A computer-aided design method for affinity maturation of nanobodies

Abstract

The present application relates to the field of biotechnology, in particular to a computer-aided design nanobody affinity improvement method. Through a series of bioinformatics tools for nanobody affinity maturation obtained by literature research, the operability and improvement effect of the tools are evaluated, and some database, server and software combination are selected to construct a set of computer-aided design (Computer Aided Design, CAD) nanobody affinity improvement method. In order to verify the effectiveness and universality of the method, based on the method, Anti-Nectin-4 camel nanobody NBNT-1 and Anti-PD-L1 shark nanobody NBP4 screened from the synthetic library are screened through model construction, model evaluation, site prediction, molecular docking, structure analysis, mutation prediction and mutation evaluation to obtain high-affinity mutant nanobodies. Traditional in vitro affinity maturation methods, such as error-prone PCR, have low screening efficiency, large randomness of mutation introduction, long experimental period, large workload and difficulty in accurately optimizing the affinity of target molecules. The present application overcomes many disadvantages of traditional in vitro affinity maturation methods and improves the success rate and efficiency of affinity maturation.

Description

Technical Field

[0001] This invention relates to the field of biotechnology, specifically to a method for enhancing the affinity of computer-aided design nanoantibodies, and more particularly to a strategy for enhancing the affinity of computer-aided design (CAD) nanoantibodies. Background Technology

[0002] The discovery of nanobodies dates back to 1989, when Belgian scholars Hamers et al. discovered naturally occurring heavy-chain antibodies (HCAs) lacking the light chain in the immune systems of camel-dwelling animals (such as camels and alpacas). In 1995, American scholars Greenberg et al. discovered immunoglobulin new antigen receptors (IgNARs) in cartilaginous fish (such as nurse sharks and bamboo sharks) with structures similar to camel-derived heavy-chain antibodies. The variable region (VHH) of the heavy-chain antibody derived from camels and the variable region (VNAR) of the heavy-chain antibody derived from cartilaginous fish have a molecular weight of approximately 12-15 kDa, only 1 / 10 that of traditional monoclonal antibodies, with a diameter of approximately 2.5 nm and a height of approximately 4 nm. Therefore, they are called nanobodies (Nbs) and are the smallest antigen-binding domains discovered to date. Summary of the Invention

[0003] The first objective of this invention is to provide a convenient, effective, and somewhat universal computer-aided method for enhancing the affinity of nanobodies.

[0004] The second objective of this invention is to provide a mutant nanobody with enhanced affinity designed by computer, specifically mutant nanobodies NBNT-M1, NBNT-M2, NBP-M1 and NBP-M2 based on NBNT-1 and NBP4.

[0005] This study utilizes computer-aided design (CAD) to construct a strategy for enhancing the affinity of nanobodies. By optimizing and verifying databases, servers, and software reported in the literature, a CAD-based method for enhancing nanobodies affinity is constructed. Furthermore, multiple algorithms and servers are employed to increase the accuracy of virtual screening. The process includes the following steps: The framework region and complementarity-determining region of the nanobody were determined using the Antibody Region-Specific Alignment (AbRSA) antibody numbering and CDR partitioning tool server. Two methods, homology modeling and ab initio computational modeling, were employed to construct the nanobody model. Using the Basic Local Alignment Search Tool (BLAST), structures with high homology to the antibody sequence were selected as templates. The original antibody amino acid sequence was input, and searches were performed in the Protein Data Bank (pdb) and blastp (protein-protein BLAST) databases. Six templates with homology greater than 70% were selected for homology modeling. Structure prediction and model construction were performed using the AlphaFold2 ab initio computational modeling server. Candidate models from both homology modeling and ab initio computational modeling were uploaded to the Structure Validation Server (SAVES v6.0) model evaluation server, and the stereochemical, geometric, and physicochemical plausibility of the models were evaluated using ERRAT, VERIFY 3D, and PROCHECK algorithms.

[0006] Possible antigen-antibody binding sites were predicted using the Protein Protein Interaction Server (InterProSurf) and the Protein Interaction Calculator (PIC) docking site prediction servers. In the GRAMM docking rigid docking server, the antigen and antibody PDB files with minimized energy were uploaded without specifying a site. The structure file of the antigen-antibody docking complex with the highest score was downloaded. Antibody amino acids within 5 Å of the antigen were analyzed in PyMOL as candidate sites. Sites were selected for flexible docking at specified sites based on the output results from InterProSurf and PIC. Molecular docking was performed using the HighAmbiguity Driven protein-protein DOCKing (HADDOCK) molecular docking server.

[0007] After HADDOCK molecular docking, 10 antigen-antibody complex structures and corresponding parameter scores were output. The best antigen-antibody complex model was selected for further analysis based on the Z-score. The selected best antigen-antibody complex model was opened in PyMOL, and the antibody amino acid sites within 5 Å of the antigen molecule were counted. Combined with the prediction results of InterProSurf and PIC server, the amino acid sites located in the CDR region that do not contain cysteine ​​(Cys) and proline (Pro) were selected as mutation sites.

