Biophysical and structural analysis of protein-protein interactions: from encounter complexes to computational design and directed evolution

Lead Research Organisation: University of York
Department Name: Biology


Proteins are the workhorses in a cell that are involved in carrying out most if not all of its functions. A central feature of how proteins do this is to interact with other proteins, with these protein-protein complexes conveying information that is interpreted by the cell in the context of all the other interactions going on simultaneously. This application aims to use a model protein-protein interaction (PPI) of a bacterial toxin that degrades DNA (a DNase) with its inhibitor protein, known as an immunity protein, to address fundamental questions concerning PPIs that remain unanswered and which are applicable to most PPIs. One of the questions concerns the mechanism by which proteins form complexes with each other. All protein complexes go through what is termed an 'encounter complex', a transient species en route to the final complex. The properties of these encounter complexes have been discussed and speculated on for many years but there has never been a description of the structure for such a complex because by their nature they are transient species where no one state is heavily populated under equilibrium conditions. Any information on such a complex would be a valuable addition to our accumulating knowledge on how proteins form complexes with each other. We have stumbled on a mechanism of association in DNase-immunity protein complexes where one protein pivots and rotates against the other, which offers us an opportunity of 'freezing-out' the numerous orientations and leaving just one that can be structurally defined. This will be accomplished by inserting a stable (covalent) bond between the two proteins in a position that traps one of these transient encounter complexes and stops it reorienting. We will then try to obtain the three dimensional structure of this covalent complex using X-ray crystallography. The aim is to do this for a number of complexes and so gain a broader picture of what this encounter complex looks like. The other question we will address relates to how these DNase-immunity protein complexes recognise they have formed the correct association (i.e. specificity). This is a generic problem in PPIs since cells often contain many versions of proteins that are very similar to each other; how do their binding partners distinguish right from wrong given the abundance of possible partnerships? We will address this question from the perspective of newly described complexes, which originate from laboratories of collaborators, where new methods have been devised to reconstruct the specificity of the PPI. In one set of complexes, specificity was design by a computer algorithm (computational design), while in another specificity was engineered by Darwinian selection in a test-tube (directed evolution). Our aim is to characterise the properties of these novel complexes through a series of biophysical and structural methods that my lab have developed over a number of years so that we can understand how these novel complexes differ from their natural counterparts. This in turn will help us understand what governs specificity in these complexes and how close we are to tailoring PPI specificity at will.

Technical Summary

1.Structure of a protein-protein interaction (PPI) encounter complex. All protein-protein associations go through an encounter complex prior to forming the final complex but as yet there has been no atomic-level resolution of such an intermediate. Our finding that encounter complexes of colicin DNases with Im proteins are rotamers suggests a way towards achieving this difficult goal by freezing-out (non-cognate) complexes by disulfide bond engineering. Covalent complexes will be crystallized, their structures solved and analysed. 2. Dissection of the kinetics and thermodynamics of computationally redesigned protein-protein complexes. One of the major advances in this area has come from David Baker (US) whose lab have developed a novel second site suppressor strategy for engineering novel PPI specificities, using colicin DNase-Im protein complexes as the model system. These novel complexes, many of which remain to be characterised, have been provided to us and will be analysed in terms of binding affinity, specificity and energetics. 3. Structural and biophysical analysis of directed evolution-derived PPIs. Directed evolution is used for investigating the evolution of protein function and for generating new functions. The methods that are generally in use however tend only to select for improvements in binding. In vitro compartmentalisation is a relatively new method where other functions such as enzyme inhibition can also be selected. Using this method novel colicin DNase immunity protein variants have been generated by Dan Tawfik's lab (Israel) that have been provided to us. We will conduct biophysical and structural studies on these novel complexes that will provide the ultimate test of the current specificity models we have proposed for these and other divergently evolved protein complexes.


