Hierarchical protein folding pathways: a computational study of protein fragments

We have previously presented a building block folding model. The model postulates that protein folding is a hierarchical top-down process. The basic unit from which a fold is constructed, referred to as a hydrophobic folding unit, is the outcome of combinatorial assembly of a set of "building blocks." Results obtained by the computational cutting procedure yield fragments that are in agreement with those obtained experimentally by limited proteolysis. Here we show that as expected, proteins from the same family give very similar building blocks. However, different proteins can also give building blocks that are similar in structure. In such cases the building blocks differ in sequence, stability, contacts with other building blocks, and in their 3D locations in the protein structure. This result, which we have repeatedly observed in many cases, leads us to conclude that while a building block is influenced by its environment, nevertheless, it can be viewed as a stand-alone unit. For small-sized building blocks existing in multiple conformations, interactions with sister building blocks in the protein will increase the population time of the native conformer. With this conclusion in hand, it is possible to develop an algorithm that predicts the building block assignment of a protein sequence whose structure is unknown. Toward this goal, we have created sequentially nonredundant databases of building block sequences. A protein sequence can be aligned against these, in order to be matched to a set of potential building blocks.     

Transient, highly populated, building blocks folding model

Protein folding is a hierarchical event, in which transiently formed local structural elements assemble to yield the native conformation. In principle, multiple paths glide down the energy landscape, but, in practice, only a few of the paths are highly traveled. Here, the literature is reviewed in this light, and, particularly, a hierarchical, building block protein-folding model is presented, putting it in the context of a broad range of experimental and theoretical results published over the past few years. The model is based on two premises: First, although the local building block elements may be unstable, they nevertheless have higher population times than all alternate conformations; and, second, protein folding progresses through a combinatorial assembly of these elements. Through the binding of the most favorable building block conformers, there is a redistribution of the conformers in solution, propagating the protein-folding reaction. We describe the algorithm, and illustrate its usefulness, then we focus on its utility in assigning simple vs complex folding pathways, on chaperonin-assisted folding, on its relevance to domain-swapping processes, and on its relevance and relationship to disconnectivity graphs and tree diagrams. Considering protein folding as initiating from local transient structural elements is consistent with available experimental and theoretical results. Here, we have shown that, early in the folding process, sequential interactions are likely to take place, even if the final native fold is a complex, nonsequential one. Such a route is favorable kinetically and entropically. Through the construction of anatomy trees, the model enables derivation of the major folding pathways and their bumps, and qualitatively explains the kinetics of protein folding.     

Protein structure prediction via combinatorial assembly of sub-structural units

Following the hierarchical nature of protein folding, we propose a three-stage scheme for the prediction of a protein structure from its sequence. First, the sequence is cut to fragments that are each assigned a structure. Second, the assigned structures are combinatorially assembled to form the overall 3D organization. Third, highly ranked predicted arrangements are completed and refined. This work focuses on the second stage of this scheme: the combinatorial assembly. We present CombDock, a combinatorial docking algorithm. CombDock gets an ordered set of protein sub-structures and predicts the inter-contacts that define their overall organization. We reduce the combinatorial assembly to a graph-theory problem, and give a heuristic polynomial solution to this computationally hard problem. We applied CombDock to various examples of structural units of two types: protein domains and building blocks, which are relatively stable sub-structures of domains. Moreover, we tested CombDock using increasingly distorted input, where the native structural units were replaced by similarly folded units extracted from homologous proteins and, in the more difficult cases, from globally unrelated proteins. The algorithm is robust, showing low sensitivity to input distortion. This suggests that CombDock is a useful tool in protein structure prediction that may be applied to large target proteins.     

