Results

Part I: Design of active monoclonal antibodies targeting specific receptor epitopes

Phase 1: De novo antibody generation based on the localization of antigenic epitope hotspot residues

In the initial stage, our goal was to design a bispecific fusion protein-drug conjugate targeting CD47 and HER2, which are upregulated in tumor cells, especially in cases of gemcitabine and trastuzumab resistance. We aimed to design a construct composed of two nanobodies—one for CD47 and one for HER2. Therefore, we first needed to design and screen active antibody fragments for HER2 and CD47 respectively (as shown in Figure 1).

Using AI tools (RFdiffusion, RoseTTAFold2, ProteinMPNN), we designed 10 active antibody fragments targeting HER2 and CD47. For HER2, we drew on the binding modes of trastuzumab and pertuzumab, selecting different hotspot residue regions to guide the design. For CD47, the design was based on the binding site of Lemzoparlimab. By specifying hotspot residues, we successfully guided the designed antibody fragments to the intended antigenic epitopes (Figure 2).

Phase 2: Screening of active fragments based on protein-protein docking and molecular dynamics simulations

To select the optimal antibody fragments, we integrated AI design with existing databases. We utilized Alphafold3 to predict structures, followed by an initial screening via protein docking. The top 10% of candidate fragments, ranked by Glide Score, were selected. A secondary screening was then conducted using molecular dynamics (MD) simulations and MM-GBSA calculations. For HER2, because the two candidate fragments (HER2_001, HER2_002) had similar scores, we performed a 500 ns MD simulation. Free energy landscape analysis confirmed that HER2_002 had more stable binding (Figure 4), as it exhibited a narrower and deeper free energy landscape valley, indicating that its low-energy conformations were more clustered and stable. For CD47, the optimal fragment, CD47_001, was directly selected via MM-GBSA. The finally selected HER2_002 and CD47_001 both demonstrated stable binding with their targets in computational simulations (Figure 3) and can be used for the subsequent design of the bispecific antibody.

Part II: Design of the Bispecific Antibody Fusion Protein

Phase 1: BioBrick Design

This section aims to design a novel bispecific antibody, NanosphinX, using the active antibody fragments screened in the first step (structures shown in Figure 5) to simultaneously bind both CD47 and HER2. This design is expected to exhibit a dual receptor blockade effect on CD47 and HER2 in cancer cells, thereby synergistically eliminating cancer cells and addressing the problem of trastuzumab and gemcitabine resistance in tumors. We attempted to connect the CD47-targeting active fragment to the light chain of the HER2-targeting antibody fragment via a flexible linker (Figure 6). The light chain is located on the outer side of the antibody's "Y" shape. Placing the CD47-binding fragment here allows it to extend more freely, reducing the potential for steric hindrance with the core antibody structure (especially the Fc region), thus facilitating easier binding to the CD47 target on the cell membrane. After designing the BioBrick, whole gene synthesis was performed.

Phase 2: Plasmid Construction and Verification

After whole gene synthesis of the light & heavy chains, they were constructed into the pCDNA3.4 plasmid using SphI and BamHI, followed by plasmid amplification, purification, and positive clone selection. The light chain (HER2nb light chain + CD47nb) and the heavy chain (HER2nb heavy chain) were synthesized as whole genes and constructed into the intermediate plasmids pUC-HER2nb light chain + CD47nb and pUC-HER2nb heavy chain, respectively. (This step was carried out by General Biosystems' whole gene synthesis service). The pCDNA3.4 target empty vector, as well as the pUC-HER2nb light chain + CD47nb and pUC-HER2nb heavy chain vectors, were digested with SphI and BamHI restriction enzymes to generate compatible ends. The restriction digestion products were then analyzed by DNA agarose gel electrophoresis. The results showed that the plasmids with the genes of interest should display two bands: one large, linearized empty plasmid backbone band, and another smaller band for the gene of interest (HER2nb-LC+CD47nb and HER2nb-HC). This indicates that the double digestion was successful and the corresponding plasmids are likely correct. The empty pCDNA3.4 vector should appear as a single, clear linear band migrating slightly slower than the supercoiled plasmid form.

Purification of restriction enzyme digestion products by DNA gel extraction: After separating DNA fragments by size using agarose gel electrophoresis, the gel slice containing the target DNA band is excised under UV light. The gel is then dissolved using heat or a high-salt buffer. A silica-membrane spin column is used to specifically adsorb the DNA under high-salt conditions. After washing away impurities, the pure DNA is eluted in a low-salt buffer.

Ligation of light/heavy chain genes and pCDNA3.4: T4 DNA ligase is used to catalyze the formation of phosphodiester bonds between the 5'-phosphate and 3'-hydroxyl ends of the double-stranded DNA. Since both the vector and the inserts were prepared by double digestion with SphI and BamHI, they possess complementary sticky ends, allowing the ligase to efficiently "stitch" them together to reform a complete, circular plasmid molecule.