[0008] Single-point saturation mutagenesis was performed on the identified mutation sites, mutating the specified amino acid at each site to one of the remaining 17 amino acids excluding cysteine, proline, and proline itself. The changes in antigen-antibody binding free energy and relative affinity before and after mutation were assessed using the Antibody-Antigen Affinity Changes Upon Mutation (mCSM-AB) and Protein-Protein Affinity Predictor (PPA_Pred) online servers. Sites where mutation to any other amino acid resulted in an increased affinity score were selected as affinity-sensitive sites. Multiple-point mutagenesis and evaluation were then performed, ultimately yielding a series of high-scoring triple mutants. Two high-scoring mutants were selected for molecular experiments to verify the affinity enhancement effect.

[0009] Preferably, the target nanobody is Anti-Nectin-4 camel-derived nanobody NBNT-1 and Anti-PD-L1 shark-derived nanobody NBP4, and their corresponding mutant nanobodies are NBNT-M1, NBNT-M2 and NBP-M1, NBP-M2, respectively.

[0010] To achieve the first objective of this invention, this invention constructs a method using a combination of bioinformatics tools, such as databases, servers, and software, that have been validated in the literature, and includes the following steps: 1. The framework region and complementarity-determining region of the nanobody were determined using the Antibody Region-Specific Alignment (AbRSA) antibody numbering and CDR partitioning tool server. Sequence alignment was performed using the Basic Local Alignment Search Tool (BLAST) homology sequence retrieval tool to analyze the differences. Homology modeling was performed using the SWISS-MODEL homology modeling server, and structure prediction and model construction were performed using the AlphaFold2 ab initio computation modeling server.

[0011] 2. Perform a comprehensive evaluation of the candidate models in the Structure Validation Server (SAVES v6.0) model evaluation server.

[0012] 3. Output possible antigen-antibody binding sites in the Protein Protein Interaction Server (InterProSurf) and Protein Interaction Calculator (PIC) docking site prediction server.

[0013] 4. Download the highest-scoring antigen-antibody docking complex structure file from the GRAMM docking rigid docking server. Analyze the antibody amino acids within 5 Å of the antigen in PyMOL as candidate sites. Combine the output results of InterProSurf and PIC to select sites for flexible docking at designated sites. Determine possible antigen-antibody binding sites by combining different algorithms and servers. Input the determined sites into the High Ambiguity Driven protein-protein DOCKing (HADDOCK) molecular docking server for further analysis based on the Z-score to select the best antigen-antibody complex model.

[0014] 5. The changes in the relative relationship between antigen-antibody binding free energy and affinity before and after mutation were assessed using the online servers Antibody-Antigen Affinity Changes Upon Mutation (mCSM-AB) and Protein-Protein Affinity Predictor (PPA_Pred).

[0015] To achieve the second objective of this invention, the present invention provides mutation sites for NBNT-1 and NBP4. The three mutation sites for NBNT-1 are Tyr59, Tyr103, and Met105, and the three mutation sites for NBP4 are Ala27, Leu94, and Leu97. The three mutants of NBNT-1 are NBNT-M1 (Y59W&Y103G&M105G) and NBNT-M2 (Y59R&Y103N&M105G), and the three mutants of the nanobody NBP4 are NBP-M1 (A27Y&L94F&L97F) and NBP-M2 (A27W&L94F&L97E). Attached Figure Description

[0016] Figure 1 The original nanobody CDR region is defined, where A: NBNT-1 structure definition, and B: NBP4 structure definition; Figure 2 The images show the homologous sequence alignment of the original nanobody, where A represents the homologous sequence alignment of NBNT-1 and B represents the homologous sequence alignment of NBP4. Figure 3 Normalized QMEAN4 and PLST scores were used to model the homology of NBNT-1, where A: 8EMZ_B homology template, B: 5IMK_B homology template, C: 7USV_C homology template, D: 7DST_E homology template, E: 5LHR_B homology template, and F: 6N4Y_E homology template. Figure 4Scoring of the NBNT-1 ab initio modeling model, where A: Sequence coverage score map, B: PLDDT score map, and C: PAE score map; Figure 5 Normalized QMEAN4 and PLST scores were used to model the homology of NBP4, where A: 1SQ2_N homology template, B: 7S83_B homology template, C: 2I26_N homology template, D: 3MOQ_A homology template, E: 6X4G_N homology template, and F: 7S83_A homology template. Figure 6 The NBP4 ab initio modeling model is scored, where A: Sequence coverage score map, B: PLDDT score map, and C: PAE score map; Figure 7 Here are the La-format conformation diagrams of the main residues in the structural model, where A: La-format conformation diagram of NBNT-1, and B: La-format conformation diagram of NBP4. Figure 8 For the prediction of docking sites in the three-dimensional model of antigen and antibody, A: prediction of docking sites between NBNT-1 and Nectin-4, and B: prediction of docking sites between NBP4 and PD-L1. Figure 9 To match the Z-score scores of the docking models and the prediction of mutation sites, where A: NBNT-1 and Nectin-4 model scores, B: NBNT-1 mutation site prediction, C: NBP4 and PD-L1 model scores, and D: NBP4 mutation site prediction. Figure 10 For the determination of three mutation sites, A: determination of the NBNT-1 three mutation site, B: determination of the NBP4 three mutation site; Figure 11 For the analysis of the three-point mutation structure of nanobodies, where A: analysis of the three-point mutation site of NBNT-1, and B: analysis of the three-point mutation site of NBP4; Figure 12 The results of PCR identification of Anti-PD-L1 nanobody positive clone bacterial culture; Figure 13 Purification of PD-L1 extracellular domain antigen; Figure 14 The purification results of Anti-Nectin-4 nanobody are shown in the figures: A: NBNT-1 purification result, B: NBNT-M1 purification result, and C: NBNT-M2 purification result. Figure 15 The purification results for Anti-PD-L1 nanobody are shown below. A: NBP4 purification results, B: NBP-M1 purification results, and C: NBP-M2 purification results. Figure 16 The affinity assays were performed for each nanobody, with A representing the affinity assay for Anti-Nectin-4 nanobody and B representing the affinity assay for Anti-PD-L1 nanobody. Detailed Implementation

