Predictive Bcl-2 family binding models rooted in experiment or structure

DDMut-PPI: predicting effects of mutations on protein-protein interactions using graph-based deep learning

Protein-protein interactions (PPIs) play a vital role in cellular functions and are essential for therapeutic development and understanding diseases. However, current predictive tools often struggle to balance efficiency and precision in predicting the effects of mutations on these complex interactions. To address this, we present DDMut-PPI, a deep learning model that efficiently and accurately predicts changes in PPI binding free energy upon single and multiple point mutations. Building on the robust Siamese network architecture with graph-based signatures from our prior work, DDMut, the DDMut-PPI model was enhanced with a graph convolutional network operated on the protein interaction interface. We used residue-specific embeddings from ProtT5 protein language model as node features, and a variety of molecular interactions as edge features. By integrating evolutionary context with spatial information, this framework enables DDMut-PPI to achieve a robust Pearson correlation of up to 0.75 (root mean squared error: 1.33 kcal/mol) in our evaluations, outperforming most existing methods. Importantly, the model demonstrated consistent performance across mutations that increase or decrease binding affinity. DDMut-PPI offers a significant advancement in the field and will serve as a valuable tool for researchers probing the complexities of protein interactions. DDMut-PPI is freely available as a web server and an application programming interface at https://biosig.lab.uq.edu.au/ddmut_ppi.

Epistasis facilitates functional evolution in an ancient transcription factor

A protein's genetic architecture - the set of causal rules by which its sequence produces its functions - also determines its possible evolutionary trajectories. Prior research has proposed that the genetic architecture of proteins is very complex, with pervasive epistatic interactions that constrain evolution and make function difficult to predict from sequence. Most of this work has analyzed only the direct paths between two proteins of interest - excluding the vast majority of possible genotypes and evolutionary trajectories - and has considered only a single protein function, leaving unaddressed the genetic architecture of functional specificity and its impact on the evolution of new functions. Here, we develop a new method based on ordinal logistic regression to directly characterize the global genetic determinants of multiple protein functions from 20-state combinatorial deep mutational scanning (DMS) experiments. We use it to dissect the genetic architecture and evolution of a transcription factor's specificity for DNA, using data from a combinatorial DMS of an ancient steroid hormone receptor's capacity to activate transcription from two biologically relevant DNA elements. We show that the genetic architecture of DNA recognition consists of a dense set of main and pairwise effects that involve virtually every possible amino acid state in the protein-DNA interface, but higher-order epistasis plays only a tiny role. Pairwise interactions enlarge the set of functional sequences and are the primary determinants of specificity for different DNA elements. They also massively expand the number of opportunities for single-residue mutations to switch specificity from one DNA target to another. By bringing variants with different functions close together in sequence space, pairwise epistasis therefore facilitates rather than constrains the evolution of new functions.

Surfaces: a software to quantify and visualize interactions within and between proteins and ligands

SUMMARY

Computational methods for the quantification and visualization of the relative contribution of molecular interactions to the stability of biomolecular structures and complexes are fundamental to understand, modulate and engineer biological processes. Here, we present Surfaces, an easy to use, fast and customizable software for quantification and visualization of molecular interactions based on the calculation of surface areas in contact. Surfaces calculations shows equivalent or better correlations with experimental data as computationally expensive methods based on molecular dynamics.

AVAILABILITY AND IMPLEMENTATION

All scripts are available at https://github.com/NRGLab/Surfaces. Surface's documentation is available at https://surfaces-tutorial.readthedocs.io/en/latest/index.html.

Mutational scan inferred binding energetics and structure in intrinsically disordered protein CcdA

Unlike globular proteins, mutational effects on the function of Intrinsically Disordered Proteins (IDPs) are not well-studied. Deep Mutational Scanning of a yeast surface displayed mutant library yields insights into sequence-function relationships in the CcdA IDP. The approach enables facile prediction of interface residues and local structural signatures of the bound conformation. In contrast to previous titration-based approaches which use a number of ligand concentrations, we show that use of a single rationally chosen ligand concentration can provide quantitative estimates of relative binding constants for large numbers of protein variants. This is because the extended interface of IDP ensures that energetic effects of point mutations are spread over a much smaller range than for globular proteins. Our data also provides insights into the much-debated role of helicity and disorder in partner binding of IDPs. Based on this exhaustive mutational sensitivity dataset, a rudimentary model was developed in an attempt to predict mutational effects on binding affinity of IDPs that form alpha-helical structures upon binding.

Hypericin, a potential new BH3 mimetic

Many types of cancer such as prostate cancer, myeloid leukemia, breast cancer, glioblastoma display strong chemo resistance, which is supported by enhanced expression of multiple anti-apoptotic Bcl-2, Bcl-XL and Mcl-1 proteins. The viable anti-cancer strategies are based on developing anti-apoptotic Bcl-2 proteins inhibitors, BH3 mimetics. Our focus in past years has been on the investigating a new potential BH3 mimetic, Hypericin (Hyp). Hyp is a naturally occurring photosensitive compound used in photodynamic therapy and diagnosis. We have demonstrated that Hyp can cause substantial effects in cellular ultrastructure, mitochondria function and metabolism, and distribution of Bcl2 proteins in malignant and non-malignant cells. One of the possible mechanisms of Hyp action could be the direct interactions between Bcl-2 proteins and Hyp. We investigated this assumption by in silico computer modelling and in vitro fluorescent spectroscopy experiments with the small Bcl2 peptide segments designed to correspond to Bcl2 BH3 and BH1 domains. We show here that Hyp interacts with BH3 and BH1 peptides in concentration dependent manner, and shows the stronger interactions than known BH3 mimetics, Gossypol (Goss) and ABT-263. In addition, interactions of Hyp, Goss and ABT263, with whole purified proteins Bcl-2 and Mcl-1 by fluorescence spectroscopy show that Hyp interacts stronger with the Bcl-2 and less with Mcl-1 protein than Goss or ABT-263. This suggest that Hyp is comparable to other BH3 mimetics and could be explore as such. Hyp cytotoxicity was low in human U87 MG glioma, similar to that of ABT263, where Goss exerted sufficient cytotoxicity, suggesting that Hyp acts primarily on Bcl-2, but not on Mcl-1 protein. In combination therapy, low doses of Hyp with Goss effectively decreased U87 MG viability, suggesting a possible synergy effect. Overall, we can conclude that Hyp as BH3 mimetic acts primarily on Bcl-2 protein and can be explored to target cells with Bcl-2 over-expression, or in combination with other BH3 mimetics, that target Mcl-1 or Bcl-XL proteins, in dual therapy.

