TPP-45142-an Anti-HER2 T-cell Engager-Designed for Selective HER2-Low Cancer Immunotherapy.

Introduction
Advancements in immunotherapy have significantly transformed the clinical landscape of cancer treatment, with considerable attention on bispecific antibodies that redirect T cell–mediated cytotoxicity (1, 2). The first-generation bispecific T cell–redirecting antibody, blinatumomab, has demonstrated remarkable clinical efficacy and is approved for the treatment of relapsed or refractory B-cell precursor acute lymphoblastic leukemia (3), establishing a precedent for the clinical advancement of cluster of differentiation (CD)3-based bispecific antibodies. Recently, numerous bispecific T-cell engagers (TCE) have been developed that target B-cell maturation antigen or G protein–coupled receptor family C group 5 member D) for the treatment of relapsed or refractory multiple myeloma (4–6) and CD20 for the treatment of relapsed or refractory diffuse large B-cell lymphoma (7, 8).
Although the clinical proof of concept for T-cell redirection has been well established in hematologic malignancies, acquisition of definitive evidence of its clinical efficacy in solid tumors is challenging. This is attributed to impediments in T-cell trafficking, access to tumor sites, and the potentially insufficient number of T cells present within the tumor and its tumor microenvironment (TME). Furthermore, the immunosuppressive TME, influenced by various immune and stromal cell types, including myeloid-derived suppressor cells, regulatory T cells, and cancer-associated fibroblasts, may reduce the effectiveness of redirecting T-cell cytotoxicity toward solid tumors. TCE-based treatment of solid tumors is also challenging due to the absence of tumor-restricted antigens. Instead, the focus is on targeting antigens overexpressed on tumor cells compared with healthy tissue. This approach can result in on-target, off-tumor adverse effects due to T-cell reactivity against antigen-expressing organs (9, 10).
HER2 (ERBB2), a member of the human epidermal growth factor (EGF) receptor tyrosine kinase family, is expressed in various malignancies, including breast and gastric cancers; physiologic levels of HER2 have been reported in certain cancers, such as triple-negative breast cancer, which resist HER2-targeted monoclonal antibody (mAb) therapies (11, 12). HER2-low breast cancers [defined by a HER2 immunohistochemistry (IHC) score of 1+, or 2+ without gene amplification] represent a group of tumors with significant biological heterogeneity and predominant luminal molecular subtypes. These cancers are more prevalent in hormone receptor–positive breast cancers, exhibit reduced Ki-67 proliferation indices, and are less responsive to neoadjuvant chemotherapy [pathologic complete response (pCR) rates: 0%–29.2%; ref. 13]). Whether HER2-low breast cancer constitutes a distinct biologic or clinical group within breast cancers, particularly when compared with HER2-zero breast cancer, is still a matter of debate; nevertheless a defined algorithm guides its assessment of HER2-low breast cancers (13). Patients with HER2-low breast cancer are considered ineligible for standard-of-care therapies for HER2-positive cancers, namely trastuzumab and pertuzumab. Furthermore, due to low HER2 expression in various essential normal tissues, its tolerability remains a significant concern for HER2-targeting therapies. Clinical trials investigating HER2-targeting therapeutics with similar mechanisms of action, such as CD3-based TCEs and chimeric antigen receptor T (CAR T) cells, have reported dose-limiting toxicities at low dosages (14–16). Trastuzumab deruxtecan (T-DXd) is an approved HER2-targeting drug with a distinct mechanism of action. It is a trastuzumab-based antibody–drug conjugate (ADC) that delivers a cytotoxic payload, effectively targeting HER2-low tumor cells, but it can lead to serious side effects, such as interstitial lung disease or pneumonitis that require active monitoring and intervention (17).
In this study, we aimed to develop a next-generation HER2-targeting TCE that enhances selectivity for HER2-low tumor cells while minimizing its effects on normal tissues that express physiologic levels of HER2. Therefore, we engineered a NANOBODY domain–based bivalent HER2/T-cell receptor (TCR) TCE, designated as TPP-45142, against a unique HER2 epitope. TPP-45142 shows a promising potential for the treatment of HER2-low and HER2-positive tumors that are refractory to standard care, particularly trastuzumab-based therapeutics.

Materials and Methods

Cells and reagents
Breast cancer cell lines (female) were obtained from the American Type Culture Collection (ATCC; HCC1954 (RRID: CVCL_1259, banked January 2022), ZR-75-1 (RRID: CVCL_0588, banked December 2020), BT-20 (RRID: CVCL_BT20, banked September 2015), and BT-549 (RRID: CVCL_1092, banked June 2021) and cultured as recommended in growth medium containing 10% fetal bovine serum (FBS) at 37°C and under 5% CO2 in a humidified incubator (Binder model CB 150), as per the manufacturer’s instructions. After banking, including both master cell bank and working cell bank (WCB), the cell lines were checked for Mycoplasma contamination using MycoAlert Mycoplasma Detection Kit (Lonza, cat. #LT07-318) and MycoAlert Assay Control Set (Lonza, cat. # LT07-518; latest dates of testing as above). Assays were performed with cells having a passage number not higher than 10 passages starting from WCB, and their morphology was verified under a microscope. HER2 expression was verified using the QIFIKIT analysis (described below). No further cell line authentication was done. Human cardiac myocytes (HCM) were obtained from PromoCell (Germany) and cultured as recommended in myocyte growth medium (Promocell, cat. #C-22070) at 37°C and under 5% CO2. Human T cells were isolated in house from buffy coats (Red Cross Belgium) using RosetteSep (cat. #15021, STEMCELL Technologies) and cryopreserved until further use. Catalog numbers of cells and proteins are listed in Supplementary Table S1.

Analysis of HER2 expression
The number of HER2 molecules on the surface of different target cell lines and primary cells was determined using QIFIKIT (Agilent Dako, cat. #K0078). QIFIKIT contains (i) calibration beads—a series of beads coated with different, well-defined quantities of mouse mAb molecules, (ii) set-up beads, and (iii) fluorescein isothiocyanate (FITC) conjugate. All steps were performed using flow cytometry (FC) buffer (DPBS, Life Technologies, Gibco, cat. #14190-094) + 2% FBS (Sigma-Aldrich, cat. #F7524) + 0.05% NaN3). On the day of the assay, 50,000 target cells were transferred to a 96-well V-bottom plate (Corning, cat. #3894) and stained using 50 nmol/L mouse anti-human CD340 antibody (BioLegend, cat. #324402, RRID: AB_756118) or immunoglobulin (Ig) G1 isotype control antibody (BioLegend, cat. #400102, RRID: AB_2891079) for 30 minutes at 4°C. The plates were washed twice using FC buffer (intermediate centrifugation steps at 300 × g for 2 minutes). Before the second wash, 100 µL each of the calibration and set-up beads was added to the plate. Subsequently, 100 µL of 1/50-diluted goat anti-mouse FITC-conjugate antibody was added to both beads and target cells. The plate was incubated for 30 minutes at 4°C and washed twice using FC buffer (intermediate centrifugation steps at 300 × g for 2 minutes). Finally, the cells were resuspended in FC buffer supplemented with 5 nmol/L TO-PRO-3 iodide (Life Technologies–Molecular Probes, cat. #T3605). Fluorescence intensity was measured using the FC MACSQuant X (Miltenyi Biotec). Set-up beads were used to adjust the voltage settings. The mean fluorescence intensity of each population of the calibration beads was used to construct a calibration curve. The measured fluorescence correlated with the concentration of mouse mAb on the cells and the beads. The number of HER2 molecules on the surface of the cell lines was determined using the calibration curve. The obtained value for the isotype control (background) was subtracted from that obtained for HER2 molecules, which are presented as antibody-binding sites (ABS).

Molecule generation and in vitro assays

Molecule generation
NANOBODY domains—27A05, 29E09, and 47D05—were expressed in Escherichia coli and purified using anti-His beads (18). TPP-45142, TPP-45131, and TPP-45161 are NANOBODY domain-fragment crystallizable (Fc) molecules with different bivalent NANOBODY HER2-binding domains [27A05 and 29E09 (Fig. 1A), 2 × 27A05, and none (control), respectively]; the NANOBODY domain binding to TCRαβ and IgG4(FALA) Fc domain was present in all molecules. These molecules were expressed in Chinese hamster ovary K1 cells (RRID: CVCL_0214) and purified using protein A in a size-exclusion chromatography system (Evitria). The sequence of TPP-45142 is provided in Supplementary Data S1.

