Detecting hidden sequence propensity for amyloid fibril formation

Forced amyloidogenic cooperativity of structurally incompatible peptide segments: Fibrillization behavior of highly aggregation-prone A-chain fragment of insulin coupled to all-L, and alternating L/D octaglutamates

Aggregation of proteins into amyloid fibrils is driven by interactions between relatively small amyloidogenic segments. The interplay between aggregation-prone and aggregation-resistant fragments within a single polypeptide chain remains obscure. Here, we examine fibrillization behavior of two chimeric peptides, ACC_1-13E₈ and ACC_1-13E_8(L/D), in which the highly amyloidogenic fragment of insulin (ACC_1-13) is extended by an octaglutamate segment composed of all-L (E₈), or alternating L/D residues (E_8(L/D)). As separate entities, ACC_1-13 readily forms fibrils with the infrared features of parallel β-sheet while E₈ forms antiparallel β-sheets with the distinct infrared characteristics. This contrasts with the profoundly aggregation-resistant E_8(L/D), although L/D patterns have been hypothesized as compatible with aggregated α-sheets. ACC_1-13E₈ and ACC_1-13E_8(L/D) are found to be equally prone to fibrillization at low pH, or in the presence of Ca²⁺ ions. Fibrillar states of both ACC_1-13E₈ and ACC_1-13E_8(L/D) reveal the infrared features of highly ordered parallel β-sheet without evidence of β₂-aggregates (ACC_1-13E₈) or α-sheets (ACC_1-13E_8(L/D)). Hence, the preferred structural pattern of ACC_1-13 overrides the tendency of E₈ to form antiparallel β-sheets and enforces the fibrillar order in E_8(L/D). We demonstrate how the powerful amyloid stretch determines the overall amyloid structure forcing non-amyloidogenic fragments to participate in its native amyloid pattern.

DES-Amyloidoses “Amyloidoses through the looking-glass”: A knowledgebase developed for exploring and linking information related to human amyloid-related diseases

More than 30 types of amyloids are linked to close to 50 diseases in humans, the most prominent being Alzheimer’s disease (AD). AD is brain-related local amyloidosis, while another amyloidosis, such as AA amyloidosis, tends to be more systemic. Therefore, we need to know more about the biological entities’ influencing these amyloidosis processes. However, there is currently no support system developed specifically to handle this extraordinarily complex and demanding task. To acquire a systematic view of amyloidosis and how this may be relevant to the brain and other organs, we needed a means to explore "amyloid network systems" that may underly processes that leads to an amyloid-related disease. In this regard, we developed the DES-Amyloidoses knowledgebase (KB) to obtain fast and relevant information regarding the biological network related to amyloid proteins/peptides and amyloid-related diseases. This KB contains information obtained through text and data mining of available scientific literature and other public repositories. The information compiled into the DES-Amyloidoses system based on 19 topic-specific dictionaries resulted in 796,409 associations between terms from these dictionaries. Users can explore this information through various options, including enriched concepts, enriched pairs, and semantic similarity. We show the usefulness of the KB using an example focused on inflammasome-amyloid associations. To our knowledge, this is the only KB dedicated to human amyloid-related diseases derived primarily through literature text mining and complemented by data mining that provides a novel way of exploring information relevant to amyloidoses.

Heterotypic amyloid interactions: Clues to polymorphic bias and selective cellular vulnerability?

The number of atomic-resolution structures of disease-associated amyloids has greatly increased in recent years. These structures have confirmed not only the polymorphic nature of amyloids but also the association of specific polymorphs to particular proteinopathies. These observations are strengthening the view that amyloid polymorphism is a marker for specific pathological subtypes (e.g. in tauopathies or synucleinopathies). The nature of this association and how it relates to the selective cellular vulnerability of amyloid nucleation, propagation and toxicity are still unclear. Here, we provide an overview of the mechanistic insights provided by recent patient-derived amyloid structures. We discuss the framework organisation of amyloid polymorphism and how heterotypic amyloid interactions with the physiological environment could modify the solubility and assembly of amyloidogenic proteins. We conclude by hypothesising how such interactions could contribute to selective cellular vulnerability.

Challenges for design of aggregation-resistant variants of granulocyte colony-stimulating factor

Non-native protein aggregation is a long-standing issue in pharmaceutical biotechnology. A rational design approach was used in order to identify variants of recombinant human granulocyte colony-stimulating factor (rhG-CSF) with lower aggregation propensity at solution conditions that are typical of commercial formulation. The approach used aggregation-prone-region (APR) predictors to select single amino acid substitutions that were predicted to decrease intrinsic aggregation propensity (IAP). The results of static light scattering temperature-ramps and chemical unfolding experiments demonstrated that none of the selected variants exhibited improved aggregation resistance, and the apparent conformational stability of each variant was lower than that of WT. Aggregation studies under partly denaturing conditions suggested that the IAP of at least one variant remained unaltered. Overall, this study highlights a general challenge in designing aggregation resistance for proteins, due to the need to accurately predict both APRs and conformational stability.