Selective Affimers Recognise the BCL-2 Family Proteins BCL-xL and MCL-1 through Noncanonical Structural Motifs*

The BCL-2 family is a challenging group of proteins to target selectively due to sequence and structural homologies across the family. Selective ligands for the BCL-2 family regulators of apoptosis are useful as probes to understand cell biology and apoptotic signalling pathways, and as starting points for inhibitor design. We have used phage display to isolate Affimer reagents (non-antibody-binding proteins based on a conserved scaffold) to identify ligands for MCL-1, BCL-xL , BCL-2, BAK and BAX, then used multiple biophysical characterisation methods to probe the interactions. We established that purified Affimers elicit selective recognition of their target BCL-2 protein. For anti-apoptotic targets BCL-xL and MCL-1, competitive inhibition of their canonical protein-protein interactions is demonstrated. Co-crystal structures reveal an unprecedented mode of molecular recognition; where a BH3 helix is normally bound, flexible loops from the Affimer dock into the BH3 binding cleft. Moreover, the Affimers induce a change in the target proteins towards a desirable drug-bound-like conformation. These proof-of-concept studies indicate that Affimers could be used as alternative templates to inspire the design of selective BCL-2 family modulators and more generally other protein-protein interaction inhibitors.

Efficient flexible backbone protein-protein docking for challenging targets

Motivation

Binding-induced conformational changes challenge current computational docking algorithms by exponentially increasing the conformational space to be explored. To restrict this search to relevant space, some computational docking algorithms exploit the inherent flexibility of the protein monomers to simulate conformational selection from pre-generated ensembles. As the ensemble size expands with increased flexibility, these methods struggle with efficiency and high false positive rates.

Results

Here, we develop and benchmark RosettaDock 4.0, which efficiently samples large conformational ensembles of flexible proteins and docks them using a novel, six-dimensional, coarse-grained score function. A strong discriminative ability allows an eight-fold higher enrichment of near-native candidate structures in the coarse-grained phase compared to RosettaDock 3.2. It adaptively samples 100 conformations each of the ligand and the receptor backbone while increasing computational time by only 20-80%. In local docking of a benchmark set of 88 proteins of varying degrees of flexibility, the expected success rate (defined as cases with ≥50% chance of achieving 3 near-native structures in the 5 top-ranked ones) for blind predictions after resampling is 77% for rigid complexes, 49% for moderately flexible complexes and 31% for highly flexible complexes. These success rates on flexible complexes are a substantial step forward from all existing methods. Additionally, for highly flexible proteins, we demonstrate that when a suitable conformer generation method exists, the method successfully docks the complex.

Availability and implementation

As a part of the Rosetta software suite, RosettaDock 4.0 is available at https://www.rosettacommons.org to all non-commercial users for free and to commercial users for a fee.

Supplementary information

Supplementary data are available at Bioinformatics online.

Importance of Hypericin-Bcl2 interactions for biological effects at subcellular levels

Hypericin (Hyp) is a naturally occurring compound used as photosensitizer in photodynamic therapy and diagnosis. Recently, we have shown that Hyp presence alone, without illumination, resulted in substantial biological effects at several sub-cellular levels. Hyp induced changes in cellular ultrastructure, mitochondria function and metabolism, and distribution of Bcl2 proteins in malignant and non-malignant cells. The molecular mechanisms that underlie Hyp light-independent effects are still elusive. We have hypothesized that Bcl2-Hyp interactions might be one possible mechanism. We performed molecular docking studies to determine the Hyp-Bcl2 interaction profile. Based on the interaction profiles small Bcl2 peptide segments were selected for further study. We designed small peptides corresponding to Bcl2 BH3 and BH1 domains and tested the binding of Hyp and Bcl2 known inhibitor, ABT263, to the peptides in computer modeling and in vitro binding studies. We employed endogenous tryptophan and tyrosine in the BH3 and BH1 peptides, respectively, and their fluorescent properties to show interaction with Hyp and ABT263. Overall, our results indicate that Hyp can interact with Bcl2 protein at its BH3-BH1 hydrophobic groove, and this interaction may trigger changes in intracellular distribution of Bcl2 proteins. In addition, our computer modeling results suggest that Hyp also interacts with other anti-apoptotic members of Bcl2 family similar to the known BH3 mimetics. Our findings are novel and might contribute to understanding Hyp light-independent effects. In addition, they may substantiate the therapeutic use of Hyp as a BH3 mimetic molecule to enhance other cancer treatments.

Tertiary Structural Motif Sequence Statistics Enable Facile Prediction and Design of Peptides that Bind Anti-apoptotic Bfl-1 and Mcl-1

Understanding the relationship between protein sequence and structure well enough to design new proteins with desired functions is a longstanding goal in protein science. Here, we show that recurring tertiary structural motifs (TERMs) in the PDB provide rich information for protein-peptide interaction prediction and design. TERM statistics can be used to predict peptide binding energies for Bcl-2 family proteins as accurately as widely used structure-based tools. Furthermore, design using TERM energies (dTERMen) rapidly and reliably generates high-affinity peptide binders of anti-apoptotic proteins Bfl-1 and Mcl-1 with just 15%-38% sequence identity to any known native Bcl-2 family protein ligand. High-resolution structures of four designed peptides bound to their targets provide opportunities to analyze the strengths and limitations of the computational design method. Our results support dTERMen as a powerful approach that can complement existing tools for protein engineering.

