- Email: [email protected]
Unraveling Structural Polymorphism of Amyloid Fibers
Structure
Previews Unraveling Structural Polymorphism of Amyloid Fibers Doryen Bubeck1,* 1Department of Life Sciences, Sir Ernst Chain Building, South Kensington Campus, Imperial College London, London SW7 2AZ, UK *Correspondence: [email protected] http://dx.doi.org/10.1016/j.str.2014.12.005
Amyloid fibers are self-assembling cross b sheet structures whose overall molecular architecture is highly variable. In this issue of Structure, Torreira and colleagues use electron microscopy to resolve ultrastructural polymorphisms of RepA-WH1 amyloids. Amyloid fibers form from abnormal assembly of soluble proteins and are associated with a number of disease pathologies, ranging from Alzheimer’s disease to type II diabetes (Chiti and Dobson, 2006). Fibrils are composed of intertwined protofilaments that have a predominantly cross b sheet quaternary structure. Self-assembly of b strand component building blocks generate a wide range of molecular level polymorphism. A detailed molecular mechanism underpinning amyloidogenesis is important for developing therapeutic strategies to treat human amyloid proteinopathies. RepA is a DNA initiation factor encoded in the Pseudomonas pPS10 replicon. Similar to other bacterial Rep proteins, monomeric RepA initiates replication while its constitutive dimeric form acts as a transcriptional autorepressor (Giraldo et al., 1998). RepA is composed of two ‘‘winged-helix’’ domains (WH1 and WH2). WH1 is involved in the dimerization interface, while WH2 binds specific double-stranded DNA sequences called interons. Interon-binding at replication origins converts the complex to its monomeric active form and enables the recruitment of replication factors. Once replication is completed, RepA proteins from each plasmid copy remain associated and sterically inhibit initiation of premature replication rounds (Das and Chattoraj, 2004). The crystal structure of RepA-WH1 identified residues involved in the dimer interface and revealed a hydrophobic core composed of a five-stranded antiparallel b sheet (Giraldo et al., 2003). However, structural information for how RepA molecules might oligomerize remained lacking.
A truncated version of RepA, containing the first WH1 domain (RepA-WH1), forms amyloid fibers in vitro. Although it remains to be seen if these occur in vivo, in Escherichia coli, RepA-WH1 amyloids impact bacterial fitness and shorten cell division cycles (Ferna´ndez-Tresguerres et al., 2010). Many bacteria produce amyloid fibers that share structural and biochemical similarities with pathogenic eukaryotic assemblies. While most bacterial amyloids are secreted from the plasma membrane and are involved in host cell adhesion or biofilm formation (Bednarska et al., 2013), RepA-WH1 fibers are nonfunctional intracellular assemblies. Thus, RepA-WH1 aggregates are an example of an amyloid proteinopathy in bacteria that may serve as a model system for characterizing human disease. In this issue of Structure, Torreira et al. (2015) report the first 3D ultrastructure of a RepA-WH1 assembly. The authors found that soluble RepA-WH1 dimers can be converted to fibers by seeding with amyloid