Macromolecular Modeling Blog ™

Although numerous structures of peptide bound MHC-II molecules were solved, no one knows how does the peptide free MHC look like. Painter et al. elegantly use molecular dynamics to model the conformational changes upon peptide removal. Most interestingly a helix from the peptide binding domain adopts the binding mode of the antigen peptide. They successfully validate their model using antibodies and superantigens, predicted to differentially bind peptide-bound/free molecules according to their model. We take the validation one step further and propose mutations based on Painter’s model that would stabilize the free MHC. Will it work? Who will pick up the gauntlet?

Corrie A. Painter, Anthony Cruz, Gustavo E. López, Lawrence J. Stern, and Zarixia Zavala-Ruiz published by PLoS ONE

Overview:

Class II major histocompatibility complex (MHC) are heterodimeric proteins which bind antigenic peptides as part of the adaptive immune response to foreign pathogens. The intact MHC II-peptide complex is displayed at the cell surface of antigen presenting cells (APC) for surveillance by T-cells. Interaction between the APC and its cognate T-cell leads to an effector response which then clears the body of the invading pathogen.

Peptides bind to the MHC II in an extended fashion along a binding groove contributed to by both the alpha and beta subunits. Previous crystal studies revealed a conserved hydrogen bonding network which exists between the peptide backbone and main chain residues along the helices of the alpha and beta binding domain. Additionally, binding energy is created by the interaction of peptide side chains and pockets within the binding groove of the MHC II binding domain. Although there is little structural variation observed among crystal structures determined for MHC II-peptide complexes, previous studies reported an unclear conformational change between peptide-bound and  peptide-free MHC II molecules (Hansen et al, Zarutskie et al). Previous work using conformationally sensitive monoclonal antibodies raised against the beta-chain of DR1, or differential chemical modification,  helped to define regions within the structure that change upon the peptide occupancy state, however, not enough information to generate a working model of the peptide-free DR1.  

What’s new ?

In this study, the authors performed a molecular dynamics simulation of HLA-DR1 in both the peptide-free and peptide-loaded states. Several regions of DR1 were predicted by this analysis to change conformation substantially upon loss of bound peptide. Differential binding of conformationally-specific antibody and superantigen probes to the peptide-free and peptide-loaded forms of DR1 provides experimental support for this model.

The model for the peptide-free structure of DR1 was obtained by molecular dynamics simulation starting from the X-ray coordinates of a DR1-peptide complex (PDB Code: 1SJE) from which peptide was removed before the start of the simulation. A parallel simulation was started from the same coordinates but without the removal of peptide. During the simulation, large changes in total root mean square deviation (RMSD) to the starting structure, were observed over the first 5–10 ns, with much smaller fluctuations occurring thereafter.

In the peptide-binding site, significant movements can be seen in the alpha-helices of the peptide-free form, as compared to those of the peptide-loaded form, which do not fluctuate or move as much. The principle differences between the peptide-loaded and peptide-free simulations were in the region of the peptide binding domain, corresponding to the last two strands of the beta sheet “floor” and the first half of the alpha-helical region forming one side of the peptide binding site. Changes were also noticed in the beta subunit lower immuoglobulin domain, but different orientations of this domain relative to the peptide binding domain have already been observed in different HLA-DR1 crystal structures.

During the simulation, the alpha50-59 region (helical side of binding pocket) of DR1 moves to fill the amino-terminal end of the peptide-binding site occupying, in part, the area where the antigenic peptide is usually found. A sharp kink forms at Gly(alpha)58, allowing the region alpha50-59 to fold into the binding site, taking the place of the bound peptide. After moving into the peptide-binding site, the main chain of the alpha50-59 region is able to satisfy essentially all of the hydrogen bonds in this region lost upon removal of the peptide. This narrowing of the site occurred early during the dynamics simulations run in the absence of peptide, and persisted throughout the entire 60ns time course. Such narrowing, of course, was not observed during dynamics simulations of the peptide-loaded form. 

