Pushing the limits of what is achievable in protein–DNA docking: benchmarking HADDOCK’s performance

The intrinsic flexibility of DNA and the difficulty of identifying its interaction surface have long been challenges that prevented the development of efficient protein–DNA docking methods. We have demonstrated the ability our flexible data-driven docking method HADDOCK to deal with these before, by using custom-built DNA structural models. Here we put our method to the test on a set of 47 complexes from the protein–DNA docking benchmark. We show that HADDOCK is able to predict many of the specific DNA conformational changes required to assemble the interface(s). Our DNA analysis and modelling procedure captures the bend and twist motions occurring upon complex formation and uses these to generate custom-built DNA structural models, more closely resembling the bound form, for use in a second docking round. We achieve throughout the benchmark an overall success rate of 94% of one-star solutions or higher (interface root mean square deviation ≤4 Å and fraction of native contacts >10%) according to CAPRI criteria. Our improved protocol successfully predicts even the challenging protein–DNA complexes in the benchmark. Finally, our method is the first to readily dock multiple molecules (N > 2) simultaneously, pushing the limits of what is currently achievable in the field of protein–DNA docking.     

HADDOCK versus HADDOCK: new features and performance of HADDOCK2.0 on the CAPRI targets

Here we present version 2.0 of HADDOCK, which incorporates considerable improvements and new features. HADDOCK is now able to model not only protein-protein complexes but also other kinds of biomolecular complexes and multi-component (N > 2) systems. In the absence of any experimental and/or predicted information to drive the docking, HADDOCK now offers two additional ab initio docking modes based on either random patch definition or center-of-mass restraints. The docking protocol has been considerably improved, supporting among other solvated docking, automatic definition of semi-flexible regions, and inclusion of a desolvation energy term in the scoring scheme. The performance of HADDOCK2.0 is evaluated on the targets of rounds 4-11, run in a semi-automated mode using the original information we used in our CAPRI submissions. This enables a direct assessment of the progress made since the previous versions. Although HADDOCK performed very well in CAPRI (65% and 71% success rates, overall and for unbound targets only, respectively), a substantial improvement was achieved with HADDOCK2.0.     

A Unified Conformational Selection and Induced Fit Approach to Protein-Peptide Docking

Protein-peptide interactions are vital for the cell. They mediate, inhibit or serve as structural components in nearly 40% of all macromolecular interactions, and are often associated with diseases, making them interesting leads for protein drug design. In recent years, large-scale technologies have enabled exhaustive studies on the peptide recognition preferences for a number of peptide-binding domain families. Yet, the paucity of data regarding their molecular binding mechanisms together with their inherent flexibility makes the structural prediction of protein-peptide interactions very challenging. This leaves flexible docking as one of the few amenable computational techniques to model these complexes. We present here an ensemble, flexible protein-peptide docking protocol that combines conformational selection and induced fit mechanisms. Starting from an ensemble of three peptide conformations (extended, a-helix, polyproline-II), flexible docking with HADDOCK generates 79.4% of high quality models for bound/unbound and 69.4% for unbound/unbound docking when tested against the largest protein-peptide complexes benchmark dataset available to date. Conformational selection at the rigid-body docking stage successfully recovers the most relevant conformation for a given protein-peptide complex and the subsequent flexible refinement further improves the interface by up to 4.5 Å interface RMSD. Cluster-based scoring of the models results in a selection of near-native solutions in the top three for ∼75% of the successfully predicted cases. This unified conformational selection and induced fit approach to protein-peptide docking should open the route to the modeling of challenging systems such as disorder-order transitions taking place upon binding, significantly expanding the applicability limit of biomolecular interaction modeling by docking.     

WHISCY: What information does surface conservation yield? Application to data‐driven docking

Protein-protein interactions play a key role in biological processes. Identifying the interacting residues is a first step toward understanding these interactions at a structural level. In this study, the interface prediction program WHISCY is presented. It combines surface conservation and structural information to predict protein-protein interfaces. The accuracy of the predictions is more than three times higher than a random prediction. These predictions have been combined with another interface prediction program, ProMate [Neuvirth et al. J Mol Biol 2004;338:181-199], resulting in an even more accurate predictor. The usefulness of the predictions was tested using the data-driven docking program HADDOCK [Dominguez et al. J Am Chem Soc 2003;125:1731-1737] in an unbound docking experiment, with the goal of generating as many near-native structures as possible. Unrefined rigid body docking solutions within 10 A ligand RMSD from the true structure were generated for 22 out of 25 docked complexes. For 18 complexes, more than 100 of the 8000 generated models were correct. Our results demonstrates the potential of using interface predictions to drive protein-protein docking.     

