DOTAM-TCB: Universal Small Molecule-guided Hapten- and T Cell-bispecific Antibodies for Cancer Immunotherapy

Abstract

Background

Tumor heterogeneity has been identified as a major roadblock for cancer immunotherapy. To overcome this, universal effector cell engagers with interchangeable tumor-targeting adaptors have been developed. Current concepts in this field consist of antibody-based adaptors. Small-molecule (SM) ligands, on the other hand, infiltrate tissues rapidly, have short half-lives, and are potentially orally available. Therefore, we hypothesized that utilizing SM adaptors, combined with effector antibodies, could represent an attractive off-the-shelf therapy.

Methods

Here, we introduce the development of target-agnostic, small-molecule-guided hapten- and T cell-bispecific (TCB) antibodies with high affinity between the adaptor-effector pair. Specifically, we designed SM adaptors based on known tumor-targeting ligands with specificity to the antigens folate receptor 1 (FOLR1), prostate-specific membrane antigen (PSMA) and carbonic anhydrase IX (CAIX), and conjugated them to Ca²⁺-loaded 1,4,7,10-tetrakis(carbamoylmethyl)-1,4,7,10-tetraazacyclododecane (DOTAM) as a hapten. As an effector antibody, we designed DOTAM-specific T cell bispecific antibodies (DOTAM-TCB) with femtomolar (fM) affinity for the hapten.

Results

In vitro experiments showed in-solution assembly of stable adaptor-effector complexes. Additionally, T cell-mediated tumor killing was initiated by the universal DOTAM-TCB when combined with adaptors across various tumor cells. Ex vivo study of CAIX-directed Ca-DOTAM-acetazolamide (Ca-DOTAM-AAZ) with DOTAM-TCB showed potent activation of T cells against murine-engrafted human HT-29 tumor.

Conclusions

The studies described here demonstrate proof-of-concept for using hapten-containing small molecules as adaptors for effective universal T cell engager-based cancer immunotherapy.

Statement of Significance This research introduces a novel approach to T cell redirection to tumors using hapten-based adaptors and antibody-based T cell engagers, with fM affinity between the components. This addresses current limitations in cancer immunotherapy, including tumor heterogeneity, limited stability of the adaptor-engager complexes, or complexity in computational design of tumor ligands.

Introduction

Cancer immunotherapy has transformed the landscape of oncology [1] with various therapeutics being utilized, including mono- and bispecific antibodies and cell-based therapies such as chimeric antigen receptor T cells (CAR-T cells) [2]. Adaptor-based strategies, described in the CAR-T cell field, tackle antigen escape, intra- and inter-patient antigen heterogeneity, and high manufacturing costs of personalized treatments. These strategies utilize universal target-agnostic adaptor-specific CAR-T cells, which can bind a range of different tumor-targeting adaptors, facilitating multi-antigen targeting with a single CAR construct [3].

Antibody-based adaptors are used with adaptor-specific CAR-T cells [3] or adaptor-specific bispecific antibodies as immune effectors, such as a P329G-engager platform developed recently by our group [4]. However, antibody adaptors can pose inherent challenges, such as a relatively long half-life of ~21 days [5], which may exacerbate on-target toxicities.

In contrast to antibodies, small molecules (SM) offer a distinct set of favorable characteristics, such as absence of immunogenicity risk [6], cost-effective production process [7], and short half-life – often ranging between several minutes to several hours [8]. Small molecules have traditionally been utilized in cancer therapy primarily for modulating the activity of intracellular targets, such as receptor tyrosine kinases. Unlike antibodies, SM can penetrate cell membranes [9]. However, recently, SM are increasingly researched as ligands for cell surface antigens on tumor cells, for example as SM-drug conjugates [10], diagnostic or intraoperative imaging agents [11], and delivery ligands for radiotherapy [12].

Small molecules can also act as haptens, which are compounds of low molecular weight that, when combined with a protein carrier, are capable of eliciting an antibody response against them [13]. This enables the development of recombinant anti-hapten antibodies, utilized, for instance, in pretargeted radioimmunotherapy (PRIT) [14].

With the developments in the fields of adaptor CAR-T cells and surface-targeted SM ligands as separate research domains, the concept of SM-guided adaptor CAR-T cells has emerged. Such advancements involve tumor-binding hapten-containing (haptenated) small molecule adaptors, paired with anti-hapten antibody binders in CAR-T cells. For instance, in 2015, Kim et al. introduced anti-FITC CAR-T cells for use with folate-FITC small molecule adaptors against folate receptor 1 (FOLR1) [15]. An alternative approach involves the use of SM adaptors with antibody effectors, developed for folate-FITC by Lu from the Low group [16]. Alternatively, in 2009, the Spiegel group pioneered a modular system with SM-based antibody recruiting molecules (ARMs), consisting of a PSMA ligand conjugated to a hapten tag DNP for use with anti-DNP antibodies to induce antibody-dependent cell-mediated cytotoxicity (ADCC) [17]. Since then, there have been further reports of ARM development, targeting tumor-associated antigens such as PSMA with DUPA molecule [18] or FOLR1 with folate molecule [19]. For T cell recruitment, FITC-labeled scFv-based adaptors have been recently introduced for preclinical use with FITC- and CD3-bispecific T cell engagers [20].