Simplifying the Design of Protein-Peptide Interaction Specificity with Sequence-Based Representations of Atomistic Models

Computationally designed peptides targeting protein-protein interaction interfaces are of great interest as reagents for biological research and potential therapeutics. In recent years, it has been shown that detailed structure-based calculations can, in favorable cases, describe relevant determinants of protein-peptide recognition. Yet, despite large increases in available computing power, such accurate modeling of the binding reaction is still largely outside the realm of protein design. The chief limitation is in the large sequence spaces generally involved in protein design problems, such that it is typically infeasible to apply expensive modeling techniques to score each sequence. Toward addressing this issue, we have previously shown that by explicitly evaluating the scores of a relatively small number of sequences, it is possible to synthesize a direct mapping between sequences and scores, such that the entire sequence space can be analyzed extremely rapidly. The associated method, called Cluster Expansion, has been used in a number of studies to design binding affinity and specificity. In this chapter, we provide instructions and guidance for applying this technique in the context of designing protein-peptide interactions to enable the use of more detailed and expensive scoring approaches than is typically possible.

Enriching Peptide Libraries for Binding Affinity and Specificity Through Computationally Directed Library Design

Peptide reagents with high affinity or specificity for their target protein interaction partner are of utility for many important applications. Optimization of peptide binding by screening large libraries is a proven and powerful approach. Libraries designed to be enriched in peptide sequences that are predicted to have desired affinity or specificity characteristics are more likely to yield success than random mutagenesis. We present a library optimization method in which the choice of amino acids to encode at each peptide position can be guided by available experimental data or structure-based predictions. We discuss how to use analysis of predicted library performance to inform rounds of library design. Finally, we include protocols for more complex library design procedures that consider the chemical diversity of the amino acids at each peptide position and optimize a library score based on a user-specified input model.

Using Evolution to Guide Protein Engineering: The Devil IS in the Details

For decades, protein engineers have endeavored to reengineer existing proteins for novel applications. Overall, protein folds and gross functions can be readily transferred from one protein to another by transplanting large blocks of sequence (i.e., domain recombination). However, predictably fine-tuning function (e.g., by adjusting ligand affinity, specificity, catalysis, and/or allosteric regulation) remains a challenge. One approach has been to use the sequences of protein families to identify amino acid positions that change during the evolution of functional variation. The rationale is that these nonconserved positions could be mutated to predictably fine-tune function. Evolutionary approaches to protein design have had some success, but the engineered proteins seldom replicate the functional performances of natural proteins. This Biophysical Perspective reviews several complexities that have been revealed by evolutionary and experimental studies of protein function. These include 1) challenges in defining computational and biological thresholds that define important amino acids; 2) the co-occurrence of many different patterns of amino acid changes in evolutionary data; 3) difficulties in mapping the patterns of amino acid changes to discrete functional parameters; 4) the nonconventional mutational outcomes that occur for a particular group of functionally important, nonconserved positions; 5) epistasis (nonadditivity) among multiple mutations; and 6) the fact that a large fraction of a protein's amino acids contribute to its overall function. To overcome these challenges, new goals are identified for future studies.

Epistatic mutations in PUMA BH3 drive an alternate binding mode to potently and selectively inhibit anti-apoptotic Bfl-1

Overexpression of anti-apoptotic Bcl-2 family proteins contributes to cancer progression and confers resistance to chemotherapy. Small molecules that target Bcl-2 are used in the clinic to treat leukemia, but tight and selective inhibitors are not available for Bcl-2 paralog Bfl-1. Guided by computational analysis, we designed variants of the native BH3 motif PUMA that are > 150-fold selective for Bfl-1 binding. The designed peptides potently trigger disruption of the mitochondrial outer membrane in cells dependent on Bfl-1, but not in cells dependent on other anti-apoptotic homologs. High-resolution crystal structures show that designed peptide FS2 binds Bfl-1 in a shifted geometry, relative to PUMA and other binding partners, due to a set of epistatic mutations. FS2 modified with an electrophile reacts with a cysteine near the peptide-binding groove to augment specificity. Designed Bfl-1 binders provide reagents for cellular profiling and leads for developing enhanced and cell-permeable peptide or small-molecule inhibitors.

Computational design of self-assembling cyclic protein homo-oligomers

Self-assembling cyclic protein homo-oligomers play important roles in biology and the ability to generate custom homo-oligomeric structures could enable new approaches to probe biological function. Here we report a general approach to design cyclic homo-oligomers that employs a new residue pair transform method for assessing the design ability of a protein-protein interface. This method is sufficiently rapid to enable systematic enumeration of cyclically docked arrangements of a monomer followed by sequence design of the newly formed interfaces. We use this method to design interfaces onto idealized repeat proteins that direct their assembly into complexes that possess cyclic symmetry. Of 96 designs that were experimentally characterized, 21 were found to form stable monodisperse homo-oligomers in solution, and 15 (4 homodimers, 6 homotrimers, 6 homotetramers and 1 homopentamer) had solution small angle X-ray scattering data consistent with the design models. X-ray crystal structures were obtained for five of the designs and each of these were shown to be very close to their design model.

Computationally designed high specificity inhibitors delineate the roles of BCL2 family proteins in cancer

Many cancers overexpress one or more of the six human pro-survival BCL2 family proteins to evade apoptosis. To determine which BCL2 protein or proteins block apoptosis in different cancers, we computationally designed three-helix bundle protein inhibitors specific for each BCL2 pro-survival protein. Following in vitro optimization, each inhibitor binds its target with high picomolar to low nanomolar affinity and at least 300-fold specificity. Expression of the designed inhibitors in human cancer cell lines revealed unique dependencies on BCL2 proteins for survival which could not be inferred from other BCL2 profiling methods. Our results show that designed inhibitors can be generated for each member of a closely-knit protein family to probe the importance of specific protein-protein interactions in complex biological processes.