Potential folding-function interrelationship in proteins

The possibility is addressed that protein folding and function may be related via regions that are critical for both folding and function. This approach is based on the building blocks folding model that describes protein folding as binding events of conformationally fluctuating building blocks. Within these, we identify building block fragments that are critical for achieving the native fold. A library of such critical building blocks (CBBs) is constructed. Then, it is asked whether the functionally important residues fall in these CBB fragments. We find that for over two-thirds of the proteins in our library with available functional information, the catalytic or binding site residues lie within the CBB regions. From the evolutionary standpoint, a folding-function relationship is advantageous, since the need to guard against mutations is limited to one region. Furthermore, conformationally similar CBBs are found in globally unrelated proteins with different functions. Hence, substituting CBBs may lead to designed proteins with altered functions. We further find that the CBBs in our library are conformationally unstable.     

Structure by design: from single proteins and their building blocks to nanostructures

Nanotechnology realizes the advantages of naturally occurring biological macromolecules and their building-block nature for design. Frequently, assembly starts with the choice of a "good" molecule that is synthetically optimized towards the desired shape. By contrast, we propose starting with a pre-specified nanostructure shape, selecting candidate protein building blocks from a library and mapping them onto the shape and, finally, testing the stability of the construct. Such a shape-based, part-assembly strategy is conceptually similar to protein design through the combinatorial assembly of building blocks. If the conformational preferences of the building blocks are retained and their interactions are favorable, the nanostructure will be stable. The richness of the conformations, shapes and chemistries of the protein building blocks suggests a broad range of potential applications; at the same time, it also highlights their complexity. In this Opinion article, we focus on the first step: validating such a strategy against experimental data.     

In silico protein design by combinatorial assembly of protein building blocks

Utilizing concepts of protein building blocks, we propose a de novo computational algorithm that is similar to combinatorial shuffling experiments. Our goal is to engineer new naturally occurring folds with low homology to existing proteins. A selected protein is first partitioned into its building blocks based on their compactness, degree of isolation from the rest of the structure, and hydrophobicity. Next, the protein building blocks are substituted by fragments taken from other proteins with overall low sequence identity, but with a similar hydrophobic/hydrophilic pattern and a high structural similarity. These criteria ensure that the designed protein has a similar fold, low sequence identity, and a good hydrophobic core compared with its native counterpart. Here, we have selected two proteins for engineering, protein G B1 domain and ubiquitin. The two engineered proteins share approximately 20% and approximately 25% amino acid sequence identities with their native counterparts, respectively. The stabilities of the engineered proteins are tested by explicit water molecular dynamics simulations. The algorithm implements a strategy of designing a protein using relatively stable fragments, with a high population time. Here, we have selected the fragments by searching for local minima along the polypeptide chain using the protein building block model. Such an approach provides a new method for engineering new proteins with similar folds and low homology.     

A permissive secondary structure-guided superposition tool for clustering of protein fragments toward protein structure prediction via fragment assembly

MOTIVATION: Secondary-Structure Guided Superposition tool (SSGS) is a permissive secondary structure-based algorithm for matching of protein structures and in particular their fragments. The algorithm was developed towards protein structure prediction via fragment assembly. RESULTS: In a fragment-based structural prediction scheme, a protein sequence is cut into building blocks (BBs). The BBs are assembled to predict their relative 3D arrangement. Finally, the assemblies are refined. To implement this prediction scheme, a clustered structural library representing sequence patterns for protein fragments is essential. To create a library, BBs generated by cutting proteins from the PDB are compared and structurally similar BBs are clustered. To allow structural comparison and clustering of the BBs, which are often relatively short with flexible loops, we have devised SSGS. SSGS maintains high similarity between cluster members and is highly efficient. When it comes to comparing BBs for clustering purposes, the algorithm obtains better results than other, non-secondary structure guided protein superimposition algorithms.     

Distinguishing between sequential and nonsequentially folded proteins: implications for folding and misfolding.

