Combination of CD117xCD3 TCE and CD33xCD28 IgG4-scFv2 mediates faster and more selective T-cell lysis of MOLM-14 target cells. A, Schematic of the experimental workflow of time‐lapse imaging process to study effector–target cell interactions over time. Cells were cocultured at an E:T ratio of 1:1 for up to 72 hours on microgrids which are glued to chamber slides to allow for quantification of single attachment and lysis events. After the required criteria are met for imaging (i.e., good movie quality allowing for cell traceability and high cell viability of the cells over the course of imaging), individual time‐lapse movies were acquired in those wells in which an E:T ratio of 3:1 to 1:3 was achieved. B, Representative time-lapse images of MOLM-14 CD117high cells (gray) and T cells (red), incubated either without binders, with 500 ng/mL CD117xCD3, with CD117xCD3 in combination with CD33xCD28 IgG4-scFv2 (0.5 nmol/L), or with CD117xCD3 and CD28 IgG4 (0.5 nmol/L), are shown. Target cell lysis is marked by PI influx (green). C and D, Time to target cell attachment and time from target cell attachment to lysis, i.e., PI influx between different conditions. T cells with CD117xCD3 in combination with CD33xCD28 IgG4-scFv2 showed fastest target cell attachment, whereas time from attachment to killing was statistically different between conditions (all results from two healthy donors, statistical analysis conducted using one-way ANOVA or unpaired Student t test; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001). E, A total of 240 individual time-lapse imaging movies were analyzed for quantification of effector–target cell interactions (n = 80, 80, 80, and 80 for each group). Quantification of target-cell interaction showed complete T-cell attachment to target cells with CD117xCD3 in combination with CD33xCD28 IgG4-scFv2. F, Following attachment, target cells in individual wells were tracked for T cell–induced lysis. Maximum lysis was observed with the combination of CD117xCD3 and CD33xCD28 IgG4-scFv2 (right; n analyzed = 29, 54, 80, 74 for each group). All results are from two healthy donors. Statistical analysis was conducted using χ2 test; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. G, Representation of the experimental setup of the specificity assay generated with BioRender.com. An equal mixture of MOLM-14 cells (either CD117+CD33+, CD117+CD33−, or CD117−CD33+) were coincubated with unexpanded healthy donor–derived T cells at an E:T ratio of 1:1 and either CD117xCD3 alone or in combination with 0.5 nmol/L CD33xCD28 IgG4-scFv2. Cocultures were evaluated by flow cytometry at 4, 8, 12, 24, 48, 72, and 96 hours. All experiments executed with T cells from three healthy donors, each analyzed in duplicates. H, Percentage specific lysis of MOLM-14 mix treated with either CD117xCD3 alone at several concentrations or in combination with 0.5 nmol/L CD33xCD28 IgG4-scFv2. I, Percentage specific lysis of mixed MOLM-14 cells and individual MOLM-14 subpopulations (J), incubated with T cells and the combination of 1.168 nmol/L CD117xCD3 and 0.5 nmol/L CD33xCD28 IgG4-scFv2 at indicated time points. Statistical analysis was conducted using two-way ANOVA; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. [A, Created in BioRender. Caiado, F. (2026) https://BioRender.com/sbcghjr; G, Created in BioRender. Caiado, F. (2026) https://BioRender.com/1gojldm.]

CD33xCD28 IgG4-scFv2 TCE does not activate T cells in the absence of TCR stimulation. A, Schematic representation of superagonism observed with the CD28 TGN1412, in which membrane-proximal binding and immobilization facilitate cross-linking of CD28 receptors leading to T-cell activation in the absence of TCR engagement (left) and of agonism observed with CD33xCD28 IgG4-scFv2 with no T-cell activation (right). B, Graphical legend indicating the different molecules added in the assay in C–I. Plates were initially coated with the respective antibody constructs at a concentration of 10 μg/mL, followed by incubation with PBMCs. After 96 hours, the different conditions were analyzed by flow cytometry (C–H) or ELISA (I). Data are represented as the mean ± SD from three healthy donor–derived PBMCs, each plated in duplicates. Each donor PBMC is represented with a different symbol. C and D, Percentage of CD69+ in CD4+ T cells (C) and in CD8+ T cells (D). E and F, Percentage of CD25+ in CD4+ T cells (E) and in CD8+ T cells (F). G and H, Absolute CD4+ T-cell (G) and CD8+ T-cell counts (H). I, IFNγ quantification in the supernatant of the different conditions. Statistical analysis for all the graphs was conducted using one-way ANOVA; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. [A, Created in BioRender. Caiado, F. (2026) https://BioRender.com/5d04w2s.]