[0017] The present invention will be further illustrated below with reference to specific embodiments. Unless otherwise specified, the experimental methods used in the following embodiments are conventional methods. Unless otherwise specified, the materials and reagents used in the following embodiments are commercially available. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention.

[0018] Example 1: Structural modeling and molecular docking of Anti-Nectin-4 with Anti-PD-L1 original nanobodies NBNT-1 and NBP4 Anti-Nectin-4 camel-derived nanobody NBNT-1 and Anti-PD-L1 shark-derived nanobody NBP4 were screened from synthetic libraries. Since different lengths of amino acid randomization strategies were used for the CDR regions in the libraries, the complete amino acid sequences of NBNT-1 and NBP4 were uploaded to the Antibody Region-Specific Alignment (AbRSA) antibody numbering and CDR partitioning tool server. The Kabat mode was selected for sequence submission to determine the frame region and complementarity-determining region of the antibody. To improve the accuracy of model construction, this study used both homology modeling and de novo computational modeling methods. The results are as follows: Figure 1 As shown, CDR1, CDR2, and CDR3 regions are represented in red, yellow, and green, respectively, and some amino acid sequences are kept confidential.

[0019] For homology modeling, structures with known crystal structures and high homology to the antibody sequence were selected from the database as templates. The original antibody amino acid sequence was entered into the ProteinBLAST mode using the Basic Local Alignment Search Tool (BLAST) (HA, 6×His, etc. tags were removed), and searches were performed using the Protein Data Bank (pdb) and blastp (protein-protein BLAST) databases. The output results were sorted in descending order of sequence similarity, and six templates with homology greater than 70% were selected as candidates and their FASTA format files were downloaded. The original sequence and homology sequence files were opened with UniproUGENE for sequence alignment, and their differences were analyzed. The sequence alignment results are shown below. Figure 2 As shown, the frame regions of the target sequence and the template sequence are basically the same, with the main difference concentrated in the CDR3 region.

[0020] The original antibody sequence was input into the SWISS-MODEL homology modeling server, and six templates overlapping with those in the BLAST server were selected for homology modeling. The output file was saved for subsequent model evaluation. The results are as follows: Figure 3 As shown, the Normalized QMEAN4 scores of all NBNT-1 models are between 0.5 and 1, meeting the scoring requirements. The local differences between the amino acid residues and the template in each model are mainly reflected in the CDR region, especially the CDR3 region, which is consistent with theoretical expectations.

[0021] The original antibody sequence was input into the AlphaFold2 ab initio computational modeling server. Parameters were set to 3 loops and 5 output models for structure prediction and model construction. The output file was saved for subsequent model evaluation. The Sequencecovery plot shows whether different structural regions are predicted based on templates or ab initio computation, which is important for assessing model reliability. The PLDDT plot shows the confidence level of the predicted structure of each local residue. Generally, a score above 70 indicates that the prediction for that part of the structure is relatively reliable, but there may be local uncertainties. A score below 70 indicates that the structure of that region is unstable or highly flexible. The PAE plot shows the confidence level of the relative positional relationships between amino acid residues in the model, which can help assess the prediction accuracy of the interaction interfaces between different protein chains. Figure 4 As shown, the NBNT-1 model's prediction scores for most regions are above 70, with the main local uncertainties concentrated in the CDR3 region, which is not based on template predictions. This result is consistent with theoretical expectations. Figure 5 As shown, the Normalized QMEAN4 scores of all NBP4 models are between 0.5 and 1, meeting the scoring requirements. The local differences between the amino acid residues and the template in each model are mainly reflected in the CDR region, especially the CDR3 region, a result consistent with theoretical expectations. Figure 6 As shown, the prediction scores of most parts of the NBP4 model are above 70, and the main local uncertainty is concentrated in the CDR3 region which is not based on template prediction. This result is in line with theoretical expectations.