Molecular Determinants of Mutant Phenotypes, Inferred from Saturation Mutagenesis Data

Understanding how mutations affect protein activity and organismal fitness is a major challenge. We used saturation mutagenesis combined with deep sequencing to determine mutational sensitivity scores for 1,664 single-site mutants of the 101 residue Escherichia coli cytotoxin, CcdB at seven different expression levels. Active-site residues could be distinguished from buried ones, based on their differential tolerance to aliphatic and charged amino acid substitutions. At nonactive-site positions, the average mutational tolerance correlated better with depth from the protein surface than with accessibility. Remarkably, similar results were observed for two other small proteins, PDZ domain (PSD95^pdz3) and IgG-binding domain of protein G (GB1). Mutational sensitivity data obtained with CcdB were used to derive a procedure for predicting functional effects of mutations. Results compared favorably with those of two widely used computational predictors. In vitro characterization of 80 single, nonactive-site mutants of CcdB showed that activity in vivo correlates moderately with thermal stability and solubility. The inability to refold reversibly, as well as a decreased folding rate in vitro, is associated with decreased activity in vivo. Upon probing the effect of modulating expression of various proteases and chaperones on mutant phenotypes, most deleterious mutants showed an increased in vivo activity and solubility only upon over-expression of either Trigger factor or SecB ATP-independent chaperones. Collectively, these data suggest that folding kinetics rather than protein stability is the primary determinant of activity in vivo. This study enhances our understanding of how mutations affect phenotype, as well as the ability to predict fitness effects of point mutations.

Computational protein design for given backbone: recent progresses in general method-related aspects

To achieve high success rate in protein design requires a reliable sequence design method to find amino acid sequences that stably fold into a desired backbone structure. This problem is addressed by computational protein design through the approach of energy minimization. Here we review recent method progresses related to improving the accuracy of this approach. First, the quality of the energy model is a key factor. Second, with structure sensitive energy functions, whether and how backbone flexibility is considered can have large effects on design accuracy, although usually only small adjustments of the backbone structure itself are involved. Third, the effective accuracy of design results can be boosted by post-processing a small number of designed sequences with complementary models that may not be efficient enough for full sequence optimization. Finally, computational method development will benefit greatly from increasingly efficient experimental approaches that can be applied to obtain extensive feedbacks.

Designing helical peptide inhibitors of protein-protein interactions

Short helical peptides combine characteristics of small molecules and large proteins and provide an exciting area of opportunity in protein design. A growing number of studies report novel helical peptide inhibitors of protein-protein interactions. New techniques have been developed for peptide design and for chemically stabilizing peptides in a helical conformation, which frequently improves protease resistance and cell permeability. We summarize advances in peptide crosslinking chemistry and give examples of peptide design studies targeting coiled-coil transcription factors, Bcl-2 family proteins, MDM2/MDMX, and HIV gp41, among other targets.

AB-Bind: Antibody binding mutational database for computational affinity predictions

Antibodies (Abs) are a crucial component of the immune system and are often used as diagnostic and therapeutic agents. The need for high-affinity and high-specificity antibodies in research and medicine is driving the development of computational tools for accelerating antibody design and discovery. We report a diverse set of antibody binding data with accompanying structures that can be used to evaluate methods for modeling antibody interactions. Our Antibody-Bind (AB-Bind) database includes 1101 mutants with experimentally determined changes in binding free energies (ΔΔG) across 32 complexes. Using the AB-Bind data set, we evaluated the performance of protein scoring potentials in their ability to predict changes in binding free energies upon mutagenesis. Numerical correlations between computed and observed ΔΔG values were low (r = 0.16-0.45), but the potentials exhibited predictive power for classifying variants as improved vs weakened binders. Performance was evaluated using the area under the curve (AUC) for receiver operator characteristic (ROC) curves; the highest AUC values for 527 mutants with |ΔΔG| > 1.0 kcal/mol were 0.81, 0.87, and 0.88 using STATIUM, FoldX, and Discovery Studio scoring potentials, respectively. Some methods could also enrich for variants with improved binding affinity; FoldX and Discovery Studio were able to correctly rank 42% and 30%, respectively, of the 80 most improved binders (those with ΔΔG < -1.0 kcal/mol) in the top 5% of the database. This modest predictive performance has value but demonstrates the continuing need to develop and improve protein energy functions for affinity prediction.

Generating High-Accuracy Peptide-Binding Data in High Throughput with Yeast Surface Display and SORTCERY

Library methods are widely used to study protein-protein interactions, and high-throughput screening or selection followed by sequencing can identify a large number of peptide ligands for a protein target. In this chapter, we describe a procedure called "SORTCERY" that can rank the affinities of library members for a target with high accuracy. SORTCERY follows a three-step protocol. First, fluorescence-activated cell sorting (FACS) is used to sort a library of yeast-displayed peptide ligands according to their affinities for a target. Second, all sorted pools are deep sequenced. Third, the resulting data are analyzed to create a ranking. We demonstrate an application of SORTCERY to the problem of ranking peptide ligands for the anti-apoptotic regulator Bcl-xL.

Design of Specific Peptide-Protein Recognition

Selective targeting of protein-protein interactions in the cell is of great interest in biological research. Computational structure-based design of peptides to bind protein interaction interfaces could provide a potential means of generating such reagents. However, to avoid perturbing off-target interactions, methods that explicitly account for interaction specificity are needed. Further, as peptides often retain considerable flexibility upon association, their binding reaction is computationally demanding to model-a stark limitation for structure-based design. Here we present a protocol for designing peptides that selectively target a given peptide-binding domain, relative to a pre-specified set of possibly related domains. We recently used the method to design peptides that discriminate with high selectivity between two closely related PDZ domains. The framework accounts for the flexibility of the peptide in the binding site, but is efficient enough to quickly analyze trade-offs between affinity and selectivity, enabling the identification of optimal peptides.