Despite many advancements, the use of small molecule tumor ligands as adaptors has been limited to tumor redirection of CAR-T cells or endogenous antibodies, or as conjugates to scFv-based adaptors. As such, the potential of engineered antibodies as effector molecules together with purely SM-based adaptors has been underutilized within this space. Here, we describe the redirection of universal adaptor-binding T cell bispecific antibodies (TCBs) to various tumor cells by a set of interchangeable hapten-containing SM adaptors, with the aim of combining advantages of small molecules with desirable off-the-shelf features and biological flexibility of T cell bispecific antibodies. Importantly, the adaptor-based character of this platform aims to enable universality in terms of targeting a tumor antigen of choice, without the need for separate production and development of an immune cell engager for each target.

The DOTAM-TCB adaptor platform described here utilizes an array of interchangeable small molecule-based adaptors, consisting of published SM ligands to tumor antigens FOLR1, PSMA, and CAIX, each conjugated to the same hapten, Ca2⁺-loaded 1,4,7,10-tetrakis(carbamoylmethyl)-1,4,7,10-tetraazacyclododecane (DOTAM) molecule. The adaptors are bound by a universal hapten- and T cell-bispecific antibody (DOTAM-TCB), forming a functional T cell engager complex, capable of activating T cells against chosen antigen-expressing cells. The mode of action of standard, direct TCBs is depicted in Fig. 1A, and contrasted with the platform depicted in Fig. 1B, along with the structures of the effector antibodies (Fig. 1C and D) and the SM adaptors (Fig. 1E–G). In this work, we present the development, together with functional and biochemical characterization of the platform in the context of cancer immunotherapy drug discovery.

Figure 1.

Mode of action and design of the universal DOTAM-TCB 2+1 with hapten-containing small molecule adaptors. (A) Mode of action of standard, direct T cell bispecific antibodies. Selected tumor associated antigen (TAA) and CD3ε are bound simultaneously, triggering CD3ε-mediated TCR crosslinking, leading to signal transduction, activation of T cells, cytokine release and tumor-directed cytotoxicity. Targeting each new TAA requires a new TAA-tailored antibody. (B) Mode of action of the DOTAM-TCB platform. Various tumor-associated antigens are targeted by interchangeable small-molecule ligands which are covalently bound to a Ca²⁺-loaded DOTAM hapten, irrespective of the targeted TAA. The adaptors are then bound by the universal DOTAM- and T cell-bispecific antibody (DOTAM-TCB) in the 2+1 format, together forming a functional T cell engager. This triggers T cell-mediated cytotoxicity. The same DOTAM-TCB can be used with various interchangeable adaptors, rendering it universal in terms of antigen targeting. (C) Different DOTAM-TCB antibody formats used in this study: 2+1 (two binders against CD3ε, one binder against DOTAM), 1+1 (one binder against CD3ε, and one binder against DOTAM on two different arms), and 1+1 one-armed (OA; one binder against CD3ε, and one binder against DOTAM on the same arm, connected by a flexible linker). (D) Antibody engineering principles incorporated in the design of the DOTAM-TCB. (E–G) Structures of the small molecule adaptors for use with DOTAM-TCB: (E) Ca-DOTAM-DUPA for PSMA targeting, (F) Ca-DOTAM-AAZ for CAIX targeting, and (G) Ca-DOTAM-folate for FOLR1 targeting. Pictograms created with BioRender.com.

Materials and methods

Detailed methods are provided in Supplementary Methods.

Molecule synthesis

Small molecule adaptors (Ca-DOTAM-DUPA, Ca-DOTAM-AAZ, and Ca-DOTAM-folate) were synthesized using standard Fmoc chemistry and purified via prep-HPLC. Antibodies were produced via transient transfection in HEK-293T cells, followed by a triple purification process. Complexes of SM-TCB were prepared by mixing the components in a protein buffer and purifying via centrifugal filtration with molecular weight cut-off.

Analytical and biochemical methods

Binding affinities were evaluated using KinExA and Surface Plasmon Resonance (SPR). KinExA was performed using beads coated with biotinylated Pb-DOTAM-Isomer Mix to determine the K_D of anti-DOTAM binder to Ca-DOTAM-ligands. SPR experiments assessed the binding of SM-TCB complexes to recombinant human CAIX, FOLR1, and PSMA using a Biacore T200. SM-TCB complexing efficiency was assessed by native SEC-MS with Q Exactive UHMR Mass Spectrometer (Thermo Scientific #0726090). For X-ray crystallography, the anti-DOTAM Fab fragment was mixed with Ca-DOTAM-AAZ compound and tested in sitting drop vapor diffusion, with grown clusters used as nucleation seeds. Three-dimensional crystals were analyzed by X-ray diffraction. Data were collected at a wavelength of 1.0005 Å using an Eiger2X 16M detector (Dectris) at the beamline X10SA of the Swiss Light Source (Villigen, Switzerland). The crystals diffracted to a resolution of 1.45 Å.

Cell-based in vitro assays

Functional activity was assessed using various human tumor cell lines cultured according to ATCC or DSMZ protocols. Tumor cell binding was measured by flow cytometry after a 1-h incubation with premixed molecules in saturating conditions. Target-dependent T cell activation was evaluated using Jurkat reporter cell lines (CD3/NFAT, Promega, custom) after 6 h of incubation of tumor cells, reporter cells, and molecules. For primary cell assays, cancer cells and molecules were incubated with primary human PBMCs or pan T cells from healthy donors. Following 48–72 h co-culture, cytotoxicity, cytokine release, and T cell activation markers (CD4, CD8, CD25, CD69) were quantified by luminescence, Luminex and flow cytometry, respectively. Real-time imaging of fluorescent tumor cell killing was monitored over 5–7 days using an Incucyte S3 system.

