CD3-Bispecific Antibody Therapy Turns Solid Tumors into Inflammatory Sites but Does Not Install Protective Memory

The idea of using the cytotoxic capacity of T cells through CD3-targeting bispecific antibodies (CD3 bsAb) to kill tumor cells dates back to 1985 (1), and it took 30 years before both the FDA and European Medicines Agency approved blinatumomab, a CD19-directed T-cell engager (BiTE), for the treatment of refractory B-cell precursor acute lymphoblastic leukemia (ALL) patients (2). The first clinically registered CD3 bsAb was actually catumaxomab, binding the epithelial adhesion molecule EpCAM on carcinomas (3, 4), which was recently voluntarily withdrawn from the market for commercial reasons. Bridging T cells and tumor cells by immunoglobulin (Ig)-based biologicals can induce dramatic regression of advanced-stage malignancies and even lead to complete remission (2, 5–7). CD3 bsAb trigger the CD3 surface receptor on T cells by binding to their second target protein expressed on tumors. Consequently, all available T cells can bind to target-expressing cells via the CD3 bsAb, irrespective of the peptide/MHC specificity of their T-cell receptor (TCR).

Currently, more than 25 different CD3 bsAb are in clinical development for the treatment of hematologic malignancies, for example, by targeting CD19, CD20, CD33, and CD123, or solid cancers, for example, by targeting EpCAM, HER2, PSMA, and CEA (8). A multitude of other tumor-associated targets are explored in preclinical investigations. Many different bispecific formats have been developed over the years, ranging from Ab fragment–based bispecifics lacking an immunoglobulin constant region (Fc), to IgG-like molecules, in which additional Ab fragments are fused to regular antibody molecules, or bsAbs with regular IgG architecture (9, 10). The CD3 bsAbs in clinical trials either lack an Fc region, such as blinatumomab, or contain an engineered Fc domain to minimize interaction with Fc receptors (8, 11). The presence of an Fc domain increases the in vivo half-life through binding to the neonatal Fc receptor (FcRn; refs. 12–15).

The activity and sensitivity of CD3 bsAbs are commonly evaluated in vitro using human cancer cell lines or 3D spheroid cultures (16). Further efficacy studies are subsequently performed in small laboratory animals, often xenograft mouse models, in which established human tumor cell lines or fresh tumor samples from patients are transplanted into immune-deficient mice (12, 13, 16–18). Coengraftment of human peripheral blood mononuclear cells or purified T cells is required in these models to supply a source of immune effector cells. Alternatively, immune-deficient mice can be transplanted with human hematopoietic progenitor cells, which generate human leukocytes, the so-called human immune system mice (19). Although these models have been valuable for the study of tumor growth inhibition and for comparison of the potency of various bsAb clones, they fail to integrate the complex interplay between tumor cells and immune cells, which is of particular importance in the microenvironment of solid cancers. Moreover, long-term effects of CD3 bsAb cannot be evaluated due to induction of graft-versus-host responses in these immune-deficient mice (20). Recently, transgenic mice that coexpress human CD3ϵ (hCD3ϵ) on their T cells, in addition to mouse CD3ϵ, were exploited in this research field, enabling the use of endogenous T cells in immune-competent mice (21, 22). Although hCD3ϵ transgenic mice harbor a relatively normal immune system and allow analysis of immune–tumor cell interaction, they still lack a natural distribution of the selected TAA in mouse tissues, hampering examination of “on-target, off-tumor” effects via target expression in host tissues (21–24). Moreover, induction of neutralizing anti-human Ig antibodies prevents long-term evaluation in these mouse models.

Here, we report, for the first time, the immunologic effects of a CD3 bsAb in a completely syngeneic mouse tumor model. We performed immunotherapy of B16F10 melanomas in fully immune-competent wild-type C57BL/6 mice with an effector function silenced (inert) mouse IgG2a bsAb, targeting mouse CD3ϵ and mouse tyrosinase-related protein 1 (TRP1 or gp75), that is expressed on the surface of transformed melanoma cells and healthy melanocytes of the skin, just like in humans (25). We generated these mouse bsAbs with regular IgG architecture using controlled Fab-arm exchange (cFAE) which, apart from applications for human antibody therapeutics, can also be applied to efficiently generate rodent bispecifics (26–28).

