Triathlon for energy functions: who is the winner for design of protein-protein interactions?

Engineered variants of the Ras effector protein RASSF5 (NORE1A) promote anticancer activities in lung adenocarcinoma

Within the superfamily of small GTPases, Ras appears to be the master regulator of such processes as cell cycle progression, cell division, and apoptosis. Several oncogenic Ras mutations at amino acid positions 12, 13, and 61 have been identified that lose their ability to hydrolyze GTP, giving rise to constitutive signaling and eventually development of cancer. While disruption of the Ras/effector interface is an attractive strategy for drug design to prevent this constitutive activity, inhibition of this interaction using small molecules is impractical due to the absence of a cavity to which such molecules could bind. However, proteins and especially natural Ras effectors that bind to the Ras/effector interface with high affinity could disrupt Ras/effector interactions and abolish procancer pathways initiated by Ras oncogene. Using a combination of computational design and in vitro evolution, we engineered high-affinity Ras-binding proteins starting from a natural Ras effector, RASSF5 (NORE1A), which is encoded by a tumor suppressor gene. Unlike previously reported Ras oncogene inhibitors, the proteins we designed not only inhibit Ras-regulated procancer pathways, but also stimulate anticancer pathways initiated by RASSF5. We show that upon introduction into A549 lung carcinoma cells, the engineered RASSF5 mutants decreased cell viability and mobility to a significantly greater extent than WT RASSF5. In addition, these mutant proteins induce cellular senescence by increasing acetylation and decreasing phosphorylation of p53. In conclusion, engineered RASSF5 variants provide an attractive therapeutic strategy able to oppose cancer development by means of inhibiting of procancer pathways and stimulating anticancer processes.

A KLK4 proteinase substrate capture approach to antagonize PAR1

Proteinase-activated receptor-1 (PAR1), triggered by thrombin and other serine proteinases such as tissue kallikrein-4 (KLK4), is a key driver of inflammation, tumor invasiveness and tumor metastasis. The PAR1 transmembrane G-protein-coupled receptor therefore represents an attractive target for therapeutic inhibitors. We thus used a computational design to develop a new PAR1 antagonist, namely, a catalytically inactive human KLK4 that acts as a proteinase substrate-capture reagent, preventing receptor cleavage (and hence activation) by binding to and occluding the extracellular R41-S42 canonical PAR1 proteolytic activation site. On the basis of in silico site-saturation mutagenesis, we then generated KLK4S207A,L185D, a first-of-a-kind 'decoy' PAR1 inhibitor, by mutating the S207A and L185D residues in wild-type KLK4, which strongly binds to PAR1. KLK4S207A,L185D markedly inhibited PAR1 cleavage, and PAR1-mediated MAPK/ERK activation as well as the migration and invasiveness of melanoma cells. This 'substrate-capturing' KLK4 variant, engineered to bind to PAR1, illustrates proof of principle for the utility of a KLK4 'proteinase substrate capture' approach to regulate proteinase-mediated PAR1 signaling.

EvoEF2: accurate and fast energy function for computational protein design

MOTIVATION

The accuracy and success rate of de novo protein design remain limited, mainly due to the parameter over-fitting of current energy functions and their inability to discriminate incorrect designs from correct designs.

RESULTS

We developed an extended energy function, EvoEF2, for efficient de novo protein sequence design, based on a previously proposed physical energy function, EvoEF. Remarkably, EvoEF2 recovered 32.5%, 47.9% and 22.3% of all, core and surface residues for 148 test monomers, and was generally applicable to protein-protein interaction design, as it recapitulated 30.9%, 42.4%, 31.3% and 21.4% of all, core, interface and surface residues for 88 test dimers, significantly outperforming EvoEF on the native sequence recapitulation. We further used I-TASSER to evaluate the foldability of the 148 designed monomer sequences, where all of them were predicted to fold into structures with high fold- and atomic-level similarity to their corresponding native structures, as demonstrated by the fact that 87.8% of the predicted structures shared a root-mean-square-deviation less than 2 Å to their native counterparts. The study also demonstrated that the usefulness of physical energy functions is highly correlated with the parameter optimization processes, and EvoEF2, with parameters optimized using sequence recapitulation, is more suitable for computational protein sequence design than EvoEF, which was optimized on thermodynamic mutation data.

