CompAllo: Towards a Unified Approach to the Analysis and Design of Allostery
- Franca Fraternali (King's College London, United Kingdom)
- Jens Kleinjung (Heptares Therapeutics, United Kingdom)
- Matteo Dal Peraro (Swiss Federal Institute of Technology Lausanne (EPFL) , Switzerland)
Registration for this Workshop is now closed!
What is Allostery?
Allostery denotes the feature of biomolecular structures to transfer signals between distant sites. The classical definition of allostery refers to the communication between a functionally active (orthosteric) site and a regulatory allosteric site. Typically, an allosteric ligand binds to the allosteric site and modulates the activity of the orthosteric site.
The allosteric coupling between regulatory and functional sites allows proteins to sense environmental cues, for example metabolic substrates, and to fine-tune their biological activity accordingly. The in vivo capacity of an enzyme to sustain a particular catalytic flux is determined by its concentration and kinetic parameters. Of the two mechanisms of enyzme regulation, gene expression acts only on enzyme abundance, while allostery modulates kinetic and thermodynamic parameters through ligand binding or post-translational modifications to orthosteric sites (Gerosa and Sauer, 2011).
Allosteric regulation is the most direct and efficient mechanism to sense concentration
changes in small molecules and to modulate pathway activity to maintain metabolic homeostasis. This is, in part, because allosteric regulation of enzyme activity is considerably faster than expression-level control. Therefore,allosteric control of protein function is an essential part of protein interaction networks and signalling pathways and it offers opportunities for drug targeting.
State of the Art
The fact that allostery often occurs on fast time scales (ns to ms) and involves changes in local conformations and slow vibrations has hampered our understanding of the underlying mechanisms since the seminal works on allostery between 1963 and 1966 (Changeux, 2012). The following sections outline how recent theoretical and experimental work has enabled us to access the tempo-spatial time scales required to study allostery at atomic resolution. These studies have shown that allostery can be conveyed through multiple pathways across the protein structure, that conformational
selection plays an important role and that we can compute the contributions of individual residues to the allosteric signal transmission. There is a substantial body of literature on allostery. For the purpose of the workshop we limit the citations in the sections below to publications associated with the invited speakers. For a more comprehensive overview on allostery and related subjects, see the following reviews: Kern and Zuiderweg (2003);
Leitner (2008); Cui and Karplus (2008); del Sol et al. (2009); Zhuravlev and Papoian (2010); Bu and Callaway (2011); Peracchi and Mozzarelli (2011); Hilser et al. (2012); Berezovsky (2013); Collier and Ortiz (2013); Nussinov (2013); DeLaBarre et al. (2014); Feher et al. (2014); Motlagh et al. (2014); Raman et al. (2014); Di Paola and Giuliani (2015); Grutsch et al. (2016); Dokholyan (2016); Hertig et al. (2016); Lisi and Loria (2016); Liu and Nussinov (2016); Ribeiro and Ortiz (2016); Tompa (2016); Vega et al. (2016).
The ASD database (Shen et al., 2016) lists 1473 allosteric target proteins, 1930 allosteric sites and 56 allosteric pathways. This set could be regarded as a starting point to create a quantitative and time-resolved compendium of allosteric signal transmission, ideally including mutational effects.
Fast computational methods like normal mode analysis (Guarnera and Berezovsky, 2016) and Gaussian networks (Haliloglu et al., 2010), both using a Cff representation, are suitable for the detection of allostery in large sets of proteins. An atomistic and time-resolved picture of signal transmission can be obtained from long (ns to ~s) Molecular Dynamics (MD) simulations combined with structural alphabets and network analysis (Pandini et al., 2012; Craveur et al., 2015). One central discussion point (see'Aim 1') would be, which resolution and CPU power is needed to achieve a given accuracy in the enthalpic, entropic, conformational and temporal changes that are associated with allosteric signals.
It has been proposed that allostery can spread beyond the limits of protein domains and obligate complexes to neighbours in the allo-network (Karet al., 2010; Nussinov et al., 2011). To predict suitable allosteric ligands, an accurate estimate of the free energy of ligand binding needs to be computed (de Ruiter et al., 2013; Oostenbrink and van Gunsteren, 2005), in particular entropic changes induced by the protein-ligand interactions (Polyanskyet al., 2012a,b). Considering the flexibility of proteins, the relatively small energetic changes involved in allosteric transitions and the fact that water
plays a role in binding processes, it is clear that allosteric ligand or protein design are non-trivial tasks, but methods have been developed to estimate ligand affinities (Pires et al., 2016). We plan to discuss strategies to design allosteric ligands and mutants (see 'Aim 2'), for example by establishing energetic boundary conditions within which one could expect a particular design to be successful.
Allosteric mechanisms can be studied by NMR relaxation experiments and residual dipolar couplings (~s to ms) and interpreted in combination with MD simulations (ns to ~s) (Kern and Zuiderweg, 2003; Salvi et al.,2016). The complementary time scales of NMR and MD provide a solid basis to reliably detect fast and slow allosteric motions. Our computational methods generally do not include interfaces for experimental data and we
will evaluate in the workshop (see 'Aim 3') whether and how NMR and other spectroscopic data can be incorporated into the analysis of simulation trajectories. This would provide an opportunity to study the dependency between the slow motions inferred from NMR and the fast motions sampled by MD. Additionally, joint code development and data sharing will be addressed ('Aim 4') in the workshop.