The TSLP amino acid sequence was uploaded into the DS 4.5 platform (NeoTrident Technology Ltd.). The template (NCBI Gene ID: 85450) was selected from gene sequences with the highest homology, performed using the Basic Local Alignment Search Tool (BLAST). Energy minimizations were performed as described by the TSLP antigen homology modelling method (7,8). The rationality of the model was evaluated using a Ramachandran plot and the Profile-3D program using DS software.

A diverse and natural full-human scFvs antibody library, with a capacity of 2.5×10⁸, was constructed using mRNA of PBMC from healthy volunteer in a previous study (9). By phage display, biotinylated TSLP protein was used as an antigen to select human anti-TSLP-scFv from the constructed human scFv library following the protocol previously described (9). The selected anti-TSLP-scFv-84 amino acid sequence was uploaded into the DS 4.5 system. Antibody domain and complementarity determining region (CDR) sequences were identified and annotated using the Annotate Antibody Sequence protocol (Discovery studio visualizer version 16.1.0.15350) (10). The sequences of variable heavy chain (VH) and variable light chain (VL) were set according to similarity, surface, light and heavy chain templates, resulting in the greatest degree of sequence similarity. The optimal model V_H_V_L_M0004_M was selected using probability density function (PDF) total energy, PDF physical energy and discrete optimized protein energy (DOPE) Score. Finally, a three-dimensional model of anti-TSLP-scFv-84 was constructed and the rationality of the model evaluated using Profile-3D and a Ramachandran plot.

Optimized anti-TSLP-scFv-84 as a receptor and TSLP as a ligand, with docked poses and clusters, were obtained using the ZDOCK operation (Discovery studio visualizer v16.1.0.15350). Scanning alanine and saturation mutations in the docking region were used to predict amino acid mutation sites for increased affinity. Based on mutational energy, amino acid sites were selected to increase the affinity of the antibody after mutation.

Based on DS system predictive mutation sites (Table SI), single-point mutation primers were designed (Table SII) and synthetized by Invitrogen, Thermo Fisher Scientific, Inc. Using the method of overlap extension PCR, the recombinant plasmid anti-TSLP-scFv-84 (pre-mutated) served as the PCR template in the two first stage reactions. The first and second PCR amplifications were performed using 32 cycles at 98°C for 10 sec, 60°C for 30 sec and 72°C for 1 min; The first and second PCR products were mixed as templates for the third PCR. The third PCR amplification was performed 10 cycles at 98°C for 10 sec, 60°C for 30 sec and 72°C for 1 min; Then, Forward primer scFv84F-TSLP and reverse primer KM168 were added for 20 cycles. Finally, single point mutations products of anti-TSLP-scFv were obtained (M1, M2, M3, M4 and M5). The amplified PCR products was electrophoresed on a 1.5% agarose gel in 1X Tris borate EDTA buffer with pH 8.3 and 0.03 µl ethidium bromide. The gel was visualized under ultraviolet light.

A TSLP-pLZ16 plasmid was constructed in our previous study based on the pUC vector (11). The mutated anti-scFv fragment and the pLZ16 vector were digested with NotI and NocI restriction enzymes at 37°C for 4 h. The products were ligated using 1 µl T4 ligase (Takara Bio, Inc.) at 16°C for 12 h and transformed into DH5αF′ competent cells (Tiangen Biotech Co., Ltd.). Transformed positive scFvs were identified by PCR using 30 cycles at 98°C for 10 sec, 55°C for 30 sec and 72°C for 1 min. High-fidelity thermostable ExTaq enzyme (Takara Biotechnology Co., Ltd.) and forward primer scfv84F-TSLP: 5′-CATGCCATGGCCGGCCCAGCCGGCCC-3′, reverse primer KM168: 5′-CTGAGTAGAAGAACTCAAACTA-3′ were employed. The positive scFvs clones were cultured on a shaker at 37°C, 200 rpm for overnight, and the plasmids were extracted with QuickPure Plasmid Mini Kit (CW Biotech) and Sanger sequenced by Shanghai Sangon Biotech Co., Ltd. The sequences were aligned in BLAST (Fig. S1). The pre-mutation scFv84 and mutant scFvs (M1, M2, M4 and M5) were expressed at a constant temperature of 32°C for 5 h with 1 mmol/l lisopropyl β-D-1-thiogalactopyranoside induction. The expressed scFvs were assessed for binding to human TSLP using indirect ELISA with 200 ng/ml HRP-labelled anti-FLAG monoclonal antibody (Thermo Fisher Scientific, Inc.) as secondary antibody.