We describe here an algorithm for distinguishing sequential from nonsequentially folding proteins. Several experiments have recently suggested that most of the proteins that are synthesized in the eukaryotic cell may fold sequentially. This proposed folding mechanism in vivo is particularly advantageous to the organism. In the absence of chaperones, the probability that a sequentially folding protein will misfold is reduced significantly. The problem we address here is devising a procedure that would differentiate between the two types of folding patterns. Footprints of sequential folding may be found in structures where consecutive fragments of the chain interact with each other. In such cases, the folding complexity may be viewed as being lower. On the other hand, higher folding complexity suggests that at least a portion of the polypeptide backbone folds back upon itself to form three-dimensional (3D) interactions with noncontiguous portion(s) of the chain. Hence, we look at the mechanism of folding of the molecule via analysis of its complexity, that is, through the 3D interactions formed by contiguous segments on the polypeptide chain. To computationally splice the structure into consecutively interacting fragments, we either cut it into compact hydrophobic folding units or into a set of hypothetical, transient, highly populated, contiguous fragments ("building blocks" of the structure). In sequential folding, successive building blocks interact with each other from the amino to the carboxy terminus of the polypeptide chain. Consequently, the results of the parsing differentiate between sequentially vs. nonsequentially folded chains. The automated assessment of the folding complexity provides insight into both the likelihood of misfolding and the kinetic folding rate of the given protein. In terms of the funnel free energy landscape theory, a protein that truly follows the mechanism of sequential folding, in principle, encounters smoother free energy barriers. A simple sequentially folded protein should, therefore, be less error prone and fold faster than a protein with a complex folding pattern.     

Building blocks, hinge-bending motions and protein topology.

Here we show that the locations of molecular hinges in protein structures fall between building block elements. Building blocks are fragments of the protein chain which constitute local minima. These elements fold first. In the next step they associate through a combinatorial assembly process. While chain-linked building blocks may be expected to trial-associate first, if unstable, alternate more stable associations will take place. Hence, we would expect that molecular hinges will be at such inter-building block locations, or at the less stable, unassigned regions. On the other hand, hinge-bending motions are well known to be critical for protein function. Hence, protein folding and protein function are evolutionarily related. Further, the pathways through which proteins attain their three dimensional folds are determined by protein topology. However, at the same time the locations of the hinges, and hinge-bending motions are also an outcome of protein topology. Thus, protein folding and function appear coupled, and relate to protein topology. Here we provide some results illustrating such a relationship.     

A phage display system for studying the sequence determinants of protein folding.

We have developed a phage display system that provides a means to select variants of the IgG binding domain of peptostreptococcal protein L that fold from large combinatorial libraries. The premise underlying the selection scheme is that binding of protein L to IgG requires that the protein be properly folded. Using a combination of molecular biological and biophysical methods, we show that this assumption is valid. First, the phage selection procedure strongly selects against a point mutation in protein L that disrupts folding but is not in the IgG binding interface. Second, variants recovered from a library in which the first third of protein L was randomized are properly folded. The degree of sequence variation in the selected population is striking: the variants have as many as nine substitutions in the 14 residues that were mutagenized. The approach provides a selection for "foldedness" that is potentially applicable to any small binding protein.     

Building Blocks,Hinge-Bending Motions and Protein Topology

Abstract

Here we show that the locations of molecular hinges in protein structures fall between building block elements. Building blocks are fragments of the protein chain which constitute local minima. These elements fold first. In the next step they associate through a combinatorial assembly process. While chain-linked building blocks may be expected to trial-associate first, if unstable, alternate more stable associations will take place. Hence, we would expect that molecular hinges will be at such inter-building block locations, or at the less stable, ‘unassigned’ regions.

On the other hand, hinge-bending motions are well known to be critical for protein function. Hence, protein folding and protein function are evolutionarily related. Further, the pathways through which proteins attain their three dimensional folds are determined by protein topology. However, at the same time the locations of the hinges, and hinge-bending motions are also an outcome of protein topology. Thus, protein folding and function appear coupled, and relate to protein topology. Here we provide some results illustrating such a relationship.     

Protein folding: binding of conformationally fluctuating building blocks via population selection.