A theoretical study of polymorphism in VQIVYK fibrils

The VQIVYK fragment from the Tau protein, also known as PHF6, is essential for aggregation of Tau into neurofibrillary lesions associated with neurodegenerative diseases. VQIVYK itself forms amyloid fibrils composed of paired β-sheets. Therefore, the full Tau protein and VQIVYK fibrils have been intensively investigated. A central issue in these studies is polymorphism, the ability of a protein to fold into more than one structure. Using all-atom molecular simulations, we generate five stable polymorphs of VQIVYK fibrils, establish their relative free energy with umbrella sampling methods, and identify the side chain interactions that provide stability. The two most stable polymorphs, which have nearly equal free energy, are formed by interdigitation of the mostly hydrophobic VIY "face" sides of the β-sheets. Another stable polymorph is formed by interdigitation of the QVK "back" sides. When we turn to examine structures from cryo-electron microscopy experiments on Tau filaments taken from diseased patients or generated in vitro, we find that the pattern of side chain interactions found in the two most stable face-to-face as well as the back-to-back polymorphs are recapitulated in amyloid structures of the full protein. Thus, our studies suggest that the interactions stabilizing PHF6 fibrils explain the amyloidogenicity of the VQIVYK motif within the full Tau protein and provide justification for the use of VQIVYK fibrils as a test bed for the design of molecules that identify or inhibit amyloid structures.

Structure-based machine-guided mapping of amyloid sequence space reveals uncharted sequence clusters with higher solubilities

The amyloid conformation can be adopted by a variety of sequences, but the precise boundaries of amyloid sequence space are still unclear. The currently charted amyloid sequence space is strongly biased towards hydrophobic, beta-sheet prone sequences that form the core of globular proteins and by Q/N/Y rich yeast prions. Here, we took advantage of the increasing amount of high-resolution structural information on amyloid cores currently available in the protein databank to implement a machine learning approach, named Cordax (https://cordax.switchlab.org), that explores amyloid sequence beyond its current boundaries. Clustering by t-Distributed Stochastic Neighbour Embedding (t-SNE) shows how our approach resulted in an expansion away from hydrophobic amyloid sequences towards clusters of lower aliphatic content and higher charge, or regions of helical and disordered propensities. These clusters uncouple amyloid propensity from solubility representing sequence flavours compatible with surface-exposed patches in globular proteins, functional amyloids or sequences associated to liquid-liquid phase transitions.

An increasing number of amyloid structures are determined. Here, the authors present the structure-based amyloid core sequence prediction method Cordax that is based on machine learning and allows the detection of aggregation-prone regions with higher solubility, disorder and surface exposure, and furthermore predicts the structural topology, orientation and overall architecture of the resulting putative fibril core.

In vivo localization of human acetylcholinesterase-derived species in a β-sheet conformation at the core of senile plaques in Alzheimer's disease

Many neurodegenerative diseases are characterized by amyloid deposition. In Alzheimer's disease (AD), β-amyloid (Aβ) peptides accumulate extracellularly in senile plaques. The AD amyloid cascade hypothesis proposes that Aβ production or reduced clearance leads to toxicity. In contrast, the cholinergic hypothesis argues for a specific pathology of brain cholinergic pathways. However, neither hypothesis in isolation explains the pattern of AD pathogenesis. Evidence suggests that a connection exists between these two scenarios: the synaptic form of human acetylcholinesterase (hAChE-S) associates with plaques in AD brains; among hAChE variants, only hAChE-S enhances Aβ fibrillization in vitro and Aβ deposition and toxicity in vivo Only hAChE-S contains an amphiphilic C-terminal domain (T40, AChE_575-614), with AChE_586-599 homologous to Aβ and forming amyloid fibrils, which implicates T40 in AD pathology. We previously showed that the amyloid scavenger, insulin-degrading enzyme (IDE), generates T40-derived amyloidogenic species that, as a peptide mixture, seed Aβ fibrillization. Here, we characterized 11 peptides from a T40-IDE digest for β-sheet conformation, surfactant activity, fibrillization, and seeding capability. We identified residues important for amyloidogenicity and raised polyclonal antibodies against the most amyloidogenic peptide. These new antisera, alongside other specific antibodies, labeled sections from control, hAChE-S, hAPPswe, and hAChE-S/hAPPswe transgenic mice. We observed that hAChE-S β-sheet species co-localized with Aβ in mature plaque cores, surrounded by hAChE-S α-helical species. This observation provides the first in vivo evidence of the conformation of hAChE-S species within plaques. Our results may explain the role of hAChE-S in Aβ deposition and aggregation, as amyloidogenic hAChE-S β-sheet species might seed Aβ aggregation.

RFAmyloid: A Web Server for Predicting Amyloid Proteins

Amyloid is an insoluble fibrous protein and its mis-aggregation can lead to some diseases, such as Alzheimer’s disease and Creutzfeldt–Jakob’s disease. Therefore, the identification of amyloid is essential for the discovery and understanding of disease. We established a novel predictor called RFAmy based on random forest to identify amyloid, and it employed SVMProt 188-D feature extraction method based on protein composition and physicochemical properties and pse-in-one feature extraction method based on amino acid composition, autocorrelation pseudo acid composition, profile-based features and predicted structures features. In the ten-fold cross-validation test, RFAmy’s overall accuracy was 89.19% and F-measure was 0.891. Results were obtained by comparison experiments with other feature, classifiers, and existing methods. This shows the effectiveness of RFAmy in predicting amyloid protein. The RFAmy proposed in this paper can be accessed through the URL http://server.malab.cn/RFAmyloid/.