In this tumor model, we observed a systemic T-cell activation that was selectively sustained in the tumor microenvironment for at least 4 days, coinciding with influx of inflammatory macrophages. Strikingly, the strong initial elimination of melanomas was not accompanied by induction of protective long-term immunity, as surviving mice were still vulnerable to a rechallenge with B16F10 melanoma cells. Interestingly, our syngeneic mouse model revealed a fast tumor influx of innate immune cells such as inflammatory macrophages before the arrival of additional CD4⁺ and CD8⁺ T-cell subsets. Our data demonstrate that antitumor activity of CD3 bsAb in vivo is associated with direct tumor kill and a complex interplay between immune cell subsets.

C57BL/6J mice were purchased from The Charles River Laboratories. All mouse experiments were performed at the animal facility of the Leiden University Medical Center (LUMC), the Netherlands. The health status of the animals was monitored over time, and all animals were tested negative for agents listed in the Federation of European Laboratory Animal Science Associations (FELASA) guidelines for specific pathogen-free mouse colonies (29). All mouse studies were approved by the Dutch animal ethics committee (CCD) and the local Animal Welfare Body of the LUMC on the permit number AVD116002015271. Experiments were performed in accordance with the Dutch Act on Animal Experimentation and EU Directive 2010/63/EU (“On the protection of animals used for scientific purposes”).

Chimeric antibodies were produced as described previously (26). In short, the VH and VL regions of Armenian hamster mAb 145-2C11 (mouse CD3ϵ-specific; ref. 30), murine mAb TA99 (human and mouse TYRP1/gp75-specific) (25), or human mAb b12 (HIV-1 gp120-specific; ref. 31) were cloned into mouse IgG2a and kappa constant region backbones, that contained the appropriate mutations to enable controlled Fab-arm exchange (cFAE; either F405L-R411T or T370K-K409R; ref. 26) and silence effector functions (L234A-L235A; LALA; refs. 32, 33). Exact amino acid sequences of the 2C11xTA99 bsAb are provided in Supplementary Fig. S1a. All antibodies were transiently produced under serum-free conditions in FreeStyle 293-F or Expi293F cells (Life Technologies), purified by protein A affinity chromatography (MabSelect SuRe; GE Health Care), dialyzed to phosphate-buffered saline (PBS) and sterilized over 0.2-μm filters. The purity was determined by sodium-dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) or capillary electrophoresis sodium-dodecyl sulfate (CE-SDS), and the concentration was measured by absorbance at 280 nm. Aggregates and degradation products were measured by high-performance size-exclusion chromatography (HP-SEC). Purified antibody batches with low endotoxin levels (<0.1 units/mg IgG) were stored at 2–8°C.

Equimolar amounts of relevant mouse IgG2a-LALA-F405L-R411T and IgG2a-LALA- T370K-K409R antibodies were mixed and incubated with 2-mercaptoethylamine (2-MEA; Sigma) at a final concentration of 1 mg/mL per antibody. The final concentration of 2-MEA was 75 mmol/L. The mixtures were typically incubated at 31°C for 5 hours. To remove 2-MEA, the mixtures were buffer-exchanged against PBS using PD-10 desalting columns (5 kDa molecular weight cutoff; GE Healthcare) or dialysis using Slide-A-Lyzer cassettes (10 kDa molecular weight cutoff; Pierce). Samples were stored at 4°C overnight, to allow for the reoxidation of the disulfide bonds. The efficacy of cFAE was assessed by hydrophobic interaction chromatography and depicted in Supplementary Fig. S1b.