[0022] Candidate models constructed using both homology modeling and ab initio computational modeling algorithms were uploaded to the StructureValidation Server (SAVES v6.0) model evaluation server. The ERRAT, VERIFY 3D, and PROCHECK algorithms were used to comprehensively evaluate the stereochemical, geometric, and physicochemical plausibility of each model. The ERRAT algorithm was used to calculate the number of non-bonded interactions between different types of atomic pairs; the VERIFY 3D algorithm was used to assess whether the protein's three-dimensional structure was compatible with the amino acid sequence of the primary structure; and the PROCHECK algorithm was used to check the stereochemical quality of the protein structure, judging the plausibility of the structure by analyzing the geometric parameters between residues (such as bond lengths, bond angles, and dihedral angles). The server compared the calculation results with a database of known reliable structures to obtain a score; a higher score indicated stronger model reliability. The ERRAT and VERIFY 3D scores for each model are shown in Table 1. All models achieved ERRAT scores above 75 and VERIFY 3D scores mostly above 70%, indicating a high overall quality of model construction. Based on the scoring results and sequence homology, the homology modeling models 6N4Y_E and 6X4G_N were finally selected as the structural models for the nanobodies NBNT-1 and NBP4 for further evaluation.

[0023] Table 1 SAVES Model Scoring Results

[0024] The optimal model was selected as the original antibody model based on the SAVES v6.0 model score and template sequence similarity. After dehydration and hydrogenation in PyMOL to minimize energy, it was ready for use. The resolved crystal structures of Nectin-4 and PD-L1 extracellular domain antigens were uploaded to the NCBI database. The pdb files were downloaded and depolymerized, dehydrated, hydrogenated, and deligated in PyMOL to minimize energy, then saved. The energy-minimized antigen and antibody pdb files were uploaded to the Protein Protein Interaction Server (InterProSurf) and Protein Interaction Calculator (PIC) docking site prediction servers, outputting possible antigen-antibody binding sites under different algorithms. The NBNT-1 model structure in the La-form conformation diagram contains 125 amino acids, and its distribution in the La-form conformation diagram is as follows: Figure 7 As shown in (A), the proportions of amino acids in the optimal, suboptimal, permissible, and unreasonable regions are 91.3%, 8.7%, 0%, and 0%, respectively. All residues are within the reasonable region, indicating that the model construction is reasonable. The NBP4 model structure in the La conformation diagram contains a total of 108 amino acids, and their distribution in the La conformation diagram is as follows: Figure 7(B) The vast majority of sites are within the reasonable range, indicating that the model is reasonably constructed.

[0025] In the GRAMM docking rigid docking server, the antigen and antibody PDB files with minimized energy were uploaded without specifying a site. The structure file of the antigen-antibody docking complex with the highest score was downloaded. Antibody amino acids within 5 Å of the antigen were analyzed in PyMOL as candidate sites. Sites were selected for flexible docking at specified sites based on the output results from InterProSurf and PIC. Results are as follows: Figure 8 As shown, the binding sites of NBNT-1 and Nectin-4 are mainly concentrated in the CDR2 and CDR3 regions, while the binding sites of NBP4 and PD-L1 are mainly concentrated in the CDR1 and CDR3 regions, which is consistent with theoretical expectations. Combining the InterProSuf and PIC prediction results, the overlapping sites mainly concentrated in the CDR region were used for flexible docking at the designated sites in HADDOCK.

[0026] By combining different algorithms and servers, potential binding sites for antigens and antibodies were determined. These sites were then input into the HighAmbiguity Driven protein-protein DOCKing (HADDOCK) molecular docking server for docking. HADDOCK, by combining rigid, semi-flexible, and flexible docking characteristics, can flexibly adjust the docking strategy at different stages, thereby improving the accuracy and reliability of docking results while ensuring computational efficiency. After HADDOCK molecular docking, it outputs 10 antigen-antibody complex structures and corresponding parameter scores, such as Root Mean Square Deviation (RSMD), van der Waals energy, and Z-score. The Z-score represents the standard deviation between the HADDOCK score of a cluster and the average HADDOCK score of all clusters. A smaller Z-score indicates that the HADDOCK score of that cluster is significantly lower than the average, meaning that the docking model for that cluster is more energy-efficient. Therefore, the optimal antigen-antibody complex model is selected based on the Z-score for further analysis. Its Z-score is shown below. Figure 9 As shown, the docking complex model with the lowest Z-score was selected for structural analysis in PyMOL. It was observed that the CDR2 and CDR3 regions of the NBNT-1 nanobody and the CDR1 and CDR3 regions of the NBP4 nanobody were the main binding regions. Combining InterProSuf and PIC prediction results, cysteine ​​and proline residues were excluded to determine the final mutation sites. Ten mutation sites were identified in the CDR2 and CDR3 regions of NBNT-1: 52S, 55G, 56G, 59Y, 100R, 101R, 103Y, 105M, 106R, and 107W. Figure 2As shown in red in .10(B), NBP4 identified 10 sites in the CDR1 and CDR3 regions: 27A, 28S, 29Y, 91L, 92V, 93Y, 94L, 96R, 97L, and 98F. Figure 9 (D) The blue part is shown.

[0027] In PyMOL, open the selected optimal antigen-antibody complex model and count the antibody amino acid sites within 5 Å of the antigen molecule. Analyze the results using predictions from InterProSurf and the PIC server. The CDR region of the nanobody is the main region involved in antigen-specific binding, with the CDR3 region playing a dominant role. Introducing mutations into these sites is more likely to improve affinity.