AVAILABILITY AND IMPLEMENTATION

The source code of EvoEF2 and the benchmark datasets are freely available at https://zhanglab.ccmb.med.umich.edu/EvoEF.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Recent advances in automated protein design and its future challenges

INTRODUCTION

Protein function is determined by protein structure which is in turn determined by the corresponding protein sequence. If the rules that cause a protein to adopt a particular structure are understood, it should be possible to refine or even redefine the function of a protein by working backwards from the desired structure to the sequence. Automated protein design attempts to calculate the effects of mutations computationally with the goal of more radical or complex transformations than are accessible by experimental techniques. Areas covered: The authors give a brief overview of the recent methodological advances in computer-aided protein design, showing how methodological choices affect final design and how automated protein design can be used to address problems considered beyond traditional protein engineering, including the creation of novel protein scaffolds for drug development. Also, the authors address specifically the future challenges in the development of automated protein design. Expert opinion: Automated protein design holds potential as a protein engineering technique, particularly in cases where screening by combinatorial mutagenesis is problematic. Considering solubility and immunogenicity issues, automated protein design is initially more likely to make an impact as a research tool for exploring basic biology in drug discovery than in the design of protein biologics.

Converting a broad matrix metalloproteinase family inhibitor into a specific inhibitor of MMP-9 and MMP-14

MMP-14 and MMP-9 are two well-established cancer targets for which no specific clinically relevant inhibitor is available. Using a powerful combination of computational design and yeast surface display technology, we engineered such an inhibitor starting from a nonspecific MMP inhibitor, N-TIMP2. The engineered purified N-TIMP2 variants showed enhanced specificity toward MMP-14 and MMP-9 relative to a panel of off-target MMPs. MMP-specific N-TIMP2 sequence signatures were obtained that could be understood from the structural perspective of MMP/N-TIMP2 interactions. Our MMP-9 inhibitor exhibited 1000-fold preference for MMP-9 vs. MMP-14, which is likely to translate into significant differences under physiological conditions. Our results provide new insights regarding evolution of promiscuous proteins and optimization strategies for design of inhibitors with single-target specificities.

Identifying Residues that Determine SCF Molecular-Level Interactions through a Combination of Experimental and In silico Analyses

The stem cell factor (SCF)/c-Kit receptor tyrosine kinase complex-with its significant roles in hematopoiesis and angiogenesis-is an attractive target for rational drug design. There is thus a need to map, in detail, the SCF/c-Kit interaction sites and the mechanisms that modulate this interaction. While most residues in the direct SCF/c-Kit binding interface can be identified from the existing crystal structure of the complex, other residues that affect binding through protein unfolding, intermolecular interactions, allosteric or long-distance electrostatic effects cannot be directly inferred. Here, we describe an efficient method for protein-wide epitope mapping using yeast surface display. A library of single SCF mutants that span the SCF sequence was screened for decreased affinity to soluble c-Kit. Sequencing of selected clones allowed the identification of mutations that reduce SCF binding affinity to c-Kit. Moreover, the screening of these SCF clones for binding to a structural antibody helped identify mutations that result in small or large conformational changes in SCF. Computational modeling of the experimentally identified mutations showed that these mutations reduced the binding affinity through one of the three scenarios: through SCF destabilization, through elimination of favorable SCF/c-Kit intermolecular interactions, or through allosteric changes. Eight SCF variants were expressed and purified. Experimentally measured in vitro binding affinities of these mutants to c-Kit confirmed both the yeast surface display selection results and the computational predictions. This study has thus identified the residues crucial for c-Kit/SCF binding and has demonstrated the advantages of using a combination of computational and combinatorial methods for epitope mapping.