Ex vivo studies

The study protocol was approved by The Institutional Animal Care and Use Committee of Roche Innovation Center Zürich as well as the Cantonal Veterinary Office of Zürich under the Veterinary License ZH 180/2020 in agreement with The Swiss Animal Welfare Act.

Human HT-29 cells were injected subcutaneously into humanized BRGS-CD47 mice (Jackson Laboratory). Upon tumor growth, mice were sacrificed, and tumor tissues were collected and analyzed via standard histology and immunohistochemistry staining (H&E, CD3, CAIX). For the ex vivo assay, HT-29 tumor was collected and processed mechanically and with enzymatic digestion. After obtaining single-cell suspension, PBMCs from healthy human donors and the molecules were co-cultured and incubated for 48 h. T cell activation was assessed by flow cytometry.

Statistical analysis

Data were analyzed and visualized using Tibco Spotfire and GraphPad Prism 10.2.3. Statistical significance was determined using paired or unpaired ANOVA (one-way or two-way) with Dunnett’s or Tukey’s multiple comparison tests as appropriate. P-values in the figures were represented by the following: ^*P < .05, ^**P < .01, ^***P < .001, and ^****P < .0001. ns - non-significant.

Use of artificial intelligence

Large language models were used for minor language refinement.

Results

DOTAM-TCB design

In this work, we aimed to design a platform using an array of interchangeable small molecule-based tumor ligands, each conjugated to the same hapten, redirecting a universal hapten-specific T cell engaging antibody. In order to retain properties of low molecular weight adaptors, we focused on conceptualizing small molecule haptens, as opposed to protein or peptide-based adaptors, mediating the binding to hapten-specific TCBs. Since T cell activating capacity relies in part on the stability of the synapse [21], careful consideration was given to this aspect. Notably, several high affinity hapten-binding antibodies have been developed previously [22–25]. Therefore, despite the non-covalent binding between our conceptual adaptors and an effector antibody, we postulated that the hapten – anti-hapten antibody interaction is possible with sufficiently high affinity for effective synaptic activity. Consequently, the tumor-targeting adaptors, bearing haptens, would be recognized by an anti-hapten T cell bispecific antibody, and form a functional T cell engager, capable of stable synapse formation.

As a T cell engaging moiety, an anti-DOTAM x anti-CD3ε T cell bispecific antibody was used, henceforth referred to as DOTAM-TCB. The 2+1 TCB structure was derived from the FDA-approved CD20-TCB antibody glofitamab [26]. Briefly, the format consisted of a human IgG1, with a humanized DOTAM binder on one arm, and humanized CD3ε with another DOTAM binder on another arm, connected by a flexible linker (Fig. 1D). To ensure proper assembly during antibody production, several technologies were incorporated into the protein structure, such as CrossMab [27], charge pairs, and knobs-into-holes. For Fc silencing, P329G LALA mutations were used [28] (Fig. 1D). To expand the possibility of valency and format assessment on the TCB activity, two more formats were designed later in the study: TCB 1+1 with one binder against CD3ε, and one binder against DOTAM on two different arms, and TCB 1+1 one-armed (1+1 OA), with one binder against CD3ε, and one binder against DOTAM on the same arm, connected by a flexible linker (Fig. 1C).

DOTAM is a known chelator of bivalent cations used in radiotherapy applications [29]. Therefore, for this research, Ca²⁺-loaded DOTAM, in conjunction with a DOTAM-specific antibody binder was selected as the hapten-anti-hapten binding pair [29]. In detail, the design of the small molecule adaptors consisted of: (1) Ca²⁺-loaded DOTAM, acting as a hapten, (2) a linker, and (3) a small-molecule ligand with specificity to a tumor associated cell surface antigen, with design based on published chemical structures. Three different tumor surface antigens were selected for the studies described herein: PSMA, CAIX, and FOLR1, due to their tumor association, and the availability of the structure of the surface-bound small-molecule ligands in the literature. More specifically, DUPA, AAZ, and folate binding to PSMA, CAIX, and FOLR1, respectively, were explored [30–34]. As a result, three Ca-DOTAM-ligand adaptors were designed: Ca-DOTAM-DUPA, Ca-DOTAM-AAZ, and Ca-DOTAM-folate, with their chemical structures depicted in Fig. 1E–G. The ligands and their linkers were derived from FITC-conjugated adaptors published by the Low group [30]. The interaction between the hapten-containing small-molecule ligand and the DOTAM-TCB antibody was designed to result in a functional T cell engager complex which represents an active therapeutic entity. With this approach, our aim was to combine the benefits of small molecules, hapten binding, and antibody-mediated T cell engagement, while retaining target-agnostic character of the effector antibody.

DOTAM-TCB binds Ca-DOTAM-ligands with femtomolar-range affinity

Initially, Ca-DOTAM-DUPA and Ca-DOTAM-folate, were designed and synthesized, alongside two DOTAM-TCBs: DOTAM-TCB 1+1 and DOTAM-TCB 1+1 OA. To validate binding of the DOTAM-TCB to the ligands, KinExA affinity measurements were performed. Equilibrium dissociation constants (K_D) were found to be in the double-to-triple digit femtomolar range (51–263 fM), as depicted in Fig. 2A. The values fell within an expected range, as previous experiments performed for the DOTAM antibody binder to Ca-DOTAM (lacking the tumor ligand part) resulted in similar observed affinity [29]. To expand the platform potential of this concept, an additional ligand, Ca-DOTAM-AAZ was designed and synthesized.

Figure 2.