The scFv-84 (pre-mutated) and scFv-M4 (mutated) sequences in pLZ16 were amplified by PCR with scFv84-KpnI-F (5′-CGGGGTACCATGGCCGGCCCAGCCGGCCCA-3′) and scFv84-BamHI-R-C (5′-CGCGGATCCACGTTTGATCTCCAGCTTGGTCC-3′) (12). PCR amplifications were performed with ExTaq enzyme, using 30 cycles at 98°C for 10 sec, 55°C for 30 sec and 72°C for 1 min. The target DNA fragment was digested with KpnI at 37°C for 4 h and BamHI at 30°C for 4 h, and then ligated with T4 ligase at 16°C for 12 h into pcDNA3.1-sp-Fc [constructed in previous work (12), sp and IgG1 Fc were insert into pcDNA3.1] to generate the pcDNA3.1-sp-scFv-Fc recombinant vectors. There combinant vectors were transformed into E. coli TOPO10 competent cells (Takara Bio, Inc.). Positive clones with the correct sp-scFv-Fc were identified by PCR using 30 cycles at 98°C for 10 sec, 55°C for 30 sec and 72°C for 90 sec with ExTaq enzyme (Takara Bio, Inc.). The constructed plasmids were sequenced by Shanghai Sangon Biotech Co., Ltd.

The sp-scFv-Fc-84 (Pre-mutated) and sp-scFv-Fc-M4 (mutated) in pcDNA3.1 were amplified by PCR using 30 cycles at 98°C for 10 sec, 55°C for 30 sec and 72°C for 90 sec with ExTaq enzyme (Takara Bio, Inc.) (11). Forward primers pcDNA3.1-F (5′-CTAGAGAACCCACTGCTTAC-3′) and pcDNA3.1-R (5′-TAGAAGGCACAGTCGAGG-3′) were employed. The target DNA fragment was digested with HindIII and NotI at 37°C for 4 h, and then ligated with T4 ligase (16°C, 12 h) to PMH3^EN (Hangzhou Amprotein BioEngineering Co., Ltd.) to generate the PMH3^EN-sp-scFv-Fc recombinant vectors. There combinant vectors were transformed and positive clones were identified as afore mentioned.

Transformed 293F cells were obtained from Abace Biotechnology Co., Ltd. and maintained in serum-free FreeStyle™ 293 expression medium (Gibco; Thermo Fisher Scientific, Inc.). Cells were cultured as described previously (13,14). Filter-sterilized plasmid DNA (PMH3^EN, PMH3^EN-sp-scFv-Fc-84/M4 vector) was prepared for transfection as described previously (14). Briefly, 293F cells were cultured at 1×10⁶ cells/ml to a final volume of 90 ml with Serum-free Freestyle™ 293 expression medium in roller bottle (Corning). Filter-sterilized DNA (100 µg) was added to 10 ml of 1XPBS and vortexed for 5 sec. Then, 313 µl of 1 mg/ml filter-sterilized polyethylenimine PEI (Polysciences, Inc.) solution was pipetted into the DNA/PBS mixture. The DNA/PEI/PBS mixture was added to the cells, which were incubated in an orbital shaker incubator at 37°C, 120 rpm and 8% CO₂ for 72 h. After centrifuging at 4°C, 350 g for 10 min, the cells pellet was collected and stored at −80°C for ELISA and western blotting.

After centrifuging, cells were resuspended in 1XPBS buffer and sonicated. The sonicated cell solutions were centrifuged and the supernatant containing the expressed protein was used to ELISA. scFv-Fc-84/M4 protein solutions were dilute with 1XPBS (at various concentrations 1:1, 1:10, 1:100, 1:1,000 and 1:10,000) derived from the PMH3^EN vector were used to coat 96-well plates overnight at 4°C in 50 mM NaHCO₃/Na₂CO₃ (pH 9.6). 200 ng/ml goat pAb to Hu IgG (HRP) (Abcam, ab97225) was used to detect PMH3^EN-scFv-Fc as described previously (15). The absorbance of each well was detected at 450 nm.

As described previously (9), 10 µg/ml TSLP (Sino Biological Inc.) was coated in 96-well plates overnight at 4°C. Following this, a scFv-Fc-84/M4 protein solution (1:1, 1:10, 1:100, 1:1,000, 1:10,000) expressed in PMH3^EN was added to the plates, then 200 ng/ml goat pAb to Hu IgG (HRP, 1:5,000; Abcam, ab97225) was added to measure binding between antigen TSLP (Sino Biological) and antibody PMH3^EN-scFv-Fc-84/M4. The absorbance of each well was detected at 450 nm.