VLITL is a major cross-β-sheet signal for fibrinogen Aα-chain frameshift variants

The first case of hereditary fibrinogen Aα-chain amyloidosis was recognized >20 years ago, but disease mechanisms still remain unknown. Here we report detailed clinical and proteomics studies of a French kindred with a novel amyloidogenic fibrinogen Aα-chain frameshift variant, Phe521Leufs, causing a severe familial form of renal amyloidosis. Next, we focused our investigations to elucidate the molecular basis that render this Aα-chain variant amyloidogenic. We show that a 49-mer peptide derived from the C-terminal part of the Phe521Leufs chain is deposited as fibrils in the patient's kidneys, establishing that only a small portion of Phe521Leufs directly contributes to amyloid formation in vivo. In silico analysis indicated that this 49-mer Aα-chain peptide contained a motif (VLITL), with a high intrinsic propensity for β-aggregation at residues 44 to 48 of human renal fibrils. To experimentally verify the amyloid propensity of VLITL, we generated synthetic Phe521Leufs-derived peptides and compared their capacity for fibril formation in vitro with that of their VLITL-deleted counterparts. We show that VLITL forms typical amyloid fibrils in vitro and is a major signal for cross-β-sheet self-association of the 49-mer Phe521Leufs peptide identified in vivo, whereas its absence abrogates fibril formation. This study provides compelling evidence that VLITL confers amyloidogenic properties to Aα-chain frameshift variants, yielding a previously unknown molecular basis for the pathogenesis of Aα-chain amyloidosis.

Mechanism of Initiation, Association, and Formation of Amyloid Fibrils Modeled with the N-Terminal Peptide Fragment, IKYLEFIS, of Myoglobin G-Helix

Some peptides and proteins undergo self-aggregation under certain conditions, leading to amyloid fibrils formation, which is related to many disease conditions. It is important to understand such amyloid fibrils formation to provide mechanistic detail that governs the process. A predominantly α-helical myoglobin has been reported recently to readily form amyloid fibrils at a higher temperature, similar to its G-helix segment. Here, we have investigated the mechanism of amyloid fibrils formation by performing multiple long molecular dynamics simulations (27 μs) on the N-terminal segment of the G-helix of myoglobin. These simulations resulted in the formation of a single-layered tetrameric β-sheet with mixed parallel and antiparallel β-strands and this is the most common event irrespective of many different starting structures. Formation of the single-layered tetrameric β-sheet takes place following three distinctive pathways. The process of fibril initiation is dependent on temperature. Further, this study provides mechanistic insights into the formation of multilayered fibrilar structure, which could be applicable to a wider variety of peptides or proteins to understand the amyloidogenesis.

Prediction of Peptide and Protein Propensity for Amyloid Formation

Understanding which peptides and proteins have the potential to undergo amyloid formation and what driving forces are responsible for amyloid-like fiber formation and stabilization remains limited. This is mainly because proteins that can undergo structural changes, which lead to amyloid formation, are quite diverse and share no obvious sequence or structural homology, despite the structural similarity found in the fibrils. To address these issues, a novel approach based on recursive feature selection and feed-forward neural networks was undertaken to identify key features highly correlated with the self-assembly problem. This approach allowed the identification of seven physicochemical and biochemical properties of the amino acids highly associated with the self-assembly of peptides and proteins into amyloid-like fibrils (normalized frequency of β-sheet, normalized frequency of β-sheet from LG, weights for β-sheet at the window position of 1, isoelectric point, atom-based hydrophobic moment, helix termination parameter at position j+1 and ΔG° values for peptides extrapolated in 0 M urea). Moreover, these features enabled the development of a new predictor (available at http://cran.r-project.org/web/packages/appnn/index.html) capable of accurately and reliably predicting the amyloidogenic propensity from the polypeptide sequence alone with a prediction accuracy of 84.9 % against an external validation dataset of sequences with experimental in vitro, evidence of amyloid formation.

Predicting the aggregation propensity of prion sequences

The presence of prions can result in debilitating and neurodegenerative diseases in mammals and protein-based genetic elements in fungi. Prions are defined as a subclass of amyloids in which the self-aggregation process becomes self-perpetuating and infectious. Like all amyloids, prions polymerize into fibres with a common core formed of β-sheet structures oriented perpendicular to the fibril axes which form a structure known as a cross-β structure. The intermolecular β-sheet propensity, a characteristic of the amyloid pattern, as well as other key parameters of amyloid fibril formation can be predicted. Mathematical algorithms have been proposed to predict both amyloid and prion propensities. However, it has been shown that the presence of amyloid-prone regions in a polypeptide sequence could be insufficient for amyloid formation. It has also often been stated that the formation of amyloid fibrils does not imply that these are prions. Despite these limitations, in silico prediction of amyloid and prion propensities should help detect potential new prion sequences in mammals. In addition, the determination of amyloid-prone regions in prion sequences could be very useful in understanding the effect of sporadic mutations and polymorphisms as well as in the search for therapeutic targets.

Intrinsically semi-disordered state and its role in induced folding and protein aggregation

Intrinsically disordered proteins (IDPs) refer to those proteins without fixed three-dimensional structures under physiological conditions. Although experiments suggest that the conformations of IDPs can vary from random coils, semi-compact globules, to compact globules with different contents of secondary structures, computational efforts to separate IDPs into different states are not yet successful. Recently, we developed a neural-network-based disorder prediction technique SPINE-D that was ranked as one of the top performing techniques for disorder prediction in the biannual meeting of critical assessment of structure prediction techniques (CASP 9, 2010). Here, we further analyze the results from SPINE-D prediction by defining a semi-disordered state that has about 50% predicted probability to be disordered or ordered. This semi-disordered state is partially collapsed with intermediate levels of predicted solvent accessibility and secondary structure content. The relative difference in compositions between semi-disordered and fully disordered regions is highly correlated with amyloid aggregation propensity (a correlation coefficient of 0.86 if excluding four charged residues and proline, 0.73 if not). In addition, we observed that some semi-disordered regions participate in induced folding, and others play key roles in protein aggregation. More specifically, a semi-disordered region is amyloidogenic in fully unstructured proteins (such as alpha-synuclein and Sup35) but prone to local unfolding that exposes the hydrophobic core to aggregation in structured globular proteins (such as SOD1 and lysozyme). A transition from full disorder to semi-disorder at about 30-40 Qs is observed in the poly-Q (poly-glutamine) tract of huntingtin. The accuracy of using semi-disorder to predict binding-induced folding and aggregation is compared with several methods trained for the purpose. These results indicate the usefulness of three-state classification (order, semi-disorder, and full-disorder) in distinguishing nonfolding from induced-folding and aggregation-resistant from aggregation-prone IDPs and in locating weakly stable, locally unfolding, and potentially aggregation regions in structured proteins. A comparison with five representative disorder-prediction methods showed that SPINE-D is the only method with a clear separation of semi-disorder from ordered and fully disordered states.