Supplementary Table 2. Colorectal patient characteristics.

Figure S7 shows the increased T-cell proliferation at 96h upon addition of CD33xCD28 IgG4-scFv2 to CD117xCD3 TCE in T-cell and primary AML cell co-culture.

Supplementary Figure 3. Quantification of CEA surface antigen density on LS174T, C51.CEA, and MC38.CEA cells.

CD33xCD28 IgG4-scFv2 in combination with CD117xCD3 TCE mediates more effective lysis of primary AML cells. A, Representative flow cytometry plots showing CD117 and CD33 expression on four different primary human AML blast populations (CD45dim) upon coculture with healthy donor–derived T cells at an E:T ratio of approximately 1:1. Cells were incubated with antibody constructs as indicated (CD117xCD3 at the indicated concentrations, in combination with 0.5 nmol/L CD33xCD28 IgG4-scFv2 or 0.5 nmol/L CD28 IgG4). Plots are shown for 0 and 48 hours. B, Percentage of specific lysis of CD45dimCD3− AML blasts of the individual patient samples after 48 hours. C, Combined percentage of CD25+ T cells. D, Combined IFNγ in supernatants of coculture. E, Combined proliferation of T cells. Data represent the mean ± SEM from two independent healthy donor–derived T-cell samples, each plated in duplicate. Statistical significance was determined using two-way ANOVA; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Supplementary Figure 8. Effect of L19-mTNF 1 h post administration on the tumor microenvironment in C51.CEA immunocompetent mouse model by immunofluorescence.

Table S1 shows the list of antibodies used for flow cytometry.

Supplementary Figure 7. Effect of L19-mTNF 24 h post administration on macrophage and neutrophil tumor infiltration in C51.CEA immunocompetent mouse model by immunofluorescence.

Figure S3 shows the dose-dependent effects of CD33xCD28 IgG4-scFv2 on MOLM-14 cell lysis, induced by CD117xCD3 and T-cells.

Figure S1 shows the amino acid sequence of CD33xCD28 IgG4-scFv2.

Supplementary Figure 6. Proliferation and phenotypes of CD4+ T cells in killing assay with CEAxCD3 as a single agent and in combination with L19-TNF at day 5.

Supplementary Figure 4. The in vitro activity of CEAxCD3 as a single agent and in combination with L19-TNF at day 1.

