and on all measurements past day 13 based on caliper measurements (figure 5A). The extremely low CD3 affinity for this BsAb, even with high HER-2 affinity, likely impacted the BsAb’s potency. Overall, however, the in vivo efficacy did correlate with the attenuated-affinity BsAb results in vitro.

Interferon (IFN)γ levels were measured in the sera of mice for those groups whose BsAbs achieved significant reductions in tumor growth. IFNγ levels were high in many (but not all) mice in the high affinity HER-2/high affinity CD3 BsAb group 2 days after dosing, but returned to baseline by the end of study (online supplemental table 2). The high affinity HER-2/low affinity CD3 BsAb group showed low IFNγ in serum at both time points (online supplemental table 2). Interestingly, the low affinity HER-2/high affinity CD3 BsAb group had low IFNγ 2 days after dosing, but still high IFNγ in multiple mice at the end of study (online supplemental table 2). However, given the potential for high cytokine release with both IgG BsAbs with high affinity to CD3, we decided to focus on a low CD3 affinity BsAb for further studies.

A dosing study was initiated to determine the relationship between efficacy and dose for a CD3 affinity-attenuated IgG BsAb in the HER-2^HI SKOV3-Luc in vivo model. We used an IgG BsAb with high affinity to HER-2 (8 nM) and low affinity to CD3: the SP34_VH_V100cT mutation (280 nM IC₅₀), because the SP34_VH_F100fH mutation (600 nM IC₅₀) did not fully inhibit tumor growth at the relatively high dose of 5 mg/kg (figure 5A). Overall, the study demonstrated strong antitumor activity for this low CD3 affinity CD3 BsAb with significant decreases in tumor growth observed at doses as low as 0.12 mg/kg and just missing significance at 0.024 mg/kg (p=0.052) (figure 5B). Interestingly, maximal efficacy was observed at the highest dose of 15 mg/kg, but with near maximal efficacy observed for both 3 mg/kg and 0.6 mg/kg. No bell-shaped efficacy curve was observed as previously described for high affinity CD3 mAbs or BsAbs.19 34

Molecular immunology aspects of the HER-2/CD3 IgG BsAbs functions were assessed from multiple angles in vitro. First, we tested for upregulation of various immune markers during redirected lysis assays with HER-2^HI SKOV3 cells. Control mAbs, the HER-2/CD3 IgG BsAbs with wild-type affinity and 13-fold reduced affinity (SP34_VH_F100fH) to CD3 were assessed for their activation potential across the critical concentration range (0.1, 0.01, and 0.001 nM). T cells from these cultures were collected at 48 hours and tested for the upregulation of activation markers CD69 and CD25, immune checkpoint inhibitors PD-1 and membrane CTLA-4, and the costimulatory protein 4-1BB. No upregulation of any of the five markers was apparent in cultures containing either the IgG1 control or pertuzumab IgG (table 1, online supplemental figures 6 and 7). SP34 led to upregulation of CD69, fractional upregulation of CD25 and 4-1BB, and no upregulation of PD-1 or CTLA-4 (table 1 and online supplemental figure 7). Both the high and low CD3 affinity BsAbs induced strong upregulation of CD69, CD25, PD-1, and 4-1BB that diminished as the concentration was lowered (table 1, online supplemental figures 5 and 6). Only the high affinity HER-2/CD3 IgG BsAb upregulated CTLA-4 on a small fraction of T cells (table 1, online supplemental figure 7).

Next, both HER-2^HI SKOV3 and HER-2^LO OVCAR3 tumor cells were investigated for upregulation of PD-L1 during T cell redirected lysis assays using 1 nM levels of both high CD3 affinity (45 nM IC₅₀) and low CD3 affinity (F100fH, 600 nM IC₅₀) BsAbs. Both cell lines expressed PD-L1 endogenously at a low level; therefore, BD Quantibrite beads and flow cytometry were used to quantify PD-L1 receptor numbers. Both cell lines express PD-L1 endogenously at 2000–4000 receptors per cell (table 2 and online supplemental figure 8). After a 24-hour incubation with T cells and the IgG BsAbs (both high and low affinity to CD3), PD-L1 was upregulated approximately eightfold on SKOV3 cells, but only marginally on OVCAR3 cells (table 2 and online supplemental figure 8) even though redirected lysis was observed on both cell lines.

T cell activation by major histocompatibility complex (MHC)/antigen recognition is known to induce their release of cytokines, such as IFNγ, that can upregulate PD-L1 on tumor cells35; therefore, cytokine release into the SKOV3/T cell culture supernatants when treated with controls or the HER-2/CD3 BsAbs (both high and low affinity) was assessed in vitro. Upregulation of multiple cytokines, most predominately IFNγ, was observed in the cell culture supernatants (online supplemental figure 9). Cytokine release correlated with the level of tumor cell killing observed for the high and low affinity BsAbs (figure 4 and online supplemental figure 9). Supernatants from the redirected lysis cultures were collected by centrifugation and applied to fresh SKOV3 or OVCAR3 cells. The culture supernatants induced PD-L1 upregulation on SKOV3 cells, and to a lesser extent on OVCAR3 cells, similar to that observed in the redirected lysis assays with T cells (table 2 and online supplemental figure 7) suggesting that soluble factors in the media, likely IFNγ, are inducing PD-L1 upregulation.

Given the upregulation of both PD-1 and 4-1BB on T cells and PD-L1 on SKOV3 cells observed in vitro with the HER-2/CD3 IgG BsAbs, an in vivo study was performed to investigate the impact of combining the HER-2/CD3 (SP34_V100cT, IC₅₀=280 nM) IgG BsAb with either in-house produced nivolumab (anti-PD-1 antagonist) or urelumab (anti-4-1BB agonist), which were confirmed to bind the soluble PD-1 and 4-1BB, respectively (online supplemental figure 10). IgG BsAb treatment began 3 days after tumor inoculation, where the IgG BsAb showed reduced efficacy, allowing a window to observe improved efficacy when combining with other immunotherapies. Anti-PD-1 or anti-4-1BB were administered on day 7. In this study, neither anti-PD-1 nor anti-4-1BB alone induced antitumor activity, possibly because the allogeneic T cells response was insufficient to recognize SKOV3 tumor cells (figure 5B). IgG BsAb alone demonstrated weak, but significant efficacy (figure 5B). BsAb combination with nivolumab or urelumab led to significant and near-complete tumor eradication. The majority of mice treated with BsAb/anti-PD-1 (six of seven mice) or BsAb/anti-4-1BB (six of eight mice) showed no sign of tumor outgrowth at day 38, the end of the study (figure 5B). Overall, combination of T cell checkpoint inhibition or costimulation with a T cell engaging BsAb led to significantly improved efficacy over the individual modes of treatment.