Sequence complexity of amyloidogenic regions in intrinsically disordered human proteins

An amyloidogenic region (AR) in a protein sequence plays a significant role in protein aggregation and amyloid formation. We have investigated the sequence complexity of AR that is present in intrinsically disordered human proteins. More than 80% human proteins in the disordered protein databases (DisProt+IDEAL) contained one or more ARs. With decrease of protein disorder, AR content in the protein sequence was decreased. A probability density distribution analysis and discrete analysis of AR sequences showed that ∼8% residue in a protein sequence was in AR and the region was in average 8 residues long. The residues in the AR were high in sequence complexity and it seldom overlapped with low complexity regions (LCR), which was largely abundant in disorder proteins. The sequences in the AR showed mixed conformational adaptability towards α-helix, β-sheet/strand and coil conformations.

Computational and theoretical methods for protein folding

A computational approach is essential whenever the complexity of the process under study is such that direct theoretical or experimental approaches are not viable. This is the case for protein folding, for which a significant amount of data are being collected. This paper reports on the essential role of in silico methods and the unprecedented interplay of computational and theoretical approaches, which is a defining point of the interdisciplinary investigations of the protein folding process. Besides giving an overview of the available computational methods and tools, we argue that computation plays not merely an ancillary role but has a more constructive function in that computational work may precede theory and experiments. More precisely, computation can provide the primary conceptual clues to inspire subsequent theoretical and experimental work even in a case where no preexisting evidence or theoretical frameworks are available. This is cogently manifested in the application of machine learning methods to come to grips with the folding dynamics. These close relationships suggested complementing the review of computational methods within the appropriate theoretical context to provide a self-contained outlook of the basic concepts that have converged into a unified description of folding and have grown in a synergic relationship with their computational counterpart. Finally, the advantages and limitations of current computational methodologies are discussed to show how the smart analysis of large amounts of data and the development of more effective algorithms can improve our understanding of protein folding.

Short tandem repeats in the inhibitory domain of the mineralocorticoid receptor: prediction of a β-solenoid structure

Background

The human mineralocorticoid receptor (MR) is one of the main components of the renin-angiotensin-aldosterone system (RAAS), the system that regulates the body exchange of water and sodium. The evolutionary origins of this protein predate those of renin and the RAAS; accordingly it has other roles, which are being characterized. The MR has two trans-activating ligand independent domains and one inhibitory domain (ID), which modulates the activity of the former. The structure of the ID is currently unknown.

Results

Here we report that the ID contains at least 15 tandem repeats of around 10 amino acids, which we computationally characterize in the human MR and in selected orthologs. This ensemble of repeats seems to have emerged around 450 million years ago, after the divergence of the MR from its close homolog, the glucocorticoid receptor, which does not possess the repeats. The region would have quickly expanded by successive duplication of the repeats stabilizing at its length in human MR shortly after divergence of tetrapoda from bony fishes 400 million years ago. Structural predictions, in combination with molecular dynamics simulations suggest that the repeat ensemble forms a β-solenoid, namely a β-helical fold with a polar core, stabilized by hydrogen-bonded ladders of polar residues. Our 3D-model, in conjunction with previous experimental data, implies a role of the β-helical fold as a scaffold for multiple intra-and inter-molecular interactions and that these interactions are modulated via phosphorylation-dependent conformational changes.

Conclusions

We, thus, propose that the structure of the repeat ensemble plays an important role in the coordination and sequential interactions of various MR partners and therefore in the functionality and specificity of MR.

Energetic domains and conformational analysis of human serum albumin upon co-incubation with sodium benzoate and glucose

Sodium benzoate (SB), a powerful inhibitor of microbial growth, is one of the most commonly used food preservative. Here, we determined the effects of SB on human serum albumin (HSA) structure in the presence or absence of glucose after 35 days of incubation under physiological conditions. The biochemical, biophysical, and molecular approaches including free amine content assay (TNBSA assay), fluorescence, and circular dichroism spectroscopy (CD), differential scanning calorimetry (DSC), and molecular docking and LIGPLOT studies were utilized for structural studies. The TNBSA results indicated that SB has the ability to bind Lys residues in HSA through covalent bonds. The docking and LIGPLOT studies also determined another specific site via hydrophobic interactions. The CD results showed more structural helicity for HSA incubated with SB, while HSA incubated with glucose had the least, and HSA incubated with glucose + SB had medium helicity. Fluorescence spectrophotometry results demonstrated partial unfolding of HSA incubated with SB in the presence or absence of glucose, while maximum partial unfolding was observed in HSA incubated with glucose. These results were confirmed by DSC and its deconvoluted thermograms. The DSC results also showed significant changes in HSA energetic structural domains due to HSA incubation with SB in the presence or absence of glucose. Together, our studies showed the formation of three different intermediates and indicate that biomolecular investigation are effective in providing new insight into safety determinations especially in health-related conditions including diabetes.