Affinity measurements between DOTAM-TCB and Ca-DOTAM ligands and crystal structure of anti-DOTAM Fab and Ca-DOTAM-AAZ (PDB 30GR). (A) Tested TCB formats and K_D values derived from KinExA measurements of the interactions between DOTAM-TCBs and Ca-DOTAM-ligands. Error bars indicate 95% confidence interval. (B) Crystal structure-based representation of anti-DOTAM Fab with Ca-DOTAM-AAZ, displaying the antibody pictogram, the crystal structure of the Fab with the bound hapten-containing small molecule adaptor. (C) Localization of the Ca²⁺ ion and van der Waals and hydrogen bonds observed between the side chains of the amino acids in the anti-DOTAM Fab and the Ca-DOTAM-AAZ. (D) 2Fo-Fc electron density map of the bound Ca²⁺-loaded DOTAM. Ca²⁺ shown as a gray star. (E) Observed electron density map of the Ca-DOTAM-AAZ, with defined Ca²⁺ in the DOTAM cage and the exit vector. Remaining part of the small molecule structure with no defined localization was overlayed on the observed electron density map. Pictograms created with BioRender.com.

For further experiments, traditional direct TCBs were produced for use as positive controls for all ligands. The format was based on the DOTAM-TCB 2+1, replacing the DOTAM binders with the corresponding antigen binders. The direct PSMA- and T cell-bispecific antibody (PSMA-TCB) utilized J591 binders, the direct CAIX- and T cell-bispecific antibody (CAIX-TCB) utilized G250 binders, and the direct FOLR1- and T cell-bispecific antibody (FOLR1-TCB) utilized 16D5 binders.

To verify binding of the Ca-DOTAM-ligands to their tumor associated antigens, SPR and cell binding experiments were performed and are shown in Supplementary Figs 1 and 2, respectively. While both Ca-DOTAM-DUPA and Ca-DOTAM-AAZ displayed association with their respective binding partners (PSMA and CAIX), Ca-DOTAM-folate showed very weak binding to recombinant FOLR1 and no binding to FOLR1-expressing cells. These results resulted in deprioritizing Ca-DOTAM-folate for further experiments.

DOTAM binder crystal structure reveals Ca-DOTAM-ligand binding pocket

Next, X-ray crystallography was employed to elucidate the spatial organization of the bound molecules. Ca-DOTAM-AAZ was selected as an example ligand for this purpose, and tested with an anti-DOTAM Fab. Obtained crystals (PDB 30GR) revealed Fab-bound small molecules (represented in Fig. 2B) with multiple van der Waals and hydrogen bonds contributing to the high-affinity interaction with the hapten-based adaptor (Fig. 2C). An electron density map of the Ca²⁺ in the DOTAM cage is depicted in Fig. 2D. Importantly, the small molecules were observed with incomplete structures, wherein only Ca²⁺-loaded DOTAM with an exit vector was visibly discernible from the electron density map (Fig. 2E). The remaining part of the molecule had no defined localization based on the observed electron density in the crystals. This observation likely stems from the flexibility of the free/unbound part of the adaptor molecule and lack of a single defined space occupied by it.

The Ca-DOTAM portion was observed to be bound within a pocket situated between the VH and VL of the anti-DOTAM Fab. The binding pocket was formed by amino acid residues from CDR1 and CDR3 of the light and CDR2 and CDR3 of the heavy chain. Van der Waals interactions of the DOTAM cage were observed for side chains of TYR29L, TYR100L, PHE49H, and TYR57H. The amide groups of the DOTAM moiety coordinated the binding of the central calcium atom with three of the amide groups also being involved in hydrogen bonding with residues ASP55H, GLU97H, and ASP33L of the antibody (Fig. 2C and Supplementary Table 1). The Ca-DOTAM portion of the ligands bound within a pocket situated between the VH and VL domains of the anti-DOTAM Fab illustrates the high shape complementarity as the likely basis for the high affinity binding. The data collection and refinement statistics are presented in Supplementary Table 2.

Universal DOTAM-TCB shows in vitro proof-of-concept of T cell-mediated tumor cell killing with comparable kinetics to direct TCBs

For functional analysis, potential of universal DOTAM-TCB to crosslink CD3ε in a target-dependent manner was assessed, and compared to positive controls, i.e. corresponding traditional direct TCBs. For this purpose, target-expressing tumor cell lines were co-cultured with CD3/NFAT Jurkat reporter cells. As illustrated in Fig. 3A and B, Ca-DOTAM-DUPA or Ca-DOTAM-AAZ combined with DOTAM-TCB in a 2+1 format (dosed in molar ratio SM:TCB 2:1) resulted in dose-dependent Jurkat cell activation against target-expressing cells while remaining inactive with target non-expressing cells. While both molecules were characterized by a higher EC₅₀ than the direct TCBs, Ca-DOTAM-AAZ with DOTAM-TCB provided higher activity plateau. Both the adaptors alone and DOTAM-TCB alone, which served as negative controls, were inactive.

Figure 3.

T cell activation, kinetics and potency of tumor killing induced by Ca-DOTAM-ligands with DOTAM-TCB 2+1. (A and B) CD3/NFAT reporter Jurkat cell activation capacity of DOTAM-TCB combined with: (A) Ca-DOTAM-AAZ against CAIX+ or CAIX- cell lines, and (B) Ca-DOTAM-DUPA against PSMA+ or PSMA- cell lines. (C and D) Representative images at 0 h and 132 h of NLR+ red fluorescent tumor cell growth with no molecules (0 nM) and (C) Ca-DOTAM-AAZ with 0.05 nM DOTAM-TCB against CAIX+ NLR+ tumor cells, or (D) Ca-DOTAM-DUPA with 50 nM DOTAM-TCB against PSMA+ NLR+ tumor cells. (E and F) Kinetics of tumor cell killing induced by DOTAM-TCB 2+1 with (E) Ca-DOTAM-AAZ against CAIX+ NLR+ tumor cells, and (F) Ca-DOTAM-DUPA against PSMA+ NLR+ tumor cells. Graphs show representative data of three healthy pan T cell donors. Data collected in technical triplicates. Error bars indicate standard error.

