Overview

In silico modeling in science is often employed to predict experimental results, identify putative targets and verify a specific hypothesis. By harnessing the power of various computational models, we are able to obtain a great deal of data within a short period of time, and at a same time, save a lot of laboratory resources and manpower. On the other hand, computational models can handle and execute millions of parameters at a time. These unique features also allow us to decipher the highly complex and variable biological systems and processes that can never be performed in typical benchtop wet-labs. Consequently, new insights are created and thereby, directs new hypotheses.

Our team PolYneer is devoted to developing a novel bispecific antibody (BsAb) to overcome Osimertinib resistant non-small cell lung carcinoma (NSCLC). To successfully engineer our novel BsAb, we have to consider the molecular targets that are simultaneously over-expressed in NSCLC cells as well as the binding affinity between our engineered BsAb and the targets.

First of all, one of the unique features of BsAb is to simultaneously target two cell surface markers on cancer cells. We used MaxQuant software to analyze the proteomes of Osimertinib-sensitive and -resistant lung cancer models. MaxQaunt is an integrated proteomics software suite for analysis and identification of peptides and proteins in biological samples. By quantitatively comparing the proteins expression in Osimertinib-sensitive and -resistant NSCLC cells, we are able to identify our BsAb targets.

Secondly, given that our novel BsAb recognizes its antigens by the conjugated peptide, it is crucial to screen the most robust peptide from a panel of peptide candidates. We understand that the 3D structure of the peptide determines the binding affinity between the BsAb and the targets, therefore, we employed BIOVIA Discovery Studio Visualizer to create the 3D structure of the conjugated peptide. Subsequently, the molecular docking simulation was performed using AutoDock Vina. The docked complexes were visualized by BIOVIA Discovery Studio Visualizer and PyMol.

The aim of these models is to identify the targets in the very beginning of our BioBrick design and help us to evaluate the binding efficiency of our bispecific antibody. All of the results guide us in the whole process of designing our BioBrick constructs.

Target identification by mass spectrometry analysis

MaxQuant is one of the most popular in silico tools for mass spectrometry (MS)-based proteomics data analysis With its own peptide search engine, Andromeda, tandem spectra from collision-induced dissociation (CID), high-energy collision dissociation (HCD) and electron transfer dissociation (ETD) can be easily analyzed with MaxQuant. For each method, a custom multi-level scoring scheme is used to optimize peptide identification for each specific fragmentation technique.

SHOW MORE

To investigate the proteome of the Osimertinib-resistant NSCLS cells, we first developed an Osimertinib-resistant lung cancer cell line, HCC827, by a prolonged incubation of increasing dose of Osimertinib up to 2 µM. Osimertinib resistance is developed by cells during the treatment and survival cells were selected for subsequent culture (Fig. 1).

Figure 1 Development of Osimertinib-resistant HCC827 cells.

Since kinase activity plays a vital role in the development of resistance in NSCLC, a kinase dependent probe, XO44 were used to selectively label all active kinases in cells. The kinase labelling by XO44 is done by covalent modification of active sites of the kinase (Fig. 2).

Figure 2 Mechanism of active kinase labelling by XO44 probe.

In tandem with subsequent enrichment and mass spectrometric analysis, the labelled kinases can be identified and quantified. The workflow of the mass spectrometry analysis has been summarized in Fig. 3.

Figure 3 Graphical Summary of the target identification by mass spectrometry analysis.

Methodology

XO44 Labeling was performed by treating 20 µL SRC (final concentration 5 μM) in kinase buffer (50 mM Tris, pH 8, 100 mM NaCl, 5 % glycerol) with probe XO44 (15 μM). After 1 hour, the reaction was quenched by adding 20 µL acetonitrile/0.1 % formic acid. Samples were analyzed by LC-MS (Waters Acquity UPLC/ESI-TQD, 2.1 × 50 mm Acquity UPLC BEH300 C4 column).

MS Raw data were searched against the Homo sapiens UniProt database (Version June 2020, 20368 entries) using the MaxQuant algorithm (Version 1.6.7.0). Carbamidomethylation of cysteine residues was set as a static modification. Protein N-terminal acetylation and oxidation of methionine residues were set as variable modifications. The maximum number of modifications was three. Specify digestion mode was used with trypsin, and the maximum number of missing cleavages was two. The precursor mass tolerance was set to 10 ppm with a fragment tolerance of 0.02 Da. Unique and razor peptides are used for protein quantification. The identified proteins were filtered with a false discovery rate of 1%.

MaxQuant provided output tables in the combined_txt folder. File ProteinGroups.txt contains information including identified protein groups, intensity values, protein properties, sequence coverage, peptide counts, MS/MS spectra counts and molecule weight. Filtering out contaminations and decoy proteins can be performed by setting ‘Only identified by side’, ‘Potential contaminant’ and ‘Reverse’ as negative. Protein with unique peptide number above two were used for protein quantification. The intensity values were used for fold change calculation of each protein. The workflow of data analysis is shown in Fig.4.