A consensus method for the prediction of 'aggregation-prone' peptides in globular proteins

The purpose of this work was to construct a consensus prediction algorithm of ‘aggregation-prone’ peptides in globular proteins, combining existing tools. This allows comparison of the different algorithms and the production of more objective and accurate results. Eleven (11) individual methods are combined and produce AMYLPRED2, a publicly, freely available web tool to academic users (http://biophysics.biol.uoa.gr/AMYLPRED2), for the consensus prediction of amyloidogenic determinants/‘aggregation-prone’ peptides in proteins, from sequence alone. The performance of AMYLPRED2 indicates that it functions better than individual aggregation-prediction algorithms, as perhaps expected. AMYLPRED2 is a useful tool for identifying amyloid-forming regions in proteins that are associated with several conformational diseases, called amyloidoses, such as Altzheimer's, Parkinson's, prion diseases and type II diabetes. It may also be useful for understanding the properties of protein folding and misfolding and for helping to the control of protein aggregation/solubility in biotechnology (recombinant proteins forming bacterial inclusion bodies) and biotherapeutics (monoclonal antibodies and biopharmaceutical proteins).

Breaking the amyloidogenicity code: methods to predict amyloids from amino acid sequence

Numerous studies have shown that the ability to form amyloid fibrils is an inherent property of the polypeptide chain. This has lead to the development of several computational approaches to predict amyloidogenicity by amino acid sequences. Here, we discuss the principles governing these methods, and evaluate them using several datasets. They deliver excellent performance in the tests made using short peptides (~6 residues). However, there is a general tendency towards a high number of false positives when tested against longer sequences. This shortcoming needs to be addressed as these longer sequences are linked to diseases. Recent structural studies have shown that the core element of the majority of disease-related amyloid fibrils is a β-strand-loop-β-strand motif called β-arch. This insight provides an opportunity to substantially improve the prediction of amyloids produced by natural proteins, ushering in an era of personalized medicine based on genome analysis.

Two-step nucleation of amyloid fibrils: omnipresent or not?

Amyloid protein fibrils feature in various diseases and nanotechnological products. Currently, it is debated whether they nucleate in one step (i.e., directly from the protein solution) or in two steps (step one being the appearance of nonfibrillar oligomers in the solution and step two being the oligomer conversion into fibrils). We employ nucleation theory to gain insight into the idiosyncrasy of two-step fibril nucleation and to determine the conditions under which this process can take place. Presenting an expression for the rate of two-step fibril nucleation, we use it to qualitatively describe experimental data for two-step nucleated amyloid-β fibrils. Our analysis helps in understanding why, in some experiments, oligomers rather than fibrils form and remain structurally unchanged and why, in others, the oligomers convert into fibrils.

Motif mining: an assessment and perspective for amyloid fibril prediction tool

Amyloid fibril forming regions in protein sequences are associated with a number of diseases. Experimental evidences compel in favor of the hypothesis that short motif regions are responsible for its amyloidogenic behavior. Thus, identifying these short peptides is critical in understanding the cause of diseases associated with aggregation of proteins and developing sequencetargeted anti-aggregation drugs. Owing to the constraints of wet lab molecular techniques for the identification of amyloid fibril forming targets, computational methods are implemented to offer better and affordable in silico predictions. The present study takes into consideration an assessment and perspective of the recent tools available for predicting a peptide status: amyloidogenic or non-amyloidogenic. To the best of our knowledge, the existing review articles on amyloidogenic prediction tools have not touched upon their effectiveness in terms of true positive rates or false positive rates. In this work, we compare few tools such as Aggrescan, Amylpred and FoldAmyloid to evaluate the performance of their predictability based on the experimentally proved data in terms of specificity, sensitivity, Matthews Correlation Coefficient and Balanced accuracy. As evident from the results, a significant reduction of sensitivity associated with a gain in specificity is noted in all the tools considered under the present study.

Computational methods to predict therapeutic protein aggregation

Protein based biotherapeutics have emerged as a successful class of pharmaceuticals. However, these macromolecules endure a variety of physicochemical degradations during manufacturing, shipping, and storage, which may adversely impact the drug product quality. Of these degradations, the irreversible self-association of therapeutic proteins to form aggregates is a major challenge in the formulation of these molecules. Tools to predict and mitigate protein aggregation are, therefore, of great interest to biopharmaceutical research and development. In this chapter, a number of such computational tools developed to understand and predict the various steps involved in protein aggregation are described. These tools can be grouped into three general classes: unfolding kinetics and native state thermal stability, colloidal stability, and sequence/structure based aggregation liabilities. Chapter sections introduce each class by discussing how these predictive tools provide insight into the molecular events leading to protein aggregation. The computational methods are then explained in detail along with their advantages and limitations.

Size distribution of amyloid nanofibrils

We consider the size distribution of amyloid nanofibrils (protofilaments) in nucleating protein solutions when the nucleation process occurs by the mechanism of direct polymerization of β-strands (extended peptides or protein segments) into β-sheets. Employing the atomistic nucleation theory, we derive a general expression for the stationary size distribution of amyloid nanofibrils constituted of successively layered β-sheets. The application of this expression to amyloid β(1-40) (Aβ(40)) fibrils allows us to determine the nanofibril size distribution as a function of the protein concentration and temperature. The distribution is most remarkable with its exhibiting a series of peaks positioned at "magic" nanofibril sizes (or lengths), which are due to deep local minima in the work for fibril formation. This finding of magic sizes or lengths is consistent with experimental results for the size distribution of aggregates in solutions of Aβ(40) proteins. Also, our approach makes it possible to gain insight into the effect of point mutations on the nanofibril size distribution, an effect that may play a role in experimentally observed substantial differences in the fibrillation lag-time of wild-type and point-mutated amyloid-β proteins.