[0028] First, single-point saturation mutagenesis was performed on the identified mutation sites. To avoid introducing new disulfide bonds and causing uncontrollable effects, the amino acids at the designated sites were mutated to the remaining 17 amino acids excluding cysteine, proline, and proline itself. The changes in antigen-antigen binding free energy and relative affinity before and after mutation were assessed using online servers such as Antibody-Antigen Affinity Changes Upon Mutation (mCSM-AB) and Protein-Protein Affinity Predictor (PPA_Pred). Based on the server prediction results and the selection of mutation diversity, three mutation sites were ultimately identified, resulting in 15 single-point mutants for subsequent multi-point mutation combinations and evaluation. The results are as follows: Figure 10 As shown, the three mutation sites identified by NBNT-1 are Tyr59, Tyr103, and Met105, while the three mutation sites identified by NBP4 are Ala27, Leu94, and Leu97. The positional relationship of the three selected mutation sites in the docking complex is shown below. Figure 11 As shown in the figure, the gray area at the top represents the antigen structure, and the green area at the bottom represents the antibody structure. The figure shows that all three selected sites are within 5 Å of the antigen molecule, which is consistent with the theoretical expectation of them being affinity-sensitive sites. Subsequent analysis will focus on the permutation and combination of mutation types at these three sites, as well as the evaluation of two-point and three-point mutations.

[0029] Based on the results of the first round of mutations, sites where mutations to any other amino acid would enhance affinity scores were selected as affinity-sensitive sites. While preserving the diversity of point mutations, the impact of multiple point mutations on nanobody specificity was also evaluated. Finally, 15 mutation types from 3 affinity-sensitive sites located in the CDR region were selected for permutation and combination to determine two-point mutation types. One site was fixed, and a second point mutation was performed and evaluated based on the original homologous template. Mutants with significant affinity enhancement effects were selected from the two-point mutations for permutation and combination to determine three-point mutation types. The above evaluation process was repeated, resulting in a series of high-scoring three-mutants. Two high-scoring mutants from each mutant were selected for molecular experiments to verify the affinity enhancement effect. Finally, a series of high-scoring mutants were obtained, and the top two candidate mutants with the highest virtual affinity evaluation and screening scores were selected for subsequent molecular experiments to verify the actual affinity enhancement effect. The three mutants identified for the nanobody NBNT-1 are NBNT-M1 (Y59W&Y103G&M105G) and NBNT-M2 (Y59R&Y103N&M105G). The three mutants identified for the nanobody NBP4 are NBP-M1 (A27Y&L94F&L97F) and NBP-M2 (A27W&L94F&L97E).

[0030] Example 2: Construction of mutant nanobody vectors NBP-M1 and NBP-M2 Reagents and kits: Isopropyl-β-D-thiogalactopyranoside (IPTG) was purchased from Maclean's, ampicillin sodium from Shanghai Sangon Biotech Co., Ltd., and protein molecular weight standards (MW Marker, 14.3-97.2 kDa) from Taraka. Other biochemical reagents were domestically produced, standard analytical grade reagents.

[0031] Strains and plasmids: E. coli BL21 and TransB(DE3) (Shanghai Sangon Biotech and Beijing TransGen Biotech) were used as plasmid expression strains. Recombinant plasmids pET24a(+)-HA-NBNT-1 and ppET22b(+)-NBP4 were constructed in the aforementioned stages of this invention. pET24a(+)-HA-NBNT-M1, pET24a(+)-HA-NBNT-M2, pET22b(+)-NBP-M1, and pET22b(+)-NBP-M2 were synthesized by Qingke Biotechnology.

[0032] Nanobodies expressing an N-terminus HA tag were constructed based on existing pET22b(+)-NBP4, pET22b(+)-NBP-M1, and pET22b(+)-NBP-M2 vectors. Reverse PCR amplification of the pET22b(+) empty vector fragment yielded a pET22b(+) vector fragment containing a homologous arm. Forward PCR amplification of the pET22b(+)-NBP4, pET22b(+)-NBP-M1, and pET22b(+)-NBP-M2 target fragments containing a homologous arm and an HA tag yielded HA-NBP4, HA-NBP-M1, and HA-NBP-M2 target fragments. Homologous recombination technology was used to ligate the target fragments HA-NBP4, HA-NBP-M1, and HA-NBP-M2 to the pET22b(+) vector fragment, constructing three HA-tagged nanobodies. In addition, the C-terminus of the pET22b(+) vector carries a histidine tag (6×His Tag) and a stop codon (TGA) to ensure the complete and correct expression of the nanobody, which can then be obtained by nickel column affinity chromatography.

[0033] Based on the gene sequences of pET22b(+) empty vector, pET22b(+)-NBP4, pET22b(+)-NBP-M1 and pET22b(+)-NBP-M2, upstream primers pET22b(+)-F, HA-NBP4-F, HA-NBP-M1-F, and HA-NBP-M2-F and downstream primers pET22b(+)-R and NBP-R were designed using SnapGene software.

