Programmable multispecific DNA-origami-based T-cell engagers

Abstract

Multispecific antibodies have emerged as versatile therapeutic agents, and therefore, approaches to optimize and streamline their design and assembly are needed. Here we report on the modular and programmable assembly of IgG antibodies, F(ab) and scFv fragments on DNA origami nanocarriers. We screened 105 distinct quadruplet antibody variants in vitro for the ability to activate T cells in the presence of target cells. T-cell engagers were identified, which in vitro showed the specific and efficient T-cell-mediated lysis of five distinct target cell lines. We used these T-cell engagers to target and lyse tumour cells in vivo in a xenograft mouse tumour model. Our approach enables the rapid generation, screening and testing of bi- and multispecific antibodies to facilitate preclinical pharmaceutical development from in vitro discovery to in vivo proof of concept.

Main

Programmable self-assembly with DNA origami enables fabricating discrete nanoscale objects with structurally well-defined two-dimensional and three-dimensional shapes from DNA molecules^1,2,3,4,5, including nanoscale devices^6,7,8, functional materials^9,10 and higher-order objects^11,12. DNA origami objects are addressable and can be modified with various biomolecules in a site-specific fashion^13,14. Previous studies have demonstrated the attachment of antibodies to DNA origami objects^15,16, the binding of antibody-conjugated DNA origami objects to cell surfaces^14,17,18 and the modulation of T-cell function^19,20. Recent developments such as the cost-efficient mass production of DNA origami raw materials and stabilization approaches for in vivo application^21,22,23 may enable the clinical translation of diverse therapeutical concepts such as DNA origami biomedical nanorobots^10,24.

In parallel to the advances in DNA nanotechnology, cancer immunotherapies have contributed to a paradigm shift in oncological treatment landscapes²⁵. In particular, T-cell-centred immunotherapies (for example, immune-checkpoint-inhibiting antibodies) are now established in clinical practice in various cancer entities²⁶. In addition, in B-cell-derived haematological malignancies (such as acute lymphoblastic leukaemia (ALL) or B-cell lymphomas), T-cell-engaging antibodies have led to clinical responses even in otherwise treatment-refractory patients²⁷. The Federal Drug Agency (FDA) and European Medicine Agency (EMA) have granted approval for blinatumomab, a CD19-CD3-bispecific T-cell engager (BiTe), prolonging the overall survival of patients.

In B-cell malignancies, using B-cell-associated antigens such as CD19 and CD20 has proven feasible, efficacious and manageable from a safety perspective, partly owing to established clinical treatments to manage induced B-cell aplasia²⁸. However, this is unique to B-cell-targeting agents and cannot be expected in other diseases. In solid cancers, tumour-associated antigens are often co-expressed on vital epithelial tissues, creating the risk for severe on-target off-tumour toxicities²⁹. Increasing cell-type specificity, for example, by the simultaneous targeting of multiple-tumour-associated antigens has the potential to minimize the risk for severe on-target off-tumour toxicities³⁰. In addition, targeting more than one antigen on a target cell may prove beneficial in preventing antigen-negative relapse, as sequential or simultaneous multiple targeting will enhance therapeutic pressure and counter the development of negative variants.

Thus, multispecific molecules targeting cancer vulnerabilities are needed to leverage immune cell potential in oncology. A large number of drug candidates are currently in preclinical and clinical development, with the focus shifting from bispecific antibodies and BiTe formats (four on the market and more than 100 in clinical development) towards formats with increased specificities or enhanced pharmacokinetic properties (eight candidates in clinical development)^31,32,33. Various approaches have been developed to produce multispecific antibodies, most of which rely on fusing engineered antibody domains³⁴. Although these approaches support controlling the degree of valency, the spatial geometric arrangements of the domains are restricted by the structural constraints of the protein scaffold used.

Here we used programmable self-assembly with DNA origami to create a synthetic antibody carrier platform called programmable T-cell engager (PTE). PTEs offer desirable properties for T-cell engagement, including the capability to modularly position antibodies (IgG, F(ab) or scFv) with control over valency, orientation and spatial arrangement. We provide the proof of concept of specific T-cell engaging in vitro and validate the PTE functionality in leukaemia models in vivo.