Prediction of amyloid aggregation in vivo

Many human diseases owe their pathology, to some degree, to the erroneous conversion of proteins from their soluble state into fibrillar, β-structured aggregates, often referred to as amyloid fibrils. Neurodegenerative diseases, such as Alzheimer and spongiform encephalopathies, as well as type 2 diabetes and both localized and systemic amyloidosis, are among the conditions that are associated with the formation of amyloid fibrils. Several mathematical tools can rationalize and even predict important parameters of amyloid fibril formation. It is not clear, however, whether such algorithms have predictive powers for in vivo systems, in which protein aggregation is affected by the presence of other biological factors. In this review, we briefly describe the existing algorithms and use them to predict the effects of mutations on the aggregation of specific proteins, for which in vivo experimental data are available. The comparison between the theoretical predictions and the experimental data obtained in vivo is shown for each algorithm and experimental data set, and statistically significant correlations are found in most cases.

An evolutionary trade-off between protein turnover rate and protein aggregation favors a higher aggregation propensity in fast degrading proteins

We previously showed the existence of selective pressure against protein aggregation by the enrichment of aggregation-opposing ‘gatekeeper’ residues at strategic places along the sequence of proteins. Here we analyzed the relationship between protein lifetime and protein aggregation by combining experimentally determined turnover rates, expression data, structural data and chaperone interaction data on a set of more than 500 proteins. We find that selective pressure on protein sequences against aggregation is not homogeneous but that short-living proteins on average have a higher aggregation propensity and fewer chaperone interactions than long-living proteins. We also find that short-living proteins are more often associated to deposition diseases. These findings suggest that the efficient degradation of high-turnover proteins is sufficient to preclude aggregation, but also that factors that inhibit proteasomal activity, such as physiological ageing, will primarily affect the aggregation of short-living proteins.

In order to carry out their biological function, proteins need to fold into well-defined three-dimensional structures. Protein aggregation is a process whereby proteins misfold into inactive and often toxic higher order structures, which is implied in about 30 human diseases such as Alzheimer's disease, Parkinson's disease and systemic amyloidosis. In earlier work it has been shown that although protein aggregation is an intrinsic property of polypeptide chains that cannot be entirely avoided, evolution has optimized protein sequences to minimize the risk of aggregation in a proteome. Here we show that this pressure is not uniform, but that proteins with a short lifetime have on average a higher aggregation propensity than long-living proteins. In addition, we show that high turnover proteins also make fewer interactions with chaperones. Taken together, these observations suggest that under normal physiological conditions the aggregation propensity of short-lived proteins does not represent a significant treat for the biochemistry of the cell. Presumably the strong dependence of these proteins on proteasomal degradation is sufficient to preclude the accumulation of aggregates. As proteasomal activity declines with age this would also explain why we observe a higher association of high turnover proteins with age-dependent aggregation-related diseases.

The induction of α-helical structure in partially unfolded HypF-N does not affect its aggregation propensity

The conversion of proteins into structured fibrillar aggregates is a central problem in protein chemistry, biotechnology, biology and medicine. It is generally accepted that aggregation takes place from partially structured states of proteins. However, the role of the residual structure present in such conformational states is not yet understood. In particular, it is not yet clear as to whether the α-helical structure represents a productive or counteracting structural element for protein aggregation. We have addressed this issue by studying the aggregation of pH-unfolded HypF-N. It has previously been shown that the two native α-helices of HypF-N retain a partial α-helical structure in the pH-unfolded state and that these regions are also involved in the formation of the cross-β structure of the aggregates. We have introduced mutations in such stretches of the sequence, with the aim of increasing the α-helical structure in the key regions of the pH-unfolded state, while minimizing the changes of other factors known to influence protein aggregation, such as hydrophobicity, β-Sheet propensity, etc. The resulting HypF-N mutants have higher contents of α-helical structure at the site(s) of mutation in their pH-unfolded states, but such an increase does not correlate with a change of aggregation rate. The results suggest that stabilisation of α-helical structure in amyloidogenic regions of the sequence of highly dynamic states does not have remarkable effects on the rate of protein aggregation from such conformational states. Comparison with other protein systems indicate that the effect of increasing α-helical propensity can vary if the stabilised helices are in non-amyloidogenic stretches of initially unstructured peptides (accelerating effect), in amyloidogenic stretches of initially unstructured peptides (no effect) or in amyloidogenic stretches of initially stable helices (decelerating effect).

Atomistic theory of amyloid fibril nucleation

We consider the nucleation of amyloid fibrils at the molecular level when the process takes place by a direct polymerization of peptides or protein segments into β-sheets. Employing the atomistic nucleation theory (ANT), we derive a general expression for the work to form a nanosized amyloid fibril (protofilament) composed of successively layered β-sheets. The application of this expression to a recently studied peptide system allows us to determine the size of the fibril nucleus, the fibril nucleation work, and the fibril nucleation rate as functions of the supersaturation of the protein solution. Our analysis illustrates the unique feature of ANT that the size of the fibril nucleus is a constant integer in a given supersaturation range. We obtain the ANT nucleation rate and compare it with the rates determined previously in the scope of the classical nucleation theory (CNT) and the corrected classical nucleation theory (CCNT). We find that while the CNT nucleation rate is orders of magnitude greater than the ANT one, the CCNT and ANT nucleation rates are in very good quantitative agreement. The results obtained are applicable to homogeneous nucleation, which occurs when the protein solution is sufficiently pure and/or strongly supersaturated.