The scFv-Fc-84/M4 protein solutions were prepared by sonicating the transfected 293F cells in 1×PBS as afore mentioned and the BCA method was used to determine the protein concentration. Total cellular proteins [30 µg: Control (cells only), PMH3^EN, scFv-Fc-84/M4] expressed in the PMH3^EN were subjected to SDS-PAGE with 10% gels and transferred to nitrocellulose membranes. After blocking with 5% non-fat milk for 1 h at room temperature, the membranes were incubated with 200 ng/ml goat pAb to Hu IgG (HRP, 1:5,000; Abcam, ab97225) for overnight at 4°C. A chemiluminescence detection kit (GE Healthcare) was used to detect the expression of scFv-Fc-84/M4. According to previously described methods (16), 5 µg TSLP (16135-H08H; Sino Biological, Inc.) was subjected to electrophoresis and membranes were subsequently incubated with scFv-Fc-84/M4 (400 ng/ml) antibodies overnight at 4°C, a secondary anti-human IgG Fc HRP-conjugated antibody was used to detect the binding ability of antigen (TSLP) and antibody (scFv-Fc-84/M4).

The scFv-Fc-84/M4 protein solutions were prepared as afore mentioned. In total, 4 µl protein supernatant of the single-chain antibody scFv-Fc-84/M4 was coupled to a protein A biosensor (Pall ForteBio). Subsequently, 4 µl standard recombinant human (rHu) TSLP protein (5, 10 and 20 nM; Sino Biological Inc.) was used to assess sensor interaction. By systematically analysing antigen-antibody binding and dissociation, affinity dissociation constant (KD) values were calculated. The interaction between antigen (TSLP) and antibody (anti-TSLP-scFv-Fc) were analyzed by Octet RED96 (Pall ForteBio). Acquired data were analyzed by the custom ForteBio software Data Acquisition 9.0.

All experiments were repeated three times. Data are expressed as the mean ± SEM. One-way ANOVA and Tukey's Multiple Comparison Test were used to determine significance with GraphPad Prism 5 software (GraphPad Software, Inc.). P<0.05 was considered to indicate a statistically significant difference.

After aligning the TSLP protein sequence (NCBI Gene ID: 85450) with the NCBI crystal library sequence, the DS 4.5 software package was used to detect the template gene sequence with the highest homology. With homology constraints, 20 optimized models were obtained. TSLP.M0020 (Fig. 1A) exhibited the lowest PDF total energy; this model was evaluated using a Ramachandran plot (Fig. 1B). The blue region indicates the optimal zone. The more amino acids (indicated by green circles) in the optimal region, the more reliable the structure. The purple region indicates the permitted zone. The pink circles are amino acids with ψ-φ conformations that are unreasonable; these regions require optimization. A Profile-3D score >0.0 indicated that the amino acids were compatible (Fig. 1C).

The DS system predicted domains and CDRs of the anti-TSLP-scFv-84 antibody. In total, 25 templates were obtained as ‘identify framework templates’. 4P59 LH was selected as the overall template, 4HS6 A as the light chain template and 4KFZ C as the heavy chain template. The model antibody framework protocol and the model antibody loops protocol simulated models of the scFv framework region and CDR. Finally, V_H_V_L_M0004_M. M0002 was selected as the best 3D model through PDF total energy, PDF physical energy and DOPE Score (Fig. 2A). This reasonable model was evaluated using a Ramachandran plot; nearly all amino acids were within the optimal zone (blue area). Only a few amino acids required optimization within the permitted region (pink area; Fig. 2B).A Profile-3D score >0.0 indicated that the amino acids were compatible (Fig. 2C).

Anti-TSLP-scFv-84 as the receptor and TSLP as the ligand were docked with docking proteins. Refine docked protein optimization was performed based on CHARMm force field energy. The range of root mean square deviation of POSE50 was 0.2–1.0A as measured by molecular dynamics. This result indicated that the dynamic change of the model in solution was reasonable (Fig. 3A). The binding regions of TSLP and anti-TSLP-scFv-84 were analyzed using the DS system and the CDRs were mutated. In total, five key amino acid residues that could increase antibody affinity after mutation were predicted by alanine scanning. The five amino acid combinations were as follows: TRP (109)-ARG, TRP (109)-LYS, TRP (109)-TYR, TRP (109)-LEU, TRP (109)-GLU (Table SI; Fig. 3B; 109 is the number of amino acid in the variable region), with the largest negative energy value (<-4) identified by saturation mutagenesis and mutation energy (Fig. 3C).