Affinity determination
The affinity of 27A05 and 29E09 with human, cynomolgus monkey (cyno) and mouse HER2 proteins was determined using surface plasmon resonance (SPR) on a ProteOn XPR36 protein interaction array system (BioRad Laboratories, Inc) at 25°C. Human, cyno, and mouse HER2 Fc proteins were immobilized on a ProteOn GLC sensor chip (BioRad Laboratories, Inc. cat. #176-5011; 20 µg/mL, 10 mmol/L acetate pH 4.0, 120 seconds, 30 µL/minute). Purified NANOBODY domain proteins were injected at six different concentrations (0.73–750 nmol/L) for 120 seconds, and dissociation was monitored for 600 to 1,800 seconds. Data were double-referenced by subtracting a reference ligand lane and a blank buffer injection. Processed sensor grams were analyzed based on a 1:1 interaction model (Langmuir-binding model) using ProteOn manager (version 3.1.0.6) software. Affinity of TPP-45131 and TPP-45142 with target proteins HER2 and TCR α and β (TCRαβ) was determined on a Biacore 8K SPR system (Cytiva) at 25°C. The target proteins were immobilized on different spots of a C1 chip (Cytiva) using Cytiva’s amine coupling kit according to the manufacturer’s instructions. Different concentrations between 800 and 45 nmol/L of TPP-45131 and TPP-45142 molecules were injected for 2 minutes and allowed to dissociate for 30 minutes at a 30 µL/minute flow rate. Between sample injections, the surfaces were regenerated using 10 mmol/L glycine (pH 1.5), and HBS-P+ buffer (Cytiva) was used as the running buffer. The kinetic rate constants (ka and kd) were calculated by fitting the sensorgrams with Biacore 8K+ Evaluation software (RRID: SCR_015936) using a 1:1 Langmuir-binding model, with mass transport limitations. The equilibrium dissociation constant KD was calculated as the kd/ka ratio. The binding of TPP-45131 and TPP-45142 was evaluated for human HER1 (ERBB1), HER3 (ERBB3), HER4 (ERBB4), mouse HER2, and rat HER2 using bio-layer interferometry (BLI) on an Octet HTX instrument (Pall ForteBio Corp). The proteins were immobilized on Octet AR2G biosensors (Pall ForteBio Corp., cat. #18-5092) till high densities (2.4–3.3 nm) were achieved. To verify the binding of the compounds, the sensors were first dipped into running buffer (1× HBS-EP+, pH 7.4) for 60 seconds to determine the baseline and then into wells containing TPP-45131 or TPP-45142 (1 µmol/L) in running buffer or into wells containing the respective positive control Ab (100 nmol/L) for 300 seconds. Data were analyzed using ForteBio software (version 11.1.0.4).

Epitope binning
In-tandem epitope binning was performed to map binding sites of the HER2-binding NANOBODY domains against each other as well as against trastuzumab and pertuzumab (in-house produced) via BLI on an Octet RED384 instrument (Pall ForteBio Corp.). The in-house biotinylated anti-human IgG-Fc NANOBODY molecule was immobilized on streptavidin sensors (ForteBio). Sensors were first dipped into a running buffer (1x HBS-EP+, pH 7.4) to determine the baseline setting. Subsequently, the sensors were dipped into wells containing purified NANOBODY domains, or positive controls, or trastuzumab and pertuzumab for the first association step (180 seconds), followed by dipping into wells containing a panel of different purified NANOBODY domains, or positive controls, or trastuzumab and pertuzumab. NANOBODY domain pairs were assessed for competitive binding. Additional binding by the second NANOBODY domain was determined with a report point at 20 seconds of association and via visual inspection. Additional binding indicated an unoccupied epitope, i.e., a noncompetitor, whereas no binding indicated epitope blocking, i.e., a competitor.

T cell–dependent cellular cytotoxicity assay (2D)
T cell–dependent lysis by TPP-45142, TPP-45131, and TPP-45161 (control molecule) was determined using the xCELLigence T cell–dependent cellular cytotoxicity (TDCC) assay; the assay was performed using the culture medium of the corresponding target cell line supplemented with 1× penicillin/streptomycin (Life Technologies, cat. #15140-122; hereafter referred as “assay medium”). On day 0, 50 µL of assay medium was added to a 96-well E-plate (Agilent Technologies, cat. #5232368001). The plates were placed in the xCELLigence RTCA device (Agilent), and any background noise was measured. Subsequently, 50 µL of the target cells were seeded in the plate at the correct seeding density (HCC1954 10,000 cells/well, ZR-75-1 15,000 cells/well, BT20 12,500 cells/well, BT-549 10,000 cells/well, and HCM 6,000–10,000 cells/well). The plates were placed back in the xCELLigence RTCA device, and impedance was measured every 15 minutes for 24 hours. On day 1, 8× concentrated compound dilution series and 8× concentrated (240 µmol/L) Alburex 20 human serum albumin (HSA; CSL Behring, cat. #2160-679) were prepared in the assay medium; 25 µL of HSA solution was added to each well of the plate, and 25 µL of diluted compounds were added to the appropriate wells. Human T cells were thawed and counted using the CASY TT cell counter (Cell Microsystems); 50 µL of T-cell suspension was added to the plate, with a final volume of 200 µL. T cells and target cells were coincubated in the xCELLigence RTCA device in an effector-to-target ratio (E:T) of 5:1. Impedance was measured every 15 minutes for 4 to 5 days. To account for well-to-well seeding variation, the cell indexes of the selected time point were normalized to the cell index right before compound addition. The following equation was used to calculate the % cytotoxicity: [NCI E + T – (NCI E + T + NANOBODY molecule)]/(NCI E + T) × 100; NCI E + T = normalized cell index (effector cells + target cells) and NCI E + T + NANOBODY molecule = normalized cell index (effector cells + target cells + NANOBODY molecule).

BT20 spheroid killing assay (3D)
BT20 Cytolight Green cells (in-house transduced) were subcultured in minimum essential medium, GlutaMAX Supplement (Life Technologies, Gibco, cat. #41090-028) supplemented with 10% FBS (Sigma, cat. #F7524), 1 mmol/L sodium pyruvate (Life Technologies, Gibco, cat. #11360-039), and 5 µg/mL puromycin (Life Technologies, Gibco, cat. #A11138-03) before use in the assay. The spheroid assay was conducted in the culture medium supplemented with 1× penicillin–streptomycin without puromycin (hereafter referred to as assay medium). On day 0 of the assay, BT-20 Cytolight Green cells were detached using TrypLE Express (Life Technologies, Gibco, cat. #12605), resuspended in assay medium, and counted. The cells were seeded at a density of 2,000 cells/well in 50 µL of assay medium in a 384 ULA spheroid microplate (Corning, cat. #4516). The assay plate was incubated in Incucyte S3 Live-Cell Analysis Instrument at 37°C under 5% CO2 for 2 days to facilitate spheroid formation. On day 2 of the assay, an 8×-concentrated IncuCyte Cytotox red reagent (Essen BioScience, cat. #4632) solution was prepared in assay medium. Ten microliters of the Cytotox Red solution was added to the assay plate, yielding a final concentration of 250 nmol/L. A dilution series of 16×-concentrated compound and 16×-concentrated HSA (CSL Behring, cat. #2160-679) solutions were prepared in assay medium. As a preincubation step, an equal amount of HSA was added to the compound dilution plates to yield an 8×-concentrated HSA/compound mix. Subsequently, 10 µL of this mixture was added to the assay plate. The final HSA concentration in the assay was 30 µmol/L. NANOBODY compounds were serially diluted starting from 0.01 µmol/L (final concentration) using a ¼ dilution step. Frozen human T cells (previously isolated from buffy coat) were thawed at 37°C, transferred to a 50-mL tube containing assay medium, and centrifuged at 250 g for 5 minutes. After discarding the supernatant, the cells were resuspended in assay medium and counted. T-cell concentration was adjusted to 0.04 E06 cells/mL, and 10 μL T-cell suspension was added to the assay plate (=400 T cells/well, E:T 1:5), resulting in a final assay volume of 80 µL. The assay plate was incubated for 12 days in Incucyte S3. Images were captured every 6 hours. Analysis was performed by plotting the Total Green Object Integrated Intensity against compound concentration. The data were normalized to the last time point just before compound addition, to account for variation in spheroid size. The % cytotoxicity was calculated similarly to the 2D assay with the following formula: [NTGO E + T– (NTGO E + T + NANOBODY molecule)]/(NTGO E + T) × 100
NTGO E + T = normalized Total Green Object Integrated Intensity (effector cells + target cells) and NTGO E + T + NANOBODY molecule = normalized Total Green Object Integrated Intensity (effector cells + target cells + NANOBODY molecule).