The B16F10 melanoma cell line was obtained from the American Type Culture Collection (CRL-6475) and maintained in tissue culture as described (34). KPC3 is a cell line isolated from the genetic pancreatic ductal adenocarcinoma “KPC” mouse model with K-ras^G12D/+, p53^R172H/+ and pancreatic and duodenal homeobox 1 (Pdx-1)-Cre transgenes (35). KPC3-TRP1 cells were generated by transfection of the TRP1/gp75-coding plasmid using lipofectamine (Invitrogen), as described (36). This plasmid was provided by Gestur Vidarsson (Academic Medical Center, Amsterdam) and optimized by exchanging the zeocin selection gene with a neomycin selection gene and by replacing the cytomegalovirus (CMV) promoter with the CMV enhancer, chicken beta-actin promoter and rabbit beta-globin splice acceptor site (CAG) promoter. Transfected cells were selected with 400 μg neomycin for 7 days and cell sorting by fluorescence-activated cell sorting on TRP1/gp75 expression using the TA99 antibody and a secondary Alexa Fluor–labeled anti-mouse IgG (BioLegend).

For every mouse tumor experiment, liquid nitrogen stored vials from the same cell batch were used and passaged once before injection. Mycoplasma testing was performed every 2 months by PCR on in vitro propagated cultures, which were maintained for maximal 6 months. No additional authentication method was performed. All cells were cultured in Iscove's modified Dulbecco's medium (Invitrogen) supplemented with 8% fetal calf serum (Gibco), 2 mmol/L l-glutamine (Gibco), and 100 μg/mL penicillin and 100 μg/mL streptomycin (Gibco). All cell lines were incubated at 37°C in a humidified atmosphere containing 5% CO₂.

B16F10, KPC3, and KPC3-TRP1 cells were irradiated at 6,000 rad and plated in 48-well plates with 30,000 cells per well in 100 μL. CD4⁺ or CD8⁺ T cells were isolated from naïve C57BL/6J mice spleens by nylon wool passage and MACS purification columns (Miltenyi Biotech). The T-cell populations were more than 95% pure, as validated by flow cytometry, and were added to the wells in 50 μL and then bsAb were added in 50 μL. Triton X-100 (20 μL) to wells containing tumor cells alone served as positive control. After 48 hours, 50 μL of supernatant was harvested and transferred to 50 μL of lactate dehydrogenase reaction mix (LDH, Life Technologies). After gently mixing, the plates were incubated at room temperature (RT) in the dark for 30 minutes, and 50 μL stop solution was added before absorbance was measured at 490 and 655 nm. Another 50 μL supernatant from these tests was subjected to interferon gamma (IFNγ) sandwich ELISA according to the manufacturer's instructions (BD Biosciences). Activation of T cells from these tests was measured by flow cytometry.

C57BL/6J mice were subcutaneously (s.c.) injected with 1 × 10⁵ B16F10, KPC3, or KPC3-TRP1 cells in 200 μL PBS and treated with i.p. administered 12.5 μg (0.5 mg/kg) of bsAb. CD8⁺ TCR-transgenic T cells specific for the gp100 tumor antigen (“pmel”) were transferred into tumor-bearing mice in combination with gp100 long synthetic peptide (human variant) and Aldara adjuvant (3M Pharmaceuticals) and IL2 (Proleukin, Novartis), as described previously (34). Tumor sizes were measured by caliper twice weekly, and mice were euthanized when tumors reached a volume of 1,000 mm³. To rechallenge animals, 1 × 10⁵ B16F10 tumor cells were injected s.c. at the opposite flank.

For depletion of immune cell subsets, C57BL/6J mice were injected i.p. with 100 μg of anti-NK1.1 antibody [mouse IgG2a, clone PK136 (own stock)] to deplete NK cells, 50 μg of anti-CD8 antibody [rat IgG2b, clone 2.43 (own stock)] to deplete CD8⁺ T cells, or 50 μg anti-CD4 [rat IgG2b, clone GK1.5 (own stock)] to deplete CD4⁺ T cells. Depletions were started at days 2 and 5 after tumor inoculation and continued weekly up to the end of the experiment.

When tumors reached a volume of 50 to 250 mm³, mice were treated with bsAb. Tumors and spleens were harvested and analyzed with IHC and flow cytometry. Single-cell suspensions for flow cytometry were made by physically fragmentation and 30 μg/mL Liberase (Roche) for 30 minutes, after which the cells were minced through a cell strainer (BD Biosciences). LIVE-DEAD Fixable yellow dead cell stain kit (Life Technologies) was included in all stainings together with antibodies to Ly6C (HK1.4), Ly6G (1A8), F4/80 (Bm8), NK1.1 (PK135), CD25 (PC61.5) CD45.2 (104), CD11b (M1/70), CD4 (RM4-4), CD8 (53-6.7), CD69 (H1.2F3), and CD3 (145-2C11). Intracellular FoxP3 (FJK-16s) staining was performed according to the manufacturer's instructions (eBioscience). All antibodies were purchased from BD Biosciences, BioLegend/Sony or eBioscience. Samples were measured in a Fortessa cytometer (BD Bioscience) and analyzed with FlowJo software (Treestar).

