Design and engineering of allosteric communications in proteins

Abstract

Allostery in proteins plays an important role in regulating protein activities and influencing many biological processes such as gene expression, enzyme catalysis, and cell signaling. The process of allostery takes place when a signal detected at a site on a protein is transmitted via a mechanical pathway to a functional site and, thus, influences its activity. The pathway of allosteric communication consists of amino acids that form a network with covalent and non-covalent bonds. By mutating residues in this allosteric network, protein engineers have successfully established novel allosteric pathways to achieve desired properties in the target protein. In this review, we highlight the most recent and state-of-the-art techniques for allosteric communication engineering. We also discuss the challenges that need to be overcome and future directions for engineering protein allostery.

Introduction

Allostery is a common phenomenon in biological macromolecules, where distant parts of a molecule are energetically coupled to produce a functional response [1–3]. The original view of allostery is based on conformational change of the functional site of a protein when perturbation is introduced at a distal site (allosteric site) [4–6]. Cooper and Dryden [7] further proposed a dynamics-based model of allostery, in which the active and inactive states are structurally similar but the dynamics of the active site is changed. Their model extends the original theory of allostery as it shifts the model of two distinguishable states to probabilistic description of ensembles that represent these states. Using the ensemble-based theory, allostery can be formulated regarding the conformational free energies of each domain in a protein and the negative or positive coupling interactions between them [8,9]. In addition to the above-mentioned mechanisms of allostery, Banerjee-Ghosh et al. [10] reported a novel allosteric mechanism based on charge. They demonstrated that charge injection at the distal site redistributed the charge within the antibody, which affected its interaction with the antigen [10].

Allosteric communication plays a pivotal role in various essential biological processes including gene regulation, cell signaling, and metabolism [11]. Genetic mutations that disrupt allosteric communications within a protein may lead to a loss of function and, in some cases, to disease. For instance, mutations in the S-loop of human glutathione synthetase that affect allosteric communication disrupt normal biological processes, resulting in hemolytic anemia, metabolic acidosis, premature death, and neurological disorders [12]. Presenilin 1 catalyzes the cleavage of transmembrane proteins including the amyloid precursor protein [13]. Genetic mutations in presenilin 1 that alter allosteric communication can alter the cleavage pattern and lead to aberrant β-amyloid peptides that have been linked to Alzheimer’s disease [13]. Therefore, understanding allosteric communications and the change in protein function caused by perturbations is critical for clarifying disease etiology and developing effective therapeutics.

Understanding of allosteric communication makes it feasible to design and engineer novel allosteric networks to serve desired purposes. In the past two years, scientists have developed innovative methods to engineer allosteric networks to improve enzyme activity, confer biosensors with novel ligand binding, and construct complex regulatory systems for synthetic biology [14]. In this review, we summarize recent advances in the design and engineering of allosteric communications by addressing the following three topics: (i) the experimental and computational approaches that we can use to elucidate allosteric communication networks, (ii) how do we use this information to engineer hybrid long-range allosteric routes, (iii) the general methods developed to design alternative allosteric communications.

The state-of-the-art methods developed to map allosteric communication pathways

To rationally design and engineer allosteric communications, a prerequisite is to decode the signal transduction pathways and critical residues within the pathway for individual proteins. In the past two years, we have seen great advances in experimental, computational, and integrative methods to meet this challenge.

Site-directed mutagenesis is a commonly used experimental approach to identify essential residues for allosteric communication [15]. Mutations that disrupt allosteric signaling pathways can affect protein structure and function, which can be distinguished from mutated residues that are not important for allosteric signaling. Based on this, Leander et al. [15] used deep mutational scanning, a systematic mutagenesis method, to study the functional landscape of allostery. Solution NMR is another popular experimental method applied to identify residues on the allosteric communication pathway. Residues with changes in chemical shift when the signaling pathway is perturbed are potential targets. Rennella et al. [16] applied methyl-transverse relaxation optimized spectroscopy (TROSY)-based NMR to investigate an extensive allosteric communication pathway initiating from the end of the barrel-like structure to the catalytic center of the proteasome. This methodology can be used to dissect allosteric communication pathways in homo-oligomeric complexes with high molecular weight [16]. Solution NMR has been combined with other experimental methods to enhance the discovery of allosteric communications. Through the combination of NMR and surface plasmon resonance, Kohler et al. [17] demonstrated that ligands with subtle changes in structure share the same allosteric pathways, while ligands with distinct pharmacophores rewire the allosteric communication network.