Proliferation assay
Proliferation assays for TPP-45142, TPP-45131, and TPP-45161 were performed using ZR-75-1 cells in an IncuCyte live-cell analysis system. At the beginning of the assay, 100 µL/well of ZR-75-1 cell suspension was seeded at 7,000 cells/well in 96-well clear flat-bottom polystyrene TC-treated microplates (Corning, cat. #3596). The cell suspension was prepared in assay medium [Roswell Park Memorial Institute Medium (RPMI) 1640 medium, Life Technologies, cat. #72400] supplemented with 10% heat-inactivated FBS (Sigma-Aldrich, cat. #F7524), 1 mmol/L sodium pyruvate (Life Technologies, cat. #11360), 1% penicillin–streptomycin (Life Technologies, cat. #15140), and 30 µmol/L HSA (Alburex, CSL Behring, cat. #2160-679). The wells on the edge of the plate were filled with 200 µL of the assay medium and not used in the assay. The plates were incubated for 15 minutes at room temperature in the IncuCyte system, which was placed in an incubator (37°C, 5% CO2, high humidity). After seeding for 24 hours, the medium was removed, and 100 µL/well of starvation medium (assay medium without FBS) was added. After 10 minutes of incubation at room temperature, the plates were placed back in the IncuCyte system. After 24 hours, the starvation medium was removed, and 200 µL of molecules serially diluted in the assay medium were added to the appropriate wells. Six 1/10 serially diluted curves were applied, starting at 100 nmol/L. Human EGF (R&D Systems, cat. #236-EG-200) was used as a positive control. At least four zero-dose (0 nmol/L compound) wells were included in each plate. The plates were sealed using Breathe Easy sealing film (Diversified Biotech, cat. #BEM-1). After 10 minutes of incubation at room temperature, the plates were placed back in the IncuCyte system until the time of analysis. For analysis, Phase Object Confluence was plotted against compound concentration.

T-cell activation assay
T-cell activation was examined by analyzing the expression of CD25 and CD69 on both CD8+ and CD4+ T cells using FC in ZR-75-1 and BT20 cells. On day 0, adherent cell lines (2 × 104 cells/well) were plated in 50 mL of R10 medium [RPMI 1640 (Gibco, cat. #11875-085) with 10% heat-inactivated FBS (Gibco, cat. #10082-147)] in 96-well flat-bottom plates (Corning Costar, cat. #3596); the cells were allowed to adhere for 30 minutes at 37°C and under 5% CO2. After incubation, 50-µL serial dilutions of NANOBODY molecules (starting at 20 nmol/L, 5-fold dilution, nine concentration points, diluted in R10 medium) were added to each well of the 96-well flat-bottom plate. Subsequently, 10 × 104 cells/well of human pan T cells (Stemcell, cat. #70024) were supplemented with 16 µL/mL HSA (OctaPharma, cat. #68982-0643-02) in 100 µL of R10 medium and added to the 96-well plate. Next, 200 µL of R10 medium were added to all unused wells, and the plates were centrifuged at 1,000 rpm for 1 minute and incubated at 37°C and under 5% CO2 for 2 days. On day 2, the plates were removed from the incubator and centrifuged at 1,000 rpm for 5 minutes, and the supernatant was used for subsequent cytokine analysis. The plates were washed and incubated in Fixable Viability Dye eFluor 780 (eBioscience, cat. #65-0865-14) for 30 minutes at 4°C in the dark according to the manufacturer’s instructions. The cells were washed and stained using the following antibodies, according to suppliers’ instructions: anti-CD4 (BioLegend, cat. #344648, RRID: AB_2734350), anti-CD8 (BD Biosciences, cat. #563795, RRID: AB_2722501), anti-CD25 (BioLegend, cat. #302610, RRID: AB_314280), and anti-CD69 (BioLegend, cat. #310906, RRID: AB_314841) for 30 minutes at 4°C in the dark. Following fixation with 4% paraformaldehyde (PFA; Thermo Fisher Scientific, cat. #28908), the cells were analyzed on an LSR Fortessa system (BD Biosciences). Data were analyzed using FlowJo (FlowJo LLC, version 10, RRID: SCR_008520) and GraphPad Prism (RRID: SCR_002798, GraphPad Software Inc., version 9), and the curves were fitted using a four-parameter dose–response curve (variable slope) model in GraphPad Prism.
The supernatants obtained on day 2 of the T-cell activation experiment were evaluated via cytokine analysis using the V-Plex Human Proinflammatory Panel 1 custom kit (Meso Scale Diagnostics, cat. #K15049D-4). Ten-fold dilutions of each supernatant were prepared using Diluent 2 from the kit. The plates were washed, and the samples and calibrators were added to appropriate wells. The plates were sealed and incubated at room temperature in the dark under shaking conditions for 2 hours. Next, the plates were washed, and detection antibody solution was added to each well; the plates were sealed again and incubated as in the previous step. Finally, the plates were washed, read buffer was added to each well, and analyzed on a Meso Sector S 600 plate imager (Meso Scale Diagnostics). The data were analyzed using Discovery Workbench software (RRID: SCR_019192, version 4) and GraphPad Prism (version 9), and the curves were fitted using four-parameter logistic regression in GraphPad Prism.

Confocal microscopy
Super-resolution confocal microscopy was performed to differentiate RRID: SCR_002798 between HER2 distribution in BT-549 cancer cells and HCM; the cells were plated on poly-D-lysine–treated glass coverslips, seeded overnight, fixed with 1% PFA for 10 minutes, and immunolabeled using a HER2 mAb (2 µg/mL; Abnova #H00002064-M02, RRID: AB_875497). HER2 was detected using goat anti-mouse IgG conjugated to Alexa Fluor 488 (4 µg/mL), and 4′,6-diamidino-2-phenylindole (4 µg/mL) was used to stain cell nuclei. Optical z-stacks were acquired on an LSM 880 confocal microscope (Carl Zeiss) using an Airyscan super-resolution detector array; the images were deconvolved to improve resolution and signal-to-noise ratio. Maximum intensity projections of the z-planes were generated and used for quantification. The differences in HER2 distribution between BT-549 cells and HCM-c were analyzed using ImageJ/Fiji software (RRID: SCR_003070, NIH Image, version v.1.54h) with the 3D Objects Counter function. The data were plotted to show the frequency distribution of the cumulative number of objects sorted by area (µm2).

Cytokine release assay
For the Modular IMmune In vitro Construct (MIMIC) assay, fresh leukocytes [red blood cells (RBC)-depleted whole-blood cells] were acquired from blood samples of 23 normal healthy donors who provided informed consent and were enrolled in the Sanofi Pasteur VaxDesign donor program (Chesapeake Research Review, Inc., protocol 0906009). The blood components of each sample were isolated. Briefly, whole-blood leukocytes and their corresponding autologous platelet-poor plasma were separated by centrifugation at 2,100 rpm for 10 minutes. The separated plasma was centrifuged at 2,100 rpm for 10 minutes and passed through a 0.2-μm filter (polyethersulfone; Nalgene) to remove platelets and platelet particles. Next, the leukocyte pellets were mixed with a sterile solution of 5% (w/v) dextran (Sigma) to allow for sedimentation of the erythrocytes. Thereafter, supernatants were collected and washed three times using DPBS buffer (Lonza) by centrifugation at 2,100 rpm for 10 minutes. Then, the RBC-depleted cells were counted via trypan blue exclusion staining. Briefly, cells were diluted 1:100 using acetic acid (R&D systems), and 20 μL of the mixture was add to the hemocytometer (Hauser Scientific) and counted at 10× magnification using a clinical microscope (Nikkon eclipse E200). The procedure for the automated MIMIC cytokine release assay (CRA) began on day 0 by coating 96-well plates with a collagen solution. After 1 day, an endothelial cell line was seeded onto the collagen cushion in a medium–serum solution, and after the cells become confluent (∼4 days), the medium was replaced with a nonserum medium. Fresh RBC-depleted whole-blood cells and test agents were applied to the construct after 1 day. Next, after 20 to 22 hours, culture supernatants were collected and analyzed for cytokine/chemokine production using EMD Millipore’s MILLIPLEX MAP Human Cytokine/Chemokine Magnetic Bead Panel (MiliporeSigma, cat. #HCYTOMAG-60K). This kit was used for the quantification of the following human cytokines and chemokines: granulocyte–macrophage colony-stimulating factor(GM-CSF), IFN-α2, IFN-γ, IL10, IL12p40, IL12p70, IL1β, IL2, IL4, IL5, IL6, IL8, IFN-γ–induced protein 10, monocyte chemoattractant protein-1, macrophage inflammatory protein-1 β, regulated on activation normal T-cell expressed and secreted, and TNF-α. For a selected group of cytokines (GM-CSF, IL2, IL6, IL10, IFN-γ, and TNF-α), the fold change from baseline was determined at each tested concentration for each cytokine and donor. The highest concentration that resulted in <250-fold change across all donors was reported. For detection of HER2 expression in endothelial cells, EA. hy926 cells were harvested, washed using PBS, and stained using Live-Dead Aqua (Invitrogen) for 20 minutes on ice. After washing and blocking using IgG-Fc (mouse normal serum, Jackson ImmunoResearch Inc. cat. #015-000-120, RRID: AB_2337194), the cells were incubated with anti-HER2/neu FITC antibody (BD Biosciences, cat. #340880, RRID: AB_400166). Thereafter, the cells were washed with buffered medium and analyzed on a BD Fortessa flow cytometer (BD Biosciences) equipped with BD FACS Diva software (version 9). Data were analyzed using FlowJo software (version 10, Tree Star). Cytokine heatmaps were prepared using geometric mean values for all donors, treatments, and analytes, in R studio version 4.1.2 with heatmap package (https://cran.r-project.org/package=pheatmap), using log2 (fold change) values relative to geometric mean values. The heatmap package was downloaded from the Comprehensive R Archive Network repository.

