Project for Molecular Immunology (AIM 620):
Modeling a Humanized Version of a Mouse Antibody
Introduction
In this project, you will make a homology model of a the variable
regions of a humanized mouse antibody, using a human immunoglobulin-G
(IgG) model as a template. Then you will compare your homology model
to the actual structure determined by X-ray crystallography.
What is a humanized mouse antibody?
A scientist at Flybynight Pharmaceuticals has found a mouse
monoclonal antibody that triggers lasting protection against a fatal
disease in mice. A similar disease occurs in humans, so the scientist
would like to make a human antibody with the same function (the mouse
antibody is rejected by humans). One way to begin designing a human
version is to replace the complementarity-determining regions (CDRs)
in a human antibody with those from the promising mouse antibody. The
result is called a humanized antibody -- all parts are
human except the CDRs, to minimize the chances that the human immune
system will recognize the antibody as foreign. The scientist hopes
that the humanized antibody will have a similar effect in humans, or
at least that it can serve as a starting point for design of such an
antibody.
The scientist uses modern techniques in molecular biology to
produce useful quantities of this hybrid antibody. It has low levels
of the desired function. The scientist now wishes to modify specific
residues to improve the function. Which residues to alter? One
approach is to make a model of the new antibody by homology modeling
and to see which non-CDR residues are in contact with the CDRs. These
residues are candidates from modifications that might improve the
function of the humanized version. A plausible template for the
modeling is the human antibody used as the framework for the
humanized version -- if it's structure is known from crystallography
or NMR spectroscopy.
Follow the instructions below to build a model of the functional
regions of this humanized antibody.
Files
Download and save the following three files into a folder named
Project:
- Target.fst: A
plain-text file containing the sequence of the humanized mouse
antibody in FastA format (one letter abbreviations of the
amino-acid residues). You will build a three-dimensional homology
model of the target using an appropriate three-dimensional as a
template. If this file appears in your browser, File: Save
As... , set Format to Text, and save the
file.
- Template.pdb: X-ray
crystallographic model of a generic human IgG, which will serve as
your template for constructing a homology model of the target.
This template is derived from PDB file 8FAB.
- Actual.pdb: X-ray
crystallographic model of the target protein. You will compare
your homology model of the target with this structure, which was
determined experimentally. This model is taken from PDB file
1BJ1.
Instructions
These instructions are not detailed. They presume that you have
worked through sections 1-11 of the Deep
View Tutorial, and are familiar with the workings of the program.
Conventions for designating commands and keystrokes are the same as
those in the tutorial.
- Open Target.fst in a word-processing program, just to
look at its contents. The letters beginning in the second line are
abbreviations for the amino-acid residues in the target protein.
The semicolon tells Deep View where to separate the residues into
two chains: the light-chain-variable and heavy-chain-variable
regions of the target antibody. Close the file without making any
changes.
- Start Deep View and cancel the opening dialog.
- Swiss-Model: Load Raw Sequence to Model. Load
Target.fst.
- Color: Chain. You will see two colors, indicating two
chains.
- File: Open PDB File... Find and open
Template.pdb. Press the zoom button (or <insert> on
PC, or <=> on Mac) to include both proteins in the view.
Color: Layer to help you distinguish them after the next
operations.
- Make sure that Target is the active layer (red in
Alignment window). Fit: Magic Fit. Deep View threads
the target onto the template. Display only the target.
- Select: aa Making Clashes. Note in the Layer
Infos window how many of these problem residues there are. To
attempt to fix them, Tools: Fix Selected Sidechains: Quick and
Dirty (other Fix options will probably take too long).
Again Select: aa Making Clashes to see how many were fixed.
Alternate fixing and selecting until the number of problem
residues does not decrease further. Now your model is about as
good as you can get with Deep View alone.
- Select: Visible Groups. Then Save: Selected Residues
Only... Name your new model 1stModel.pdb.
- Close all files. Now open Actual.pdb. This is a model
of the humanized antibody as determined by X-ray crystallography.
The CDRs are colored red and the rest of the model is colored
CPK.
- Open 1stModel.pdb. Zoom to include both models.
- Fit: Best (with Struct. Align.)... to align 1stModel
onto Actual and to align their sequences in the Align window.
- Color the CDRs in 1stModel. To make this easy, use to Align
window to see the colors in Actual and to select corresponding
residues in 1stModel. Then click the colr heading in the
Control Panel to pick a color for them.
- Now blink between the two models. How good is the
correspondence between the CDRs in 1stModel and in Actual?
- Make 1stModel the active layer. Color: RMS to color
1stModel according to how well it superimposes onto Actual. Note
that agreement is good for the core structures, but not so good
for the part we are most interested in: the CDRs.
- To guide the scientist in picking residues to modify, find a
way to color residues that are in contact with the CDRs, but that
are NOT contiguous with them in sequence. Save: Layer to
save your colored model as 2ndModel.pdb.
It turns out that modeling one antibody from another is very
difficult. As is often the case in homology modeling, if the target
and template have different functions (in this case, they bind
different antigens), then the functional regions are poor templates
for modeling. Even this relatively simple modeling problem confounds
sophisticated modeling programs. That's why, in this exercise, I did
not instruct you to send the model for optimization by
Swiss-Model.
Top of Page
To Biochemistry
Resources
HOME