Redefining the BH3 Death Domain as a 'Short Linear Motif'

B cell lymphoma-2 (BCL-2)-related proteins control programmed cell death through a complex network of protein–protein interactions mediated by BCL-2 homology 3 (BH3) domains. Given their roles as dynamic linchpins, the discovery of novel BH3-containing proteins has attracted considerable attention. However, without a clearly defined BH3 signature sequence the BCL-2 family has expanded to include a nebulous group of nonhomologous BH3-only proteins, now justified by an intriguing twist. We present evidence that BH3s from both ordered and disordered proteins represent a new class of short linear motifs (SLiMs) or molecular recognition features (MoRFs) and are diverse in their evolutionary histories. The implied corollaries are that BH3s have a broad phylogenetic distribution and could potentially bind to non-BCL-2-like structural domains with distinct functions.

BCL-2 family interactions are mediated by evolutionarily diverse BH3 motifs to regulate apoptosis. Given their key roles, BH3 mimetics are in clinical trials as cancer therapies.

The discovery of novel BH3-only proteins represents a major endeavor in the cell death field. As a result, BH3 motifs are reportedly present in a nebulous conglomerate of different proteins, both structured and intrinsically disordered.

There is no rigorous definition of a BH3 motif. Currently available BH3 signatures are diverse and elusive for predicting new functional BH3-containing proteins.

Redefining the BH3 motif as a new type of short linear motif (SLiM) or molecular recognition feature (MoRF) reconciles many puzzling features of this motif and opens up new avenues for research.

Locating Herpesvirus Bcl-2 Homologs in the Specificity Landscape of Anti-Apoptotic Bcl-2 Proteins

Viral homologs of the anti-apoptotic Bcl-2 proteins are highly diverged from their mammalian counterparts, yet they perform overlapping functions by binding and inhibiting BH3 (Bcl-2 homology 3)-motif-containing proteins. We investigated the BH3 binding properties of the herpesvirus Bcl-2 homologs KSBcl-2, BHRF1, and M11, as they relate to those of the human Bcl-2 homologs Mcl-1, Bfl-1, Bcl-w, Bcl-xL, and Bcl-2. Analysis of the sequence and structure of the BH3 binding grooves showed that, despite low sequence identity, M11 has structural similarities to Bcl-xL, Bcl-2, and Bcl-w. BHRF1 and KSBcl-2 are more structurally similar to Mcl-1 than to the other human proteins. Binding to human BH3-like peptides showed that KSBcl-2 has similar specificity to Mcl-1, and BHRF1 has a restricted binding profile; M11 binding preferences are distinct from those of Bcl-xL, Bcl-2, and Bcl-w. Because KSBcl-2 and BHRF1 are from human herpesviruses associated with malignancies, we screened computationally designed BH3 peptide libraries using bacterial surface display to identify selective binders of KSBcl-2 or BHRF1. The resulting peptides bound to KSBcl-2 and BHRF1 in preference to Bfl-1, Bcl-w, Bcl-xL, and Bcl-2 but showed only modest specificity over Mcl-1. Rational mutagenesis increased specificity against Mcl-1, resulting in a peptide with a dissociation constant of 2.9nM for binding to KSBcl-2 and >1000-fold specificity over other Bcl-2 proteins, as well as a peptide with >70-fold specificity for BHRF1. In addition to providing new insights into viral Bcl-2 binding specificity, this study will inform future work analyzing the interaction properties of homologous binding domains and designing specific protein interaction partners.

Mapping the BH3 Binding Interface of Bcl-xL, Bcl-2, and Mcl-1 Using Split-Luciferase Reassembly

The recognition of helical BH3 domains by Bcl-2 homology (BH) receptors plays a central role in apoptosis. The residues that determine specificity or promiscuity in this interactome are difficult to predict from structural and computational data. Using a cell free split-luciferase system, we have generated a 276 pairwise interaction map for 12 alanine mutations at the binding interface for three receptors, Bcl-xL, Bcl-2, and Mcl-1, and interrogated them against BH3 helices derived from Bad, Bak, Bid, Bik, Bim, Bmf, Hrk, and Puma. This panel, in conjunction with previous structural and functional studies, starts to provide a more comprehensive portrait of this interactome, explains promiscuity, and uncovers surprising details; for example, the Bcl-xL R139A mutation disrupts binding to all helices but the Bad-BH3 peptide, and Mcl-1 binding is particularly perturbed by only four mutations of the 12 tested (V220A, N260A, R263A, and F319A), while Bcl-xL and Bcl-2 have a more diverse set of important residues depending on the bound helix.

Potent and specific peptide inhibitors of human pro-survival protein Bcl-xL

The Bcl-2 family of proteins plays a critical role regulating apoptosis, and pro-survival Bcl-2 family members are important therapeutic targets due to their overexpression in different cancers. Pro-apoptotic Bcl-2 homology 3 (BH3)-only proteins antagonize pro-survival Bcl-2 protein functions by binding directly to them, and a sub-class of BH3-only proteins termed sensitizers can initiate apoptosis via this mechanism in response to diverse signals. The five pro-survival proteins Bcl-xL, Mcl-1, Bcl-2, Bcl-w and Bfl-1 differ in their binding preferences, with Bcl-xL, Bcl-2 and Bcl-w sharing similar interaction profiles for many natural sensitizers and small molecules. Peptides that bind selectively to just one or a subset of family members have shown utility in assays that diagnose apoptotic blockades in cancer cells and as reagents for dissecting apoptotic mechanism. Combining computational design, combinatorial library screening and rational mutagenesis, we designed a series of BH3 sensitizer peptides that bind Bcl-xL with sub-nanomolar affinity and selectivity up to 1000-fold over each of the four competing pro-survival proteins. We demonstrate the efficacy of our designed BH3 peptides in assays that differentiate between cancer cells that are dependent on different pro-survival proteins.