Combinatorial and Computational Approaches to Identify Interactions of Macrophage Colony-stimulating Factor (M-CSF) and Its Receptor c-FMS

The molecular interactions between macrophage colony-stimulating factor (M-CSF) and the tyrosine kinase receptor c-FMS play a key role in the immune response, bone metabolism, and the development of some cancers. Because no x-ray structure is available for the human M-CSF · c-FMS complex, the binding epitope for this complex is largely unknown. Our goal was to identify the residues that are essential for binding of the human M-CSF to c-FMS. For this purpose, we used a yeast surface display (YSD) approach. We expressed a combinatorial library of monomeric M-CSF (M-CSFM) single mutants and screened this library to isolate variants with reduced affinity for c-FMS using FACS. Sequencing yielded a number of single M-CSFM variants with mutations both in the direct binding interface and distant from the binding site. In addition, we used computational modeling to map the identified mutations onto the M-CSFM structure and to classify the mutations into three groups as follows: those that significantly decrease protein stability; those that destroy favorable intermolecular interactions; and those that decrease affinity through allosteric effects. To validate the YSD and computational data, M-CSFM and three variants were produced as soluble proteins; their affinity and structure were analyzed; and very good correlations with both YSD data and computational predictions were obtained. By identifying the M-CSFM residues critical for M-CSF · c-FMS interactions, we have laid down the basis for a deeper understanding of the M-CSF · c-FMS signaling mechanism and for the development of target-specific therapeutic agents with the ability to sterically occlude the M-CSF·c-FMS binding interface.

Crucial roles of single residues in binding affinity, specificity, and promiscuity in the cellulosomal cohesin-dockerin interface

Interactions between cohesin and dockerin modules play a crucial role in the assembly of multienzyme cellulosome complexes. Although intraspecies cohesin and dockerin modules bind in general with high affinity but indiscriminately, cross-species binding is rare. Here, we combined ELISA-based experiments with Rosetta-based computational design to evaluate the contribution of distinct residues at the Clostridium thermocellum cohesin-dockerin interface to binding affinity, specificity, and promiscuity. We found that single mutations can show distinct and significant effects on binding affinity and specificity. In particular, mutations at cohesin position Asn(37) show dramatic variability in their effect on dockerin binding affinity and specificity: the N37A mutant binds promiscuously both to cognate (C. thermocellum) as well as to non-cognate Clostridium cellulolyticum dockerin. N37L in turn switches binding specificity: compared with the wild-type C. thermocellum cohesin, this mutant shows significantly increased preference for C. cellulolyticum dockerin combined with strongly reduced binding to its cognate C. thermocellum dockerin. The observation that a single mutation can overcome the naturally observed specificity barrier provides insights into the evolutionary dynamics of this system that allows rapid modulation of binding specificity within a high affinity background.

Inferring the microscopic surface energy of protein-protein interfaces from mutation data

Mutations at protein-protein recognition sites alter binding strength by altering the chemical nature of the interacting surfaces. We present a simple surface energy model, parameterized with empirical ΔΔG values, yielding mean energies of -48 cal mol(-1) Å(-2) for interactions between hydrophobic surfaces, -51 to -80 cal mol(-1) Å(-2) for surfaces of complementary charge, and 66-83 cal mol(-1) Å(-2) for electrostatically repelling surfaces, relative to the aqueous phase. This places the mean energy of hydrophobic surface burial at -24 cal mol(-1) Å(-2) . Despite neglecting configurational entropy and intramolecular changes, the model correlates with empirical binding free energies of a functionally diverse set of rigid-body interactions (r = 0.66). When used to rerank docking poses, it can place near-native solutions in the top 10 for 37% of the complexes evaluated, and 82% in the top 100. The method shows that hydrophobic burial is the driving force for protein association, accounting for 50-95% of the cohesive energy. The model is available open-source from http://life.bsc.es/pid/web/surface_energy/ and via the CCharpPPI web server http://life.bsc.es/pid/ccharppi/.