Structural biology

Complex formation
Extracellular HER2 (UniPort code: P04626 residues 23–652) was purchased from Selvita (Poland). Anti-NANOBODY domain Fabs (TPP-37990 and TPP-37996) were produced in-house. To increase the size of the HER2-NANOBODY domain complexes, anti-NANOBODY domain Fabs were included in the complexes, i.e., HER2–27A05–47D05–Fab37996 and HER2–29E09–Fab37990. For each complex, 1 nmol HER2 was incubated with NANOBODY domains at room temperature for 1 hour in Buffer A (50 mmol/L Hepes pH 7.5, 150 mmol/L NaCl) at a ratio of 1:1.2. Next, the Fab was added to the resulting HER2-NANOBODY domain complex at a 1:1.2 ratio of complex: Fab (TPP-37990 or TPP-37996 Fab accordingly). After incubation for 1 hour at room temperature, the complexes were subjected to gel filtration using a Superdex200-increase-3.2/300 column (Sigma-Aldrich GE28-9909-46), and the purified fractions were used for grid preparation.

Grid preparation
For each complex, 3 μL of freshly prepared gel-filtered fraction was applied to glow-discharged UltrAufoil R 0.6/1 grids. Grid vitrification was performed in a Vitrobot apparatus (Thermo Fisher Scientific) at 95% humidity using ashless filter paper (Standard Vitrobot Filter Paper, Ø55/20 mm, Grade 595; Electron Microscopy Sciences) with a blot time of 5 seconds and blotting force of 5. The grids were stored in liquid nitrogen and observed on a 200 keV Thermo Fisher Scientific Glacios Cryo-Transmission Electron Microscope; images were recorded in linear mode using a Falcon IV direct electron detector camera (Supplementary Table S2). The images were collected using EPU software (version 2.10, Thermo Fisher Scientific), with a defocus range from −0.8 to −2.2 μm. The pixel size was set to 0.58 Å, and an overall dose of 67 e−/Å2 was used. The number of micrographs collected were 20.015 and 6.063 for the complexes HER2–27A05–47D05–Fab37996 and HER2–29E09–Fab37990, respectively (Supplementary Table S2).

Cryo-electron microscopy data processing and model building
Movie frame alignment and contrast transfer function (CTF) estimation were performed using CryoSPARC software (RRID: SCR_016501, v3.2). Micrographs (19) were quality curated and CTF-fit cut to 5. After a round of blob picking and several rounds of 2D classification, templates were created for picking. Particles were cleaned using several rounds of ab initio classification. Finally, uniform 3D refinement followed by no-uniformed 3D refinement was used to obtain the desired map. The resolution of the final reconstructions was estimated using the value at which the Fourier shell correlation curve was <0.143 at 2.66 Å and 3.01 Å for HER2–27A05–47D05–Fab37996 and HER2–29E09–Fab37990 complexes, respectively. A structural model of HER2-PertuzumabFab-TrastuzumabFab was obtained from the Protein Data Bank (PDB) database (code: 6OGE) and used as the template; (Supplementary Table S2). Anti-NANOBODY domain Fabs were not included in the final model. Structure refinement and model building were performed using Phenix (RRID: SCR_014224) v1.19.2-4158 (Python-based Hierarchical Environment for Integrated Xtallography; ref. 20) and COOT (RRID: SCR_014222) v0.9.6 (Crystallographic Object-Oriented Toolkit; ref. 21). Figures were generated and rendered using UCSF Chimera (22).