PaFlexPepDock: Parallel Ab-Initio Docking of Peptides onto Their Receptors with Full Flexibility Based on Rosetta

Structural information related to protein–peptide complexes can be very useful for novel drug discovery and design. The computational docking of protein and peptide can supplement the structural information available on protein–peptide interactions explored by experimental ways. Protein–peptide docking of this paper can be described as three processes that occur in parallel: ab-initio peptide folding, peptide docking with its receptor, and refinement of some flexible areas of the receptor as the peptide is approaching. Several existing methods have been used to sample the degrees of freedom in the three processes, which are usually triggered in an organized sequential scheme. In this paper, we proposed a parallel approach that combines all the three processes during the docking of a folding peptide with a flexible receptor. This approach mimics the actual protein–peptide docking process in parallel way, and is expected to deliver better performance than sequential approaches. We used 22 unbound protein–peptide docking examples to evaluate our method. Our analysis of the results showed that the explicit refinement of the flexible areas of the receptor facilitated more accurate modeling of the interfaces of the complexes, while combining all of the moves in parallel helped the constructing of energy funnels for predictions.     

Pressure dissociation of integration host factor-DNA complexes reveals flexibility-dependent structural variation at the protein-DNA interface

E. coli Integration host factor (IHF) condenses the bacterial nucleoid by wrapping DNA. Previously, we showed that DNA flexibility compensates for structural characteristics of the four consensus recognition elements associated with specific binding (Aeling et al., J. Biol. Chem. 281, 39236–39248, 2006). If elements are missing, high-affinity binding occurs only if DNA deformation energy is low. In contrast, if all elements are present, net binding energy is unaffected by deformation energy. We tested two hypotheses for this observation: in complexes containing all elements, (1) stiff DNA sequences are less bent upon binding IHF than flexible ones; or (2) DNA sequences with differing flexibility have interactions with IHF that compensate for unfavorable deformation energy. Time-resolved Förster resonance energy transfer (FRET) shows that global topologies are indistinguishable for three complexes with oligonucleotides of different flexibility. However, pressure perturbation shows that the volume change upon binding is smaller with increasing flexibility. We interpret these results in the context of Record and coworker's model for IHF binding (J. Mol. Biol. 310, 379–401, 2001). We propose that the volume changes reflect differences in hydration that arise from structural variation at IHF–DNA interfaces while the resulting energetic compensation maintains the same net binding energy.     

Data-driven docking: HADDOCK's adventures in CAPRI

We have shown previously that given high-resolution structures of the unbound molecules, structure determination of protein complexes is possible by including biochemical and/or biophysical data as highly ambiguous distance restraints in a docking approach. We applied this method, implemented in the HADDOCK (High Ambiguity Driven DOCKing) package (Dominguez et al., J Am Chem Soc 2003;125:1731-1737), to the targets in the fourth and fifth rounds of CAPRI. Here we describe our results and analyze them in detail. Special attention is given to the role of flexibility in our docking method and the way in which this improves the docking results. We describe extensions to our approach that were developed as a direct result of our participation in CAPRI. In addition to experimental information, we also included interface residue predictions from PPISP (Protein-Protein Interaction Site Predictor; Zhou and Shan, Proteins 2001;44:336-343), a neural network method. Using HADDOCK we were able to generate acceptable structures for 6 of the 8 targets, and to submit at least 1 acceptable structure for 5 of them. Of these 5 submissions, 3 were of medium quality (Targets 10, 11, and 15) and 2 of high quality (Targets 13 and 14). In all cases, predictions were obtained containing at least 40% of the correct epitope at the interface for both ligand and receptor simultaneously.     

Solvated protein–DNA docking using HADDOCK

Interfacial water molecules play an important role in many aspects of protein–DNA specificity and recognition. Yet they have been mostly neglected in the computational modeling of these complexes. We present here a solvated docking protocol that allows explicit inclusion of water molecules in the docking of protein–DNA complexes and demonstrate its feasibility on a benchmark of 30 high-resolution protein–DNA complexes containing crystallographically-determined water molecules at their interfaces. Our protocol is capable of reproducing the solvation pattern at the interface and recovers hydrogen-bonded water-mediated contacts in many of the benchmark cases. Solvated docking leads to an overall improvement in the quality of the generated protein–DNA models for cases with limited conformational change of the partners upon complex formation. The applicability of this approach is demonstrated on real cases by docking a representative set of 6 complexes using unbound protein coordinates, model-built DNA and knowledge-based restraints. As HADDOCK supports the inclusion of a variety of NMR restraints, solvated docking is also applicable for NMR-based structure calculations of protein–DNA complexes.     

