Journey to the center of the protein: allostery from multitemperature multiconformer X-ray crystallography

Abstract

Proteins inherently fluctuate between conformations to perform functions in the cell. For example, they sample product-binding, transition-state-stabilizing and product-release states during catalysis, and they integrate signals from remote regions of the structure for allosteric regulation. However, there is a lack of understanding of how these dynamic processes occur at the basic atomic level. This gap can be at least partially addressed by combining variable-temperature (instead of traditional cryogenic temperature) X-ray crystallography with algorithms for modeling alternative conformations based on electron-density maps, in an approach called multitemperature multiconformer X-ray crystallography (MMX). Here, the use of MMX to reveal alternative conformations at different sites in a protein structure and to estimate the degree of energetic coupling between them is discussed. These insights can suggest testable hypotheses about allosteric mechanisms. Temperature is an easily manipulated experimental parameter, so the MMX approach is widely applicable to any protein that yields well diffracting crystals. Moreover, the general principles of MMX are extensible to other perturbations such as pH, pressure, ligand concentration etc. Future work will explore strategies for leveraging X-ray data across such perturbation series to more quantitatively measure how different parts of a protein structure are coupled to each other, and the consequences thereof for allostery and other aspects of protein function.

Keywords: allostery; conformational heterogeneity; multiconformer modeling; multitemperature crystallography; protein flexibility.

Figure 1

Protein energy landscapes and allostery. (a) A generic framework for allostery is that the global energy landscape of the protein is altered by an allosteric effector. Here, the energy landscape is schematized as a plot of free energy versus an arbitrary collective conformational coordinate. An allosteric effector, here a small-molecule ligand binding at an allosteric site, modulates the energy landscape, which changes the conformation of the active site (*), thus altering the function of the protein. (b) However, the portrait in (a) is agnostic to the mechanisms by which the local energy landscapes of specific regions of a protein structure respond to the allosteric effector and to each other. It therefore remains unclear how the allosteric signal propagates from the allosteric site through the tertiary structure to the functional site. Although this propagation may be branching rather than linear as depicted schematically here, it must ultimately have a physical mechanistic basis that can be understood in structural terms.

Figure 2

Multitemperature multiconformer X-ray crystallography (MMX) for predicting allostery in protein structures. (a) MMX provides a way to infer how local regions of a protein structure mechanically couple to each other to facilitate allostery, as illustrated here schematically for a dynamic enzyme. At low temperature (e.g. 100 K), the active-site loop (top) and several residues linking the active site to a distal allosteric site (middle to bottom right) adopt a particular alternative conformation (blue) with higher probability or occupancy (thicker lines). As the temperature is increased (e.g. to >200 K), all of these regions concertedly shift their conformational ensemble to include increased populations of a different alternative conformation (red). This coupled behavior does not definitively prove, but is consistent with, the hypothesis that these regions are energetically coupled to each other and thereby form part of an interdependent allosteric network. By contrast, a different residue (bottom left) remains in a single conformation (purple) that is independent of temperature and thus is unresponsive to the other allosterically linked regions. The bottom-right binding site is therefore more likely to be capable of allosteric signaling to the active site than is the bottom-left binding site. (b) A molecular perturbation such as a small molecule (green) can test the hypothesis that different parts of the allosteric network are energetically coupled and that biasing the conformation of one part of the network biases the conformations of other parts. Artificial small molecules may compete with natural protein–protein interactions that play regulatory roles in cells (cyan). In addition, mutations (orange) may interfere with the energetic coupling between residues within the network. Thus, these other types of perturbations may equally well be used to interrogate allosteric networks that are predicted using MMX-based approaches.

