Embedded Files
Step 1: Mutagenesis
Step 1: Mutagenesis
PyMol was used to evaluate point mutations which might enable better binding. The conformation which resulted in the least clashes with neighboring side chains or structures was selected, and then compared to WT IgG1 and IgG3. The wild type sequences and areas of interest are shown below.
Step 2: Structural Assessment
Step 2: Structural Assessment
Input
The amino acid sequence of the Fc, and unique name for the run.
Standard settings were used for each structure generated, shown to the right.
Analysis/Use
The structures generated by AlphaFold2 provided a necessary checkpoint that the structure was not significantly which could prevent essential Fc effector functions such as complement activation. This was assessed by eye as well as by the RMSD with respect to the wild type IgG in PyMol.
The RMSD is highly influenced by larger structural changes, and minimally influenced by the small side chain changes we are targeting.
There was high confidence in each of the structures generated by AlphaFold. A typical output is shown below. They all demonstrated high confidence in all areas aside from the unstructured loops.
We utilized the rank 1 structure for all our additional steps.
Step 3: Interaction Prediction(11,12)
Step 3: Interaction Prediction(11,12)
AIR Definitions
HADDOCK2.4 uses crystallography and NMR systems for structure calculation of molecular complexes. The definition of ambiguous interaction restraints, or AIRS, is critical to this model as it translates raw data into distance restraints used in the energy functions for model ranking.
Active restraints include residues whose knockout abolishes binding interaction. This importance is highlighted by a "scoring penalty" incorporated in models which do not include these residues in the predicted interacting interface.
Passive restraints contribute weakly to interaction but do not incur a penalty if not a part of the predicted binding interface.
Critical and non-critical interaction residues were isolated from studies spanning site directed mutagenesis, random mutagenesis, alanine scanning, crystallography, and in silico modelling approaches.(6,8,9,10)
Active residues were selected based on their importance in binding from alanine scanning and mutagenesis experiments, including 3 residues, Ile253, His310, and His435. Mutations at these sites decrease binding interactions between FcRn and IgG1.
Passive restraints were defined as Fc inter-beta sheet loops in the CH2-CH3 regions (CH2:loop AB, loop DE and CH3:loop FG).
Histidine sites within the binding interface become protonated at low pH. HADDOCK2.4's residue definition module provides a "histidine protonation" toggle which can protonate or deprotonate histidines, and this feature was used to mimic "low" and "neutral" pH environments.
Energy minimization and cluster scoring
The HADDOCK 2.4 workflow involves 3 energy minimization steps in which bonds and residue partners are iteratively frozen in place, rotated, translated to optimize interactions, and made flexible in order to best define the interaction interface. Final models are then clustered and scored based on the equation:
HADDOCKscore = 1.0 * Evdw + 0.2 * Eelec + 1.0 * Edesol + 0.1 * Eair
Evdw = intermolecular van der waals energy, Eelec = intermolecular interstatic energy, Edeson = empirical desolvation energy,
Eair = AIR energy.
Our protocol generated 10,000 models during the rigid body minimization stage, then refined the best 400 models which were then clustered based on fraction of common contacts (FCC) and RMSD. A finished HADDOCK2.4 run analysis details the top 10 cluster results according to statistics including HADDOCK score, size, RMSD, Energies, BSA, and Z-score.
A total of 54 HADDOCK runs were completed for this project.
Passive and active residues on IgG-Fc (above) and FcRn-A2M (below) highlighted in green and red, respectively.
Step 4: Binding Prediction(13,14)
Step 4: Binding Prediction(13,14)
3D structure coordinates from HADDOCK are uploaded to the PRODIGY server in PDB format, with IgG defined as chain "A" and FcRn-A2m complex defined as chain "B" interactors.
PRODIGY outputs contain information regarding the predicted molar (M) dissociation constant (KD), predicted binding affinity (Delta G), predicted number of intermolecular contacts at the interface distance within 5.5 angstroms, and the percentage of charged/apolar non-interacting surfaces of the complex.
Repeat steps 1-4 until a satisfactorily optimized IgG3 can be proposed, or a deeper understanding has been developed which allows us to propose future work.
Computational Tools & Descriptions
Computational Tools & Descriptions
PyMol (paid version provided by the Johnson and Georgiou-Stone-Lavinder labs)
PyMol (paid version provided by the Johnson and Georgiou-Stone-Lavinder labs)
PyMol's wizard mutagenesis tool enables the visualization of rotamers and protonation states of amino acid point mutations. The tool also enables the export of modified PDB's. PyMol enables the overlay of structurally similar proteins and provides a root mean squared difference (RMSD) for the alignment of one or more protein structures.
ColabFold v1.5.5: AlphaFold2 using MMseqs2 (free)
ColabFold v1.5.5: AlphaFold2 using MMseqs2 (free)
AlphaFold2 predicts the 3D structure of proteins from an amino acid sequence.
Haddock2.4 PRODIGY protein-protein (free)
Haddock2.4 PRODIGY protein-protein (free)
Haddock2.4 PRODIGY (PROtein binDIng enerGY prediction) allows for prediction of binding affinity in biological complexes as well as the identification of biological interfaces from crystallographic complexes.
Page updated
Report abuse