The functionality of a therapeutic antibody candidate restored by a single mutation from proline to threonine in the variable region

ABSTRACT mAbs play an essential role in the therapeutic arsenal. Our laboratory has patented the Rendomab-B49 mAb targeting the endothelin B receptor (ETB). This G protein-coupled receptor plays a driving role in the progression of numerous cancers. We chimerized our mAb (xiRB49) to evaluate its preclinical therapeutic efficacy in different ETB+ tumor models with an antibody drug conjugate approach. As previously reported, the chimerization process of an antibody can alter its functionality. In this article, we present the chimerization of RB49. xiRB49 purified by Protein A remained perfectly soluble and did not aggregate, but it lost all its ability to recognize ETB. A detailed analysis of its variable region using IMGT tools allowed us to identify an unusual proline at position 125. In silico mAb modeling and in vitro experiments were performed for a better understanding of xiRB49 structure-function relationships. Our results show that the proline in position 125 on the heavy chain alters the xiRB49 CDR3 light chain conformation and its mutation to threonine allows complete functional recovery.


Expression of RB49, xi RB49 and xiRB49 P115T:
The rendomab B49, targeting ETB receptor, was obtained by a novel gene immunization approach according to the laboratory protocol published in 2011.The originality of this approach is to present endothelin receptors in native conformations in an in vivo context.
Gene encoding the constant and variable sequences were synthesized by Eurofins Genomics Company.Then, sequences were subcloned into the pTT5 vector. 1 Then, the vectors were added in 1 mL of chemically competent TOP10 bacteria transformed by a thermic shock (30 s at 42°C).Transformed bacteria were plated on LB agar with ampicillin and incubated at 37°C overnight.Plasmids were purified for sequencing using the Wizard® Plus SV Minipreps DNA Purification System (A1340).
Different constructions: RB49, xiRB49, xiRB49-P125T, Fab-xiRB49 and Fab-xiRB49-P125T were expressed in ExpiCHO-S cells according to the manufacturer's instructions (ThermoFisher Cat # A29127).We followed the instructions of the "Max Titer Protocol" with the addition of 300µL of ExpiFectamin TM -CHO-Enhancer and 8mL of ExpiCHO TM -feed on day 1 and day 5 post-transfection.
Cells were shifted to 32°C on day 1 post transfection.To clarify the cell culture supernatant day 12 post transfection, we centrifuged it at 4000 g for 30 minutes at 4°C and filtered the supernatant through a 0.22µm filter.

SDS-PAGE:
To assess the quality and the purification of the proteins, we analyzed samples using the 4-15%mini protean® precast TGX TM (Tris-Glycine eXtended) Gels (BIORAD #4561083).Samples were treated in a different way, either with a non-reduced-buffer (Native Sample Buffer #1610738) either with a reduced-buffer (Laemmli 4X #1610747).Samples and molecular weight markers, Precision Plus Protein TM Markers (Biorad #1610363), were denatured at 95°C during 10 minutes and loaded on precast gels.Migrations were performed at a constant voltage of 200 V during 15 minutes.