[0034] Construction of pET28a(+) recombinant plasmid and primer design: Forward primer (SEQ ID NO:1): gaattcgagctccgtcgacaagc Reverse primer (SEQ ID NO:2): catggtatatctccttcttaaagt Construction of pET24a(+)-HA-NBNT-M1 recombinant plasmid Forward primer design (SEQ ID NO:3): acaccaaacagtggcgtgacgaaaccaaaggtttccgtgatgaggccaaacgcttcaaaaacaccgctggtcaggtacagctgcaagaatccggt Reverse primer design (SEQ ID NO:4): gcttgtcgacggagctcgaattctcaggagctaacggtcacctgagtacc Construction of pET24a(+)-HA-NBNT-M2 recombinant plasmid Forward primer design (SEQ ID NO:5): actttaagaaggagatataccatgcatcatcatcatcatcacaaaaacgagtccagcaccaacgcaactaacaccaaacagtggcgtgacgaaaccaaaggtttccgtgatgaggccaaacgcttcaaaaacaccgctggtcaggtacagctgcaagaatccggt Reverse primer design (SEQ ID NO:6): cggtacgtcgtatggataggagctaacggtcacctgagtacc PCR reaction system: 1 µL of 10 µM upstream primer, 1 µL of 10 µM downstream primer, template ≤200 ng, 25 µL of 2×High-Fidelity DNA Polymerase, and ddH2O to 50 µL; PCR reaction conditions: 1. Pre-denaturation: 98℃ for 3 min, 2. Denaturation: 98℃ for 10 s, 3. Annealing: 60℃ for 20 s, 35 cycles, 4. Extension: 72℃ for 30 s / kb, 5. Final extension: 72℃ for 5 min. Agarose gel electrophoresis was used to detect the amplification of the target fragment, and the DNA fragment was recovered using an agarose gel recovery kit.

[0035] The target fragment was homologous to the linearized vector for recombinant vector construction. The reaction system consisted of: 5 µL of 2×Baxic Assembly Mix, 0.03 pmol of the linearized vector pET22b(+), 0.06 pmol of the target fragment HA-NBP4 / HA-NBP-M1 / HA-NBP-M2, and ddH2O to a final volume of 10 µL. The reaction temperature and time were 50 °C for 15 min. The ligation product was transformed into competent E. coli DH5α cells. Single bacteria were picked and incubated overnight in LB broth. Colony PCR was performed to verify the presence of the target band. The colony containing the target molecular weight band was then sent for sequencing. Agarose gel electrophoresis of the colony PCR products was performed as follows: Figure 12 As shown, the amplified bands were single and bright, and their positions matched the theoretical size. After sequencing, the results were compared with the theoretical bases using SnapGene software, and the overlap rate was 100%. The pET22b(+)-HA-NBP4, pET22b(+)-HA-NBP-M1, and pET22b(+)-HA-NBP-M2 vectors were successfully constructed.

[0036] Example 3: Expression of nanobody proteins Reagents and kits: Isopropyl-β-D-thiogalactopyranoside (IPTG) was purchased from Maclean's, ampicillin sodium from Shanghai Sangon Biotech Co., Ltd., and protein molecular weight standards (MW Marker, 14.3-97.2 kDa) from Taraka. Other biochemical reagents were domestically produced, standard analytical grade reagents.

[0037] Reagent preparation: 5x protein electrophoresis buffer (Tris-glycine): 15.1 g Tris base, 94 g glycine, 5 g SDS, add double-distilled water to a final volume of 1 L.

[0038] 15% protein electrophoresis separating gel: Measure 2.3 mL of ddH2O, 5.0 mL of 30% acrylamide, 2.5 mL of Tris-HCl (pH 8.8), 100 μL of 10% SDS, 100 μL of 10% APS, and 5 μL of TEMED, and mix thoroughly.

[0039] 5% protein electrophoresis stacking gel: Measure 3.15 mL of ddH2O, 0.75 mL of 30% acrylamide, 0.57 mL of Tris-HCl (pH 6.8), 45 μL of 10% SDS, 45 μL of 10% APS, and 4.5 μL of TEMED, and mix thoroughly.

[0040] Coomassie Brilliant Blue Staining Solution: Weigh 2 g of Coomassie Brilliant Blue (R-250), 200 mL of ethanol, 100 mL of glacial acetic acid, and 750 mL of deionized water. Protein decolorization solution: Measure 200 mL of ethanol, 750 mL of deionized water, and 50 mL of glacial acetic acid, mix well, and store at room temperature. One µL of each of the seven plasmids—pET-24a(+)-PD-L1, pET-24a(+)-HA-NBNT-1 (preserved by the research group), pET-24a(+)-HA-NBNT-M1, pET-24a(+)-HA-NBNT-M2 (synthesized by Beijing Qingke Biotechnology Co., Ltd.), and pET-22b(+)-HA-NBP4, pET-22b(+)-HA-NBP-M1, and pET-22b(+)-HA-NBP-M2 (constructed by homologous recombination)—was transformed into 50 µL of expression-type Escherichia coli competent cells. The pET-24a(+) series plasmids were transformed into BL21 cells, and the pET-22b(+) series vectors were transformed into TransB cells. Various expression strains were inoculated at 1% in LB medium containing 1‰ resistance and cultured overnight at 37°C and 220 rpm. The pET-24a(+) series plasmids were kanamycin resistant, and the pET-22b(+) series plasmids were ampicillin resistant. The cultured bacterial solution was then transferred at a ratio of 1:100 to 250 mL of LB medium containing 1‰ resistance in a shake flask and cultured at 37°C and 200 rpm. When the OD600 reached 0.6 (approximately 2.5 hours), IPTG (final concentration 0.5 mM / mL) was added, and the culture was induced at 37°C for 8 hours. After induction, the cells were centrifuged at 4°C and 8000 rpm for 10 min, the supernatant was discarded, and the bacterial cells were collected.