CPORT: a consensus interface predictor and its performance in prediction-driven docking with HADDOCK

Background

Macromolecular complexes are the molecular machines of the cell. Knowledge at the atomic level is essential to understand and influence their function. However, their number is huge and a significant fraction is extremely difficult to study using classical structural methods such as NMR and X-ray crystallography. Therefore, the importance of large-scale computational approaches in structural biology is evident. This study combines two of these computational approaches, interface prediction and docking, to obtain atomic-level structures of protein-protein complexes, starting from their unbound components.

Methodology/Principal Findings

Here we combine six interface prediction web servers into a consensus method called CPORT (Consensus Prediction Of interface Residues in Transient complexes). We show that CPORT gives more stable and reliable predictions than each of the individual predictors on its own. A protocol was developed to integrate CPORT predictions into our data-driven docking program HADDOCK. For cases where experimental information is limited, this prediction-driven docking protocol presents an alternative to ab initio docking, the docking of complexes without the use of any information. Prediction-driven docking was performed on a large and diverse set of protein-protein complexes in a blind manner. Our results indicate that the performance of the HADDOCK-CPORT combination is competitive with ZDOCK-ZRANK, a state-of-the-art ab initio docking/scoring combination. Finally, the original interface predictions could be further improved by interface post-prediction (contact analysis of the docking solutions).

Conclusions/Significance

The current study shows that blind, prediction-driven docking using CPORT and HADDOCK is competitive with ab initio docking methods. This is encouraging since prediction-driven docking represents the absolute bottom line for data-driven docking: any additional biological knowledge will greatly improve the results obtained by prediction-driven docking alone. Finally, the fact that original interface predictions could be further improved by interface post-prediction suggests that prediction-driven docking has not yet been pushed to the limit. A web server for CPORT is freely available at http://haddock.chem.uu.nl/services/CPORT.     

Condensation by DNA looping facilitates transfer of large DNA molecules into mammalian cells

Experimental studies of complete mammalian genes and other genetic domains are impeded by the difficulty of introducing large DNA molecules into cells in culture. Previously we have shown that GST–Z2, a protein that contains three zinc fingers and a proline-rich multimerization domain from the polydactyl zinc finger protein RIP60 fused to glutathione S-transferase (GST), mediates DNA binding and looping in vitro. Atomic force microscopy showed that GST–Z2 is able to condense 130–150 kb bacterial artificial chromosomes (BACs) into protein–DNA complexes containing multiple DNA loops. Condensation of the DNA loops onto the Z2 protein–BAC DNA core complexes with cationic lipid resulted in particles that were readily transferred into multiple cell types in culture. Transfer of total genomic linear DNA containing amplified DHFR genes into DHFR^– cells by GST–Z2 resulted in a 10-fold higher transformation rate than calcium phosphate co-precipitation. Chinese hamster ovarian cells transfected with a BAC containing the human TP53 gene locus expressed p53, showing native promoter elements are active after GST–Z2-mediated gene transfer. Because DNA condensation by GST–Z2 does not require the introduction of specific recognition sequences into the DNA substrate, condensation by the Z2 domain of RIP60 may be used in conjunction with a variety of other agents to provide a flexible and efficient non-viral platform for the delivery of large genes into mammalian cells.     

ParaDock: a flexible non-specific DNA--rigid protein docking algorithm

Accurate prediction of protein-DNA complexes could provide an important stepping stone towards a thorough comprehension of vital intracellular processes. Few attempts were made to tackle this issue, focusing on binding patch prediction, protein function classification and distance constraints-based docking. We introduce ParaDock: a novel ab initio protein-DNA docking algorithm. ParaDock combines short DNA fragments, which have been rigidly docked to the protein based on geometric complementarity, to create bent planar DNA molecules of arbitrary sequence. Our algorithm was tested on the bound and unbound targets of a protein-DNA benchmark comprised of 47 complexes. With neither addressing protein flexibility, nor applying any refinement procedure, CAPRI acceptable solutions were obtained among the 10 top ranked hypotheses in 83% of the bound complexes, and 70% of the unbound. Without requiring prior knowledge of DNA length and sequence, and within <2?h per target on a standard 2.0?GHz single processor CPU, ParaDock offers a fast ab initio docking solution.     

