Engineering

Overview

---

Our project began with bioinformatic analysis and flow cytometry experiments, which revealed that breast and pancreatic cancers—particularly in cases resistant to trastuzumab and gemcitabine—frequently exhibit upregulation of both CD47 and HER2 receptors. To address this challenge, we designed a bispecific fusion protein incorporating an active antibody fragment targeting HER2 and a nanobody targeting CD47. This design aims to simultaneously inhibit both receptors on resistant tumor cells, thereby inducing cell death and overcoming a major obstacle in cancer treatment: therapy resistance. In the initial design phase, we employed AI-based de novo design to generate active fragments targeting the functional epitopes of these two receptors. These designed fragments, along with those retrieved from existing databases, formed a library of candidate antibody fragments. The structures of these candidates were predicted using AlphaFold 3, followed by protein-protein docking to perform preliminary screening and evaluation. The top 10% of fragments—based on docking scores—underwent further refinement through molecular dynamics simulations and MM-PBSA binding free energy calculations. The most stable fragments with the lowest binding free energy were selected for the assembly of the active bispecific antibody, NanosphinX. Two assembly strategies were tested to link the two selected active antibody fragments into the final bispecific format. Cell viability assays demonstrated that NanosphinX exhibits cytotoxic effects against wild-type breast cancer cells and maintains strong activity even in trastuzumab-resistant cells, effectively inhibiting tumor growth. It also enhanced the cytotoxicity of gemcitabine in gemcitabine-resistant cell lines, providing a rational basis for combination therapy with gemcitabine. Further detailed studies are currently underway to fully explore the therapeutic potential of this bispecific nanobody in cancer treatment.

Design of Active Monoclonal Antibodies Targeting Specific Receptor Epitopes

---

Phase 1: De Novo Generation of Antibodies Based on Hotspot Residue Localization within Antigenic Epitopes

We utilized AI tools (RFdiffusion, RoseTTAFold2, and ProteinMPNN) to design 10 active antibody fragments targeting HER2 and CD47. For HER2, we adopted distinct strategies inspired by the binding modes of trastuzumab and pertuzumab, selecting different hotspot residue regions to guide the design process. For CD47, the design was based on the binding site of Lemzoparlimab (TJC4). By specifying these hotspot residues, we successfully directed the designed antibody fragments to bind to the intended antigenic epitopes (Fig. 2). (Refer to the Dry Lab section for detailed design methodology.)

Phase 2: Screening of Active Antibody Fragments via Protein Docking and Molecular Dynamics Simulations

To identify the optimal antibody fragments, we integrated AI-generated designs with existing antibody databases. The structures of candidate fragments were predicted using AlphaFold3, followed by initial screening through protein-protein docking. Candidates ranking in the top 10% by Glide Score were selected for further evaluation.

These candidates then underwent molecular dynamics (MD) simulations and MM-GBSA binding free energy calculations for refined screening. For HER2, since two candidates (HER2_001 and HER2_002) received similar docking scores, we conducted 500 ns MD simulations and free energy landscape analysis. The results confirmed that HER2_002 exhibited superior binding stability (Fig. 4). For CD47, the optimal candidate CD47_001 was selected directly based on MM-GBSA scores.

The finally selected fragments, HER2_002 and CD47_001, both demonstrated stable binding to their respective targets throughout computational simulations (Fig. 3) and were thus chosen for subsequent bispecific antibody development.

Design of Bispecific Antibody Fusion Protein

Phase 1: Initial BioBrick Design of the Bispecific Antibody Fusion Protein

This phase aims to design a novel bispecific antibody, NanosphinX, using the active antibody fragments selected from the initial screening (structures shown in Fig. 5). The objective is to construct a bispecific antibody capable of simultaneously binding both CD47 and HER2. In our first construction strategy, we attempted to connect the CD47-targeting active fragment to the heavy chain of the HER2-targeting antibody fragment via a flexible linker (Fig. 6).

This design was anticipated to exhibit dual-receptor blockade against both CD47 and HER2 on cancer cells, thereby synergistically eliminating cancer cells and addressing resistance to trastuzumab and gemcitabine. However, this initial BioBrick design—though demonstrating strong binding affinity toward HER2—was found to yield a bispecific antibody with insufficient binding affinity for CD47. As shown in Fig. 7 to Fig. 9, NanosphinX_01 exhibited very low dual affinity for both CD47 and HER2. Further investigation using flow cytometry confirmed that this lack of affinity was primarily due to reduced binding to CD47. Consequently, additional modifications were undertaken to improve the design.

Phase 2: BioBrick Optimization of the Bispecific Antibody Fusion Protein

To address the insufficient affinity for CD47 observed in the initial BioBrick design from Phase 1, we hypothesized that connecting the CD47-targeting active fragment to the heavy chain of the HER2-targeting antibody via a flexible linker might have caused steric hindrance between the CD47-binding module and the Fc region of the HER2 antibody. This constraint could have limited the freedom of movement and structural flexibility required for efficient CD47 binding. Therefore, we modified the design by instead connecting the CD47-targeting active fragment to the light chain of the HER2-targeting antibody using a flexible linker (Fig. 10). The light chain is located on the outer side of the characteristic Y-shaped antibody structure. Positioning the CD47-binding fragment here allows it greater freedom to extend outward, minimizing potential steric clashes with the core antibody structure—particularly the Fc region—and thereby facilitating more effective binding to the CD47 target on the cell membrane.

Accordingly, we designed NanosphinX_02 by connecting the CD47-targeting active fragment to the light chain of the HER2-targeting antibody. Molecular experiments confirmed that NanosphinX_02 exhibits significantly enhanced dual affinity for both HER2 and CD47 (Figs. 7–9).

Consequently, we selected the NanosphinX_02 design strategy, in which the CD47-targeting fragment was fused to the N-terminus of the light chain of the HER2-targeting antibody. The construct was then expressed using a CRO (Contract Research Organization): a signal peptide was added, the plasmid was constructed, and transient transfection was performed followed by expression and purification to obtain the NanosphinX_02 protein in sufficient quantity.

After obtaining the specific active bispecific antibody, we evaluated its efficacy against trastuzumab-resistant tumor cells (HCC1954 cell line), gemcitabine-resistant tumor cells (PANC-1 cell line), and triple-negative breast cancer cells (MDA-MB-231 cell line). Detailed experimental results can be found in the Experimental Results section.