Harvested tumors were fixed in 4% neutral buffered formalin and after 24 hours dehydrated in 70% alcohol and embedded in paraffin. Formalin-fixed paraffin-embedded (FFPE) 3- to 4-μm sections were stained for CD3ϵ (SP7, rabbit anti-mouse, Abcam), CD4 (EPR19514, rabbit anti-mouse, Abcam), and CD8α (D4W2Z, rabbit anti-mouse, Cell Signaling Technology) using the Ventana Discovery autostainer platform (Ventana Medical Systems). Slides were deparaffinized, subjected to antigen retrieval buffer CC1 (Ventana) at 95°C for 32 minutes, and blocked for endogenous peroxidase using Discovery Inhibitor (Ventana) at 37°C for 8 minutes. All primary antibodies were incubated in Discovery Ab Diluent (Ventana) at 37°C for 32 minutes, and subsequently incubated with Discovery OmniMap-anti-Rabbit horseradish peroxidase (HRP; Ventana) at 37°C for 16 minutes. For the anti-CD8 antibody, signal amplification was applied by subsequent incubations with Discovery AMP HQ (Ventana) and Discovery Amplification anti-HQ HRP (Ventana) at 37°C for 16 minutes. HRP was visualized with Discovery purple (Ventana) at 37°C for 4 minutes. Nuclei were counterstained with Hematoxylin II (Ventana) at 37°C for 16 minutes and Bluing reagent (Ventana) at 37°C for 8 minutes. Slides were mounted in glycergel (DAKO, Code C0563). Macrophages were manually stained by incubation with rat anti-mouse macrophage mAb F4/80 (clone CI:A3-1, IgG2b; Serotec) at RT for 30 minutes, followed by incubation with biotinylated rabbit anti-rat IgG Ab (code number E0467; DakoCytomation) at RT for 30 minutes and alkaline phosphatase-streptavidin (code number K0391; DakoCytomation) at RT for 20 minutes. The alkaline phosphatase reaction was developed using Fast Red (Scytek) in a naphthol-phosphate buffer (Scytek) with 50 mmol/L levamisole (DakoCytomation). After 20 minutes, this reaction was blocked in distilled water. The slides were counterstained with Mayer's hematoxylin and mounted in Kaiser's glycerin.

All immunostained slides were scanned on AxioScan Z1 (Zeiss), and computerized image analysis was performed by OracleBio (Biocity Scotland, UK) using the Indica Labs HALO platform. A customized classifier algorithm was developed within HALO to exclude necrotic areas and any prominent artifacts. Numbers of CD3, CD4, and CD8 cells were quantified using a customized cellular multiplex algorithm and F4/80 staining was quantified as percentage area.

Survival between groups was compared using Kaplan–Meier curves and statistical log-rank test (Mantel–Cox). Additional statistical methods are stated in the figure legends. All P < 0.05 were considered statistically significant. The different levels of significance were labeled with asterisks, where *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Previously, we showed that treatment of B16F10 melanoma with the bsAb 2C11xTA99, targeting the mouse CD3 with one arm and the surface melanocyte differentiation protein TRP1 with the other arm, resulted in significant survival rates (26). This mouse IgG2a-based bsAb contains a full Fc region with silenced (inert) effector function. We now expanded on these data by directly comparing the therapeutic efficacy with a previously established cellular immunotherapy protocol, consisting of transfer of naïve TCR-transgenic, gp100-specific, CD8⁺ T cells into melanoma-bearing animals in combination with a long synthetic peptide vaccine (34). This cellular immunotherapy protocol induced delayed tumor outgrowth, but no long-term survival of mice (Fig. 1A and B). In contrast, 3 injections of bsAb 2C11xTA99 at days 6, 9, and 12 at a dose of 0.5 mg/kg resulted in delayed tumor outgrowth in all mice and long-term survival in 50% of the group (Fig. 1A and B). In a previous study, we showed that control antibodies lacking the TA99 tumor-targeting arm failed to control tumor outgrowth (26). These results highlight the potency of CD3 bsAb for immunotherapy of solid cancers even in a T-cell–poor tumor microenvironment such as the B16F10 model (37).