Quaternary structure predictions and estimation of mutational effects on the free energy of dimerization of the OMPLA protein

This study represents an extension to the outer membrane phospholipase A protein (OMPLA) of the docking-based protocols previously developed for quaternary structure predictions of transmembrane oligomeric proteins and for estimating mutational effects on the thermodynamics of protein–protein and protein–DNA association.Predictions of the likely architecture of OMPLA homo-dimers were carried out on 31 different forms of the monomer, 30 of which were variants of the unbound state. In all the test cases but the ones characterized by combined deletions of the 98–110 and 145–153 segments (L2 and L3, respectively), native-like complexes could be predicted, independent of the bound or unbound state of the structural model, of side chain conformation and presence or absence of amino acid deletions at the putative inter-monomer interface.The protocol for estimating mutational effects on the thermodynamics of protein–protein association proved effective as well. In fact, it was possible to estimate correctly the effects of five mutants on the free energy of dimerization of the sulfonylated form of OMPLA.The integrity of L2 and either one of the L1, L3 and L4 loops turned out to be more important than sulfonylation for the achievement of the native dimeric architecture. On the other hand, sulfonylation seems to be essential for a favorable dimerization energetics.     

Ultrafast interfacial solvation dynamics in specific protein DNA recognition

An overwhelming number of structural and functional studies on specific protein–DNA complexes reveal the existence of water molecules at the interaction interface. What role does the interfacial water molecules play in determining the specificity of association is thus a critical question. Herein, we have explored the dynamical role of minor groove water molecules and DNA side chain flexibility in lambda repressor–operator DNA interaction using well-characterized DNA minor groove binder dye, Hoechst 33258. The most striking finding of our studies reveals that the solvation time scale corresponding to the minor groove water molecules (∼50 ps) and DNA side chain flexibility (∼10 ns) remain unaltered even in protein–DNA complex in comparison to unbound operator DNA. The temperature dependent study further reveals the slower exchange of minor grove water molecules with bulk water in DNA–protein complex in comparison to the unbound DNA. Detailed structural studies including circular dichroism (CD) and Förster resonance energy transfer (FRET) have also been performed to elucidate the interaction between protein and DNA.     

Benchmarking AlphaFold for protein complex modeling reveals accuracy determinants

High‐resolution experimental structural determination of protein–protein interactions has led to valuable mechanistic insights, yet due to the massive number of interactions and experimental limitations there is a need for computational methods that can accurately model their structures. Here we explore the use of the recently developed deep learning method, AlphaFold, to predict structures of protein complexes from sequence. With a benchmark of 152 diverse heterodimeric protein complexes, multiple implementations and parameters of AlphaFold were tested for accuracy. Remarkably, many cases (43%) had near‐native models (medium or high critical assessment of predicted interactions accuracy) generated as top‐ranked predictions by AlphaFold, greatly surpassing the performance of unbound protein–protein docking (9% success rate for near‐native top‐ranked models), however AlphaFold modeling of antibody–antigen complexes within our set was unsuccessful. We identified sequence and structural features associated with lack of AlphaFold success, and we also investigated the impact of multiple sequence alignment input. Benchmarking of a multimer‐optimized version of AlphaFold (AlphaFold‐Multimer) with a set of recently released antibody–antigen structures confirmed a low rate of success for antibody–antigen complexes (11% success), and we found that T cell receptor–antigen complexes are likewise not accurately modeled by that algorithm, showing that adaptive immune recognition poses a challenge for the current AlphaFold algorithm and model. Overall, our study demonstrates that end‐to‐end deep learning can accurately model many transient protein complexes, and highlights areas of improvement for future developments to reliably model any protein–protein interaction of interest.     

Full-length RAG1 promotes contact with coding and intersignal sequences in RAG protein complexes bound to recombination signals paired in cis

The RAG proteins initiate V(D)J recombination by mediating synapsis and cleavage of two different antigen receptor gene segments through interactions with their flanking recombination signal sequences (RSS). The protein–DNA complexes that support this process have mainly been studied using RAG–RSS complexes assembled using oligonucleotide substrates containing a single RSS that are paired in trans to promote synapsis. How closely these complexes model those formed on longer, more physiologically relevant substrates containing RSSs on the same DNA molecule (in cis) remains unclear. To address this issue, we characterized discrete core and full-length RAG protein complexes bound to RSSs paired in cis. We find these complexes support cleavage activity regulated by V(D)J recombination's ‘12/23 rule’ and exhibit plasticity in RSS usage dependent on partner RSS composition. DNA footprinting studies suggest that the RAG proteins in these complexes mediate more extensive contact with sequences flanking the RSS than previously observed, some of which are enhanced by full-length RAG1, and associated with synapsis and efficient RSS cleavage. Finally, we demonstrate that the RAG1 C-terminus facilitates hairpin formation on long DNA substrates, and full-length RAG1 promotes hairpin retention in the postcleavage RAG complex. These results provide new insights into the mechanism of physiological V(D)J recombination.     