Transformation of ligation products into TOP10 competent cells: The transformed bacterial culture was plated on LB agar plates containing ampicillin (as pCDNA3.4 carries the AmpR resistance gene) and incubated inverted at 37°C for 12-16 hours. The observation of multiple bacterial colonies on the LB agar plates indicated the presence of ampicillin-resistant transformants, confirming the successful transformation of TOP10 cells with the target plasmids (Figures 10-11). On plates containing antibiotics, only bacteria that have successfully taken up the recombinant plasmid can grow to form monoclonal colonies. Theoretically, each colony is a population of cells derived from a single ancestral cell and therefore has an identical genetic background. By picking and amplifying a single colony, we can obtain a pure bacterial culture containing the plasmid of interest for subsequent plasmid extraction, verification, and storage.

Phase 3: Sequencing Verification

To verify the correctness of the plasmids we designed and synthesized, we sent the positive clones to a sequencing company to obtain their sequence information. The sequencing results confirmed that our designed plasmids contained the correct sequences (Figures 12-13).

Phase 4: Plasmid Extraction

After amplifying the positive clones, plasmids were extracted using a TIANGEN Plasmid Maxi Kit to obtain high-purity and correct plasmids.

Lane 1: The undigested, original plasmid was loaded. It appeared on the gel as a primary band, which is typically the supercoiled conformation of the plasmid that migrates faster than its linear form.

Lane 2: The plasmid digested with two restriction endonucleases, XbaI and BamHI, was loaded. After digestion, the circular plasmid was cut, producing two linear DNA fragments of different sizes: a larger band, which is the vector backbone, and a smaller band, which is the excised gene fragment of interest. The results from the restriction enzyme digestion gel electrophoresis of our extracted plasmids confirmed that the gene for our fusion protein was successfully constructed into the plasmid vector.

Phase 5: Protein Expression and Purification

The purified plasmids were transiently transfected into CHO cells for protein expression, followed by purification and validation using SDS-PAGE and SEC-HPLC. We found that our synthesized antibody not only met the required molecular weight but also achieved a purity of 98.74%.

Part III: Activity Testing of the Bispecific Antibody Fusion Protein

Phase 1: Effect on trastuzumab-resistant tumor cells (HCC1954 cells)

We investigated the inhibitory effect of the bispecific antibody on trastuzumab-resistant tumor cells by testing the cell viability under different treatments. As shown in the results, HCC1954 cells exhibit a degree of resistance to trastuzumab. However, upon the addition of our designed bispecific antibody, NanosphinX, the tumor cell viability was significantly reduced. Therefore, our designed active bispecific antibody, NanosphinX, can inhibit trastuzumab-resistant tumors, thus addressing the issue of trastuzumab resistance.

Phase 2: Effect of NanosphinX on PANC-1

We explored the inhibitory effect of the bispecific antibody on gemcitabine-resistant tumor cells by testing cell viability under different treatments. As shown, PANC-1 cells are resistant to gemcitabine. However, when our designed bispecific antibody NanosphinX was added, tumor cell viability significantly decreased. This indicates that NanosphinX can inhibit gemcitabine-resistant tumor cells, offering a solution to gemcitabine resistance. Furthermore, the results from the Gemcitabine + NanosphinX treatment show that NanosphinX can be used in combination with gemcitabine to enhance its efficacy and overcome resistance.

Phase 3: Effect of NanosphinX on MDA-MB-231

Triple-negative breast cancer has long been a major challenge regarding drug resistance, as its surface lacks many biological targets, making it difficult to treat with targeted therapies. We further investigated whether our bispecific antibody could inhibit triple-negative breast cancer cells through its ability to alter the tumor immune microenvironment. As shown in Figure 18, the inhibition of MDA-MB-231 cells by NanosphinX was significantly stronger than that of trastuzumab. Additionally, Figure 19 shows that the number of MDA-MB-231 cells was markedly reduced after the addition of NanosphinX, indicating significant growth inhibition.

Phase 4: Changes in Chemokine CXCL9 and CXCL10 Secretion after NanosphinX Addition

Having understood the superiority of NanosphinX in addressing drug resistance, we wanted to further investigate whether NanosphinX could combat cancer not only by directly attacking tumor cells but also by mobilizing the patient's own immune system. Through literature research, we found that antibodies can potentially increase the secretion of chemokines CXCL9 and CXCL10 by macrophages, thereby modulating the tumor microenvironment to strongly attract immune killer cells like T cells and NK (Natural Killer) cells into the tumor tissue, achieving an anti-tumor effect. Therefore, we designed a co-culture experiment with NanosphinX, tumor cells, and macrophages. The results (Figure 20) showed that NanosphinX can effectively induce an increase in the secretion of chemokines CXCL9 and CXCL10 by macrophages in the tumor microenvironment, thereby enhancing our tumor-killing effect.

Conclusion

In summary, the bispecific antibody fusion protein we designed, NanosphinX, can effectively address the problem of drug resistance in cancer therapy by targeting two overexpressed molecular targets on the surface of resistant tumor cells. Furthermore, it can further inhibit tumor cells by modulating the tumor microenvironment. The experimental results clearly demonstrate that NanosphinX has a potent inhibitory effect on both trastuzumab-resistant and gemcitabine-resistant cells, providing a new approach for overcoming resistance to these commonly used drugs and for combination therapy. Designing rational bispecific antibody fusion proteins that target overexpressed molecules on the surface of resistant cells appears to be a promising breakthrough in solving the problem of tumor drug resistance. We will continue to validate our designed bispecific antibody, NanosphinX, in multiple animal studies.