Animal studies

Mouse studies
A nuclear BioImaging study with the radiolabeled compounds was performed to evaluate pharmacokinetics (PK) and tumor accumulation for TPP-45142 and TPP-45161 in ZR-75-1 tumor-bearing NGS mice. The studies were approved by the Sanofi’s Institutional Animal Care and Use Committee. TPP-45142 and TPP-45161 were first purified in 0.1 mol/L Na2CO3 (pH 9.5) using a 7k MWCO Zeba spin column, conjugated with a 10-fold molar excess of p-SCN-Bn-deferoxamine (pSCN-DfO; B-705, Macrocyclics), incubated for 1 hour at 37°C (gentle stirring), and then purified in DPBS buffer using a 7k MWCO Zeba spin column. Quality check of conjugation ratio was performed using high-resolution mass spectrometry (pSCN-DfO/NANOBODY molecule ratio: 1.7). A day before the injection, the compounds were radiolabeled with 89Zr (Perkin Elmer) at low specific activity (As:0.14 MBq/µg) in 0.5 mol/L HEPES buffer (pH 7.4) for 90 minutes at room temperature and then purified in DPBS buffer using a 7k MWCO Zeba spin column. Subsequently, instant thin-layer chromatography (iTLC) was performed using chromatography strips (TEC control Biodex 150-771) and acetic acid 20 mmol/L (Sigma-Aldrich) as the running buffer. The iTLC profiles were analyzed for the following two fractions: (1) 89Zr chelated by NANOBODY molecule-DfO and (2) 89Zr chelated by unconjugated DfO, using an automatic gamma counter (1480 WIZARD; Perkin Elmer). Conjugation rates (%) were calculated by dividing the fraction of 89Zr-labeled to NANOBODY molecule-DfO with the fraction of 89Zr-labeled to NANOBODY molecule-DfO + fraction of 89Zr-chelated by unconjugated DfO. The conjugation rates were 98% to 99%.
For analysis of ZR-75-1 tumor growth in mice, female NOD scid gamma (NSG) mice (n = 20, 6 weeks, Charles River Laboratory, RRID: IMSR_JAX:010636) were housed under specific pathogen-free conditions in ventilated cages (5 mice/cage) with a filter top (Tecniplast) in the presence of cage enrichment with food and 8.5 µg/mL estradiol with 0.2% ethanol beverage ad libitum. The tumors were inoculated on the right ventral surface of the mice via intramammary injection with 200 µL of 50% dPBS Ca2+and Mg2+/50% Matrigel (Corning Incorporated, cat. #354263) containing 1 × 106 ZR-75-1 cells. Tumor size was measured using a caliper twice weekly. Activated T cells (1 × 107 per mouse) were injected 4 days before compound injections. Fifteen tumor-bearing mice were block-randomized based on tumor size (mean volumes: 173–207 mm3) to include 6 animals (3 per treatment) which were injected with an equal single dose of 5.7 µmol/kg radiolabelled NANOBODY molecules 35 and 36 days after inoculation. Study endpoints were based on the general level of discomfort and tumor sizes (<1,000 mm3); animal health was checked daily. Body weight was measured on each day of the experiment, and tumor size was measured twice weekly. From these measurements, the changes in body weight and tumor volume were calculated during the 6 days of the experiment. Mean radioactivity amounts were 0.54 to 1.7 MBq per mouse. Blood PK parameters of the compounds were evaluated for 6 days by microsampling (5 µL) at 5 minutes and 1, 2, 4, 6, 24, 48, 72, and 144 hours after injection. The radioactivity of the samples was measured extemporaneously using the gamma counter; it was expressed in CPM and then converted into µg/mL and normalized by considering the 89Zr decay time and injected amounts. To longitudinally visualize the in vivo biodistribution of NANOBODY molecules in the tumor-bearing mice, photon emission computed tomography (PET)/computed tomography (CT) imaging was performed using an Inveon PET/CT scanner (Siemens). The images were acquired at 4, 24, 48, 72, and 144 hours after injection (CT and PET acquisition times: 5 and 15 minutes, respectively) using a step and shoot mouse mode; the mice were under inhalation anesthesia (2% isoflurane in medical O2) on a heated bed (38°C). CT parameters included total rotation: 360, projections: 270, effective pixel size: 42.25 µm, transaxial CDD size: 4096 px, transaxial field-of-view (FOV): 86.54 mm, axial FOV: 83.83 mm, current: 500 µA, and voltage: 80 kV. PET parameters included axial scan length: 127 mm, level discrimination: 350 to 650 KeV, and timing window: 3.4 ns. Data were reconstructed using Siemens Medical Solutions IAW Clnt software v2.0.0.1050 with the following settings: CT Method: 9, Feldcamp algorithm, Shepp–Logan filter, noise reduction: medium, and beam mouse; PET OSEM3D/SP-MAP: 6 iterations, 18 MAP iterations scatter corrected, and 1.5-mm resolution/3D Histogram protocol. Data of 2D images (axial, coronal, and sagittal planes) were analyzed using Inveon Research Workplace software v4.2.0.8 and of 3D illustrative images using IRW multimodal 3D visualization software v4.2.0.15. Regions of interest were determined on a maximal part of organs [lung, heart, liver, kidney, spleen, tumor, muscle (background control), and bone] if the radioactive signal was above the detection level. Tissue accumulation of 89Zr-NANOBODY molecule was calculated as the percentage of the injected dose (%ID) and normalized for tissue weight, which was directly expressed by the software in percentage of injected dose per gram (%ID/g). %ID/g values were normalized by taking the 89Zr decay time and injected amounts into account.
Female NSG (NOD.Cg-PrkdcscidIl2rgtm1Wjl/SzJ; NOD SCID gamma, RRID: IMSR_JAX:010636) mice ages 6 to 8 weeks were purchased from The Jackson Laboratory. They were acclimated for ≥4 days before the experiment with ad libitum access to food (Harlan 2916 rodent diet) and sterile water. The animals were housed under conditions outlined in the NIH Guide for Care and Use of Laboratory Animals in compliance with the USDA Laboratory Animal Welfare Act, in a fully accredited American Association for Accreditation of Laboratory Animal Care facility at 640 Memorial Drive, Cambridge, Massachusetts, United States 02139. The studies were approved by Sanofi’s Institutional Animal Care and Use Committee. ZR-75-1 parental breast cancer tumor cells were purchased from the ATCC (ATCC CRL-1555) and cultured in complete medium, which included RPMI + GlutaMAX (Gibco, cat. #61870-036) supplemented with 10% heat-inactivated FBS (HI-FBS; Gibco-Invitrogen, cat. #10082-147) in 5% CO2 at 37°C. The cells were harvested and resuspended in Hanks’ Balanced Salt Solution (Gibco, cat. #14175-095). In a 50% mix with Matrigel (Corning, cat. #356234), 5 × 106 cells/200 µL per mouse were surgically implanted into the mammary fat pad of female NSG mice subcutaneously bearing 17-B estradiol pellets (Innovative Research of America, cat. #NE-121 0.72 MG/P). TPP-45142 was administered intravenously on day 21 to groups of 10 mice with a median tumor volume of 112 to 120 mm3 at doses of 500, 100, 50, and 10 µg/kg after 3 hours of intraperitoneal administration of 1 × 106in vitro–expanded T cells. Tumor volume was measured on days 21, 26, 29, 33, and 36 after tumor inoculation.

Nonhuman primate studies
Tolerability and PK/pharmacodynamics (PD) study was performed in nonhuman primates (NHP) for TPP-45131, which is a TPP-45142–like molecule. Cynomolgus monkeys (Macaca fascicularis), ages 27 to 42 months at the time of dosing of the first group, were supplied by NOVEPRIM Ltd.. Two monkeys, one male and one female, were dosed once via a 30-minute intravenous infusion in the saphenous vein using an intravenous catheter at a dose volume of 2.5 mL/kg, followed by a 2-week observation period. Naïve animals were dosed via sequential administration of 0, 10, 50, 100, and 125 µg/kg TPP-45131 [batch E21482, which was provided as a stock solution (7.41 mg/mL) in PBS]. Whole-blood samples for PK evaluation were collected before the injection and at multiple time points until day 14 after administration. Serum TPP-45131 concentrations were determined using a qualified Meso Scale Discovery–based ligand-binding assay. PK parameters were estimated using Phoenix WinNonlin software (RRID: SCR_024504, version 8.2). Tolerability was determined based on clinical observations, local tolerance scoring, body temperature, body weight, food consumption, oxygen saturation, arterial blood pressure, immunophenotyping, cytokine evaluation, and clinical pathology.

Statistical analysis
Statistical analysis of the data obtained through in vitro experiments was performed using Microsoft Excel (RRID: SCR_016137, Microsoft Corporation) and GraphPad Prism software (RRID: SCR_002798, version 8). For the mouse model studies, a two-way ANOVA with factors treatment and day (repeated) using an AR(1) covariance matrix (based on Akaike Information Criterion and Bayesian Information Criterion criteria) was performed on tumor volume changes from baseline. This was followed by a contrast analysis on each day (using the Bonferroni–Holm correction for multiplicity) to compare all groups with the negative control group. ΔT/ΔC was calculated on the last day of the analysis as the ratio of the median tumor volume change from baseline in treated groups versus the control group, multiplied by 100. The probability of <5% (P < 0.05) was considered significant. Statistical analyses for mouse studies were performed using SAS (1) version 9.4 software (RRID: SCR_022199).

Results

Generation of bispecific NANOBODY domain-Fc molecule with a unique HER2 epitope
TPP-45142 is a humanized bispecific NANOBODY domain-Fc molecule comprising two variable domains (derived from camelid-derived heavy chain antibodies and the variable heavy domain of heavy chain or NANOBODY domain) that specifically bind to HER2—27A05 and 29E09, a NANOBODY domain that targets the TCRαβ chains (23, 24)—TCE02, and an Fc IgG4 (Ig) domain containing F234A/L235A mutations to inhibit effector function (Fig. 1A). The NANOBODY domains 27A05 and 29E09 were selected because of their ability to bind to human and cyno HER2 without competing with trastuzumab and pertuzumab. HER2-binding NANOBODY domains showed affinities in the low nanomolar range, which were maintained when formatted as TPP-45142 (Supplementary Table S3). TPP-45142 did not bind with other members of the HER2 family, including HER1, HER3, and HER4, nor with rat or mouse HER2 (Supplementary Table S4). The NANOBODY domain TCE02 is a sequence optimized variant of TCE01 used in other TCE molecules (23, 24). The affinity of TCE02 to human and cyno TCRαβ (the extracellular domains of TCRα and TCRβ chains fused to a zipper peptide for dimerization) is in the low nanomolar range, similar to TCE01, and is maintained when formatted as TPP-45142.
Tandem epitope binning revealed that 27A05 and 29E09 competed with each other but not with trastuzumab or pertuzumab, indicating that they belonged to the same HER2 epitope bin but were distinct from the known antibodies. To better characterize the HER2 epitopes of 27A05 and 29E09, Cryo-electron microscopy (cryo-EM) was performed to resolve the structures of the anti-HER2 NANOBODY domains of TPP-45142 bound to HER2. First, we generated the complexes containing HER2 and a single NANOBODY domain which did not yield high-resolution structures due to size limitations. The acquisition of high-resolution data was facilitated by the addition of a NANOBODY domain–binding Fab (nbFab) which increased the overall size of the complex. For the 29E09–HER2 complex, Fab37990 was used as an nbFab, and the final complex HER2–29E09–Fab37990 was used for grid preparation (Fig. 1B). In the case of HER2–27A05, high resolution was achieved by adding the NANOBODY domain 47D05 and Fab37996 as nbFab (Fig. 1C), and the resulting complex HER2–27A05–47D05–Fab37996 was used for grid preparation.
The structure of the 29E09-containing complex was resolved at 3.06 Å, whereas the 27A05-containing complex was determined at an overall resolution of 2.66 Å (Fig. 1B and C). Local resolution analysis revealed reliable density at the NANOBODY domain–HER2 interfaces in both cases (Supplementary Figs. S1 and S2). Cryo-EM results confirmed that the binding regions of both NANOBODY domains overlapped with each other but did not compete with either trastuzumab or pertuzumab (Fig. 1D). Epitope analysis of HER2 revealed that 19 residues were closely involved (less than 4 Å distance) in the interaction with 29E09 and 18 residues in the interaction with 27A05; 15 of these residues were common between both epitopes (Fig. 1E; Supplementary Table S5). The cryo-EM structures revealed consistent binding interface areas of 901 Å2 and 949 Å2 for 27A05 and 29E09, respectively.
Protein interfaces, surfaces, and assemblies (25) analysis indicated stronger interaction of 27A05 with HER2 than that of 29E09, as evidenced by the calculated binding energy (BE) and ΔG (free energy). These differences can be attributed to the greater number of hydrogen bonds formed by 27A05 as well as the presence of three salt bridges, primarily facilitated by the interaction between Asp417 in HER2 and Arg52 in the NANOBODY domain (Supplementary Fig. S3). In contrast, no salt bridges were observed for 29E09, although the interface area was slightly increased. These findings are consistent with the results of SPR analysis, which revealed the similar affinity of 27A05 and 29E09 but lower dissociation rate (Koff) of 27A05 than that of 29E09.