Affinity- and specificity-enhancing mutations are frequent in multispecific interactions between TIMP2 and MMPs

Multispecific proteins play a major role in controlling various functions such as signaling, regulation of transcription/translation, and immune response. Hence, a thorough understanding of the atomic-level principles governing multispecific interactions is important not only for the advancement of basic science but also for applied research such as drug design. Here, we study evolution of an exemplary multispecific protein, a Tissue Inhibitor of Matrix Metalloproteinases 2 (TIMP2) that binds with comparable affinities to more than twenty-six members of the Matrix Metalloproteinase (MMP) and the related ADAMs families. We postulate that due to its multispecific nature, TIMP2 is not optimized to bind to any individual MMP type, but rather embodies a compromise required for interactions with all MMPs. To explore this hypothesis, we perform computational saturation mutagenesis of the TIMP2 binding interface and predict changes in free energy of binding to eight MMP targets. Computational results reveal the non-optimality of the TIMP2 binding interface for all studied proteins, identifying many affinity-enhancing mutations at multiple positions. Several TIMP2 point mutants predicted to enhance binding affinity and/or binding specificity towards MMP14 were selected for experimental verification. Experimental results show high abundance of affinity-enhancing mutations in TIMP2, with some point mutations producing more than ten-fold improvement in affinity to MMP14. Our computational and experimental results collaboratively demonstrate that the TIMP2 sequence lies far from the fitness maximum when interacting with its target enzymes. This non-optimality of the binding interface and high potential for improvement might characterize all proteins evolved for binding to multiple targets.

Predicting affinity- and specificity-enhancing mutations at protein-protein interfaces

Manipulations of PPIs (protein-protein interactions) are important for many biological applications such as synthetic biology and drug design. Combinatorial methods have been traditionally used for such manipulations, failing, however, to explain the effects achieved. We developed a computational method for prediction of changes in free energy of binding due to mutation that bring about deeper understanding of the molecular forces underlying binding interactions. Our method could be used for computational scanning of binding interfaces and subsequent analysis of the interfacial sequence optimality. The computational method was validated in two biological systems. Computational saturated mutagenesis of a high-affinity complex between an enzyme AChE (acetylcholinesterase) and a snake toxin Fas (fasciculin) revealed the optimal nature of this interface with only a few predicted affinity-enhancing mutations. Binding measurements confirmed high optimality of this interface and identified a few mutations that could further improve interaction fitness. Computational interface scanning of a medium-affinity complex between TIMP-2 (tissue inhibitor of metalloproteinases-2) and MMP (matrix metalloproteinase) 14 revealed a non-optimal nature of the binding interface with multiple mutations predicted to stabilize the complex. Experimental results corroborated our computational predictions, identifying a large number of mutations that improve the binding affinity for this interaction and some mutations that enhance binding specificity. Overall, our computational protocol greatly facilitates the discovery of affinity- and specificity-enhancing mutations and thus could be applied for design of potent and highly specific inhibitors of any PPI.

Mapping of the binding landscape for a picomolar protein-protein complex through computation and experiment

Our understanding of protein evolution would greatly benefit from mapping of binding landscapes, i.e., changes in protein-protein binding affinity due to all single mutations. However, experimental generation of such landscapes is a tedious task due to a large number of possible mutations. Here, we use a simple computational protocol to map the binding landscape for two homologous high-affinity complexes, involving a snake toxin fasciculin and acetylcholinesterase from two different species. To verify our computational predictions, we experimentally measure binding between 25 Fas mutants and the 2 enzymes. Both computational and experimental results demonstrate that the Fas sequence is close to the optimum when interacting with its targets, yet a few mutations could further improve Kd, kon, and koff. Our computational predictions agree well with experimental results and generate distributions similar to those observed in other high-affinity PPIs, demonstrating the potential of simple computational protocols in capturing realistic binding landscapes.

Predicting the Impact of Missense Mutations on Protein-Protein Binding Affinity

The crucial prerequisite for proper biological function is the protein's ability to establish highly selective interactions with macromolecular partners. A missense mutation that alters the protein binding affinity may cause significant perturbations or complete abolishment of the function, potentially leading to diseases. The availability of computational methods to evaluate the impact of mutations on protein-protein binding is critical for a wide range of biomedical applications. Here, we report an efficient computational approach for predicting the effect of single and multiple missense mutations on protein-protein binding affinity. It is based on a well-tested simulation protocol for structure minimization, modified MM-PBSA and statistical scoring energy functions with parameters optimized on experimental sets of several thousands of mutations. Our simulation protocol yields very good agreement between predicted and experimental values with Pearson correlation coefficients of 0.69 and 0.63 and root-mean-square errors of 1.20 and 1.90 kcal mol^-1 for single and multiple mutations, respectively. Compared with other available methods, our approach achieves high speed and prediction accuracy and can be applied to large datasets generated by modern genomics initiatives. In addition, we report a crucial role of water model and the polar solvation energy in estimating the changes in binding affinity. Our analysis also reveals that prediction accuracy and effect of mutations on binding strongly depends on the type of mutation and its location in a protein complex.