A knowledge-based scoring function for protein-RNA interactions derived from a statistical mechanics-based iterative method

Protein-RNA interactions play important roles in many biological processes. Given the high cost and technique difficulties in experimental methods, computationally predicting the binding complexes from individual protein and RNA structures is pressingly needed, in which a reliable scoring function is one of the critical components. Here, we have developed a knowledge-based scoring function, referred to as ITScore-PR, for protein-RNA binding mode prediction by using a statistical mechanics-based iterative method. The pairwise distance-dependent atomic interaction potentials of ITScore-PR were derived from experimentally determined protein–RNA complex structures. For validation, we have compared ITScore-PR with 10 other scoring methods on four diverse test sets. For bound docking, ITScore-PR achieved a success rate of up to 86% if the top prediction was considered and up to 94% if the top 10 predictions were considered, respectively. For truly unbound docking, the respective success rates of ITScore-PR were up to 24 and 46%. ITScore-PR can be used stand-alone or easily implemented in other docking programs for protein–RNA recognition.     

Conformer selection and induced fit in flexible backbone protein-protein docking using computational and NMR ensembles

Accommodating backbone flexibility continues to be the most difficult challenge in computational docking of protein-protein complexes. Towards that end, we simulate four distinct biophysical models of protein binding in RosettaDock, a multiscale Monte-Carlo-based algorithm that uses a quasi-kinetic search process to emulate the diffusional encounter of two proteins and to identify low-energy complexes. The four binding models are as follows: (1) key-lock (KL) model, using rigid-backbone docking; (2) conformer selection (CS) model, using a novel ensemble docking algorithm; (3) induced fit (IF) model, using energy-gradient-based backbone minimization; and (4) combined conformer selection/induced fit (CS/IF) model. Backbone flexibility was limited to the smaller partner of the complex, structural ensembles were generated using Rosetta refinement methods, and docking consisted of local perturbations around the complexed conformation using unbound component crystal structures for a set of 21 target complexes. The lowest-energy structure contained > 30% of the native residue-residue contacts for 9, 13, 13, and 14 targets for KL, CS, IF, and CS/IF docking, respectively. When applied to 15 targets using nuclear magnetic resonance ensembles of the smaller protein, the lowest-energy structure recovered at least 30% native residue contacts in 3, 8, 4, and 8 targets for KL, CS, IF, and CS/IF docking, respectively. CS/IF docking of the nuclear magnetic resonance ensemble performed equally well or better than KL docking with the unbound crystal structure in 10 of 15 cases. The marked success of CS and CS/IF docking shows that ensemble docking can be a versatile and effective method for accommodating conformational plasticity in docking and serves as a demonstration for the CS theory—that binding-competent conformers exist in the unbound ensemble and can be selected based on their favorable binding energies.     

An S/MAR-based infectious episomal genomic DNA expression vector provides long-term regulated functional complementation of LDLR deficiency

Episomal gene expression vectors offer a safe and attractive alternative to integrating vectors. Here we describe the development of a high capacity episomal vector system exploiting human episomal retention sequences to provide efficient vector maintenance and regulated gene expression through the delivery of a genomic DNA locus. The iBAC-S/MAR vector is capable of the infectious delivery and retention of large genomic DNA transgenes by exploiting the high transgene capacity of herpes simplex virus type 1 (HSV-1) and the episomal retention properties of the scaffold/matrix attachment region (S/MAR). The iBAC-S/MAR vector was used to deliver and maintain a 135kb genomic DNA insert carrying the human low density lipoprotein receptor (LDLR) genomic DNA locus at high efficiency in CHO ldlr^−/− a7 cells. Long-term studies on CHO ldlr^−/− a7 clonal cell lines carrying iBAC-S/MAR-LDLR demonstrated low copy episomal stability of the vector for >100 cell generations without selection. Expression studies demonstrated that iBAC-S/MAR-LDLR completely restored LDLR function in CHO ldlr^−/− a7 cells to physiological levels and that this expression can be repressed by ~70% by high sterol levels, recapitulating the same feedback regulation seen at the endogenous LDLR locus. This vector overcomes the major problems of vector integration and unregulated transgene expression.