TPP-45142 exhibits cytotoxicity against HER2-low cancer cells
To assess the functionality of TPP-45142, TDCC assay was performed using donor T cells and cancer cells in a co-culture system with an E:T ratio of 5:1. Several breast cancer cell lines with varying HER2 expression were used as target cells (Table 1). The ZR-75-1, BT-20, and BT-549 cell lines were categorized as HER2-low based on their HER2 density and IHC intensity. TPP-45142 demonstrated high potency in the TDCC assay against all breast cancer cell lines evaluated, including HER2-low cell lines, wherein its expression levels were as low as 7,000 ABS bound/cell and IC50 value was in the picomolar range (Fig. 2A–D; Table 1). The negative control, a NANOBODY-Fc molecule designed to mimic TPP-45142 but lacking the HER2-binding NANOBODY domains, did not exhibit any activity, indicating that the presence of the TCRαβ binder alone was insufficient for killing the tumor cells (Fig. 2A–D). To our knowledge, no consensus exists regarding the physiologic E:T ratio in patients, as this parameter is dynamic and depends on different factors, such as T-cell infiltration, tumor-associated antigen expression, and tumor architecture. Despite the challenges in translating an in vitro assay to the clinical scenarios, we developed a TDCC assay with BT-20 3D spheroids and using an E:T ratio of 1:5. In these stringent conditions, TPP-45142 still exhibited potent activity against tumor cell spheroids, achieving an IC50 value of 62.1 pmol/L and an efficacy of 74% (Fig. 2E; Table 1). The activation by TPP-45142 of CD4+ and CD8+ T cells—as measured by CD25 and CD69 expression—demonstrated EC50 values comparable with its cytotoxicity (Fig. 2F). This T-cell activation was accompanied by the production of cytokines consistent with the expected mode of action, including IL2, IFN-γ, and IL6 (Fig. 2G). Given the bivalent nature of TPP-45142 and promotion of cell proliferation via HER2 signaling, it is essential to determine whether cancer cell proliferation can be induced in the absence of T cells. In a proliferation assay conducted using ZR-75-1 cells, unlike the positive control EGF, TPP-45142 did not induce cell proliferation (Supplementary Fig. S4). Thus, the bispecific binding of TPP-45142 with TCRαβ on T cells and HER2 on target cells effectively activated T cells and induced TDCC against HER2-low tumor cells.

TPP-45142 accumulates in tumor cells and induces tumor regression in a HER2-low breast cancer mouse model
The PK characteristics and tumor accumulation of radiolabeled 89Zr-TPP-45142 and 89Zr-TPP-45161 [non-HER2 negative control, a molecule comprising NANOBODY domain binding to TCRαβ and IgG4(FALA) Fc domains but not HER2 NANOBODY binding domains] were evaluated in a ZR-75-1 xenograft model in the presence of T cells. Both 89Zr-TPP-45142 and 89Zr-TPP-45161 exhibited similar PK behavior with a half-life of approximately 4 days (Supplementary Table S6; Fig. 3A). At all time points, significant accumulation of 89Zr-TPP-45142 was observed in the tumor compared with that of 89Zr-TPP-45161, driven by the binding of TPP-45142 to HER2 on ZR-75-1 tumor cells (Supplementary Table S7; Fig. 3B). No adverse uptake for TPP-45142 compared with TPP-45161 was observed overtime in the heart, kidney, liver, and spleen (Supplementary Fig. S5).
To assess the in vivo efficacy of TPP-45142, the same ZR-75-1 mouse model was used. After letting the tumor grow until around 110 mm3, TPP-45142 and TPP-45161 were administered intravenously on day 21 at 3 hours after a single intraperitoneal administration of 10 × 106in vitro expanded T cells. Antibody treatment was repeated on day 29, and the study ended on day 36. TPP-45142 monotherapy induced significant antitumor activity in this model from day 29 to day 36, resulting in strong tumor stasis when dosed at 100 µg/kg (DT/DC = 0 at day 36, P < 0.0001); significant growth inhibition was also observed at 500 µg/kg from day 33 to day 36 (DT/DC = 9 at day 36 P= 0.0006; Fig. 3C; Supplementary Tables S8 and S9).

TPP-45142 does not exhibit cytotoxicity against HCM and demonstrates a favorable in vitro safety profile
Cardiotoxicity caused by HER2-targeted therapies could be because HER2 is expressed in normal tissues, such as in the heart (26). The assessment of cytotoxic activity of TPP-45142 on primary HCM revealed that TPP-45142 exhibited no cytotoxic activity toward HCM (6,000 ABS/cell; Fig. 4A) even though HER2 expression of HCM was similar to that of breast cancer BT-549 cells (7,000 ABS/cell; Table 1). To support this observation, super-resolution confocal microscopy was conducted to examine the differences in HER2 distribution between BT-549 cells and HCM via immunofluorescence. Although both cell types displayed comparable HER2 expression on the cell surface, distinct HER2-enriched areas were observed on the surface of BT-549 cells compared with that on HCM (Fig. 4B and C). The differences in HER2 distribution between these two cell types were assessed by counting the number of objects detected on the cell surface, which were sorted by area. The analysis confirmed that larger objects were detected on BT-549 cells than on HCM (Fig. 4D). This suggests that T cell–dependent lysis by TPP-45142 is facilitated in HER2-enriched areas, leading to more effective T cell–dependent killing than that in nonenriched areas.
As cytokine release syndrome (CRS) is a major safety concern for TCE therapies, MIMIC CRA was performed to evaluate the cytokine/chemokine response profiles induced by TPP-45142. This system was used to retrospectively evaluate the cytokine secretion profiles induced by TGN1412, which targets CD28 on T-lymphocytes and caused severe CRS during a first-in-human clinical trial (27), various underdevelopment TCE molecules (28, 29), and recently developed CD19-CAR T cells (30). The presence of the HER2 target antigen on the surface of endothelial cells (Supplementary Fig. S6) makes the MIMIC (Supplementary Fig. S7) CRA platform a suitable model for evaluating the cytokine profile of TCE molecules being developed against HER2-expressing tumors. TPP-45142 significantly induced the release of majority of the 17 cytokines and chemokines tested as compared with the negative control TPP-45161 at 0.4 and 10 nmol/L. Moreover, no difference was observed at the concentration of 0.0032 nmol/L, which is similar to that of cytotoxic IC50 value on HER2-low BT-549 (0.0023 nmol/L) cells (Fig. 4E). The in vitro toxic threshold (TT), defined as the maximum concentration that induces <250-fold change from baseline, is based on one of the diagnostic criteria for severe CRS in CAR T therapy in the clinic (31), and it was calculated for cytokines IL2, IL6, IFN-γ, IL10, and TNF-α (Table 2). The ratio between the in vitro TT and the cytotoxic IC50 value of HER2-low BT-549 cells provided a positive therapeutic index ranging from 7 to 174 for the major cytokines involved in CRS.