To further substantiate these results, the efficacy of the CD3 bsAb was tested in a syngeneic pancreatic adenocarcinoma model, as these lesions are notoriously enriched in fibroblastic microenvironment with little T-cell infiltration (38). Stable expression of TRP1 by KPC3 pancreatic tumor cells was achieved after transfection (Supplementary Fig. S2a). The KPC3 line was derived from the tumor-prone KPC mouse with activated K-ras and mutant p53 genes (35). CD3 bsAb therapy indeed induced a significant delay in tumor outgrowth of KPC3-TRP1 tumors, but not of parental KPC3 tumors, while half of the mice with B16F10 tumors displayed long-term survival in this 2 injection protocol (days 6 and 9; Fig. 1C), showing therapeutic efficacy in a second, T-cell–poor, tumor model (38).

We then took advantage of the fact that our mouse tumor model is completely syngeneic, including immune cells, tumor cells, and therapeutic compound, and investigated long-term immunity. Long-term surviving mice that successfully eliminated B16F10 melanomas after treatment with the CD3 bsAb were rechallenged with the same tumor cells in the opposite flank at day 70. These secondly injected melanomas grew progressively and at a similar pace as in a second cohort of naïve mice, implying that successful treatment with CD3 bsAb did not install protective immunity after tumor clearance (Fig. 1D). Together, these results indicate that CD3 bsAb induce a potent antitumor response, but do not induce immunologic memory in this particular mouse tumor model.

The rationale behind CD3 bsAb is the activation of T cells residing in conjunction with tumor cells, irrespective of their peptide/MHC specificity. Direct cytotoxic killing is predominantly expected from CD8⁺ T cells, which classically express perforin and granzymes, but a subset of cytotoxic CD4⁺ T cells exists as well (39, 40). We therefore tested tumor cell killing capacity of purified CD8⁺ and CD4⁺ T cells from spleens of naïve mice in coculture assays. Surprisingly, even though the bsAb 2C11xTA99 induced strong antitumor effects against the B16F10 melanoma in vivo, it was not so potent in mediating cell lysis of B16F10 cells in vitro (Fig. 2A). High numbers of effector cells isolated from the splenocytes were required (E:T ratio of 20:1) in combination with high bsAb concentration (10 μg/mL) to reach killing percentages around 35% (Supplementary Fig. S2b). When the CD8⁺ T cells were analyzed at the end of the assay by flow cytometry, clear signs of activation were detectable as measured by increased expression of CD25 and CD69 molecules on the T cells (Supplementary Fig. S2c). In addition, high concentrations of IFNγ were observed in the supernatants of these cocultures (Supplementary Fig. S2d), indicating that CD8⁺ T cells were activated by bsAb 2C11xTA99, despite the lack of efficient melanoma cell killing.

KPC3-TRP1 cells displayed much higher levels of TRP1 protein on their surface than B16F10 cells (Supplementary Fig. S2a) and were accordingly more efficiently killed (Fig. 2A). Parental KPC3 tumor cells were not lysed, showing dependency on the tumor antigen. Control bsAb 2C11xb12 lacking the tumor-binding arm also failed to induce cytolysis. Finally, purified CD4⁺ T cells induced some killing of KPC3-TRP1 tumor cells, but not of B16F10 (Fig. 2A), in line with earlier findings that report inefficiency of this subset (41).