Supplementary Figure 1 shows (A) EDB expression in F9 teratocarcinoma cell line wild-type (red) and EDB transduced (blue). (B) EDB expression in WEHI-164 sarcoma cell line wild-type (red) and EDB transduced (blue). (C) F9 teratocarcinoma tumor sections of the wild-type model (top) and the EDB transfected model (bottom). Left panel stained with 2C11xL19 and right panel stained with 2C11x7NP2. (D) WEHI-164 Sarcoma tumor sections of the wild-type model (top) and the EDB transfected model (bottom). Left panel stained with 2C11xL19 and right panel stained with 2C11x7NP2. Green: detection of bispecific antibody, Red: CD31 detection, Blue: DAPI staining. Scale bars: 100 µm. Supplementary Figure 2 shows representative images of EDB expression in WEHI-164 (A) or F9 (B) tumor sections from wild-type and membrane-bound EDB transduced cell lines stained with L19 IgG-FITC or KSF IgG-FITC together with anti-CD31 (top). Green: detection of antibody, Red: CD31 detection, Blue: DAPI staining. Scale bars: 100 µm. Integrated intensity signal of the stained area (bottom). n = 3-5. (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001). Supplementary Figure 3 shows Collagen 1 expression in F9 (A) or WEHI-164 (B) tumor sections from wild-type and membrane-bound EDB transduced cell lines. Green: detection of collagen 1, Blue: DAPI staining. Scale bars: 100 µm. Supplementary Figure 4 shows the amino acid sequence of the 2C11xL19 and 2C11x7NP2 bispecific antibodies. Supplementary Figure 5 shows stability data of the 2C11xL19 (left) and 2C11x7NP2 (right) measured by size-exclusion chromatography. (A) Profile of the proteins after Protein A purification. (B) Profile of the proteins after fractionation by size-exclusion chromatography. (C) Protein after one freeze-thaw cycle. (D) Protein after storage at 4°C for 3 days. (E) Protein after storage at 4°C for 5 days. Supplementary Figure 6 shows IL2 release in the supernatant of murine T cells cross-linked with the BsAb on antigen coated plates. Data represented as mean ± SEM. Supplementary Figure 7 shows the Tumor targeting properties of 2C11xL19. Ex vivo microscopic fluorescence analysis of EDB expression in F9WT and F9EDB tumor sections and other organs of the same mouse (left) and in WEHI-164WT and WEHI-164EDB and other organs of the same mouse (right). Green: 2C11xL19 detection, Red: CD31 detection, Blue: DAPI. Scale bar 100 µm. Supplementary Figure 8 shows the body weight monitoring of immunocompetent mice during the therapy study. Data represent the percentage of body weight change ±SEM based on a baseline measured before the start of the therapy. Arrows indicate injection of 50 µg of bispecific antibody or saline as negative control. (A) Body weight change in 129/SvEv mice with F9WT tumor model. (B) Body weight change in 129/Sv mice with F9EDB tumor model. (C) Body weight change in BALB/c mice with WEHI-164WT tumor model. (D) Body weight change in BALB/c mice with WEHI-164EDB tumor model. Supplementary Figure 9 shows the flow cytometry based immune assay at different timepoints including day 2, day 4 and day 6 after the start of the therapy with a saline group as negative control. Data represented as mean ± SEM. (A) IFNg positive immune cells (left CD4+; middle CD8+, right NK cells) per mL of blood. (B) IFNg positive immune CD4+ cells (left), Granzyme B+ CD4+ cells (middle) and Perforin + CD4+ cells (right) in tumor as percent of total live tumor cells. (C) IFNg positive immune NK cells (left), Granzyme B + NK cells (middle) and Perforin + NK cells (right) in tumor as percent of total live tumor cells. (D) IFNg positive immune CD8+ cells in tumor as percent of total live tumor cells. (E) CD69+ CD8+ cells (left) and CD25+ CD8+ cells (right) as percent of total live tumor cells. (F) IL2 concentration measured by ELISA. *p < 0.05; **p < 0.01; ***p < 0.001. Supplementary Figure 10 shows the gating strategy of blood immune cells (A). (B) Gating strategy of IFNg levels in blood sample for different types of immune cells. (C) Gating strategy of activation markers in tumor sample. (D) Gating strategy of immune cells in tumor sample. (E) Gating strategy of IFNg, Granzyme and Perforin levels in different immune cells in tumor sample. Supplementary Figure 11 shows H&E staining of tumors in the F9EDB tumor model treated with 2C11xL19 over different timepoints. Representative H&E images at the different timepoints with saline as negative control. Scale bar 2.5 mm. Supplementary Figure 12 shows NK cell infiltrate assessment in the F9EDB tumor model treated with 2C11xL19 over different timepoints. Representative IHC images with NK cell staining at the different timepoints with saline as negative control. Summary results plotted as number of NK cells (DAB) elements / mm2 at different timepoint. Data plotted as mean ± SEM. No statistical difference was observed. Scale bar 100 µm.

Supplementary Table 1. The amino acid sequences of TCEs and L19-TNF.

Supplementary Figure 1. Gating strategy of intracellular NF-kB in human PBMC samples.

Bispecific antibodies (BsAb) are a rapidly advancing class of biopharmaceuticals with substantial potential for cancer immunotherapy. Although BsAbs have shown notable success in treating certain hematologic malignancies, their application for solid tumors remains limited. The extra domain B (EDB) of fibronectin represents a promising pan-tumoral stromal target, offering an attractive alternative to conventional cellular tumor antigens, which often face limitations with respect to specificity in solid tumors. In this study, we describe the generation and characterization of a T cell-engaging BsAb that targets murine CD3 using the 2C11 clone and EDB with the L19 clone. Specifically, the BsAb consists of a Fab fragment (targeting CD3) fused with two single-chain Fv fragments (targeting EDB) at the C-terminus. The BsAb was produced in Chinese hamster ovary cells and purified to homogeneity. To compare stromal and cellular targeting, two murine tumor cell lines naturally secreting EDB in the stroma were transduced to express the target on the cell surface. In both cell lines, biodistribution analysis revealed increased tumor uptake in the cellular model compared with the stromal one. Similarly, treating immunocompetent cellular EDB tumor-bearing mice with the BsAb improved anticancer activity. By contrast, no significant therapeutic benefit was observed in the stromal model. These findings underscore the importance of direct tumor cell targeting compared with stromal targeting for effective BsAb therapy.