Next, tests were conducted to evaluate the cancer cell killing capacity and kinetics via DOTAM-TCB-adaptor-directed human primary T cells. The kinetics was of particular interest, due to the adaptor and TCB being separate molecules, as opposed to classical direct TCBs. For this purpose, the Incucyte imaging platform was utilized. NucLightRed (NLR)-expressing target-expressing cancer cells were co-cultured with the molecules and human primary T cells isolated from healthy donors (NLR-negative). Cell growth and T cell-mediated killing kinetics were monitored by imaging every 4 h for 5+ days, with the red fluorescent area quantified for each image. Representative images from the assay depicting tumor cells with and without treatment by Ca-DOTAM-AAZ or Ca-DOTAM-DUPA with DOTAM-TCB at 0 h and 132 h are presented in Fig. 3C and D. As shown in Fig. 3E and F depicting quantified imaging data, Ca-DOTAM-AAZ or Ca-DOTAM-DUPA with DOTAM-TCB 2+1 (molar ratio SM:TCB 2:1) led to cancer cell growth inhibition in a kinetic manner similar to that of the direct CAIX-TCB or PSMA-TCB controls, respectively. However, responses across multiple doses revealed high potency of Ca-DOTAM-AAZ with DOTAM-TCB, with strong tumor killing activity outperforming positive control observed at the concentration of ≥0.05 nM TCB (Fig. 3E). In case of Ca-DOTAM-DUPA with DOTAM-TCB, the SM-TCB condition reached tumor killing comparable to the positive control at 0.5 nM TCB (Fig. 3F).

As part of the in vitro assessment of the molecule’s functionality, non-image-based killing assays were conducted, allowing evaluation of target cell killing alongside T cell activation and cytokine release. The results, shown in Supplementary Fig. 3 (Ca-DOTAM-AAZ with DOTAM-TCB) and Supplementary Fig. 4 (Ca-DOTAM-DUPA with DOTAM-TCB), provided evidence of dose-dependent activity of the SM-TCB treatments with comparable activity to positive controls.

DOTAM-TCB displays anti-tumor activity in vitro in various antibody formats

To explore the influence of antibody valency on the activity of adaptor-TCB complexes, DOTAM-TCBs in the following formats were produced and tested: 2+1, 1+1, and 1+1 OA (Fig. 4A). The cytotoxicity and T cell activation potential of the DOTAM-TCBs combined with Ca-DOTAM-DUPA (all dosed in molar ratio SM:TCB 2:1) were assessed. All formats exhibited dose-dependent cytotoxicity and CD8+ T cell activation capacity against PSMA+ LNCaP cells (Fig. 4B) and PSMA+ 22Rv1 cells (Fig. 4C). The 2+1 TCB format demonstrated the lowest EC₅₀, while the 1+1 OA format achieved the highest activity plateau at saturating concentrations, likely due to higher occupancy resulting from monovalent binding. Of note, the 1+1 and 1+1 OA formats, despite having the same valency for both DOTAM binder and CD3ε binder, displayed different activity profiles with the 1+1 format being less active. Considering its low EC₅₀, and potential avidity benefits, the optimal TCB format for further experiments was determined to be the 2+1 TCB.

Figure 4.

DOTAM-TCB antibody format comparison upon combination with Ca-DOTAM-DUPA, on two different PSMA+ cell lines: LNCaP and 22Rv1. (A) Tested DOTAM-TCB formats: 2+1, 1+1 and 1+1 OA. (B) LNCaP-directed cytotoxicity and T cell activation marker expression upon treatment by 2+1, 1+1 and 1+1 OA TCB formats. (C) 22Rv1-directed cytotoxicity and T cell activation marker expression upon treatment by 2+1, 1+1 and 1+1 OA TCB formats. Representative experiment of three healthy PBMC donors. Data collected in technical triplicates. Error bars indicate standard error. Pictograms created with BioRender.com.

Ca-DOTAM-ligands with DOTAM-TCB can be complexed into a functional TCB in a non-covalent manner due to affinity of Ca-DOTAM-anti-Ca-DOTAM interaction

Based on affinity data, we hypothesized that the small molecule adaptors might form stable non-covalent complexes through a simple process of co-incubation. Obtaining such preformed complexes was of special interest for future in vivo studies. Specifically, preformed SM-TCB complexes with antibody-like PK/PD properties could serve as a positive control of a separately administered SM and TCB. This could provide information on maximal drug activity. The separately injected small molecule adaptor and DOTAM-TCB would be used in an experimental test group to verify complex formation in vivo.

To assess complex formation in a simple preincubation process, a complexing protocol was developed (Fig. 5A). In this procedure, the SM adaptor was incubated with DOTAM-TCB 2+1 at a molar ratio of SM:TCB 4:1 or 100:1 (with SM in excess) for 1 h at room temperature, followed by removal of unbound SM using centrifugal filters with a molecular weight cut-off (30 kDa). The retained molecules in the protein solution after this procedure theoretically contain only molecules above 30–50 kDa, including fully complexed SM-TCB (TCB + 2 SM), half-complexed SM-TCB (TCB + 1 SM), or empty TCB (TCB + 0 SM). The relative amounts of these forms in the protein solution would depend on the complexing efficiency.