This prompted us to examine the contribution of the 2 T-cell subsets to tumor clearance in the B16F10 mouse model. T-cell subsets were successfully removed from B16F10-bearing mice by depleting antibodies (Supplementary Fig. S3a). Importantly, depletion of CD8⁺ T cells did not diminish the treatment effect of bsAb 2C11xTA99, indicating that this subset was redundant and suggesting that CD4⁺ T cells could replace the role of CD8⁺ T cells in vivo (Fig. 2B). Vice versa, CD4⁺ T cells were also redundant. Depletion of both T-cell populations, however, completely neutralized the antitumor response induced by the CD3 bsAb (Fig. 2B; Supplementary Fig. S3b). NK cells were not required, because depletion of these innate lymphocytes did not change the percentage of surviving mice and the NK cells in the T-cell–deficient animals were incapable to delay tumor outgrowth (Fig. 2C). Thus, the presence of CD4⁺ or CD8⁺ T cells is sufficient to eradicate B16F10 tumors using CD3 bsAb in this syngeneic mouse model, despite the fact that their killing capacity, as tested in coculture assays, was very low. This demonstrates that in vitro cytolysis assays may be poor predictors of tumor control in vivo and that activation of T cells and release of IFNγ should be regarded as well.

We then studied T-cell activation in vivo. Already 6 hours after injection of the CD3 bsAb, the frequencies of activated CD4⁺ and CD8⁺ T cells in the tumor microenvironment more than doubled, as measured by expression of CD25 and CD69 markers by flow cytometry (Fig. 3). CD3-positive, NK1.1-positive NKT cells in the tumor were also activated, whereas the CD3-negative, NK1.1-positive NK cells were not (Fig. 3). Strong activation of intratumoral CD4⁺ T cells was sustained for at least 2 days and started to decline after 4 days, as observed by a decrease in CD69-expressing cells (Fig. 4A). Importantly, a high frequency of intratumoral CD8⁺ T cells still displayed the CD69 activation marker at 4 days. These results indicated that the CD3 bsAb was able to sustain the T-cell response in the tumors for at least 4 days.

Interestingly, analyses of activation markers on T cells in the spleens revealed a transient response, detected only at the 6-hour time point, but not thereafter, as determined by CD69 expression (Fig. 4A). The control antibody that lacked the tumor antigen targeting arm TA99 (2C11xb12) seemed to also induce T-cell activation in spleens of some animals, but this was not statistically significant compared with untreated mice. These results prompted us to analyze T-cell activation in the absence of B16F10 melanomas (Fig. 4B). Interestingly, 6 hours after i.p. injection, bsAb 2C11xTA99 induced systemic activation to comparable extent as was observed in tumor-bearing animals, suggesting that the presence of the tumor was not related to this transient response (Fig. 4B compared with Fig. 4A; Supplementary Fig. S4a). Administration of the control bsAb 2C11xb12 in tumor-free animals did not result in T-cell activation, suggesting that TRP1 expression on normal melanocytes might have caused this transient response. The bsAb-activated CD4 T cells expressing CD25 and CD69 at 6 hours after the injection in tumor-free mice were FoxP3-negative, indicating that these cells represent effector T cells and not regulatory CD4 T cells (Supplementary Fig. S4b). Together, these results demonstrated that after a very transient systemic T-cell response, a sustained T-cell activation in the tumor microenvironment is induced by CD3 bsAb treatment. Of note, mice did not experience a cytokine release syndrome in these first hours, nor was autoimmune reactivity, like overt depigmentation, observed during long-term experiments.

Finally, we examined the cascade of immune events after the initial T-cell activation in the B16F10 melanoma model. T-cell and myeloid cell numbers in the tumors upon treatment with CD3 bsAb were evaluated (Figs. 5 and 6). Frequencies of T-cell subsets first seemed to decrease 6 hours after injection of the antibody, before strong increases were observed, as measured by tissue slide staining and flow cytometry. Four days after bsAb injection, the number of CD3-positive cells increased 4 times (Fig. 5A–5B). This increase was caused mainly by CD4⁺ cells (Fig. 5B; Supplementary Fig. S5A–S5B). Multiparameter staining and analysis by flow cytometry confirmed the findings from tissue sections and, furthermore, showed an increase of NK cells from day 2 onward (Fig. 5C and D), even though these cells were not direct targets of the bsAb (Fig. 3). The temporal decrease of intratumoral T cells at 6 hours after treatment was also observed in lymph nodes (Supplementary Fig. S5c). This T-cell redistribution has been described in patients as well and might be explained by adhesion to the vasculature via activation-induced integrins (42).