Protein design with a comprehensive statistical energy function and boosted by experimental selection for foldability

The de novo design of amino acid sequences to fold into desired structures is a way to reach a more thorough understanding of how amino acid sequences encode protein structures and to supply methods for protein engineering. Notwithstanding significant breakthroughs, there are noteworthy limitations in current computational protein design. To overcome them needs computational models to complement current ones and experimental tools to provide extensive feedbacks to theory. Here we develop a comprehensive statistical energy function for protein design with a new general strategy and verify that it can complement and rival current well-established models. We establish that an experimental approach can be used to efficiently assess or improve the foldability of designed proteins. We report four de novo proteins for different targets, all experimentally verified to be well-folded, solved solution structures for two being in excellent agreement with respective design targets.

SORTCERY-A High-Throughput Method to Affinity Rank Peptide Ligands

Uncovering the relationships between peptide and protein sequences and binding properties is critical for successfully predicting, re-designing and inhibiting protein-protein interactions. Systematically collected data that link protein sequence to binding are valuable for elucidating determinants of protein interaction but are rare in the literature because such data are experimentally difficult to generate. Here we describe SORTCERY, a high-throughput method that we have used to rank hundreds of yeast-displayed peptides according to their affinities for a target interaction partner. The procedure involves fluorescence-activated cell sorting of a library, deep sequencing of sorted pools and downstream computational analysis. We have developed theoretical models and statistical tools that assist in planning these stages. We demonstrate SORTCERY's utility by ranking 1026 BH3 (Bcl-2 homology 3) peptides with respect to their affinities for the anti-apoptotic protein Bcl-xL. Our results are in striking agreement with measured affinities for 19 individual peptides with dissociation constants ranging from 0.1 to 60nM. High-resolution ranking can be used to improve our understanding of sequence-function relationships and to support the development of computational models for predicting and designing novel interactions.

Designed BH3 peptides with high affinity and specificity for targeting Mcl-1 in cells

Mcl-1 is overexpressed in many cancers and can confer resistance to cell-death signaling in refractory disease. Molecules that specifically inhibit Mcl-1 hold potential for diagnosing and disrupting Mcl-1-dependent cell survival. We selected three peptides from a yeast-surface display library that showed moderate specificity and affinity for binding to Mcl-1 over Bfl-1, Bcl-xL, Bcl-2, and Bcl-w. Specificity for Mcl-1 was improved by introducing threonine at peptide position 2e. The most specific peptide, MS1, bound Mcl-1 with 40-fold or greater specificity over four other human Bcl-2 paralogs. In BH3 profiling assays, MS1 caused depolarization in several human Mcl-1-dependent cell lines with EC50 values of ∼3 μM, contrasted with EC50 values of >100 μM for Bcl-2-, Bcl-xL-, or Bfl-1-dependent cell lines. MS1 is at least 30-fold more potent in this assay than the previously used Mcl-1 targeting reagent NoxaA BH3. These peptides can be used to detect Mcl-1 dependency in cells and provide leads for developing Mcl-1 targeting therapeutics.

Genome-wide prediction and validation of peptides that bind human prosurvival Bcl-2 proteins

Programmed cell death is regulated by interactions between pro-apoptotic and prosurvival members of the Bcl-2 family. Pro-apoptotic family members contain a weakly conserved BH3 motif that can adopt an alpha-helical structure and bind to a groove on prosurvival partners Bcl-x_L, Bcl-w, Bcl-2, Mcl-1 and Bfl-1. Peptides corresponding to roughly 13 reported BH3 motifs have been verified to bind in this manner. Due to their short lengths and low sequence conservation, BH3 motifs are not detected using standard sequence-based bioinformatics approaches. Thus, it is possible that many additional proteins harbor BH3-like sequences that can mediate interactions with the Bcl-2 family. In this work, we used structure-based and data-based Bcl-2 interaction models to find new BH3-like peptides in the human proteome. We used peptide SPOT arrays to test candidate peptides for interaction with one or more of the prosurvival proteins Bcl-x_L, Bcl-w, Bcl-2, Mcl-1 and Bfl-1. For the 36 most promising array candidates, we quantified binding to all five human receptors using direct and competition binding assays in solution. All 36 peptides showed evidence of interaction with at least one prosurvival protein, and 22 peptides bound at least one prosurvival protein with a dissociation constant between 1 and 500 nM; many peptides had specificity profiles not previously observed. We also screened the full-length parent proteins of a subset of array-tested peptides for binding to Bcl-x_L and Mcl-1. Finally, we used the peptide binding data, in conjunction with previously reported interactions, to assess the affinity and specificity prediction performance of different models.

Bcl-2 family proteins regulate key cell death vs. survival decisions and are implicated in the development of many cancers. To understand the roles of Bcl-2 family proteins in both normal and diseased cells, it is important to map the interaction network of the family. Low sequence conservation in known Bcl-2 interaction motifs precludes easy identification of possible binding partners, but we developed computational models based on structure and experimental mutation data that show good predictive performance. We used our models to search the human proteome for new Bcl-2 interaction partners. We predicted and experimentally validated more than twice as many tight-binding peptides as were previously known.

Targeting γ-herpesvirus 68 Bcl-2-mediated down-regulation of autophagy

γ-herpesviruses (γHVs) are common human pathogens that encode homologs of the anti-apoptotic cellular Bcl-2 proteins, which are critical to viral reactivation and oncogenic transformation. The murine γHV68 provides a tractable in vivo model for understanding general features of these important human pathogens. Bcl-XL, a cellular Bcl-2 homolog, and the murine γHV68 Bcl-2 homolog, M11, both bind to a BH3 domain within the key autophagy effector Beclin 1 with comparable affinities, resulting in the down-regulation of Beclin 1-mediated autophagy. Despite this similarity, differences in residues lining the binding site of M11 and Bcl-XL dictate varying affinities for the different BH3 domain-containing proteins. Here we delineate Beclin 1 differential specificity determinants for binding to M11 or Bcl-XL by quantifying autophagy levels in cells expressing different Beclin 1 mutants and either M11 or Bcl-XL, and we show that a G120E/D121A Beclin 1 mutant selectively prevents down-regulation of Beclin 1-mediated autophagy by Bcl-XL, but not by M11. We use isothermal titration calorimetry to identify a Beclin 1 BH3 domain-derived peptide that selectively binds to M11, but not to Bcl-XL. The x-ray crystal structure of this peptide bound to M11 reveals the mechanism by which the M11 BH3 domain-binding groove accommodates this M11-specific peptide. This information was used to develop a cell-permeable peptide inhibitor that selectively inhibits M11-mediated, but not Bcl-XL-mediated, down-regulation of autophagy.