aggregates. Analyzed by both electron microscopy and atomic force microscopy, these fibers were found to be largely heterogeneous and polymorphic. However, by characterizing frayed ends, the authors discovered that fibers were constructed from several filament units. Single-particle image processing methods, applied to isolated filaments, were used to define two filament subtypes. Conformational heterogeneity within each subtype was addressed, and their 3D structures revealed single and double constituent RepA-WH1 helical nanotubes. Segmentation of the reconstructions and fitting of the RepA-WH1 crystal structure showed that both filament subtypes were composed of
10 Structure 23, January 6, 2015 ª2015 Elsevier Ltd All rights reserved
RepA-WH1 monomers. Thus, the structures shown here report a different arrangement of RepA-WH1 than was previously observed in the crystal structure of the dimer. The authors speculate that this dissociation of dimers and the rearrangement of monomeric RepAWH1 may reflect structural transitions of RepA that could occur during initiation of DNA replication. Structural studies of amyloids have demonstrated that both fibers and their protofilament components have a highly variable ultrastructure. Amyloid b protein, associated with Alzheimer’s disease, assembles into nanotubes with a unique triple helix molecular architecture (Nicoll et al., 2013). In contrast, a truncated fragment of transthyretin organizes into three distinct fibril types composed of multiple linear protofilaments (Fitzpatrick et al., 2013). While these studies characterize self-assembly of peptides, Torreira et al., (2015) present the hierarchical organization of a protein domain. All three studies report the variability of amyloids at both the intra- and interprotofilament levels. This polymorphic molecular organization may be reflective of the diversity in clinical subphenotypes for amyloidbased disease. ACKNOWLEDGMENTS I thank S. Matthews for discussions. D.B. is supported by a Career Establishment Award from Cancer Research UK (C26409/A16099). REFERENCES Bednarska, N.G., Schymkowitz, J., Rousseau, F., and Van Eldere, J. (2013). Microbiology 159, 1795–1806. Chiti, F., and Dobson, C.M. (2006). Annu. Rev. Biochem. 75, 333–366.
Structure
Previews Das, N., and Chattoraj, D.K. (2004). Mol. Microbiol. 54, 836–849.
et al. (2013). Proc. Natl. Acad. Sci. USA 110, 5468–5473.
Ferna´ndez-Tresguerres, M.E., de la Espina, S.M., Gasset-Rosa, F., and Giraldo, R. (2010). Mol. Microbiol. 77, 1456–1469.
Giraldo, R., Andreu, J.M., and Dı´az-Orejas, R. (1998). EMBO J. 17, 4511–4526.
Fitzpatrick, A.W., Debelouchina, G.T., Bayro, M.J., Clare, D.K., Caporini, M.A., Bajaj, V.S., Jaroniec, C.P., Wang, L., Ladizhansky, V., Mu¨ller, S.A.,
Giraldo, R., Ferna´ndez-Tornero, C., Evans, P.R., Dı´az-Orejas, R., and Romero, A. (2003). Nat. Struct. Biol. 10, 565–571.
Nicoll, A.J., Panico, S., Freir, D.B., Wright, D., Terry, C., Risse, E., Herron, C.E., O’Malley, T., Wadsworth, J.D., Farrow, M.A., et al. (2013). Nat. Commun. 4, 2416. Torreira, E., Moreno-del A´lamo, M., FuentesPerez, M.E., Ferna´ndez, C., Martı´n-Benito, J., Moreno-Herrero, F., Giraldo, R., and Llorca, O. (2015). Structure 23, this issue, 183–189.