We then investigated whether the CD3 bsAb induced recruitment of myeloid cells (Fig. 6). Staining for the common macrophage marker F4/80 was evaluated as a percentage positive area, by IHC, because myeloid cells were difficult to exactly enumerate due to their veiled morphology. A significant increase in macrophages was observed in the tumor area, whereas their high numbers at the surrounding tumor rim did not alter in time (Fig. 6A and B). Flow cytometry allowed us to discern inflammatory macrophages from tumor-associated stationary macrophages and neutrophilic granulocytes using the markers CD45, CD11b, F4/80, Ly6G, and Ly6C. Interestingly, the frequencies of neutrophils (Ly6G⁺, Ly6C^low, F4/80⁻) and stationary macrophages (F4/80⁺, Ly6C⁻) were not significantly altered, whereas numbers of inflammatory macrophages (F4/80⁺, Ly6C⁺) were increased, already 6 hours after start of treatment (Fig. 6C and D; Supplementary Fig. S5d). The fact that the early increase of this macrophage subset was not detected in tissue slide staining might be due to the broader tissue area we resected for flow cytometry.

In summary, systemic injection of the bsAb 2C11xTA99 leads to a fast activation of intratumoral T cells and a fast and continued influx of inflammatory macrophages. Within a few days, this culminates in an additional increased number of CD4⁺ T cells, CD8⁺ T cells, and NK cells. The examination of immune cell composition in the tumor microenvironment was possible because we applied a completely syngeneic mouse model system and this might form the basis for studying combinatorial approaches to further improve CD3 bsAb therapies.

Here, we show immunologic effects of immunotherapy with a CD3 bsAb in a completely immune-competent mouse melanoma model. Treatment with the TRP1-targeting bsAb resulted in a strong and sustained activation of intratumoral T-cell subsets (CD4⁺ and CD8⁺ T cells, as well as NKT cells), but not of CD3-lacking NK cells. From day 2 onward, a strong increase of innate and adaptive immune cells was observed in the tumor microenvironment. This remodeling was especially pronounced for NK cells, inflammatory macrophages, and CD4⁺ T cells. A brief systemic T-cell activation in lymphoid organs was observed only up to 6 hours. Although the capacity of T cells to kill melanoma cells in vitro was very modest, melanoma control in vivo was rather efficient and mediated by both CD4⁺ and CD8⁺ T cell subsets. These data suggest that cytokine-mediated effector functions of CD4⁺ T cells, which rarely contain cytolytic granules, might be important contributors to tumor control, as previously observed in other systems (43–46). Finally, the CD3 bsAb led to a fast and strong induction of antitumor immunity, but failed to install long-term protective memory responses, indicating a lack of memory T-cell formation in the B16F10 tumor model.

Most of the immune features observed in our study could be detected due to the presence of a fully intact immune system in our immune-competent mouse model, in which we applied mouse bsAb targeting the mouse CD3 protein and the mouse TAA to treat mouse melanoma in a syngeneic host. Most PDX interspecies models have used bispecifics targeting human CD3 and human tumor antigens, not expressed on mouse host tissues or immune cells (12, 13, 16–18) and thus fail to monitor the full so-called on-target, off-tumor effects. Even the more elegant models in this research field, in which human CD3E is transgenically expressed in mouse T cells and human TAA are expressed in mouse tumor cell lines under heterologous promoters, fall short, due to artificial expression levels of the CD3 and TAA (21, 22). Indeed, the expression levels and frequency of CD3E in these transgenic mouse models are lower than those in wild-type mice (24). Recently, a triple knock-in mouse was reported in which human CD3E, -D, and -G genes replace the mouse homologous genes CD3ϵ, -δ, and -γ. In this knock-in mouse, T-cell numbers and CD3 expression levels are comparable to those in wild-type mice and treatment with CD3 bsAb binding to human CD3 can representatively be evaluated (23, 24). In our syngeneic mouse model, we report a strong increase of innate immune cells, such as inflammatory macrophages and natural killer cells, and a fast intratumoral activation of CD4⁺ and CD8⁺ T-cell subsets (Figs. 5 and 6).