A. Peptide loaded Xray structure. B. Peptide Free Model of MHC-II molecule (Painter et al. Fig. 3)

Experimental Validation:

Superantigens are soluble bacterial toxins that bind to class II MHC proteins and T cell receptors, independent of peptide antigen. Superantigen binding sites on the DR alpha subunit have been determined by X-ray crystallography of superantigen-MHC-peptide complexes. One such case is Staphylococcal Enterotoxin C3 (SEC3). The contact residues between SEC3 and peptide-loaded DR1 have been mapped by crystallography and by mutagenesis. None of the defined contact residues move appreciably during molecular dynamics simulation of the empty protein. Hence, similar binding is expected to peptide-bound and peptide-free DR1. Indeed SPR measurements shows that when immobilizing either the superantigen or the DR1, SEC3 shows identical binding behavior for peptide-free and peptide-loaded DR1. This data demonstrates that SEC3 binding does not distinguish between peptide-loaded and peptide-free DR1. Because this epitope is not predicted to move upon release of peptide, these observations are consistent with the model.

Monoclonal antibody MEM-264 previously has been shown to bind specifically to the peptide-free conformation of DR1. The MEM-264 epitope has been mapped by overlapping peptides and alanine scanning mutagenesis, and corresponds to a discontinuous region on the beta subunit helical region. In the molecular dynamics model for peptide-free DR1, this entire region is predicted to move significantly relative to the rest of the beta subunit. Both SPR experiments and an equilibrium sandwich ELISA assay proved that peptide-loaded DR1 indeed binds to MEM-264 at least 400-fold more weakly than does peptide-free DR1. Similar validation was performed with monoclonal antibody LB3.1. 

What now ?

Up until now – a great work by Painter et al. From here on, our very own “Macromolecular Modeling Blog” science project… 

The ability to better understand and manipulate MHC molecules will have many important applications starting from designing leads against autoimmune diseases and up to potential defenses against bio-terror utilizing superantigens. But in order to get there, an atomic model of the system is of need. The ultimate goal of an atomic model of a protein is to enable us to manipulate this protein. Be it by studying a binding pocket and designing a drug to bind to it, or by discovering the exact atoms which take part in a reaction mechanism. In order to further validate the model suggested by Painter et al. and to check whether indeed it is accurate enough to allow for molecular manipulation, we asked the authors for the reported HLA-DR1 peptide free model,  which they happily provided us with (For that we are ever so grateful.)

We used ROSETTA to perform a very quick analysis of this system. We focused on the peptide binding domain, since, as reported, this is the part that undergoes the major conformational change. On this domain, we evaluated the stabillity of three different models, the bound structure containing the peptide, the bound structure – peptide removed, and the model for the peptide free domain. The three structures were subjected to a short minimiztaion (repacking), in which the solved structures were allowed to repack their side chains (keeping the solved backbone fixed) and the model was also allowed small relaxation of the backbone and a small rigid body movement between the alpha and beta subunits. After this minimization these were the energy values for the models: Bound (w/o peptide): -330, Bound (with peptide): -353, Free: -344 (scores are in ROSETTA energy units). Indeed it seems that the removal of the peptide from the bound structures results in an energy decrease of about 23 energy units and when letting the free monomer stabilize itself it regains about 14 energy units. We thought, why stop there, if we really want to see how the free monomer looks like, let’s stabilize it even more using computational design!

We applied RosettaDesign on residues alpha50-59 (reported to adapt the place of the peptide), and on the adjecent residues from the beta subunit. The design converged to 4 mutations:

• alpha F51K – A hydrophobic PHE was sticking out into the solvent,  the mutation allows for better solvation.
• alpha S53L, E55W – These two mutations create a hydrophobic cluster that stabilizes the structure of the 50-59 loop and also perhaps creates a pi-cation interaction with beta R81.
• beta T90R – creates a new salt bridge between the two subunits.

Left - a hydrophobic cluster that stabilizes the free loop structure and facilitates pi-cation interaction with the beta chain. Right - a newly created salt bridge between the two subunits.

The designed structure energy was -357, a bit better then that of the original, peptide bound structure. Suggesting that with these 4 mutations, the MHC molecule would be stable even without the peptide and perhaps could be solved by x-ray crystallography. 

We call for a brave experimentalist, to pick up the gauntlet, and try to validate these mutations! apart from validating Painter’s model it would also allow us a first glimpse into the unbound form of MHC molecules.  

Know of any such brave experimentalists? How would you use a model of the peptide free MHC molecule? What are your thoughts of this approach? Tell us in the comments.

By Nir London.

 
* A special post celebrating “PLoS ONE @ Two” the second birthday of PLoS ONE

Corrie A. Painter, Anthony Cruz, Gustavo E. López, Lawrence J. Stern, Zarixia Zavala-Ruiz (2008). Model for the Peptide-Free Conformation of Class II MHC Proteins PLoS ONE, 3 (6) DOI: 10.1371/journal.pone.0002403