Here we review different aspects of the protein folding literature. We present a broad range of observations, showing them to be consistent with a general hierarchical protein folding model. In such a model, local relatively stable, conformationally fluctuating building blocks bind through population selection, to yield the native state. The model includes several components: (1) the fluctuating building blocks that constitute local minima along the polypeptide chain, which even if unstable still possess higher population times than all alternate conformations; (2) the landscape around the bottom of the funnels; (3) the consideration that protein folding involves intramolecular recognition; (4) similar landscapes are observed for folding and for binding, and that (5) the landscape is dynamic, changing with the conditions. The model considers protein folding to be guided by native interactions. The reviewed literature includes the effects of changing the conditions, intermediates and kinetic traps, mutations, similar topologies, fragment complementation experiments, fragments and pathways, focusing on one specific well-studied example, that of the dihydrofolate reductase, chaperones, and chaperonines, in vivo vs. in vitro folding, still using the dihydrofolate example, amyloid formation, and molecular "disorder". These are consistent with the view that binding and folding are similar events, with the differences stemming from different stabilities and hence population times.     

Learning about protein folding via potential functions

Over the last few years we have developed an empirical potential function that solves the protein structure recognition problem: given the sequence for an n-residue globular protein and a collection of plausible protein conformations, including the native conformation for that sequence, identify the correct, native conformation. Having determined this potential on the basis of only some 6500 native/nonnative pairs of structures for 58 proteins, we find it recognizes the native conformation for essentially all compact, soluble, globular proteins having known native conformations in comparisons with 10⁴ to 10⁶ reasonable alternative conformations apiece. In this sense, the potential encodes nearly all the essential features of globular protein conformational preference. In addition it “knows” about many additional factors in protein folding, such as the stabilization of multimeric proteins, quaternary structure, the role of disulfide bridges and ligands, proproteins vs. processed proteins, and minimal strand lengths in globular proteins. Comparisons are made with other sorts of protein folding problems, and applications in protein conformational determination and prediction are discussed. © 1994 Wiley-Liss, Inc.     

Protein folding and function: the N-terminal fragment in adenylate kinase

Three-dimensional protein folds range from simple to highly complex architectures. In complex folds, some building block fragments are more important for correct protein folding than others. Such fragments are typically buried in the protein core and mediate interactions between other fragments. Here we present an automated, surface area-based algorithm that is able to indicate which, among all local elements of the structure, is critical for the formation of the native fold, and apply it to structurally well-characterized proteins. In particular, we focus on adenylate kinase. The fragment containing the phosphate binding, P-loop (the "giant anion hole") flanked by a beta-strand and an alpha-helix near the N-terminus, is identified as a critical building block. This building block shows a high degree of sequence and structural conservation in all adenylate kinases. The results of our molecular dynamics simulations are consistent with this identification. In its absence, the protein flips to a stable, non-native state. In this misfolded conformation, the other local elements of the structure are in their native-like conformations; however, their association is non-native. Furthermore, this element is critically important for the function of the enzyme, coupling folding, and function.     

Comparison of protein fragments identified by limited proteolysis and by computational cutting of proteins

Here we present a comparison between protein fragments produced by limited proteolysis and those identified by computational cutting based on the building block folding model. The principles upon which the two methods are based are different. Limited proteolysis of natively folded proteins occurs at flexible sites and never at the level of chain segments of regular secondary structure such as alpha-helices. Therefore, the targets for limited proteolysis are locally unfolded regions. In contrast, the computational cutting algorithm considers the compactness of the fragments, their nonpolar buried surface area, and their isolatedness, that is, the surface area which was buried prior to the cutting and becomes exposed subsequently. Despite the different criteria, there is an overall correspondence between sites or regions of limited proteolysis with those identified by computational cutting. The computational cutting method has been applied to several model proteins for which detailed limited proteolysis data are available, namely apomyoglobin, cytochrome c, ribonuclease A, alpha-lactalbumin, and thermolysin. As expected, more cuts are obtained computationally than experimentally and the agreement is better when a number of proteolytic enzymes are used. For example, cytochrome c is cleaved by thermolysin at 56-57, 45-46, and at 80-81, and by proteinase K at 48-49 and 50-51. Incubation of the noncovalent and native-like complex of cytochrome c fragments 1-56 and 57-104 with proteinase K yielded the gapped protein species 1-48/57-104 and finally 1-40/57-104. Computational cutting of cytochrome c reproduced the major experimental observations, with cuts at 47, 64-65 or 65-66 and 80-81 and an unstable 32-47 region not assigned to any building block. The next step, not addressed in this work, is to probe the ability of the generated fragments to fold independently. Since both the computational algorithm and limited proteolysis attempt to dissect the protein folding problem, the general agreement between the two procedures is gratifying. This consistency allows us to propose the use of limited proteolysis to produce protein fragments that can adopt an independent folding and, therefore, to study folding intermediates. The results of the present study appear to validate the building block folding model and are in line with the proposal that protein folding is a hierarchical process, where parts constituting local minima of energy fold first, with their subsequent association and mutual stabilization to finally yield the global fold.     