Figure 4 Workflow of the MS data analysis.

Results

After processing, we have successfully identified a total of 104 kinases in Osimertinib -resistant (Resistant) and -sensitive (Mock) HCC827 cells, respectively. We ranked the kinases according to their intensity, which indicates the relative abundance for comparison and quantification (Fig. 5). For better comparison, we visualized the relative intensity of all as heatmap (Fig. 6). Results show that, the MET gene encoding the tyrosine kinase, c-MET has an elevated intensity in the Resistant group when compared with the Mock group.

Figure 5 EGFR-TKI Treatment in Lung Cancer.
Figure 6 EGFR-TKI Treatment in Lung Cancer.

To compare the fold-change (FC) of each kinase between Resistant and Mock HCC827 cells, we define FC as Intensity of Resistant group/ Intensity of Mock group. We subsequently plotted the intensity against the FC and express the result in a volcano plot [Fig. 7]. We note that the MET gene encoding the tyrosine kinase, c-MET has a high fold change with high intensity.

Figure 7 A volcano plot showing the relationship between intensity and fold-change (FC) in Osimertinib -resistant (Resistant) and -sensitive (Mock) HCC827 cells.

Conclusion

We observe that the tyrosine kinase, c-MET has an increased intensity in Resistant group when compared with the Mock group, indicating the c-MET protein has elevated during course of Osimertinib resistance development. With the high intensity with noticeable fold-change of c-MET shown in figure 7, we believe that c-MET is a robust target for our bispecific antibody (BsAb). On the other hand, we also performed a western blot analysis to confirm the over-expression of the epidermal growth factor receptor (EGFR) in Osimertinib-resistant cells. Therefore, targeting both EGFR and c-MET will be our direction in engineering the novel BsAb.

Reference

[1]
Cox, J., & Mann, M. (2008). MaxQuant enables high peptide identification rates, individualized ppb-range mass accuracies and proteome-wide protein quantification. Nature biotechnology, 26(12), 1367-1372.
[2]
Parker, C. G., Pratt, M. R. (2020). Click chemistry in proteomic investigations. Cell, 180(4), 605-632.
[3]
Schenone, M., Dančík, V., Wagner, B. K., Clemons, P. A. (2013). Target identification and mechanism of action in chemical biology and drug discovery. Nature chemical biology, 9(4), 232-240.
[4]
Tyanova, S., Temu, T., Carlson, A., Sinitcyn, P., Mann, M., Cox, J. (2015). Visualization of LC‐MS/MS proteomics data in MaxQuant. Proteomics, 15(8), 1453-1456.
[5]
Tyanova, S., Temu, T., Cox, J. (2016). The MaxQuant computational platform for mass spectrometry-based shotgun proteomics. Nature protocols, 11(12), 2301-2319.
[6]
Zhao, Q., Ouyang, X., Wan, X., Gajiwala, K. S., Kath, J. C., Jones, L. H., Taunton, J. (2017). Broad-spectrum kinase profiling in live cells with lysine-targeted sulfonyl fluoride probes. Journal of the American Chemical Society, 139(2), 680-685.
Simulation and evaluation of the peptide-target binding affinity by a docking model

Comparing the traditional bispecific antibody [BsAb] one of our innovations is the creation of the secondary antigen binding site in the BsAb. Unlike the traditional approach, which the two antigen binding sites are formed by a combination of two different light-heavy chain antibody moieties, our approach creates the secondary antigen binding site in the BsAb by conjugating a synthetic peptide to the primary antibody to (Fig. 8)

Figure 8 Structure of conventional bispecific antibody and our novel design of bispecific antibody.
SHOW MORE

Because of this unique property in our novel approach, we believed that the binding of the conjugated synthetic peptide to its target will inevitably affect the efficacy of our BsAb. Therefore, it is of great importance to choose a suitable peptide for the conjugation. We searched for literature to select published peptide sequence of the epidermal growth factor receptor (EGFR) binding peptides.

Peptide candidates
VAPRRL
LARLLT
CMYIEALDRYAF
Table 2A list of EGFR binding peptide.

To evaluate the suitability of the peptide, we adopted a molecular docking model to simulate and elucidate the binding between the peptide and the target, EGFR. Owing to the fact that we had no available tools to obtain the actual structure of the peptide conjugated to the primary antibody, we could only use the peptide for subsequent simulation. The simulation was done by AutoDock Vina which is an open-source program for performing molecular docking. It is capable to include parameters, such as Kollman and Gasteiger partial charges, torsion, electrostatic interactions and hydrogen bonds for the simulation, and calculate the binding affinity as kcal/mol. The implementation steps of the docking simulation are summarized in Figure 9.

Figure 9 Graphical Summary of the docking simulation.