Figure 5.

Assessment of the complexing protocol of Ca-DOTAM-AAZ with DOTAM-TCB 2+1. (A) Schematic representation of the complexing method. (B) Native size-exclusion chromatography-mass spectrometry assessment of the drug solution after the complexing procedures, from 4:1 or 100:1 SM:TCB premixing ratio. (C) CD3/NFAT Jurkat cell activation upon co-culture with CAIX+ HT-29 cells and complex solutions with spiked Ca-DOTAM-AAZ or DOTAM-TCB. Error bars indicate standard error. Pictograms created with BioRender.com.

To verify and quantify this complexing efficiency, native size-exclusion chromatography-mass spectrometry (native SEC-MS) was performed. The results of the analysis of the drug solution after complexing DOTAM-TCB 2+1 with Ca-DOTAM-AAZ are presented in Fig. 5B. Both half-complexed and fully complexed SM-TCB were observed, and were comparable across the SM:TCB ratio samples. Therefore, it was concluded that increasing the SM:TCB ratio to 100:1 did not improve the complexing efficiency results, and the SM:TCB ratio of 4:1 was likely sufficient to saturate the DOTAM binding pockets with the established protocol.

Importantly, it cannot be excluded that native SEC-MS analysis may impact the observed integrity of the SM-TCB complex. This may falsely lead to the observation of partially occupied TCB (TCB + 1 SM), which may either stem from its presence in the original drug solution or form from fully occupied TCB (TCB + 2 SM) during analysis through SM detachment (e.g. in-source detachment). Due to this uncertainty, we decided to expand the complex evaluation to include a functional T cell activation assay to complement our SEC-MS findings. CD3/NFAT Jurkat reporter assay and Ca-DOTAM-AAZ—DOTAM-TCB 2+1 were chosen for this purpose. The analysis involved spiking in the complexed drug solution with either additional free SM or additional free DOTAM-TCB. The SM spike was introduced to bind any residual empty DOTAM-TCBs in the complexed drug solution. Analogously, DOTAM-TCB spike was used to bind any residual empty SM. Such spikes would be expected to increase T cell activation potential due to increasing concentration of active SM-TCB complexes. The Ca-DOTAM-AAZ—DOTAM-TCB 2+1 complex elicited the same dose–response curves without spiking, with SM spike, and with TCB spike, indicating the absence of free SM and empty DOTAM binders in solution (Fig. 5C). Additionally, comparison of complexed and separately added SM and TCB were performed in tumor killing assays for Ca-DOTAM-DUPA and DOTAM-TCB in different antibody formats. As shown in Supplementary Fig. 5, comparable cytotoxic activity was observed for complexed and separate molecular formulations. These results, which were generated without processing of the complexed sample, provide evidence that the SM-TCB complex preparation may have primarily resulted in fully complexed SM-TCBs.

Ca-DOTAM-ligands with DOTAM-TCB display in vitro activity across a wide range of molarity and concentration ratios, resulting in wide therapeutic windows

In vivo, optimal local levels of both SM and TCB are needed for efficacy. Therefore, exploration of the therapeutic window across a range of different SM:TCB ratios was initiated. To this end, the activity of SM adaptors and DOTAM-TCB was evaluated in vitro across a wide range of concentrations in a two-dimensional matrix. A CD3/NFAT Jurkat assay was performed, cross-titrating SM and TCBs from saturating concentrations to 0 nM using serial dilutions. Fig. 6A depicts a resulting heatmap of Jurkat cell activation, as reflected by emitted luminescence across a range of Ca-DOTAM-AAZ and DOTAM-TCB 2+1 concentrations, against HT-29 cells. As comparative controls, SM-TCB complex and direct CAIX-TCB 2+1 were used. Fig. 6B shows a heatmap of Jurkat cell activation across a range of Ca-DOTAM-DUPA and DOTAM-TCB 2+1 concentrations, against LNCaP cells, and compared to SM-TCB complex and to direct PSMA-TCB 2+1. For both targets, CAIX and PSMA, the activity window was found to be relatively broad across a wide range of SM and TCB concentrations. The heatmaps revealed bell shape curve tendencies across both variables, with the extent of the diminished signal dependent on the molecule concentration (Fig. 6A and B). Overall, the data provided evidence that efficacy could potentially be achieved over a wide range of SM and TCB concentrations and SM:TCB ratios, indicating that DOTAM-TCB-adaptor complexes could be exerting efficacy in vivo in future studies.

Figure 6.

Therapeutic window assessment by a cross-titration of Ca-DOTAM-AAZ or Ca-DOTAM-DUPA with DOTAM-TCB 2+1 in a functional CD3/NFAT Jurkat reporter assay. Left: cross-titration of test molecules. Right: titration of positive controls - complexed SM-TCB 2+1 and corresponding direct TCB 2+1. CD3/NFAT Jurkat T cell activation upon co-culture with: (A) CAIX+ HT-29 cells and Ca-DOTAM-AAZ (with CAIX-TCB as direct TCB control), and (B) PSMA+ LNCaP cells and Ca-DOTAM-DUPA (with PSMA-TCB as direct TCB control).