[0041] Example 4: Purification of Nanobody Proteins Protein purification buffer: 0.2 M PB stock solution: Na₂HPO₄·12H₂O 58.01 g, NaH₂PO₄·H₂O 5.93 g, pH=6.8, dissolved in 1 L of ultrapure water, adjusted to pH 7.4. Inclusion body washing solution: 8 M urea, 0.02 M PB stock solution, 0.5 M NaCl, 30 mM imidazole, pH 6.8 Inclusion body eluent: 8 M urea, 0.02 M PB stock solution, 0.5 M NaCl, 100 mM imidazole, pH 6.8 Inclusion body washing buffer: 8 M urea, 0.02 M PB stock solution, 0.5 M NaCl, 600 mM imidazole, pH 6.8 Dialysis refolding A / B / C / D: 0.5 M NaCl, 0.02 M PB stock solution, 50 mL glycerol, 4 M urea / 2 M urea / 1 M urea / 0 M urea, pH 6.8 Inclusion body washing solution: 0.5 M urea, 0.02 M PB stock solution, 0.5% (v / v) Triton X-100, 0.05 M NaCl, 2.5 mM EDTA, pH 6.8 Inclusion body dissolution solution: 8 M urea, 0.02 M PB stock solution Inclusion body equilibrium solution: 8 M urea, 0.02 M PB stock solution, 0.5 M NaCl, 5 mM imidazole After resuspending the collected bacterial cells in an appropriate amount of PBS, they were lysed in a high-pressure cell disruptor at 6 °C and 700 bar for 5 min. The bacterial suspension became noticeably clear after the disruption. The disrupted bacterial suspension was centrifuged at 4 °C and 8000 r / min for 30 min, and the supernatant and precipitate were collected separately for subsequent SDS-PAGE analysis of the target protein expression. The target protein was found to be mainly present in the precipitate. The precipitate was resuspended in inclusion body washing buffer, stirred at 4 °C for 30 min, and centrifuged at 8000 r / min for 30 min. The supernatant was discarded to obtain purified inclusion bodies. The purified inclusion bodies were resuspended in dissolving buffer, stirred at 4 °C for 4–6 h, and centrifuged at 8000 r / min for 30 min. The supernatant was collected for subsequent purification. Inclusion body purification was performed using nickel column affinity chromatography. The nickel column was equilibrated with equilibration buffer, and after loading the sample, it was washed with washing buffer and Bradford working solution was used for identification. Finally, the target protein was eluted with elution buffer, the nickel column was cleaned with washing buffer, and the column was sealed with 20% ethanol and stored at 4 °C. The washing concentration of PD-L1 extracellular domain antigen was determined to be 30 mM imidazole and the elution concentration to be 100 mM imidazole by Ni-NTA affinity chromatography. Figure 13 As shown, ImageJ grayscale analysis revealed that the target protein, purified from the inclusion body precipitate, had a purity greater than 90%, meeting the requirements for subsequent molecular experiments. The purification conditions for both the original and mutant Anti-Nectin-4 antibodies were consistent: washing with 30 mM imidazole followed by elution with 100 mM imidazole. Figure 14 As shown, ImageJ grayscale analysis revealed that three target proteins with a purity greater than 90% were purified from inclusion body precipitates. The purification conditions for the original and mutant Anti-PD-L1 antibodies were consistent: washing with 50 mM imidazole and eluting with 400 mM imidazole. Figure 15 As shown, ImageJ grayscale analysis revealed that three target proteins with a purity greater than 90% were purified from inclusion body precipitates, meeting the requirements for subsequent molecular experiments.

[0042] During inclusion body refolding, the protein concentration was adjusted to below 0.3 mg / mL, added to a dialysis bag, and dialyzed sequentially with dialysis buffer AD. Finally, the protein was dialyzed three times in PBS buffer, replacing the PBS system. After protein dialysis and refolding, the protein was concentrated using a 3 kDa ultrafiltration tube at 4500 r / min at 4°C. The concentration was measured using Nanodrop to above 1 mg / mL, and stored at -20°C. The protein concentration was determined using a BCA kit, and diluted to the required concentration for subsequent experiments according to the Elabscience manual.