Compared to experimental methods, computational approaches used to investigate allosteric communications can be more time-efficient and cost-effective. Molecular dynamics (MD) simulations are powerful tools for studying allosteric communications in biological systems [18], such as CRISPR-Cas9 [19], tyrosine phosphatase [20], HBV Capsid [21], β-lactamases [22], and G protein-coupled receptor (GPCR) [23]. Besides MD simulations, another group of computational methods based on spectral graph or information theory has been developed [24]. In this method, the three-dimensional structure of a protein is converted to a network whereby each node represents an amino acid and the edge indicates a covalent or non-covalent bond between residues [25]. Combining this network modeling approach with physics and perturbation propagation algorithm, Wang et al. developed Ohm [26], a comprehensive and user-friendly platform for allosteric communication analysis. This innovative approach relies solely on protein structure and minimizes the calculation time, which is more rapid and cost-effective than MD simulations. Westerlund et al. [27] extended the framework of residue interacting network to include lipids and small molecules, which takes into account the effect of lipids on membrane protein allostery. Alfayate et al. [28] established a torsional network model to analyze dynamic couplings between amino acids in a protein. The dynamic couplings dissect a protein into different domains with correlated motion and the couplings between these domains can be applied to identify potential allosteric sites [28]. The Berezovsky group developed a structure-based statistical mechanical model of allostery to study allosteric communication between regulatory and functional sites [29]. Based on this model, they created the AlloMAPS database that can be used to analyze the causality and energetics of allosteric signaling, access the allosteric effects of perturbations, and evaluate the modulatory effects of new allosteric mutations and sites [29]. Recently, Schupfner et al. [30] developed an ancestral sequence reconstruction method that resurrects the highly probable sequences of extinct proteins based on a phylogenetic tree and multiple sequence alignment of modern proteins. Using this method and further sequence comparison, they identified four residues that contributed to the allosteric signal transduction within the tryptophan synthase complex.

Using experimental and computational methods, integrative methods have been developed that combine the two to illustrate allosteric communications. Subramanian et al. [31] revealed critical residues in the allosteric signal transduction pathway by deep mutagenesis and MD simulations. Ni et al. [32] applied MD simulations, Markov state models, and site-directed mutagenesis to identify a novel cryptic allosteric site of nicotinamide dinucleotide (NAD+)-dependent protein lysine deacetylase sirtuin 6. Sztain et al. [33] combined solution NMR spectroscopy and MD to decode the communication mechanism between acyl carrier protein and its corresponding partner enzyme. They elucidated an allosteric communication network in which the binding of the substrate within the four-helical bundle of the acyl carrier protein causes conformational changes to the exterior, and thus determined its interaction with a specific partner enzyme [33]. Using solution NMR and Gaussian-accelerated MD simulations, East et al. [34] discovered a signal transduction pathway within the CRISPR-Cas9 HNH domain that propagates from the region interacting with the RuvC nuclease to the DNA recognition site.

Construction of hybrid and long-range allosteric networks

The allosteric pathways identified in two different proteins can be rationally linked to form a long-range allosteric pathway. In this hybrid system, perturbation introduced (signal detected) at the ligand-binding site of protein A can be propagated to the functional site of protein B and then affect its activity. By engineering the hybrid allosteric network, we can generate diverse combinations of signal detection and functional outputs, where various signals can be used to manipulate different functions. This application is significant as it provides more options for synthetic biologists and makes regulation more sophisticated.

Modular design is a general strategy to create hybrid proteins and allosteric communication pathways [14]. It divides a system into smaller intact parts called modules that can be exchanged and fused with other modules to create a hybrid system. By coupling various allosteric modules and connecting different allosteric communication pathways, we can establish a synthetic allosteric communication network (Figure 1a). A candidate that has been exploited for modular design is the transcriptional repressor LacI, which consists of a DNA binding domain that interacts with specific promoters and a ligand sensing module that detects environmental signals. The ligand interaction at the environmental sensing module allosterically regulates the DNA binding with the promoter. However, not all possible combinations of DNA and ligand-binding domains will generate hybrid repressors that are functional. To address this problem and facilitate the rewiring of allosteric networks, Dimas et al. [35] developed a computational model based on coevolution theory. They propose that a mutation in the ligand binding region is coupled with a specific mutation in the DNA binding domain if these two domains are structurally or functionally related. Using direct coupling analysis, this model identifies strong coupling pairs that can result in functional hybrid repressors [35]. The generalizability of this modular design strategy has been validated by several studies and described in this review [14]. Recently, Chen et al. [36] designed de novo protein modules that can be fused to create a heterodimer (e.g. A-B) endowed with interaction cooperativity. Specifically, the interaction of subunit A (within the heterodimer A-B) with monomer A’ induces the binding between subunit B and monomer B’, which might be attributed to the hybrid allosteric network constructed by fusing A and B. This ground-breaking research demonstrated the possibility to engineer de novo allosteric communication pathways in proteins.

Figure 1.

Construction of hybrid and long-range allosteric pathways. (a) Schematic representation of the modular design strategy. Synthetic allosteric communications are established by connecting signaling pathways from different modules. (b) Design long-range allosteric networks to control the functions of proteins of interest (POI) using chemical or light responsive sensor domains. Black dots represent critical residues in the allosteric pathway and white lines indicate covalent or non-covalent interactions.