Increasing affinity of interferon-γ receptor 1 to interferon-γ by computer-aided design

We describe a computer-based protocol to design protein mutations increasing binding affinity between ligand and its receptor. The method was applied to mutate interferon-γ receptor 1 (IFN-γ-Rx) to increase its affinity to natural ligand IFN-γ, protein important for innate immunity. We analyzed all four available crystal structures of the IFN-γ-Rx/IFN-γ complex to identify 40 receptor residues forming the interface with IFN-γ. For these 40 residues, we performed computational mutation analysis by substituting each of the interface receptor residues by the remaining standard amino acids. The corresponding changes of the free energy were calculated by a protocol consisting of FoldX and molecular dynamics calculations. Based on the computed changes of the free energy and on sequence conservation criteria obtained by the analysis of 32 receptor sequences from 19 different species, we selected 14 receptor variants predicted to increase the receptor affinity to IFN-γ. These variants were expressed as recombinant proteins in Escherichia coli, and their affinities to IFN-γ were determined experimentally by surface plasmon resonance (SPR). The SPR measurements showed that the simple computational protocol succeeded in finding two receptor variants with affinity to IFN-γ increased about fivefold compared to the wild-type receptor.

Designing specific protein-protein interactions using computation, experimental library screening, or integrated methods

Given the importance of protein-protein interactions for nearly all biological processes, the design of protein affinity reagents for use in research, diagnosis or therapy is an important endeavor. Engineered proteins would ideally have high specificities for their intended targets, but achieving interaction specificity by design can be challenging. There are two major approaches to protein design or redesign. Most commonly, proteins and peptides are engineered using experimental library screening and/or in vitro evolution. An alternative approach involves using protein structure and computational modeling to rationally choose sequences predicted to have desirable properties. Computational design has successfully produced novel proteins with enhanced stability, desired interactions and enzymatic function. Here we review the strengths and limitations of experimental library screening and computational structure-based design, giving examples where these methods have been applied to designing protein interaction specificity. We highlight recent studies that demonstrate strategies for combining computational modeling with library screening. The computational methods provide focused libraries predicted to be enriched in sequences with the properties of interest. Such integrated approaches represent a promising way to increase the efficiency of protein design and to engineer complex functionality such as interaction specificity.

Design of native-like proteins through an exposure-dependent environment potential

We hypothesize that the degree of surface exposure of amino acid side chains within a globular, soluble protein has been optimized in evolution, not only to minimize the solvation free energy of the monomeric protein but also to prevent protein aggregation. This effect needs to be taken into account when engineering proteins de novo. We test this hypothesis through addition of a knowledge-based, exposure-dependent energy term to the RosettaDesign solvation potential [Lazaridis, T., and Karplus, M. (1999) Proteins 35, 133-152]. Correlation between amino acid type and surface exposure is determined from a representative set of experimental protein structures. The amino acid solvent accessible surface area (SASA) is estimated with a neighbor vector measure that increases in accuracy compared to the neighbor count measure while remaining pairwise decomposable [Durham, E., et al. (2009) J. Mol. Model. 15, 1093-1108]. Benchmarking of this potential in protein design displays a 3.2% improvement in the overall sequence recovery and an 8.5% improvement in recovery of amino acid types tolerated in evolution.

Hotspot-centric de novo design of protein binders

Protein-protein interactions play critical roles in biology, and computational design of interactions could be useful in a range of applications. We describe in detail a general approach to de novo design of protein interactions based on computed, energetically optimized interaction hotspots, which was recently used to produce high-affinity binders of influenza hemagglutinin. We present several alternative approaches to identify and build the key hotspot interactions within both core secondary structural elements and variable loop regions and evaluate the method's performance in natural-interface recapitulation. We show that the method generates binding surfaces that are more conformationally restricted than previous design methods, reducing opportunities for off-target interactions.