Nonclinical toxicology evaluation of TPP-45142–like molecule—TPP-45131
As a first exploration of the nonclinical toxicology of a HER2-targeting humanized bispecific NANOBODY-Fc molecule, we used a TPP-45142–like molecule—TPP-45131. TPP-45131 exhibited biochemical properties similar to those of TPP-45142, and it binds to the same HER2 epitope. The HER2-binding domains consisted of bivalent 27A05, replacing 27A05 and 29E09 of TPP-45142. Both TPP-45131 and TPP-45142 demonstrated comparable binding affinities to human and cyno HER2 (Supplementary Table S10), similar cytotoxicity potency toward breast cancer cell lines and primary HCM (Supplementary Table S11), and comparable cytokine release as per the MIMIC assay (Supplementary Fig. S8). An exploratory PK/PD/tolerability study was conducted to determine the tolerability and PK/PD profile of a single dose of TPP-45131, followed by a 2-week observation period. Cyno was selected as the most suitable species for preclinical toxicology studies considering the cross-reactivity of the test molecule (Supplementary Table S10). Clinical observations at the mid-high and high doses were transient and included increased body temperature, redness of the limbs and face, vomiting, hypoactivity, and/or hypotension. Clinical pathology changes included a dose-dependent transient increase in the levels of alanine aminotransferase, C-reactive protein, and fibrinogen and decrease in lymphocyte counts. These changes correlated with immunophenotyping changes, including a dose-related decrease in CD3, CD4+, and CD8+ on T cells and increase in levels of T-cell activation markers such as CD69, CD25, and PD-1 on CD4+ and CD8+ T cells. Transient dose-related increases in IL6 were noted with a peak observed 4 hours after the start of infusion. No changes were noted for IFN-γ, IL8, or TNF-α. The maximum tolerated dose for TPP-45131 was 125 μg/kg (1.13 μmol/kg). The total clearance was 4.2 to 11.8 mL/hour/kg, exhibiting a decreasing trend with increase in dose (Supplementary Fig. S9; Supplementary Table S12). TPP-45131 demonstrated a terminal half-life of approximately PK behavior with a half-life of 2 to 4 days. Given the similar properties of TPP-45131 and TPP-45142, the results of this study indicate the tolerability and PK/PD profile that can be expected for TPP-45142 and inform on eventual subsequent studies with this molecule.

Discussion
HER2 is a well-established therapeutic target in breast cancer. Trastuzumab is considered the standard-of-care treatment for patients with HER2-positive status, which is defined by a HER2 IHC score of 3+ or 2+/fluorescence in situ hybridization (+; ref. 32). However, this definition excludes patients classified as “HER2-negative” or “HER2-low,” resulting in a significant unmet need for patients classified as HER2-low (a HER2 IHC score of 1+, or 2+ without gene amplification; ref. 33). T-DXd, is an approved ADC for patients with HER2-low breast cancer, but it is associated with serious risks of interstitial lung disease or pneumonitis that require active management with prompt dose interruption and early institution of glucocorticoid treatment. A few other HER2-targeted therapies are in clinical development for HER2-low breast or gastric cancer/gastroesophageal junction (GEJ) adenocarcinoma, including zanidatamab (34) and DF1001 (NCT04143711), which function via antibody-dependent cell-mediated cytotoxicity and natural killer cell engagement, respectively. Recently, zanidatamab was approved for HER2-high biliary tract cancer based on the results from a clinical trial, wherein confirmed objective responses were observed for 41.3% participants of the HER2-high (IHC score: 2+ or 3+) cohort, but no responses were observed for the small HER2-low (IHC score: 0 or 1+) cohort (35). SAR443216 is a trispecific HER2/CD3 × CD28 TCE that couples a high-affinity anti-HER2 antibody with CD3 redirecting and CD28 costimulatory arms, and it demonstrated antitumor activity in a HER2-low tumor (ZR-75-1) mouse model (36). A major barrier to the clinical translation of T-cell therapeutics in solid tumors is on-target, off-tumor toxicity, which is often dose-limiting. For instance, a significant challenge in designing HER2-specific CAR-T cells is mitigating off-tumor cytotoxicity against normal epithelial tissues, as HER2 is also expressed on epithelial cells and is not a tumor-restricted antigen (2).
To address this therapeutic gap, we developed TPP-45142, a next-generation TCE that specifically targets HER2-low cancers while reducing the risk of on-target, off-tumor toxicity. Unlike many HER2-targeted therapy molecules based on trastuzumab, such as T-DXd and zanidatamab, TPP-45142 binds to a novel HER2 epitope, facilitating administration in combination with or following trastuzumab and trastuzumab-based therapies without a washout period. Structural analysis confirmed the novel binding site for NANOBODY domains in a region that did not overlap with that of either pertuzumab or trastuzumab. The binding regions of both NANOBODY domains were largely the same, which explains the previously observed competition between them. In vitro, TPP-45142 demonstrated high potency (in the picomolar range) against breast cancer cell lines, with HER2 expression ranging from an IHC score of 3+ (HCC-1954) to 0 (BT-549). This suggests that TPP-45142 could be an efficacious therapy for a broad population of HER2-expressing cancers, including HER2-low cancers. In vivo, tumor inhibition by TPP-45142 in a ZR-75-1 xenograft mouse model (IHC score: +1) was comparable with that by the HER2-targeting TCE, HER2/CD3 × CD28 trispecific antibody (36).
Efficacy against HER2-low tumors may extend to on-target cytotoxicity against HER2-normal cells. Trastuzumab has been associated with cardiotoxicity because of HER2 expression in cardiomyocytes and its role in maintaining normal cardiac structure and function, particularly under stress conditions (26). Therefore, we tested the cytotoxic effects of TPP-45142 on HCM, which have HER2 levels comparable with those of the cancer cell line BT-549. Our study revealed that the HCMs were spared by TPP-45142, which is likely attributable to the distinct distribution pattern of HER2 observed in cancer cells compared with that in normal cells. A comparative analysis revealed more enriched areas on the BT-549 cell line than on HCM, increasing the capacity for bivalent molecule binding. Although other molecules have exhibited avidity in differentiating between high and normal expression levels of the target (37, 38), TPP-45142 is the first molecule to effectively distinguish between low tumor and normal cell target expression, with similar HER2 expression levels. Additionally, zanidatamab can cluster HER2 and form caps on the surface of cancer cells, thereby exhibiting potent activity due to its biparatopic nature; however, it is unclear whether zanidatamab can form caps even in HER2-expressing normal cells (34). In summary, TPP-45142 can provide a therapeutic window in terms of on-target, off-tumor adverse effects.
CRS is a significant concern for TCE-based therapies (39). Clinical CRS has been associated with elevated levels of systemic IL10, IL6, and IFN-γ in peripheral blood, which can be effectively managed with anti-IL6 treatment, such as tocilizumab (39). In the present study, in vitro safety analysis revealed that TPP-45142 induced the release of a wide range of cytokines, including IL2, IL6, IL10, IFN-γ, and TNF-α, which is consistent with its mode of action. Therefore, we evaluated the therapeutic index related to each relevant cytokine and found that IL10 and IL6 had a favorable therapeutic index ranging from 7 to 174, indicating the potential for a therapeutic window. Thus, based on this observation and the dose-dependency of cytokine production, CRS risk can be reduced by lowering the first dose of infusion. In a single-dose NHP toxicology study, a TPP-45142–like molecule was well tolerated up to 125 µg/kg. TPP-45142 showed expected PK characteristics for a IgG4 LALA fusion NANOBODY in tumor-bearing NSG mice reconstituted with human T cells with half-life of approximately 4 days. Despite the fast clearance rate of TPP-45142 in blood, enhanced tumor accumulation was observed relative to the control, confirming that TPP-45142 effectively binds to HER2 expressed on tumor cells. This target engagement reduces the in vivo clearance rate of the molecule. For PK studies in NHP, the total clearance of TPP-45131 (TPP-45142-like molecule) decreased from 11.8 to 4.2 mL/hour/kg whereas the dose increased from 10 to 125 µg/kg. This nonlinear PK phenomenon can be attributed to TPP-45131’s binding to HER2 followed by internalization and potential degradation. This target-mediated disposition is nonlinear and saturable. Overall, there is decreased target mediated disposition of TPP-45131 with increased dose levels. For additional PK/PD and safety data on TPP-45142 to assess the in vivo therapeutic index and propose a safe first-in-human starting dose, a repeated-dose study in NHPs is required.
The approval of tebentafusp-tebn and tarlatamab has demonstrated that TCEs can effectively treat solid tumors when the target antigen is expressed with high selectivity in tumors compared with that in normal tissues (40, 41). Particularly, tarlatamab is a DLL3-targeting bispecific TCE approved for extensive-stage small cell lung cancer. DLL3 plays a crucial role in embryonic development by regulating the Notch signaling pathway. It has limited expression in adult tissues, thereby significantly reducing the risk of on-target, off-tumor toxicity. However, majority of tumor antigens are self-antigens, which are also physiologically expressed on normal tissues. Therefore, we adopted a NANOBODY domain–based bivalent approach to develop a TCE that selectively targets HER2-low tumors and spares normal HER2-expressing cells, particularly cardiomyocytes. Recently, a STEAP1-targeted XmAb 2 + 1 TCE (AMG 509) also demonstrated that bivalent targeting of STEAP1 enabled selectivity for cells with high STEAP1 expression compared with that for normal cells (38). Although the tumor selectivity of AMG 509 is driven by avidity, we posit that the differential binding of TPP-45142 to HER2-low breast tumors, as opposed to normal cells expressing similar levels of HER2, may be attributed to the differences in HER2 topology between the surface of tumor cells and normal cells, which warrants further investigation. Furthermore, given the intratumoral heterogeneity in patients with HER2-low tumors, we anticipate that TPP-45142 will be functional as long as HER2-enriched areas are present. The preclinical model used in our study, comprising healthy donor T cells engrafted into immunodeficient mice bearing human xenografts, presents inherent limitations. The model does not fully reflect the native breast TME, including TCE and stromal/immune interactions, nor does it reflect human HER2 expression in normal tissues. Therefore, definitive validation of the efficacy of TPP-45142 in heterogeneous tumors necessitates human clinical studies.
Following the positive DESTINY trial outcomes that established T-DXd as a benchmark for HER2-low disease, it is essential to position TPP-45142 in comparison with T-DXd and related HER2-directed modalities. T-DXd is an ADC requiring intracellular delivery of a cytotoxic payload, whereas TPP-45142 is a dual-specific NANOBODY TCE that redirects T cells through TCRαβ to a HER2 epitope that is distinct from those targeted by trastuzumab or pertuzumab. In vitro, TPP-45142 achieves picomolar cytotoxicity even in IHC 0 models (e.g., BT-549), suggesting potential activity at antigen densities below conventional ADC thresholds. In the T-cell therapeutics space for solid tumors, HER2-directed CAR-T therapies represent a competing approach (2); however, their prolonged in vivo persistence and potential for uncontrolled expansion have historically increased the risk of on-target, off-tumor toxicity. Conversely, the preclinical profile of TPP-45142, characterized by a short functional half-life enabling dose titration, a quantified cytokine release therapeutic index, and selective targeting of low-HER2 tumor cells, while sparing HER2-normal primary cells, supports a favorable risk–benefit profile for HER2-low solid tumors, pending clinical validation. Finally, the unique binding specificity and mechanistic properties of TPP-45142, which exhibits robust TDCC at low HER2 expression, support its evaluation beyond breast cancer, particularly in HER2-low gastric and GEJ adenocarcinomas. Consequently, TPP-45142 is anticipated to show activity in gastric cancer, and dedicated preclinical investigations along with biomarker assessments will further substantiate its translational potential in this indication.
In conclusion, TPP-45142 emerges as an effective next-generation TCE, demonstrating potent cytotoxic efficacy against HER2-low tumors with notable selectivity, significantly mitigating on-target off-tumor toxicity. Its distinct cytokine release therapeutic index, manageable safety profile, and tumor-selective mechanism strongly advocate for clinical advancement, positioning TPP-45142 as a potentially transformative therapeutic option for patients with HER2-low breast cancer and other HER2-low solid malignancies.