Modular design is limited to proteins with multiple distinct domains (building blocks) that can be separated and functional. Over the last decade, a more universal approach has been developed to engineer long-range allosteric networks in most proteins of interest [37–40]. In this method, a chemical or light-sensitive sensor domain is inserted at an allosteric site of the target protein using molecular cloning techniques [41,42]. Through the connected allosteric pathway, the signal detected by the sensor domain can be propagated to the functional site of the target protein and thus influence its function (Figure 1b). Gil et al. [41] designed opto-nanobodies by inserting a photoswitchable light-oxygen-voltage (LOV) domain at a solvent-exposed loop of the nanobodies. The LOV domain recognizes blue light and leads to structural perturbations that are further transmitted to the binding surface of nanobodies and then interfere with the recognition of a target protein. The reversible control of binding between nanobody and its target has many applications such as modulating intracellular signaling and controlling protein binding in cells. Recently, the Dokholyan’s group [43] engineered a complex allosteric communication network by inserting two sensor domains, a rapamycin responsive uniRapR domain and a blue light sensitive LOV2 domain in the focal adhesion kinase (FAK) (Figure 1c). By connecting allosteric pathways between the sensor domains and FAK, they built a ‘two-input logic OR gate’, where the inputs rapamycin and blue light independently control the kinase activity of FAK. This innovative work demonstrated that two allosteric communication networks can exist in one protein design to control its activity without crosstalk.

Engineering of alternative allosteric communication pathways

Allosteric communications are highly adaptable during evolution as one mutation that interrupts the original signaling pathway can be functionally compensated by another or a few mutations that create an alternative allosteric communication [15]. The cutting-edge methods used to confer potential allosteric networks include directed evolution and rational design.

Directed evolution utilizes the principle of natural selection by imposing stringent rules to screen targets with desired functions from a library of randomly mutated proteins (Figure 2a) [44]. Directed evolution does not require structural information of proteins of interest. However, a high-throughput screening methodology is necessary for efficient protein design. Using a cell growth-based assay, D’Amico et al. [45] identified a mutant involved in the allosteric network that enhances tryptophan synthase function. Substitution of this residue in the alpha subunit of tryptophan synthase allosterically affects the opening of the indole channel and stimulates the activity of the beta subunit [45]. The Wilson lab extensively engineered the allosteric communications in the LacI system based on directed evolution [14,46]. For example, to confer alternate ligand binding in the PurR scaffold, Rondon et al. [47] blocked the original allosteric communication pathway and introduced additional mutations through error-prone PCR in order to reconstruct alternate allosteric routes with alternative ligand recognition [47].

Figure 2.

Directed evolution and rational design of alternative allosteric communication pathways. (a) The process of directed evolution in engineering alternative allosteric communications. (b) The rational design steps in designing alternate allosteric communication pathways in GPCRs. Black dots and white lines represent essential residues and interactions within the allosteric network. Red dots indicate mutated residues.

The rational design of allosteric communication requires a thorough understanding of the target protein in terms of structure, function, and the signal transition pathway. To preserve allosteric regulation during evolution, allosteric networks are expected to be conserved and residues on these networks have coevolved [48]. Based on statistical coupling analysis, Jiao et al. [48] identified coevolved residues that may form important communication networks in the enzyme, 3-deoxy-D-arabino-heptulosonate 7-phosphate synthase. Through site-directed mutagenesis of a critical node within this network, they successfully blocked allosteric communication and constitutively activated the enzyme function [48]. To rationally engineer allosteric communications, Campitelli et al. [49] calculated the effects of residue substitution on flexibilities of the functional surface (the DNA-binding domain in LacI repressor), and the dynamic coupling between the mutation site and the DNA-binding domain. Based on the change in flexibility and dynamic coupling, they successfully rewired the allosteric signaling pathways to adjust DNA binding affinity by substituting non-conserved residues in the linker region of the LacI repressor [49]. To engineer allosteric functions, Chen et al. [50] developed a general approach to reprogram allosteric communications in membrane receptors, such as GPCRs (Figure 2b). First, this method predicts GPCR conformations in the active (ligand-bound) and inactive (ligand-free) states using homology modeling. Allosteric sites and communication pathways are identified through MD simulations. Then residues that mediate allosteric signaling in the transmembrane region are mutated in silico to all possible 20 amino acids. Mutations that shift conformational stability or structural coupling between allosteric sites are selected. For instance, proteins with enhanced constitutive activity can be designed by mutating residues that stabilize the active state rather than the inactive ligand-free state [50].

Conclusions

Genetic variants that disrupt allosteric communication can be detrimental, while rationally designed or directed evolved allosteric networks provide many opportunities for protein engineering and synthetic biology. Over the past two years, we have seen great strides in the identification of allosteric communications in various biological systems and the development of new approaches to the design and engineering of allosteric networks. However, a more general and effective design method, as well as the de novo design of allosteric communications, are still challenges to be addressed in the future. We highlight computationally-guided methods as they are effective means of designing and engineering allosteric communications, surpassing sequence-based deep mutagenesis in terms of efficiency, speed, and success rate. One major limitation to consider when designing allosteric communications is preserving the structural integrity of the target protein. Evolutionary conserved residues can be important in maintaining structural integrity, while non-conserved residues in the allosteric network may be targeted to reprogram allosteric communications.