An initial systemic activation of T cells was observed up to 6 hours after bsAb administration, as witnessed by the early activation markers CD25 and CD69 on T cells in the spleen (Fig. 4). This transient pulse of activation was also observed in tumor-free animals, suggesting that this might be caused by binding of the CD3 bsAb to endogenous expressed TRP1 on healthy melanocytes of the host. The fact that control bsAb 2C11xb12 did not induce this brief T-cell activation in most tumor-free mice further substantiated this notion. TRP1 expression indeed has been detected on melanocytes in the skin in humans and mice (25, 47, 48). In addition, trace amounts of monospecific parental CD3 antibodies might have been present in the product as a result of incomplete cFAE reaction. However, less than 0.3% monospecific anti-CD3 Ab were present in our bsAb (Supplementary Fig. S1b), and it is very unlikely that these small amounts would explain the observed transient systemic T-cell activation. Alternatively, low-level residual interactions between FcγR or complement and the mmIgG2a-LALA backbone could exist (26), allowing for residual cross-linking of CD3. A combination of these factors might contribute to the early peak of systemic T-cell activation.

The analyzed surface markers CD69 and CD25 are early indicators of T-cell activation and might merely reflect a brief and reversible TCR triggering and do not guarantee full-blown differentiation to memory cells. Lack of protective memory responses in our model might be explained by lack of tumor specificity of the targeted pool of T cells or lack of costimulatory signals during the treatment of the CD3 bsAb. Most likely, higher expression levels of TRP1 on the tumor or different environmental cues in tumor tissue influencing the trafficking of CD3⁺ cells contribute to the observed tumor eradication via CD3 bsAb in the tumor microenvironment.

In conclusion, our findings show that a CD3 bsAb treatment in the B16F10 mouse solid tumor model can result in antitumor efficacy and remodeling of the tumor microenvironment. The B16F10 melanoma model is known for its low immune infiltration and can be considered as immunologically cold (37, 49). The fact that bsAb treatment leads to increased numbers of intratumoral inflammatory macrophages, natural killer cells, and T cells within a few days marks this therapeutic as a remodeler of the tumor microenvironment. The TCR specificity of these expanded T cells in the tumor microenvironment and their contribution to tumor reduction remains to be investigated.

I. Altıntaş has ownership interest (including stock, patents, warrants, etc.). A. Labrijn is Principal Scientist at Genmab and has ownership interest (including stock, patents, etc.) in the same. D.H. Schuurhuis has ownership interest (including stock options, patents, etc.). M.A. Houtkamp has ownership interest (including stock, patents, warrants, etc.). J. Schuurman has ownership interest (including stock, patents, etc.) in Genmab. T. van Hall reports receiving a commercial research grant from Genmab and is a consultant/advisory board member for the same. No potential conflicts of interest were disclosed by the other authors.

Conception and design: H. Benonisson, I. Altıntaş, S. Verploegen, J.S. Verbeek, J. Schuurman, T. van Hall

Development of methodology: H. Benonisson, I. Altıntaş

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H. Benonisson, S. Verploegen, M.A. Houtkamp, J.S. Verbeek

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Benonisson, I. Altıntaş, S. Verploegen, A.F. Labrijn, M.A. Houtkamp, J.S. Verbeek, T. van Hall

Writing, review, and/or revision of the manuscript: H. Benonisson, I. Altıntaş, S. Verploegen, A.F. Labrijn, D.H. Schuurhuis, M.A. Houtkamp, J.S. Verbeek, J. Schuurman, T. van Hall

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Sluijter, S. Verploegen

Study supervision: J. Schuurman, T. van Hall

This research received funding from the People Programme (Marie Curie Actions) of the European Union's Seventh Framework Programme FP7/2007-2013 under grant agreement no. 317445 (to J.S. Verbeek). A commercial research grant was supplied by Genmab (to T. van Hall). Paul Parren, Ferry Ossendorp, and Marieke Fransen are acknowledged for advisory comments; Heng Sheng Sow, Marcel Camps, Connie Brouwers, and Cor Breukel for technical support; Henry J. Witteveen and Patrick F. Franken for IHC analysis; and Marcel Brandhorst for support with animal experiments.

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.