Binding and folding: in search of intramolecular chaperone-like building block fragments

We propose an intramolecular chaperone which catalyzes folding and neither dissociates nor is cleaved. This uncleaved foldase is an intramolecular chain-linked chaperone, which constitutes a critical building block of the structure. Macroscopically, all molecular chaperones facilitate folding reactions and manifest similar energy landscapes. However, microscopically they differ. While intermolecular chaperones catalyze folding by unfolding misfolded conformations or prevent misfolding, the chain-linked cleaved (proregion) and uncleaved intramolecular chaperone-like building blocks suggested here, catalyze folding by binding to, stabilizing and increasing the populations of native conformations of adjacent building block fragments. In both, the more stable the intramolecular chaperone fragment region, the faster is the folding rate. Hence, mechanistically, intramolecular chaperones and chaperone-like segments are similar. Both play a dual role, in folding and in protein function. However, while the functional role of the proregions is inhibitory, necessitating their cleavage, the function of the uncleaved intramolecular chaperone-like building blocks does not require their subsequent removal. On the contrary, it requires that they remain in the structure. This may lead to the difference in the type of control they are under: proteins folding with the assistance of the proregion have been shown to be under kinetic control. It has been suggested that kinetically controlled folding reactions, with the proregion catalyst removed, lend longevity under harsh conditions. On the other hand, proteins with uncleaved intramolecular chaperone-like building blocks, with their 'foldases' still attached, are largely under thermodynamic control, consistent with the control observed in most protein folding reactions. We propose that an uncleaved intramolecular chaperone-like fragment occurs frequently in proteins. We further propose that such proteins would be prone to changing conditions and in particular, to mutations in this critical building block region. We describe the features qualifying it for its proposed chaperone-like role, compare it with inter- and intramolecular chaperones and review current literature in this light. We further propose a mechanism showing how it lowers the barrier heights, leading to faster folding reaction rates. Since these fragments constitute an intergal part of the protein structure, we call these critical building blocks intramolecular, chaperone-like fragments, to clarify, distinguish and adhere to the definition of the transiently associating chaperones. The new mechanism presented here differs from the concept of 'folding nuclei'. While the concept of folding nuclei focuses on a non-sequential distribution of the folding information along the entire protein chain, the chaperone-like building block fragments proposition focuses on a segmental distribution of the folding information. This segmental distribution controls the distributions of the populations throughout the hierarchical folding processes.     

Subdomain interactions as a determinant in the folding and stability of T4 lysozyme.