10 25 50
Meenan NA (2010) The structural and energetic basis for high selectivity in a high-affinity protein-protein interaction. in Proceedings of the National Academy of Sciences of the United States of America
Description Objective 1 Structure of a protein-protein interaction encounter complex (Meenan et al (2010) PNAS)
We sought to crystallize and obtain structural information on a weak, transient protein-protein complex using a novel disulfide trapping approach. Such information is rare in the pdb because of the inherent difficulties in stabilising the complex for crystallography in a way that reflects the true solution characteristics of the complex. The strategy was based on our previous work showing that colicin DNases bind their cognate Im proteins with Kds in the femtomolar affinity range but form interactions with non-cognate complexes that are many orders of magnitude weaker (~micromolar). Using NMR-based docking restraints in conjunction with the docking algorithm HADDOCK, we determined optimal positions for cysteine residues that would readily form a trapping disulfide bond across the interface of the colicin E9 DNase-Im2 complex and then solved the crystal structure to 1.77 angstrom resolution. Remarkably, the structure revealed an entirely non-covalent interface where the non-cognate complex became trapped by rotations around the intersubunit disulfide. We speculated in the publication that this approach could be adapted for other weak and transient protein-protein interactions, their dynamic nature being the key to their entrapment within a crystal lattice and hence their structural resolution. It showed the first example of a 'frustrated' protein-protein interface where non-optimal contacts are maintained resulting in destabilisation of the complex. It also highlighted the importance of protein loops and buried water molecules in the fine-tuning of specificity.

Objective 3 Structural and biophysical analysis of directed evolution-derived protein-protein interactions (Levin et al (2009) NSMB)
One of our collaborators on this project was Dan Tawfik at the Weizmann Institute who used colicin DNase-Im complexes as a model system to study how affinity and specificity evolve in protein-protein interactions. Starting with the non-cognate complex of colicin E7 DNase-Im9 (Kd~ 10 nanomolar) his laboratory used In Vitro Compartmentalisation to evolve the Im protein to near cognate levels by using a combination of positive and negative selection. The result was an evolved complex that was over five-orders of magnitude higher affinity and over eight-orders of magnitude improved specificity. The CK lab was involved in biophysically characterising some of the evolutionary intermediates, the resulting paper, which included two structures for some of the latter evolved proteins, providing new insight into how specificity in protein-protein interactions evolves.

Finally, we embarked on a new collaboration with Venki Ramakrishnan's lab at the LMB in Cambridge. Ramakrishnan's nobel prize winning work on the bacterial ribosome, the machine that synthesis all proteins in all cells, prompted us to attempt the structural resolution of the ribosome with the killing domain of colicin E3. This enzyme is a well-known protein antibiotic that selectively cuts a single phosphodiester bond. The Ramakrishnan lab solved the structure of the 70S ribosome-E3 rRNase complex using a range of mutants we provided to them, providing the first molecular understanding of how these cytotoxins inhibit protein synthesis in bacteria.
Exploitation Route The work with the Ramakrishnan lab showed how a protein antibiotic cleaved the bacterial ribosome. This could be developed into novel antibiotics.
Sectors Healthcare,Pharmaceuticals and Medical Biotechnology
Description The focus of this grant was an in-depth analysis of how proteins interact with each other. Cells contain thousands of proteins some promiscuous in their associations, others very specific. This is an important biological problem that has relevance to the next generation of designer protein drugs being developed in the pharmaceutical industry.
Description Design of protein antibiotics 
Organisation The Hebrew University of Jerusalem
Department Weizmann Institute of Science
Country Israel, State of 
Sector Academic/University 
PI Contribution One of the key publications from this work (Meenan et al (2010) PNAS) involved a collaboration with Dr. Sarel Fleishman, then in the laboratory of Prof David Baker (Univ. Washington), on computational design of protein-protein interactions. This initiated a collaboration with Dr Fleishman (now at the Weizmann Institute, Israel) on the design of protein antibiotics. This new project, which I am currently building as a consortium bid to the Wellcome Trust, involves the design of toxins that target Gram-negative pathogens
Collaborator Contribution Protein design using through the Rosetta program
Impact See Meenan, Sharma, Fleishman et al and Kleanthous (2010) PNAS 107, 10080
Start Year 2009