The pre-mutated and DS system-predicted mutated amino acids of scFv M1, M2, M3, M4 and M5 are presented in Table SI. Primers were designed based on the predicted mutation sites (Table SII) and paired to amplify the pre- and post-fragments of the mutation sites. The fragments were connected by overlap extension PCR. The products were 750 bp, which demonstrated that the fragments were connected successfully (Fig. 4A and B).

scFv M1, M2, M3, M4 and M5 were inserted into the pLZ16 prokaryotic expression vector. The sequencing results demonstrated that the amino acids in M1, M2, M4 and M5 were mutated successfully; however, scFv-M3 had a stop codon (data not shown). Therefore, scFv M1, M2, M4 and M5 were expressed in the pLZ16 prokaryotic expression vector. The target protein was detected using HRP-labelled anti-FLAG monoclonal antibody. Indirect ELISA results demonstrated that mutated scFv-M4 exhibited significantly increased binding to antigen compared with the pre-mutant scFv-84 (Fig. 4C). The expression of scFv-84/M4 was detected by coating the wells of a 96-well plate with the expressed single-chain antibody. The results revealed that the expression of scFv-84 was significantly higher than M4 (Fig. 4D). However, when combined with the antigen, scFv-84 binding was significantly lower than that of M4 (Fig. 4E), indicating enhanced affinity for mutated M4.

In general, scFv have a low stability, low affinity and a short half-life in vivo (5). Therefore, the pre-mutated scFv-84 and affinity enhanced scFv-M4 genes were inserted into the eukaryotic expression vector pcDNA3.1-sp-Fc. The vectors were transfected into 293F cells to express the IgG-like anti-TSLP-scFv-Fc fusion protein. Based on the cloning site for the pcDNA3.1-sp-Fc vector, specific primers were designed and the scFv gene fragment of the single-chain antibody was amplified using the ExTaq enzyme. A single band of 750 bp was detected by agarose gel electrophoresis of the PCR product that contained scFv (Fig. S2A). After digestion with KpnI and BamHI, anti-TSLP-scFv was ligated into the pcDNA3.1-sp-Fc vector with T4 ligase. The results demonstrated that scFv-84 and scFv-M4 were recombined with pcDNA3.1-sp-Fc successfully (Fig. S2B).

As the expression level of scFv-Fc in pcDNA3.1 was low (Fig. 5A and B), the recombinant plasmid PMH3^EN was used to improve expression. The pcDNA3.1 vector was digested with HindIII and NotI. Expression of sp-scFv-Fc-84/M4 was detected by agarose gel electrophoresis (Fig. S3A). Specific primers were designed based on the cloning site of the PMH3^EN vector and the anti-TSLP-scFv-Fc sequence. Empty plasmid PMH3^EN was cleaved with HindIII and NotI. Then sp-scFv-Fc was ligated into the PMH3^EN vector (Fig. S3B). Sequencing results demonstrated that the full-length sp-scFv-Fc had been inserted into all clones.

The expression of PMH3^EN-scFv-Fc-84/M4 was detected by direct ELISA, which was performed using an anti-TSLP HRP-conjugated antibody. The expression of anti-TSLP-scFv-Fc was not detected from the control and empty plasmid PMH3^EN; however, PMH3^EN-scFv-Fc-84/M4 expression was obvious. Furthermore, expression of PMH3^EN-scFv-Fc-M4 was significantly higher than PMH3^EN-scFv-Fc-84 (P<0.05; Fig. 5C). In addition, the expressed anti-TSLP-scFv-Fc protein was identified by western blotting. Using anti-human IgG Fc HRP-conjugated antibody, a single band of 50 kDa was identified (Fig. 5D).

Indirect ELISA was used to assess the affinity of the antibody for the antigen. ELISA plates were coated with TSLP and incubated with anti-TSLP-scFv-Fc. An anti-TSLP HRP-conjugated antibody was used to assess the binding of PMH3^EN-scFv-Fc-84/M4 to TSLP (Fig. 5E). ELISA results indicated that the antibodies bound TSLP and that the binding of PMH3^EN-scFv-Fc-M4 was higher than that of PMH3^EN-scFv-Fc-84. Western blotting revealed that the TSLP protein bands specific binding with recombinant PMH3^EN-scFv-Fc-84/M4 was ~25 kDa (Fig. 5F). A positive association between ELISA and western blotting data confirmed the validity of the results.

To verify that the affinity of mutated scFv-Fc-M4 was higher after mutation, PMH3^EN-scFv-Fc-84/M4 was captured with an anti-Fc biosensor and evaluated with different concentrations of standard rHuTSLP protein. By analysing the process of binding and dissociation (Fig. 6A and B), the affinity KD of PMH3^EN-scFv-Fc-M4 was calculated to be 3.21×10⁻⁹, which was ~10-fold higher than before the mutation. The product was biologically active and retained high binding affinity to the antigen, as detailed by BIAcore real-time interaction analysis.