The folding of large, multidomain proteins involves the hierarchical assembly of individual domains. It remains unclear whether the stability and folding of small, single-domain proteins occurs through a comparable assembly of small, autonomous folding units. We have investigated the relationship between two subdomains of the protein T4 lysozyme. Thermodynamically, T4 lysozyme behaves as a cooperative unit and the unfolding transition fits a two-state model. The structure of the protein, however, resembles a dumbbell with two potential subdomains: an N-terminal subdomain (residues 13-75), and a C-terminal subdomain (residues 76-164 and 1-12). To investigate the effect of uncoupling these two subdomains within the context of the native protein, we created two circular permutations, both at the subdomain interface (residues 13 and 75). Both variants adopt an active wild-type T4 lysozyme fold. The protein starting with residue 13 is 3 kcal/mol less stable than wild type, whereas the protein beginning at residue 75 is 9 kcal/mol less stable, suggesting that the placement of the termini has a major effect on protein stability while minimally affecting the fold. When isolated as protein fragments, the C-terminal subdomain folds into a marginally stable helical structure, whereas the N-terminal subdomain is predominantly unfolded. ANS fluorescence studies indicate that, at low pH, the C-terminal subdomain adopts a loosely packed acid state. An acid state intermediate is also seen for all of the full-length variants. We propose that this acid state is comprised of an unfolded N-terminal subdomain and a loosely folded C-terminal subdomain.     

Analysis on folding of misgurin using two-dimensional HP model

Misgurin is an antimicrobial peptide from the loach, while the hydrophobic-polar (HP) model is a way to study the folding conformations and native states in peptide and protein although several amino acids cannot be classified either hydrophobic or polar. Practically, the HP model requires extremely intensive computations, thus it has yet to be used widely. In this study, we use the two-dimensional HP model to analyze all possible folding conformations and native states of misgurin with conversion of natural amino acids according to the normalized amino acid hydrophobicity index as well as the shortest benchmark HP sequence. The results show that the conversion of misgurin into HP sequence with glycine as hydrophobic amino acid at pH 2 has 1212 folding conformations with the same native state of minimal energy -6; the conversion of glycine as polar amino acid at pH 2 has 13,386 folding conformations with three native states of minimal energy -5; the conversion of glycine as hydrophobic amino acid at pH 7 has 2538 folding conformations with three native states of minimal energy -5; and the conversion of glycine as polar amino acid at pH 7 has 12,852 folding conformations with three native states of minimal energy -4. Those native states can be ranked according to the normalized amino acid hydrophobicity index. The detailed discussions suggest two ways to modify misgurin.     

Exploring zipping and assembly as a protein folding principle

It has been proposed that proteins fold by a process called "Zipping and Assembly" (Z&A). Zipping refers to the growth of local substructures within the chain, and assembly refers to the coming together of already-formed pieces. Our interest here is in whether Z&A is a general method that can fold most of sequence space, to global minima, efficiently. Using the HP model, we can address this question by enumerating full conformation and sequence spaces. We find that Z&A reaches the global energy minimum native states, even though it searches only a very small fraction of conformational space, for most sequences in the full sequence space. We find that Z&A, a mechanism-based search, is more efficient in our tests than the replica exchange search method. Folding efficiency is increased for chains having: (a) small loop-closure steps, consistent with observations by Plaxco et al. 1998;277;985-994 that folding rates correlate with contact order, (b) neither too few nor too many nucleation sites per chain, and (c) assembly steps that do not occur too early in the folding process. We find that the efficiency increases with chain length, although our range of chain lengths is limited. We believe these insights may be useful for developing faster protein conformational search algorithms.     

Protein structure prediction using a docking-based hierarchical folding scheme

The pathways by which proteins fold into their specific native structure are still an unsolved mystery. Currently, many methods for protein structure prediction are available, and most of them tackle the problem by relying on the vast amounts of data collected from known protein structures. These methods are often not concerned with the route the protein follows to reach its final fold. This work is based on the premise that proteins fold in a hierarchical manner. We present FOBIA, an automated method for predicting a protein structure. FOBIA consists of two main stages: the first finds matches between parts of the target sequence and independently folding structural units using profile-profile comparison. The second assembles these units into a 3D structure by searching and ranking their possible orientations toward each other using a docking-based approach. We have previously reported an application of an initial version of this strategy to homology based targets. Since then we have considerably enhanced our method's abilities to allow it to address the more difficult template-based target category. This allows us to now apply FOBIA to the template-based targets of CASP8 and to show that it is both very efficient and promising. Our method can provide an alternative for template-based structure prediction, and in particular, the docking-basedranking technique presented here can be incorporated into any profile-profile comparison based method.