Human beta-synuclein rendered fibrillogenic by designed mutations

Filamentous inclusions made of α-synuclein are found in nerve cells and glial cells in a number of human neurodegenerative diseases, including Parkinson disease, dementia with Lewy bodies, and multiple system atrophy. The assembly and spreading of these inclusions are likely to play an important role in the etiology of common dementias and movement disorders. Both α-synuclein and the homologous β-synuclein are abundantly expressed in the central nervous system; however, β-synuclein is not present in the pathological inclusions. Previously, we observed a poor correlation between filament formation and the presence of residues 73-83 of α-synuclein, which are absent in β-synuclein. Instead, filament formation correlated with the mean β-sheet propensity, charge, and hydrophilicity of the protein (global physicochemical properties) and β-strand contiguity calculated by a simple algorithm of sliding averages (local physicochemical property). In the present study, we rendered β-synuclein fibrillogenic via one set of point mutations engineered to enhance global properties and a second set engineered to enhance predominantly β-strand contiguity. Our findings show that the intrinsic physicochemical properties of synucleins influence their fibrillogenic propensity via two distinct but overlapping modalities. The implications for filament formation and the pathogenesis of neurodegenerative diseases are discussed.

Protein aggregation profile of the bacterial cytosol

BACKGROUND

Protein misfolding is usually deleterious for the cell, either as a consequence of the loss of protein function or the buildup of insoluble and toxic aggregates. The aggregation behavior of a given polypeptide is strongly influenced by the intrinsic properties encoded in its sequence. This has allowed the development of effective computational methods to predict protein aggregation propensity.

METHODOLOGY/PRINCIPAL FINDINGS

Here, we use the AGGRESCAN algorithm to approximate the aggregation profile of an experimental cytosolic Escherichia coli proteome. The analysis indicates that the aggregation propensity of bacterial proteins is associated with their length, conformation, location, function, and abundance. The data are consistent with the predictions of other algorithms on different theoretical proteomes.

CONCLUSIONS/SIGNIFICANCE

Overall, the study suggests that the avoidance of protein aggregation in functional environments acts as a strong evolutionary constraint on polypeptide sequences in both prokaryotic and eukaryotic organisms.

Simulated pathogenic conformational switch regions matched well with the biochemical findings

Pathogenic conformational conversion is a general causation of many disease, such as transmissible spongiform encephalopathy (TSE) caused by misfolding of prion, sickle cell anemia, and etc. In such structural changes, misfolding occurs in regions important for the stability of native structure firstly. This destabilizes the normal conformation and leads to subsequent errors in folding pathway. Sites involved in the first stage can be deemed switch regions of the protein, and are vital for conformational conversion. Namely it could be a switch of disease at residue level. Here we report an algorithm that can identify such sites computationally with an accuracy of 93%, by calculating the probability of the native structure of a short segment jumping to a mistake one. Knowledge of such switch sites could be used to target clinical therapy, study physiological and pathologic mechanism of protein, and etc.

Elongation dynamics of amyloid fibrils: a rugged energy landscape picture

Protein amyloid fibrils are a form of linear protein aggregates that are implicated in many neurodegenerative diseases. Here, we study the dynamics of amyloid fibril elongation by performing Langevin dynamic simulations on a coarse-grained model of peptides. Our simulation results suggest that the elongation process is dominated by a series of local minimum due to frustration in monomer-fibril interactions. This rugged energy landscape picture indicates that the amount of recycling of monomers at the fibrils' ends before being fibrilized is substantially reduced in comparison to the conventional two-step elongation model. This picture, along with other predictions discussed, can be tested with current experimental techniques.

Prediction of beta-turn in protein using E-SSpred and support vector machine

Beta-turn is a secondary protein structure type that plays an important role in protein configuration and function. Here, we introduced an approach of beta-turn prediction that used the support vector machine (SVM) algorithm combined with predicted secondary structure information. The secondary structure information was obtained by using E-SSpred, a new secondary protein structure prediction method. A 7-fold cross validation based on the benchmark dataset of 426 non-homologous protein chains was used to evaluate the performance of our method. The prediction results broke the 80% Q (total) barrier and achieved Q (total) = 80.9%, MCC = 0.44, and Q (predicted) higher 0.9% when compared with the best method. The results in our research are coincident with the conclusion that beta-turn prediction accuracy can be improved by inclusion of secondary structure information.

Self-assembly of protein amyloids: a competition between amorphous and ordered aggregation

Protein aggregation in the form of amyloid fibrils has important biological and technological implications. Although the self-assembly process is highly efficient, aggregates not in the fibrillar form would also occur and it is important to include these disordered species when discussing the thermodynamic equilibrium behavior of the system. Here, we initiate such a task by considering a mixture of monomeric proteins and the corresponding aggregates in the disordered form (micelles) and in the fibrillar form (amyloid fibrils). Starting with a model on the respective binding free energies for these species, we calculate their concentrations at thermal equilibrium. We then discuss how the incorporation of the disordered structure furthers our understanding on the various amyloid promoting factors observed empirically, and on the kinetics of fibrilization.

Pathogenic or not? And if so, then how? Studying the effects of missense mutations using bioinformatics methods

Many gene defects are relatively easy to identify experimentally, but obtaining information about the effects of sequence variations and elucidation of the detailed molecular mechanisms of genetic diseases will be among the next major efforts in mutation research. Amino acid substitutions may have diverse effects on protein structure and function; thus, a detailed analysis of the mutations is essential. Experimental study of the molecular effects of mutations is laborious, whereas useful and reliable information about the effects of amino acid substitutions can readily be obtained by theoretical methods. Experimentally defined structures and molecular modeling can be used as a basis for interpretation of the mutations. The effects of missense mutations can be analyzed even when the 3D structure of the protein has not been determined, although structure-based analyses are more reliable. Structural analyses include studies of the contacts between residues, their implication for the stability of the protein, and the effects of the introduced residues. Investigations of steric and stereochemical consequences of substitutions provide insights on the molecular fit of the introduced residue. Mutations that change the electrostatic surface potential of a protein have wide-ranging effects. Analyses of the effects of mutations on interactions with ligands and partners have been performed for elucidation of functional mutations. We have employed numerous methods for predicting the effects of amino acid substitutions. We discuss the applicability of these methods in the analysis of genes, proteins, and diseases to reveal protein structure-function relationships, which is essential to gain insights into disease genotype-phenotype correlations.