Ca-DOTAM-AAZ with DOTAM-TCB induces T cell activation against mouse-grown HT-29 cancer cells in an ex vivo assay

Next, we aimed to assess Ca-DOTAM-AAZ with DOTAM-TCB activity against in vivo grown cells expressing CAIX. First, HT-29 tumor samples from tumor-inoculated humanized BRGS-CD47 mice were stained histologically for human CD3 and human CAIX (Supplementary Fig. 6). Low to moderate-low CD3+ cell numbers were observed, mostly in stromal areas of tumors. Most of the tumor cells showed positive CAIX expression from very low to very high levels. HT-29 samples were therefore deemed suitable for ex vivo functional assays. An HT-29 tumor was collected, and a co-culture experiment was performed, including a dissociated HT-29 tumor, human primary PBMCs from healthy donors, and Ca-DOTAM-AAZ with DOTAM-TCB 2+1 or control molecules. The DOTAM-TCB-adaptor-mediated response was evaluated by flow cytometry by quantification of the T cell activation markers. The results of the ex vivo assay are shown in Fig. 7. Fig. 7A shows data from one representative PBMC donor, Fig. 7B shows area under the curve values from two healthy PBMC donors. For both donors, Ca-DOTAM-AAZ with DOTAM-TCB 2+1 displayed strong dose dependent T cell activating capacity. Interestingly, in the case of both donors and all tested activated T cell populations but CD25+ [% of CD8+], Ca-DOTAM-AAZ with DOTAM-TCB 2+1 showed significantly higher response than two positive controls – direct CAIX-TCB 2+1 and direct FOLR1-TCB 2+1. Importantly, empty DOTAM-TCB 2+1 did not elicit any T cell response. These data provided evidence that the level of CAIX expression in mouse-grown HT-29 tumor was sufficient for a potent activity of Ca-DOTAM-AAZ with DOTAM-TCB 2+1. The results provided additional proof of the molecules’ potency and their suitability for the HT-29 model targeting in future in vivo studies.

Figure 7.

Ex vivo functional assay results of mouse-grown HT-29 tumors, human healthy donor PBMCs and Ca-DOTAM-AAZ with DOTAM-TCB 2+1. T cell activation marker expression on CD4+ and CD8+ cell subsets from assay results performed with (A) individual dose response values from a healthy PBMC donor. (B) AUC values from two healthy donors.

Discussion

In this work, we designed, produced and characterized a small molecule-guided, hapten- and T cell-bispecific antibody platform using interchangeable small molecule adaptors against tumor-associated antigens. Chemical conjugation of the ligands to Ca²⁺-loaded DOTAM, which acted as a hapten, allowed formation of complexes with DOTAM- and T cell-bispecific antibodies. DOTAM-TCBs were shown to potently redirect cytotoxic T cells to adaptor-bound tumor cells in vitro and ex vivo. Importantly, the DOTAM antibody binder bound the adaptors with high affinity, and assembled into stable small molecule-antibody complexes in a non-covalent manner. In vitro proof of concept was demonstrated for TCB anti-tumor activity against several tumor associated antigens.

Ca-DOTAM in the adaptors was effectively bound with double-digit femtomolar affinity by the DOTAM-TCB. Importantly, there have been other reports of antibody-hapten binders with various affinities, such as anti-FITC with a reported K_D of 0.3 nM [35], or 30 pM [36] to FITC, or anti-DOTA scFvs that reached a K_D range of 8.2 pM [22]. The binder presented in this work, therefore, constitutes a significant advancement in binding strength, making the interaction virtually irreversible upon hapten binding.

Importantly, Ca-DOTAM-ligands retained binding to the respective tumor-associated antigens CAIX, PSMA, and FOLR1 when complexed with DOTAM-TCB. Additionally, the crystal structure of the complexed DOTAM binder and Ca-DOTAM-AAZ revealed a deep binding pocket between VH and VL domains, a result consistent with the literature on hapten-antibody interactions [37], and the most likely basis for the observed high affinity interaction.

Ca-DOTAM-DUPA binding to PSMA and Ca-DOTAM-AAZ binding to CAIX, combined with universal DOTAM-TCB, mediated effective T cell anti-tumor cytotoxicity in various in vitro assays in multiple antibody formats. Notably, neither adaptors alone nor DOTAM-TCB alone induced T cell-mediated cytotoxicity against tumor cells. However, the third adaptor, Ca-DOTAM-folate, combined with DOTAM-TCB displayed low to no activity and was consequently excluded from further analysis. Since folate is a well-known ligand of FOLR1 [38], this lack of activity may be attributed to suboptimal linker chemistry between folate and Ca-DOTAM, requiring further optimization.

The learnings from our previous antibody-based adaptor platform highlighted a necessity for optimization of both adaptor and the universal TCB in vivo dosing [4]. Thus, for this project, the goal was to compare both separate and complexed molecules. To this end, in addition to using defined SM-TCB mixtures in in vitro assays, a protocol was developed to generate purified non-covalent SM-TCB complexes, which was then evaluated via native SEC-MS. No uncomplexed TCBs were observed, and both partially and fully complexed TCBs were present. In contrast, functional results indicated very high complexing efficiency, indicating that native SEC-MS may introduce artifacts due to sample processing. To deepen the understanding of complex stability, further studies should include in vivo assembly and disassembly rates.

Cross-titration of both adaptors and the TCBs demonstrated a broad activity window in terms of absolute, as well as relative concentrations needed for target-dependent T cell activation. Additionally, human primary T cells were efficiently activated by Ca-DOTAM-AAZ with DOTAM-TCB against in vivo grown HT-29 tumors, indicating successful crosslinking of the SM-TCB complexes between the tumor cells and T cells.