[0043] Example 5: Affinity Evaluation of Original and Mutated Nanobodies Reagents and Kits: The BCA kit was purchased from Elabscience. Reagent Preparation: TBST buffer: 2.43 g Tris, 8.77 g NaCl, 1 mL Tween-20, add ultrapure water to 1 L Blocking solution: 5 g skim milk powder per 100 mL, 0.1% (v / v) TBST 10× Transfer Buffer: 30.2 g Tris, 144.13 g glycine, pH 8.3 1× Transfer Buffer: 100 mL 10× Transfer Buffer, 200 mL anhydrous methanol, add ultrapure water to 1 L After determining the antigen concentrations (Nectin-4, PD-L1) using the BCA method, dilute to 1 μg / mL with pre-chilled PBS. Add 100 μL to each well of an ELISA plate. For the blank control group, add the same concentration of BSA. Coat overnight at 4 °C, then discard the coating solution. Wash three times with TBST. Add 200 μL of skim milk blocking buffer to each well and incubate at room temperature with shaking for 2 h. Discard the blocking solution, wash three times with TBST, and add 100 μL of serially diluted nanobody in pre-chilled PBS to each well. Repeat for three wells and incubate at room temperature with shaking for 1 h. Discard the primary antibody, wash three times with TBST, and add 100 μL of Anti-HA Tag Antibody (HRP) diluted to the working concentration as the secondary antibody to each well. Incubate at room temperature in the dark with shaking for 1 h. Discard the secondary antibody, wash three times with TBST, and add 100 μL of TMB chromogenic solution (a 1:1 mixture of solutions A and B) to each well. Incubate at 37 °C in the dark with shaking for 30 min until the solution turns blue. The reaction was stopped by adding 100 μL of 1 M HCl to each well. The solution turned yellow at this point, and the OD450 value was measured using a microplate reader. Based on the measured OD450 values ​​(data expressed as mean ± SD, n=3), a four-parameter curve was fitted to the antibody concentration in GraphPad after logarithmic processing, resulting in an "S"-shaped curve with two plateau phases. The EC50 value of the antibody concentration corresponding to the midpoint of the curve, converted to the antibody molecular weight, is considered the antibody affinity value. Figure 16 As shown in Table 2, the affinity of the mutant nanobodies enhanced by computer-aided affinity enhancement was higher than that of the original nanobodies. In the Anti-Nectin-4 nanobodies, the mutant antibodies NBNT-M1 and NBNT-M2 showed 2.4-fold and 1.5-fold increased affinity, respectively, compared to the original antibody NBNT-1. In the Anti-PD-L1 nanobodies, the mutant antibodies NBP-M1 and NBP-M2 showed 4.3-fold and 2.1-fold increased affinity, respectively, compared to the original antibody NBP4.

[0044] Table 2 Affinity of each nanobody

Claims

1. A method for enhancing the affinity of computer-aided designed nanobodies, characterized in that, Includes the following steps: S1. Using the Antibody Region-Specific Alignment (ARLA) and CDR (Corrective Determinant) mapping tool server, the framework region and complementarity-determining region of the nanobody were determined. Nanobody models were constructed using both homology modeling and ab initio computational modeling. In the Basic Local Alignment Search Tool, structures with high homology to the antibody sequence were selected as templates. The original antibody amino acid sequence was input, and searches were performed using the Protein Data Bank and blastp database. Six templates with greater than 70% homology were selected for homology modeling. Structure prediction and model construction were performed on the AlphaFold2 ab initio computational modeling server. Candidate models from both homology modeling and ab initio computational modeling were uploaded to the StructureValidation Server, and the stereochemical, geometric, and physicochemical rationality of the models were evaluated using ERRAT, VERIFY 3D, and PROCHECK algorithms. S2. Predict possible antigen-antibody binding sites using the Protein Protein Interaction Server and Protein Interaction Calculator docking site prediction server. Upload the antigen and antibody PDB files with minimized energy in the GRAMM docking rigid docking server without specifying a site. Download the structure file of the antigen-antibody docking complex with the highest score. Analyze the antibody amino acids within 5 Å of the antigen as candidate sites in PyMOL. Combine the results from InterProSurf and PIC to select sites for flexible docking at specified sites. Perform molecular docking in the High Ambiguity Driven protein-protein DOCKing molecular docking server. After S3 and HADDOCK molecular docking, 10 antigen-antibody complex structures and corresponding parameter scores are output. The best antigen-antibody complex model is selected based on the Z-score for further analysis. The selected best antigen-antibody complex model is opened in PyMOL, and the antibody amino acid sites within 5 Å of the antigen molecule are counted. Combined with the prediction results of InterProSurf and PIC server, the amino acid sites located in the CDR region that do not contain cysteine ​​and proline are selected as mutation sites. S4. Single-point saturation mutations were performed on the identified mutation sites, and the amino acids at the specified sites were mutated to the remaining 17 amino acids excluding cysteine, proline, and themselves. The changes in antigen-antigen affinity changes before and after mutation were evaluated using the Antibody-Antigen Affinity Changes Upon Mutation and Protein-Protein Affinity Predictor online servers. Sites that could improve affinity scores when mutated to any other amino acid were selected as affinity-sensitive sites. Multi-point mutations and evaluations were performed, and a series of high-scoring triple mutants were finally obtained. Two high-scoring mutants were selected for molecular experiments to verify the affinity enhancement effect. The nanobodies are Anti-Nectin-4 camel-derived nanobodies NBNT-1 and Anti-PD-L1 shark-derived nanobodies NBP4. The mutant nanobodies corresponding to Anti-Nectin-4 camel-derived nanobodies NBNT-1 are NBNT-M1 and NBNT-M2, respectively, and the mutant nanobodies corresponding to Anti-PD-L1 shark-derived nanobodies NBP4 are NBP-M1 and NBP-M2, respectively.

Images