Amyloidogenic determinants are usually not buried

Background

Amyloidoses are a group of usually fatal diseases, probably caused by protein misfolding and subsequent aggregation into amyloid fibrillar deposits. The mechanisms involved in amyloid fibril formation are largely unknown and are the subject of current, intensive research. In an attempt to identify possible amyloidogenic regions in proteins for further experimental investigation, we have developed and present here a publicly available online tool that utilizes five different and independently published methods, to form a consensus prediction of amyloidogenic regions in proteins, using only protein primary structure data.

Results

It appears that the consensus prediction tool is slightly more objective than individual prediction methods alone and suggests several previously not identified amino acid stretches as potential amyloidogenic determinants, which (although several of them may be overpredictions) require further experimental studies. The tool is available at: . Utilizing molecular graphics programs, like O and PyMOL, as well as the algorithm DSSP, it was found that nearly all experimentally verified amyloidogenic determinants (short peptide stretches favouring aggregation and subsequent amyloid formation), and several predicted, with the aid of the tool AMYLPRED, but not experimentally verified amyloidogenic determinants, are located on the surface of the relevant amyloidogenic proteins. This finding may be important in efforts directed towards inhibiting amyloid fibril formation.

Conclusion

The most significant result of this work is the observation that virtually all, to date, experimentally determined amyloidogenic determinants and the majority of predicted, but not yet experimentally verified short amyloidogenic stretches, lie 'exposed' on the surface of the relevant amyloidogenic proteins, and also several of them have the ability to act as conformational 'switches'. Experiments, focused on these fragments, should be performed to test this idea.

NetCSSP: web application for predicting chameleon sequences and amyloid fibril formation

The calculation of contact-dependent secondary structure propensity (CSSP) is a unique and sensitive method that detects non-native secondary structure propensities in protein sequences. This method has applications in predicting local conformational change, which typically is observed in core sequences of protein aggregation and amyloid fibril formation. NetCSSP implements the latest version of the CSSP algorithm and provides a Flash chart-based graphic interface that enables an interactive calculation of CSSP values for any user-selected regions in a given protein sequence. This feature also can quantitatively estimate the mutational effect on changes in native or non-native secondary structural propensities in local sequences. In addition, this web tool provides precalculated non-native secondary structure propensities for over 1 400 000 fragments that are seven-residues long, collected from PDB structures. They are searchable for chameleon subsequences that can serve as the core of amyloid fibril formation. The NetCSSP web tool is available at http://cssp2.sookmyung.ac.kr/.

Identification of amyloidogenic peptide sequences using a coarse-grained physicochemical model

Cross-beta amyloid is implicated in over 20 human diseases. Experiments suggest that specific sequence elements within amyloidogenic proteins play a major role in seeding amyloid formation. Identifying these seeding sequences is important for rationalizing the molecular mechanisms of amyloid formation and for elaborating therapeutic strategies that target amyloid. Theoretical techniques play an important role in facilitating the identification and structural characterization of putative seeding sequences; most amyloid species are not amenable to high resolution experimental structure techniques. In this study we have combined a coarse-grained physicochemical protein model with a highly efficient Monte Carlo sampling technique to identify amyloidogenic sequences in four proteins for which respective experimental peptide fragmentation data exist. Peptide sequences were defined as amyloidogenic if the ensemble structure predicted for three interacting peptides described a stable and regular three-stranded beta-sheet. For such peptides, free energies were calculated to provide a measure of amyloid propensity. The overall agreement between the experimental and predicted data is good, and we correctly identify several self-recognition motifs proposed to define the cross-beta amyloid fibril architectures of two of the proteins. Our results compare very favorably with those obtained using atomistic molecular dynamics methods, though our simulations are 30-40 times faster.

Prediction of amyloid fibril-forming segments based on a support vector machine

BACKGROUND

Amyloid fibrillar aggregates of proteins or polypeptides are known to be associated with many human diseases. Recent studies suggest that short protein regions trigger this aggregation. Thus, identifying these short peptides is critical for understanding diseases and finding potential therapeutic targets.

RESULTS

We propose a method, named Pafig (Prediction of amyloid fibril-forming segments) based on support vector machines, to identify the hexpeptides associated with amyloid fibrillar aggregates. The features of Pafig were obtained by a two-round selection from AAindex. Using a 10-fold cross validation test on Hexpepset dataset, Pafig performed well with regards to overall accuracy of 81% and Matthews correlation coefficient of 0.63. Pafig was used to predict the potential fibril-forming hexpeptides in all of the 64,000,000 hexpeptides. As a result, approximately 5.08% of hexpeptides showed a high aggregation propensity. In the predicted fibril-forming hexpeptides, the amino acids--alanine, phenylalanine, isoleucine, leucine and valine occurred at the higher frequencies and the amino acids--aspartic acid, glutamic acid, histidine, lysine, arginine and praline, appeared with lower frequencies.

CONCLUSION

The performance of Pafig indicates that it is a powerful tool for identifying the hexpeptides associated with fibrillar aggregates and will be useful for large-scale analysis of proteomic data.