The structural biology of BH3-only proteins

B-cell lymphoma-2 (Bcl-2) homology-3 (BH3)-only proteins are considered members of the Bcl-2 family, though they bear little sequence or structural identity with the multi-BH motif prosurvival or proapoptotic Bcl-2 proteins like Bcl-2 or Bax. They are better considered a separate phylogenetic entity. In combination, results from biophysical, biochemical, cell biological, and animal studies in conjunction with structural investigations have elucidated the function and mechanism of action of these proteins. Either by antagonizing prosurvival Bcl-2 proteins or directly activating proapoptotic Bcl-2 proteins (Bax or Bak) they initiate apoptosis. BH3-only proteins are intrinsically disordered and fold and bind into a groove provided by their cognate receptor Bcl-2 family proteins. Our detailed molecular understanding of BH3-only protein action has aided the development of novel chemical entities that initiate cell death by mimicking the properties of BH3-only proteins.

Rheostats and toggle switches for modulating protein function

The millions of protein sequences generated by genomics are expected to transform protein engineering and personalized medicine. To achieve these goals, tools for predicting outcomes of amino acid changes must be improved. Currently, advances are hampered by insufficient experimental data about nonconserved amino acid positions. Since the property “nonconserved” is identified using a sequence alignment, we designed experiments to recapitulate that context: Mutagenesis and functional characterization was carried out in 15 LacI/GalR homologs (rows) at 12 nonconserved positions (columns). Multiple substitutions were made at each position, to reveal how various amino acids of a nonconserved column were tolerated in each protein row. Results showed that amino acid preferences of nonconserved positions were highly context-dependent, had few correlations with physico-chemical similarities, and were not predictable from their occurrence in natural LacI/GalR sequences. Further, unlike the “toggle switch” behaviors of conserved positions, substitutions at nonconserved positions could be rank-ordered to show a “rheostatic”, progressive effect on function that spanned several orders of magnitude. Comparisons to various sequence analyses suggested that conserved and strongly co-evolving positions act as functional toggles, whereas other important, nonconserved positions serve as rheostats for modifying protein function. Both the presence of rheostat positions and the sequence analysis strategy appear to be generalizable to other protein families and should be considered when engineering protein modifications or predicting the impact of protein polymorphisms.

Computational protein design quantifies structural constraints on amino acid covariation

Amino acid covariation, where the identities of amino acids at different sequence positions are correlated, is a hallmark of naturally occurring proteins. This covariation can arise from multiple factors, including selective pressures for maintaining protein structure, requirements imposed by a specific function, or from phylogenetic sampling bias. Here we employed flexible backbone computational protein design to quantify the extent to which protein structure has constrained amino acid covariation for 40 diverse protein domains. We find significant similarities between the amino acid covariation in alignments of natural protein sequences and sequences optimized for their structures by computational protein design methods. These results indicate that the structural constraints imposed by protein architecture play a dominant role in shaping amino acid covariation and that computational protein design methods can capture these effects. We also find that the similarity between natural and designed covariation is sensitive to the magnitude and mechanism of backbone flexibility used in computational protein design. Our results thus highlight the necessity of including backbone flexibility to correctly model precise details of correlated amino acid changes and give insights into the pressures underlying these correlations.

Proteins generally fold into specific three-dimensional structures to perform their cellular functions, and the presence of misfolded proteins is often deleterious for cellular and organismal fitness. For these reasons, maintenance of protein structure is thought to be one of the major fitness pressures acting on proteins. Consequently, the sequences of today's naturally occurring proteins contain signatures reflecting the constraints imposed by protein structure. Here we test the ability of computational protein design methods to recapitulate and explain these signatures. We focus on the physical basis of evolutionary pressures that act on interactions between amino acids in folded proteins, which are critical in determining protein structure and function. Such pressures can be observed from the appearance of amino acid covariation, where the amino acids at certain positions in protein sequences are correlated with each other. We find similar patterns of amino acid covariation in natural sequences and sequences optimized for their structures using computational protein design, demonstrating the importance of structural constraints in protein molecular evolution and providing insights into the structural mechanisms leading to covariation. In addition, these results characterize the ability of computational methods to model the precise details of correlated amino acid changes, which is critical for engineering new proteins with useful functions beyond those seen in nature.

Programming biological models in Python using PySB

PySB is a framework for creating biological models as Python programs using a high-level, action-oriented vocabulary that promotes transparency, extensibility and reusability. PySB interoperates with many existing modeling tools and supports distributed model development.

PySB models are programs and leverage existing programming tools for documentation, testing, and collaborative development.

Reusable functions can encode common low-level biochemical processes as well as high-level modules, making models transparent and concise.

Modeling workflow is accelerated through close integration with Python numerical tools and interoperability with existing modeling software.

We demonstrate the use of PySB to encode 15 alternative hypotheses for the mitochondrial regulation of apoptosis, including a new ‘Embedded Together' model based on recent biochemical findings.

Mathematical equations are fundamental to modeling biological networks, but as networks get large and revisions frequent, it becomes difficult to manage equations directly or to combine previously developed models. Multiple simultaneous efforts to create graphical standards, rule-based languages, and integrated software workbenches aim to simplify biological modeling but none fully meets the need for transparent, extensible, and reusable models. In this paper we describe PySB, an approach in which models are not only created using programs, they are programs. PySB draws on programmatic modeling concepts from little b and ProMot, the rule-based languages BioNetGen and Kappa and the growing library of Python numerical tools. Central to PySB is a library of macros encoding familiar biochemical actions such as binding, catalysis, and polymerization, making it possible to use a high-level, action-oriented vocabulary to construct detailed models. As Python programs, PySB models leverage tools and practices from the open-source software community, substantially advancing our ability to distribute and manage the work of testing biochemical hypotheses. We illustrate these ideas using new and previously published models of apoptosis.