Methodology

Linear peptides (CMYIEALDRYAF, LARLLT, and VAPRRL) were assessed for their binding affinity to EGFR. The two-dimensional (2D) chemical structures of peptides were sketched in ChemDraw(Fig. 10) Subsequently, their 2D structures were saved in Mol2 file format and uploaded to the BIOVIA Discovery Studio Visualizer, and their three-dimensional (3D) structures were constructed in PDB format (Fig. 11). The crystal structure of EGFR (PDB ID:5Y9T) was retrieved in the PDB format from the Protein Data Bank. The crystal structure of EGFR in PDB format was uploaded to BIOVIA Discovery Studio Visualizer, the heteroatoms (water and other ligands) were extracted, and subsequently, polar hydrogens were added. Finally, the resultant protein was saved in PDB format and imported into AutoDock Vina , and Kollman and Gasteiger partial charges were added. The formatted 3D structures of the peptides in PDB were then imported to AutoDock Vina, checked for torsion, and saved in pdbqt format. The uploaded peptides and proteins were selected as ligands and macromolecules, respectively, and later saved in pdbqt format. A grid box was constructed for each protein for blind docking using AutoDock Vina. The scripts were then written for molecular docking using a command prompt, and the acquired results were presented in the form of binding affinity. The docked complexes were further visualized using BIOVIA Discovery Studio Visualizer, and PyMol and 3D images were generated.

Figure 10 Chemical Structure of the peptide candidates.
Figure 11 3D Structure of the peptide candidates.

Results

The peptides were further analyzed for molecular docking with the EGFR. The results of the molecular docking analysis in terms of binding affinity (kcal/mol) scores are presented in the form of a heatmap, as shown in Fig. 12. The peptide sequence CMYIEALDRYAF has a scores of -7.3 kcal/mol, which is the highest among other peptide candidates. Therefore, we chose the CMYIEALDRYAF as the EGFR binding peptide for our BioBrick design.

However, protein insolubility problem of our original BioBrick occurred during the wet lab stage. To solve this problem, we included a cyclic form of CMYIEALDRYAF to perform the modeling again. Surprisingly, the binding affinity scores of the cyclic form of CMYIEALDRYAF is even higher than the linear form. Furthermore, the best two results in docked complexes are shown in 3D forms in Fig. 13.

Figure 12 A heatmap showing the binding affinity (kcal/mol) scores for linear peptides (CMYIEALDRYAF, LARLLT, and VAPRRL) and CMYIEALDRYAF cyclic peptide with the EGFR target.
Figure 13 Molecular docking results. (a) Binding of the cyclic peptide to EGFR. (b) Binding of CMYIEALDRYAF (linear peptide) to EGFR.

Conclusion

The docking model simulates the binding between the peptide and EGFR and generates a binding affinity score for us to evaluate the suitability of the peptides. Although this docking simulation did not include the actual structure of peptide-antibody and did not reflect the real scenario of the binding between our BsAb and it targets, it offers a preliminary screening of peptides, which is helpful at the beginning of our project design. On the other hand, the docked complex generated by the model can also help us to understand the interaction and bonding between the peptide and target. This piece of information can guide us to optimize the peptide by modification, addition or removal of amino acids from the peptide in the future study.

Reference

[1]
Ai, S., Duan, J., Liu, X., Bock, S., Tian, Y., Huang, Z. (2011). Biological evaluation of a novel doxorubicin− peptide conjugate for targeted delivery to EGF receptor-overexpressing tumor cells. Molecular pharmaceutics, 8(2), 375-386.
[2]
Barlesi, F., Scherpereel, A., Gorbunova, V., Gervais, R., Vikström, A., Chouaid, C., Rittmeyer, A. (2014). Maintenance bevacizumab-pemetrexed after first-line cisplatin-pemetrexed-bevacizumab for advanced nonsquamous nonsmall-cell lung cancer: updated survival analysis of the AVAPERL (MO22089) randomized phase III trial. Annals of oncology, 25(5), 1044-1052.
[3]
Han, C. Y., Yue, L. L., Tai, L. Y., Zhou, L., Li, X. Y., Xing, G. H., Pan, W. S. (2013). A novel small peptide as an epidermal growth factor receptor targeting ligand for nano-delivery in vitro. International journal of nanomedicine, 1541-1549.
[4]
J. Eberhardt, D. Santos-Martins, A. F. Tillack, and S. Forli. (2021). AutoDock Vina 1.2.0: New Docking Methods, Expanded Force Field, and Python Bindings. Journal of Chemical Information and Modeling
[5]
Yuan, S., Chan, H. S., Hu, Z. (2017). Using PyMOL as a platform for computational drug design. Wiley Interdisciplinary Reviews: Computational Molecular Science, 7(2), e1298.
[6]
Seeliger, D., de Groot, B. L. (2010). Ligand docking and binding site analysis with PyMOL and Autodock/Vina. Journal of computer-aided molecular design, 24(5), 417-422.