Supplementary Material
Supplementary Table S1Supplementary Table S1 comprises of manufacturer’s details of cells and proteins used in the study

Supplementary Table S2Supplementary Table S2 comprises of cryo-EM data collection, refinement, and validation statistics

Supplementary Table S3Supplementary Table S3 represents binding parameters of TPP-45142 with individual HER2-binding NANOBODY domains (human and cynomolgus HER2) and with TCRab-binding NANOBODY domain (human and cynomolgus TCRαβ zipper)

Supplementary Table S4Supplementary Table S4 shows binding levels of TPP-45142 with different targets as per BLI

Supplementary Table S5Supplementary Table S5 shows HER2 residues at a distance of 4 Å from the NANOBODY domains. Residues common between 29E09 and 27A05 are highlighted in bold

Supplementary Table S6Supplementary Table S6 shows PK parameters calculated from the exposure of the 89Zr-labeled NANOBODY molecules in blood samples of ZR-75-1 tumor-bearing NSG mice engrafted with in vitro expanded T cells

Supplementary Table S7Supplementary Table S7 compares statistical results for tumor accumulation of TPP-45142 and TPP-45161 at each time point

Supplementary Table S8Supplementary Table S8 shows ΔT/ΔC values on day 36

Supplementary Table S9Supplementary Table S9 comprises of statistical results for comparison of all groups with the vehicle group (daily)

Supplementary Table S10Supplementary Table S10 shows binding parameters of TPP-45131 and TPP-45142 for human and cyno HER2

Supplementary Table S11Supplementary Table S11 shows in vitro cytotoxicity of TPP-45131 and TPP-45142 in breast cancer cell lines and primary cells

Supplementary Table S12Supplementary Table S12 shows mean pharmacokinetic parameters of TPP-45131 in a non-clinical exploratory NHP study (n = 2)

Supplementary Figure S1Supplementary Figure S1A shows representative micrograph and 2D classes. Supplementary Figure S1B represents cryo-EM data analysis pipeline. Initial blob picking and particle subset selection to obtain a good and bad initial model. Particle cleaning using several rounds of heterogeneous refinement classification followed by non-uniform refinement using a mask. Supplementary Figure S1C shows local resolution estimation and Supplementary Figure S1D shows HER2 model with the residues within the 27A05 epitope highlighted in yellow (defined as all HER2 residues within 4 Å distance from 27A05).

Supplementary Figure S2Supplementary Figure S2A shows representative micrograph and 2D classes. Supplementary Figure S2B shows cryo-EM data analysis pipeline. Initial blob picking and particle subset selection to obtain a good and a bad initial model. Particle cleaning using several rounds of heterogeneous refinement classification followed by a non-uniform refinement using a mask. Supplementary Figure S2C demonstrates local resolution estimation and Supplementary Figure S2D shows HER2 model with the residues within the 29E09 epitope highlighted in yellow (defined as all HER2 residues within 4 Å distance from 29E09).

Supplementary Figure S3Supplementary Figure S3A depicts Left: interface summary of HER2-27A05; Middle: interaction radar; Right: interface parameter table. Monomers 1 and 2 correspond to HER2 and 27A05, respectively. Supplementary Figure S3A depicts: Left: interface summary of HER2-29E09; Middle: interaction radar; Right: interface parameter table. Monomers 1 and 2 correspond to HER2 and 29E09, respectively.

Supplementary Figure S4Supplementary Figure S4 shows that TPP-45142 did not induce HER2-low breast cancer cell proliferation. ZR-75-1 proliferation in the presence of TPP-45142. EGF was used as positive control and TPP-45161 as non-HER2 negative control (Neg. Ctrl), as measured by confluence using IncuCyte

Supplementary Figure S5Supplementary Figure S5 shows that there was no adverse uptake for TPP-45142 compared with TPP-45161 in different organs. Heart, liver, kidney and spleen accumulation of 89Zr-TPP-45142 or non-HER2 negative control 89Zr-TPP-45161 as measured by PET/CT imaging (n = 3) (time: post radiolabeled-compound injection).

Supplementary Figure S6Supplementary Figure S6 shows that EA. hy926 cells were cultured to confluency, harvested, and stained using anti-HER2-FITC antibody. Histograms depict HER2 expression profile. Blue histogram represents unstained control, and red histogram represents HER2 FITC-stained cells. This is a representative example of two experiments.

Supplementary Figure S7Supplementary Figure S7 depicts the automated 3D co-culture assay system consisting of fresh RBC-depleted whole-blood cells, autologous plasma, and a human endothelial cell line that is pre-cultured to confluency on a collagen bed. Culture supernatants were collected after 20 to 22 h of treatment and analyzed for cytokine/chemokine production using Luminex multiplex arrays (Figure created using BioRender).

Supplementary Figure S8Supplementary Figure S8 is a heatmap of cytokine concentration as measured using Luminex multiplex array after MIMIC CRA for comparison of TPP-45142 and TPP45131

Supplementary Figure S9Supplementary Figure S9 shows the mean pharmacokinetic profile of TPP-45131 in a non-clinical exploratory NHP study (n = 2, one male and one female)

Supplementary Data S1Supplementary Data S1. Sequence of TPP-45142