Figure 3

‘Hidden’ alternative conformations in natural and artificial protein structures. (a) Natural protein: residues Asn173 and Arg464 in a 0.88 Å resolution structure of catalase (PDB entry 1gwe; Murshudov et al., 2002 ▸) are each modeled with a single conformation. However, 2F _o − F _c (0.7σ as a light blue volume and blue mesh) and ±F _o − F _c (+3.5σ and −3.5σ in green and red, respectively; both volume and mesh) electron-density maps suggest that the existing Arg464 conformation is overmodeled and reveal evidence for a ‘hidden’ alternative conformation. Supporting this interpretation, the existing Arg464 conformation sterically clashes (red/orange/yellow spikes; Word, Lovell, LaBean et al., 1999 ▸) with several waters (red spheres) that were mistakenly modeled into that electron density. (b) A refitted and rerefined model, with the Asn173 side-chain amide flipped 180° (curved dotted arrow; Word, Lovell, Richardson et al., 1999 ▸), an alternative rotamer added for Arg464 (purple versus orange), the offending waters removed and alternative water positions that are mutually exclusive with the original Arg464 conformation added, results in a better fit to the electron density, including diminished F _o − F _c difference peaks, elimination of steric clashes and a more extensive hydrogen-bonding network (green dotted lines). Some additional partial-occupancy waters may also be present, given the remaining positive F _o − F _c density. (c) Artificial protein: residues Arg104 and Gln105 in chain B of a 2.09 Å resolution structure of a de novo designed protein (PDB entry 5e6g; Jacobs et al., 2016 ▸) are modeled with single conformations. However, 2F _o − F _c (0.7σ as a light blue volume and blue mesh) and ± F _o − F _c electron-density maps (+3.0σ and −3.0σ in green and red, respectively; both volume and mesh) reveal evidence for a ‘hidden’ alternative conformation for Arg104 and a missing partial-occupancy ordered water molecule nearby that were not specified in the design model (arrows). (d) A refitted and rerefined model, with alternative conformations (purple versus orange) for Arg104 and the partial-occupancy water added, results in a better fit to the electron density, including diminished F _o − F _c difference peaks and a more extensive hydrogen-bonding network (green dotted lines). Images were obtained using PyMOL (Schrödinger).

Figure 4

Building multiconformer models by cross-pollinating conformations. (a) In a single-conformer version of a multiconformer model for a room-temperature apo (278 K) structure of PTP1B (PDB entry 6b8x; Keedy et al., 2018 ▸), ‘loop 16’ fits the 2F _o − F _c (1.25σ as a cyan volume and blue mesh) electron-density map well, but significant positive F _o − F _c (+3.0σ and −3.0σ in green and red, respectively; both volume and mesh) peaks remain. It is difficult to visually guess the conformational change that would relate the single-conformer model to the difference density. (b) Loop 16 in another structure of PTP1B, at cryogenic temperature with a ligand bound elsewhere (PDB entry 1t49; Wiesmann et al., 2004 ▸), easily explains the difference density, allowing one to combine these states into a multiconformer model (PDB entry 6b8x; Keedy et al., 2018 ▸). Images were obtained using PyMOL (Schrödinger).

Figure 5

Outline of the intended MMX approach for identifying coupled conformational motions. This manuscript discusses a new paradigm in structural biology: multitemperature multiconformer crystallography (MMX). Future approaches based on MMX will identify residues whose conformational ensembles change concertedly with respect to temperature, which could predict energetically coupled residues that are key to allosteric communication through a protein structure. (a) To ensure that future MMX-based approaches compare related data sets in an unbiased way, it will be important to build a sufficiently complete multiconformer model at each temperature. This may be improved by ‘cross-pollinating’ conformers between models at different temperatures. Some of the occupancies of these conformers may refine to low but appreciable values, which will aid in identifying coordinated changes in mixtures of states (Smith et al., 2015 ▸). (b) Conformational changes will be monitored by changes in the electron-density map or refined occupancies as a function of temperature. In the schematic example depicted here, the side chains of two residues on adjacent helices in the tertiary structure have mutually exclusive conformations, and the helix–helix interface reconfigures as the populations of the side chains shift from one collective state to another with temperature. Similar analyses could also be performed with other experimental perturbations such as humidity, pH, pressure, ligand concentration etc. in future MMX experiments. (c) To capture the more complex conformational transitions involving subtle distributed backbone motions that occur in proteins (Deis et al., 2014 ▸), the principles of MMX can be used to superpose maps in real space based on models (Pearce, Krojer, Bradley et al., 2017 ▸) and to examine not just arbitrary volumes of space, but rather structural elements that may move as a cooperative unit – for example, the volume around (dotted rectangle) an α-helix whose conformational ensemble shifts from ordered to quasi-disordered (semi-transparent rectangle), or β-sheets, loops and other ‘fragments’ that compose protein structure (Rohl et al., 2004 ▸).

Figure 6

A complex network of coupled conformational heterogeneity. A network of alternative conformations in a cryogenic structure of catalase (PDB entry 1gwe; Murshudov et al., 2002 ▸) with diverse properties. Multiple phenomena define the network: van der Waals interactions (blue dots and line segments) between side chains, a hydrogen bond (dotted green line) through a partial-occupancy water (brown), coupling through the locally mobile backbone (black) and perhaps electrostatic forces between the Lys (green) and nearby polar residues (Glu in blue, Asp in yellow and Ser in purple). This image was obtained using KiNG (Chen et al., 2009 ▸).