Latent effects of Hsp90 mutants revealed at reduced expression levels

In natural systems, selection acts on both protein sequence and expression level, but it is unclear how selection integrates over these two dimensions. We recently developed the EMPIRIC approach to systematically determine the fitness effects of all possible point mutants for important regions of essential genes in yeast. Here, we systematically investigated the fitness effects of point mutations in a putative substrate binding loop of yeast Hsp90 (Hsp82) over a broad range of expression strengths. Negative epistasis between reduced expression strength and amino acid substitutions was common, and the endogenous expression strength frequently obscured mutant defects. By analyzing fitness effects at varied expression strengths, we were able to uncover all mutant effects on function. The majority of mutants caused partial functional defects, consistent with this region of Hsp90 contributing to a mutation sensitive and critical process. These results demonstrate that important functional regions of proteins can tolerate mutational defects without experimentally observable impacts on fitness.

Peptide ligands for pro-survival protein Bfl-1 from computationally guided library screening

Pro-survival members of the Bcl-2 protein family inhibit cell death by binding short helical BH3 motifs in pro-apoptotic proteins. Mammalian pro-survival proteins Bcl-x(L), Bcl-2, Bcl-w, Mcl-1, and Bfl-1 bind with varying affinities and specificities to native BH3 motifs, engineered peptides, and small molecules. Biophysical studies have determined interaction patterns for these proteins, particularly for the most-studied family members Bcl-x(L) and Mcl-1. Bfl-1 is a pro-survival protein implicated in preventing apoptosis in leukemia, lymphoma, and melanoma. Although Bfl-1 is a promising therapeutic target, relatively little is known about its binding preferences. We explored the binding of Bfl-1 to BH3-like peptides by screening a peptide library that was designed to sample a high degree of relevant sequence diversity. Screening using yeast-surface display led to several novel high-affinity Bfl-1 binders and to thousands of putative binders identified through deep sequencing. Further screening for specificity led to identification of a peptide that bound to Bfl-1 with K(d) < 1 nM and very slow dissociation from Bfl-1 compared to other pro-survival Bcl-2 family members. A point mutation in this sequence gave a peptide with ~50 nM affinity for Bfl-1 that was selective for Bfl-1 in equilibrium binding assays. Analysis of engineered Bfl-1 binders deepens our understanding of how the binding profiles of pro-survival proteins differ and may guide the development of targeted Bfl-1 inhibitors.

Analyses of the effects of all ubiquitin point mutants on yeast growth rate

The amino acid sequence of a protein governs its function. We used bulk competition and focused deep sequencing to investigate the effects of all ubiquitin point mutants on yeast growth rate. Many aspects of ubiquitin function have been carefully studied, which enabled interpretation of our growth analyses in light of a rich structural, biophysical and biochemical knowledge base. In one highly sensitive cluster on the surface of ubiquitin, almost every amino acid substitution caused growth defects. In contrast, the opposite face tolerated virtually all possible substitutions. Surface locations between these two faces exhibited intermediate mutational tolerance. The sensitive face corresponds to the known interface for many binding partners. Across all surface positions, we observe a strong correlation between burial at structurally characterized interfaces and the number of amino acid substitutions compatible with robust growth. This result indicates that binding is a dominant determinant of ubiquitin function. In the solvent-inaccessible core of ubiquitin, all positions tolerated a limited number of substitutions, with hydrophobic amino acids especially interchangeable. Some mutations null for yeast growth were previously shown to populate folded conformations indicating that, for these mutants, subtle changes to conformation caused functional defects. The most sensitive region to mutation within the core was located near the C-terminus that is a focal binding site for many critical binding partners. These results indicate that core mutations may frequently cause functional defects through subtle disturbances to structure or dynamics.

Computational approaches for rational design of proteins with novel functionalities

Proteins are the most multifaceted macromolecules in living systems and have various important functions, including structural, catalytic, sensory, and regulatory functions. Rational design of enzymes is a great challenge to our understanding of protein structure and physical chemistry and has numerous potential applications. Protein design algorithms have been applied to design or engineer proteins that fold, fold faster, catalyze, catalyze faster, signal, and adopt preferred conformational states. The field of de novo protein design, although only a few decades old, is beginning to produce exciting results. Developments in this field are already having a significant impact on biotechnology and chemical biology. The application of powerful computational methods for functional protein designing has recently succeeded at engineering target activities. Here, we review recently reported de novo functional proteins that were developed using various protein design approaches, including rational design, computational optimization, and selection from combinatorial libraries, highlighting recent advances and successes.

In silico and in vitro elucidation of BH3 binding specificity toward Bcl-2

Interactions between Bcl-2-like proteins and BH3 domains play a key role in the regulation of apoptosis. Despite the overall structural similarity of their interaction with helical BH3 domains, Bcl-2-like proteins exhibit an intricate spectrum of binding specificities whose underlying basis is not well understood. Here, we characterize these interactions using Rosetta FlexPepBind, a protocol for the prediction of peptide binding specificity that evaluates the binding potential of different peptides based on structural models of the corresponding peptide-receptor complexes. For two prominent players, Bcl-xL and Mcl-1, we obtain good agreement with a large set of experimental SPOT array measurements and recapitulate the binding specificity of peptides derived by yeast display in a previous study. We extend our approach to a third member of this family, Bcl-2: we test our blind prediction of the binding of 180 BIM-derived peptides with a corresponding experimental SPOT array. Both prediction and experiment reveal a Bcl-2 binding specificity pattern that resembles that of Bcl-xL. Finally, we extend this application to accurately predict the specificity pattern of additional human BH3-only derived peptides. This study characterizes the distinct patterns of binding specificity of BH3-only derived peptides for the Bcl-2 like proteins Bcl-xL, Mcl-1, and Bcl-2 and provides insight into the structural basis of determinants of specificity.