Assembly and screening of IgG-based PTEs

The ability to place IgG antibodies in a user-defined fashion on a DNA origami carrier is the prerequisite for building more complex multivalent configurations. We tested and optimized methods to meet these requirements in auxiliary experiments (Supplementary Notes 1 and 2 and Supplementary Figs. 1–3), including demonstrating multivalent cell binding and testing for cell internalization (Supplementary Note 3 and Supplementary Figs. 4 and 5). Building on these optimized methods, we created a tetravalent antibody carrier featuring four distinct antibody attachment sites (Fig. 1a). For each attachment site on the DNA origami chassis, we created a library of DNA-tagged antibodies carrying a sequence-complementary single-stranded DNA tag. To induce the activation of effector T cells, we chose anti-CD3, anti-CD28 and anti-CD137 antibodies (Fig. 1a, left). To mediate binding to the target cells, we chose antibodies against the known antigens CD19, CLL-1, CD22 and CD123. We prepared antibody–DNA conjugates and then assembled 105 unique antibody combinations. We validated the assembly of these PTEs via gel electrophoretic mobility analysis (Fig. 1b) and quantified the yield of fully assembled tetravalent combinations for each variant. The assembly yield varied between from 85% (variant 2× anti-CD123 2× anti-CD3) to 97% (variant 2× anti-CD19 2× anti-CD3).

To study PTE-induced T-cell activation, we used a nuclear factor of activated T cells (NFAT)-luciferase reporter assay³⁵. We co-cultured CD19⁺, CD22⁺, CD123⁻ and CLL-1^– NALM-6 ALL cells with NFAT-luciferase-transduced Jurkat cells in the presence or absence of PTEs (Fig. 1c,d). Bivalent aCD3 IgG antibodies can crosslink T-cell receptors causing T-cell activation in the absence of target cells. We, therefore, subtracted the background signal generated by the PTEs that carried only T-cell antibodies (Supplementary Fig. 6). Here aCD19 × aCD3 constructs lead to the strongest T-cell activation, whereas aCD22 × aCD3 constructs induced weak T-cell activation (Fig. 1c). These results are in line with the differing densities of CD19 and CD22 molecules on NALM-6 cells³⁶. Also, aCLL-1, aCD123 × aCD3 constructs did not cause target-cell-induced T-cell activation, proving the specificity of the PTEs. The inclusion of additional T-cell-activating antibodies (aCD28, aCD137) resulted in the increased activation of T cells compared with the single aCD19 × aCD3 variant, with the aCD19 × aCD3-aCD28 PTEs showing the strongest T-cell activation (Fig. 1c). Again, these results are compliant with the literature³⁷.

Next, we analysed the dual-tumour-targeting constructs for their ability to activate T cells (Fig. 1d). We observed increased activation signals for two-target variants with a single activation of T cells via CD3 compared with single-target variants (Fig. 1d). The addition of a co-stimulatory domain to the dual-targeting variants further enhanced the signal. We observed the strongest activation for variants consisting of aCD3, aCD28 and two aCD19 antibodies or one aCD19 antibody and one aCD22 antibody (Fig. 1e). The aCLL-1 control constructs did not induce the activation of T cells.

In vitro characterization of F(ab)-based PTEs

Protein-based bispecific antibodies can mediate the potent T-cell-directed lysis of tumour cells³⁸. Our approach should allow us to assemble bispecific variants with similar capabilities. However, the size of the antibody and the arrangement of the paratopes are essential factors that can influence the efficacy of target cell killing³⁹ (Supplementary Fig. 7). In this context, full-sized IgGs present limitations, such as Fc-domain-mediated binding to immune cells, which can cause undesired cell interactions⁴⁰, and crosslinking receptors through their two paratopes, leading to non-specific target cell lysis (Supplementary Fig. 8). With these aspects in mind, we designed a smaller antibody carrier chassis (20.0 × 15.0 × 7.5 nm³) that can display an anti-CD3 F(ab) fragment for T-cell binding and up to four F(ab) fragments to recognize the target cell antigens (Fig. 2a,b and Supplementary Fig. 9). We placed the F(ab) fragments at the corners of the DNA origami chassis to realize a paratope-to-paratope distance of approximately 7.5 nm, which is comparable with reported cell–cell distances in immunological synapses⁴¹. The antibody attachment concepts established with the larger chassis were directly transferable to the small chassis and enabled fabricating multispecific variants with high yields (>98%), as seen by agarose gel electrophoresis and transmission electron microscopy (TEM) (Fig. 2b,c and Supplementary Fig. 10). To orient the F(ab) fragments on the DNA origami chassis, we site-specifically coupled the adapter DNA strands to one of the thiol groups of its disulfide bond, located opposite the paratope (Supplementary Fig. 11)…