For in vivo studies with Ca-DOTAM-AAZ DOTAM-TCB complexes, off-tumor expression of CAIX remains a concern [39]. Additionally, AAZ has been described to bind to other cell-surface human carbonic anhydrases, including but not limited to CAIV [40] or CAXIV [41], which can be present in other murine organs, such as kidney [42], therefore potentially leading to off-tumor activity. Additionally, when testing small molecule-guided TCBs in humanized mouse models, species specificity needs to be taken into account, as small-molecule ligands typically display species cross-reactivity, resulting in binding to both human and murine receptors. This might result in high observed toxicity, which is not reflective of what would happen in humans.

Another toxicity concern, related to the TCB as a modality, is the risk of cytokine release syndrome (CRS). The DOTAM-TCB platform offers several ways of minimizing this risk. Firstly, as shown previously by Trinklein et al., affinity of the CD3ε binder might be optimized for minimizing cytokine release while retaining killing capacity [43]. However, it is worth noting that such strategy has been shown to be more complex than simply affinity attenuation of the T cell-engaging binder [44]. Secondly, modularity of the DOTAM-TCB and the adaptors introduces the possibility of separate administration, potentially facilitating higher control of the TCB activity. Since SM have quicker distribution into tissues than antibodies [45], as well as a shorter half-life, administration of an inactive adaptor-binding TCB can be followed by an SM acting as a rapid on-switch. In case of adverse events, SM administration may be withheld as an off-switch. Alternative dosing strategy could be pretargeting the tumor with SM, and upon sufficient enrichment in the cancerous tissue, administering the TCB. Analogous scheduling, including both SM and antibodies, is successfully utilized in PRIT [14]. Thirdly, both SM and TCB administered as separate molecules offer possibilities for step-up dosing, which is another strategy used for CRS mitigation [46].

The DOTAM-TCB platform described herein, however, poses challenges. Importantly, for small molecule-based adaptors, reaching sufficient target specificity and affinity for extracellular targets with favorable pharmacokinetics can prove difficult. Furthermore, the optimal in vivo activity requires extensive optimization. However, beyond therapeutic use, and similarly to the previously described antibody adaptor-based anti-P329G CAR-T cell platform that can be used for drug screening [47] and academic research [48] in addition to its therapeutic application [49] (NCT05199519, NCT05270928, NCT05266768), the DOTAM-TCB adaptor system is suitable for exploratory research and drug discovery applications. Importantly, the concept can be extended to other modalities such as NK cell or macrophage engagement, co-stimulation, cytokine delivery, receptor signaling blockade, antibody drug conjugates, radioimmunotherapy or CAR-T cell therapy.

The work shown here provides characterization and in vitro proof of concept data for small molecule-guided hapten and T cell bispecific universal antibody DOTAM-TCB for use against multiple cancer types. This study constitutes the first described system of fully small molecule-based hapten-containing adaptors with fM affinity for IgG-based TCB, which, upon assembly, redirect T cells against tumor cells.

Author contributions

Marlena Surowka (Conceptualization [supporting], Data curation [lead], Formal analysis [lead], Investigation [lead], Methodology [equal], Project administration [lead], Software [lead], Visualization [lead], Writing—original draft [lead], Writing—review & editing [equal]), Theresa Kober (Investigation [supporting], Methodology [supporting]), Daniela Matscheko (Investigation [supporting], Methodology [supporting]), Andreas Ehler (Investigation [supporting], Methodology [supporting]), Wilson Macedo (Investigation [supporting]), Denis Assisi (Investigation [supporting], Methodology [supporting]), John Challier (Investigation [supporting], Resources [supporting]), Andrzej Sobieniecki (Investigation [supporting]), Reto Gianotti (Investigation [supporting]), Ali Bransi (Conceptualization [supporting], Resources [supporting]), Johannes Sam (Investigation [supporting], Methodology [supporting], Resources [supporting]), Juan J. Bonfiglio (Formal analysis [supporting], Investigation [supporting], Methodology [supporting]), Moreno Wichert (Investigation [supporting], Methodology [supporting], Resources [supporting]), Antonio Ricci (Investigation [supporting], Methodology [supporting], Resources [supporting]), Joerg Benz (Data curation [supporting], Formal analysis [supporting], Investigation [supporting], Visualization [supporting]), Pablo Umaña (Funding acquisition [equal], Supervision [supporting]), Dario Venetz (Conceptualization [equal], Methodology [equal], Resources [equal]), and Christian Klein (Conceptualization [equal], Methodology [equal], Resources [equal], Supervision [lead], Writing—review & editing [lead])

Conflict of interest

All authors are current or previous Roche employees and hold ownership of Roche stock and/or patents. MS, AB, MW, AR, DV, CK are inventors on the patent WO/2026/041568 and its patent family, entitled ANTIBODIES BINDING TO CD3 AND DOTAM, applied by F. Hoffmann-La Roche AG, which covers aspects of the methods/results described in this manuscript.

Funding

The work was funded by Roche Glycart AG.

Data availability

Original data and more detailed information on methodologies and protocols are available upon reasonable request from the corresponding author.

Ethics and consent statement

Buffy coats from healthy donors were obtained from Blutspende Zürich SRK (BASEC Nr. Req 2020-0098). Blutspende Zürich SRK confirmed that all donors provided written consent for the use of the samples for research purposes.

Animal research statement

The in vivo study protocol was approved by The Institutional Animal Care and Use Committee of Roche Innovation Center Zurich as well as the Cantonal Veterinary Office of Zurich under the Veterinary License ZH 180/2020 in agreement with The Swiss Animal Welfare Act.