RB49 Fabs modeling
The RB49 variable regions were modelled by the Antibody modeling tool of the Rosetta webserver, by asking to also model the H-CDR3.The obtained model was submitted to a quality check on the MolProbity webserver.Figure S9 shows the resulting report, which indicates a model of good quality (MolProbity score of 1.28 in the 99 th percentile), without Ramachandran outliers and with a good clashscore.The only minor issues come from other geometrical features, such as bonds lengths (0.44% of bad bonds) and angles amplitudes (0.53% of bad angles).However, this kind of issues are resolved during the following classical molecular dynamics simulation.In addition, the light and heavy chain variable regions were independently submitted to a BLAST search (https://blast.ncbi.nlm.nih.gov/Blast.cgi),using the PDB database to confirm the quality of the variable regions modeled by Rosie.The BLAST search on the variable region of the heavy chain provided an alignment with a cover percentage of 100%, an identity percentage of 86.67%, and an E-value of 610 - 10 with the variable region of the heavy chain of a Fv with code PDB 1A6U, while the one on the variable of the light chain provided an alignment with 100% of cover and identity percentages and an E-value of 110 -79 with the variable region of a Fab fragment with code PDB 4TPR. Figure S10 reports the superposition of the model to the two structures retrieved by the BLAST searches, which supports the conclusions on the quality of the variable regions model.
We repeated the BLAST search for the constant regions.For the RB49 light chain constant region we found an alignment with 100% of cover and identity percentages and an E-value of 310 -77 with the constant region of the light chain of the Fab with code PDB 1FIG.In addition, the variable regions of the light chains of RB49 and 1FIG show many conserved positions (Figure S11), thus we decided to use this structure to build the RB49 light chain constant region.The BLAST search performed on the xiRB49 kappa constant region gave an alignment with 100% of cover and identity percentages and an E-value of 110 -75 with the kappa constant region of a Fab with code PDB 6U8K, and the alignment on the whole xiRB49 and 6U8K light chain showed many conserved residues also in the variable region (Figure S12), making this structure exploitable for the xiRB49 light chain modeling.
Analogously, we performed a BLAST search on the full RB49 constant region, and we obtained an alignment with 100 % of cover, 94.38% of identity and an E-value of 0 with the constant region of the heavy chain of an IgG1 with code PDB 1IGY.In addition, the alignment with the whole heavy chain showed a high number of conserved residues in the variable region also (Figure S13).Therefore, we proceeded with the homology modeling of the RB49 CH1 using the 1IGY constant region CH1 with SwissModel.We obtained a model with a GMQE of 0.82 and an QMeanDisCo Global of 0.76  0.09, indicating a model of good quality.The structural assessment provided by SwissModel indicates a model with a few issues, notably 4 Ramachandran outliers, 2 distorted bonds and 11 distorted angles.However, these geometrical issues are resolved during the classical molecular dynamics simulations.
The same procedure was applied on the xiRB49 heavy chain constant region.The BLAST search provided an alignment with the constant region of the heavy chain of an IgG (code PDB 1HZH) with 100% cover, 99.08% identity, and an E-value of 0. In addition, many conserved residues were found also in the variable region, as shown in the alignment of the xiRB49 heavy chain and 1HZH heavy chain of Figure S1.Therefore, we proceeded with the homology modeling of the xiRB49 heavy chain CH1 with SwissModel using the 1HZH heavy chain as template.In this case, we obtained a GMQE of 0.88 and a QMEANDisCo Global of 0.82  0.08, suggesting a model of good quality.In addition, the SwissModel structure assessment results indicated the absence of Ramachandran outliers and of distorted bonds, while we found 9 distorted angles.
Finally, we built the final Fab-RB49 and Fab-xiRB49 by simultaneously aligning the modelled structures to the structures used as templates, manually creating a bond between the variable and constant regions, and finally minimizing the obtained structures with a minimization consisting in 2500 cycles of steepest descent and 5000 cycles of conjugated gradient.

Molecular dynamics simulations on RB49 Fabs
The modelled mouse, chimeric and mutated chimeric RB49 Fabs have been successively submitted to classical molecular dynamics (cMD) simulations.These were performed with the pmemd.cudamodule of Amber20 package 2 using the ff14SB force field 3 .For each Fab, the total charge was neutralized by including an adequate number of Na + /Cl -ions and the systems were embedded in an octahedral TIP3P water box added up to 10 Å from the solute.Each system was then relaxed by optimizing the hydrogens geometry (1000 cycles of steepest descent and 5000 cycles of conjugated gradient), ions and water molecules (2000 cycles of steepest descent and 5000 cycles of conjugated gradient).The water box was equilibrated at 300 K by 100 ps of NVT and 100 ps of NPT simulation using a Langevin thermostat with a collision frequency of 2.0 ps -1 .Then, we performed a minimization of side chains, water and ions by applying backbone restraints of 25 kcal/mol and a total minimization with backbone restraints of 10 kcal/mol (2500 cycles of steepest descent and 5000 cycles of conjugated gradient).Each system was then gradually brought to 300K in 6 steps of 5 ps each with an temperature increase step of 50 K, while backbone restraints were progressively reduced from 10 to 5 kcal/mol.A 100 ps NVT equilibration step followed by a 200 ps NPT equilibration step (backbone restraints = 5 kcal/mol) were performed.Successively, the backbone restraints were gradually removed by 100 ps NPT equilibration steps.Finally, 10 ns of unrestrained production at 300 K were performed to collect average potentials for the following accelerated MD (aMD) simulations.During the cMD simulations an electrostatic cutoff of 8.0 Å, a Berendsen barostat, PME for long-range electrostatic interactions and the SHAKE algorithm were applied.
Finally, for RB49 and xiRB49 Fabs 3 independent aMD runs of 1 s each were run, for a total of 3 s.Conversely, in order to better consider the single point mutation, for the p125t-xiRB49 Fab 3 independent aMD runs of 1.5 s each were run, for a total of 4.5 s.

Figure S10 .
Figure S10.Superposition of the modelled RB49 variable regions with the structures selected after the

Figure S11 .
Figure S11.Sequence alignment between the RB49 and the 1FIG light chains.The alignment has

Figure S12 .
Figure S12.Sequence alignment between the xiRB49 and the 6U8K light chains.The alignment has