A New Test of Computational Protein Design: Predicting Posttranslational Modification Specificity for the Enzyme SMYD2 Kimberly A. Reynolds1,* 1Green Center for Systems Biology, University of Texas Southwestern Medical Center, 6001 Forest Park Road, ND11.136, Dallas, TX 75390-8597, USA *Correspondence: [email protected] http://dx.doi.org/10.1016/j.str.2014.12.004
In this issue of Structure, Lanouette and colleagues use a combination of computation and experiment to define a specificity motif for the lysine methyltransferase SMYD2. Using this motif, they predict and experimentally verify four new SMYD2 substrates. Protein lysine methyltransferases (PKMTs) such as SMYD2 transfer one, two, or three methyl groups to the ε-amine of a lysine side chain. This protein family methylates histones, executing a critical step in the epigenetic regulation of chromatin structure and gene expression. PKMTs also modify several nonhistone substrates (p53, Retinoblastoma protein, and estrogen receptor a) and are frequently associated with aspects of cell cycle progression and chromosome segregation. Despite the central importance of PKMTs to cell biology, the comprehensive identification of PKMT targets is currently limited by technical obstacles. In particular, the small change in molecular weight and lack of charge associated with modification by methylation makes detection by mass spectrometry nontrivial and the generation of specific antibodies difficult (Lanouette et al., 2014). In this issue of Structure, Lanouette et al. (2015) address this challenge by taking an alternate approach to the identification of SMYD2 substrates. The idea is simple: to conduct comprehensive saturation point mutagenesis for the peptide
substrate in silico and use the results to define a specificity motif for methyltransferase activity. To do this, they begin with a crystal structure of SMYD2 in complex with a p53 substrate peptide. The peptide residues 1, +1, and +2 relative to the methylated lysine are then individually mutated to all other amino acid residues with the exception of proline, and the new point mutant peptide/SMYD2 complex is energy minimized and scored using a physics-based potential energy function. The energetic scores (referred to in the article as substitution fitness values) are then used to select tolerated or beneficial mutations in the peptide, resulting in a calculated binding motif for SMYD2. For these calculations, the authors make a key strategic choice: to represent the complex between SMYD2 and peptide ligand as a conformational ensemble. This approach, often referred to as ‘‘multistate design,’’ attempts to capture the flexibility of natural proteins by calculating energy over a conformational ensemble (typically on the order of 10–102 structures) instead of for a fixed single-structure template (Friedland and
Kortemme, 2010). Indeed, prior work has shown that multistate design can improve the de novo design of small proteins (Allen et al., 2010) and the accuracy of peptide binding specificity predictions (Smith and Kortemme, 2010). In this article, Lanouette et al. (2015) directly compare the predictive capacity of multistate and single-state computational design protocols to experimental methyltransferase data. Using peptide SPOT arrays, SMYD2 methyltransferase activity is tested for a collection of p53 substrate peptides where the 1, +1, and +2 positions are individually mutated to all other amino acid residues. The results are striking: the multistate design calculations largely reproduce the experimentally determined specificity motif (Figure 1). In contrast, single-state design predicts unfavorable energies for several amino acids well tolerated by SMYD2 experimentally. The observation that multistate design calculations result in fewer false negatives for this system is consistent with the well-known observation that fixed backbone calculations often induce steric clashes that would be
Structure 23, January 6, 2015 ª2015 Elsevier Ltd All rights reserved 11
Previews Unraveling Structural Polymorphism of Amyloid Fibers Doryen Bubeck1,* 1Department of Life Sciences, Sir Ernst Chain Building, South Kensington Campus, Imperial College London, London SW7 2AZ, UK *Correspondence: [email protected] http://dx.doi.org/10.1016/j.str.2014.12.005
Amyloid fibers are self-assembling cross b sheet structures whose overall molecular architecture is highly variable. In this issue of Structure, Torreira and colleagues use electron microscopy to resolve ultrastructural polymorphisms of RepA-WH1 amyloids. Amyloid fibers form from abnormal assembly of soluble proteins and are associated with a number of disease pathologies, ranging from Alzheimer’s disease to type II diabetes (Chiti and Dobson, 2006). Fibrils are composed of intertwined protofilaments that have a predominantly cross b sheet quaternary structure. Self-assembly of b strand component building blocks generate a wide range of molecular level polymorphism. A detailed molecular mechanism underpinning amyloidogenesis is important for developing therapeutic strategies to treat human amyloid proteinopathies. RepA is a DNA initiation factor encoded in the Pseudomonas pPS10 replicon. Similar to other bacterial Rep proteins, monomeric RepA initiates replication while its constitutive dimeric form acts as a transcriptional autorepressor (Giraldo et al., 1998). RepA is composed of two ‘‘winged-helix’’ domains (WH1 and WH2). WH1 is involved in the dimerization interface, while WH2 binds specific double-stranded DNA sequences called interons. Interon-binding at replication origins converts the complex to its monomeric active form and enables the recruitment of replication factors. Once replication is completed, RepA proteins from each plasmid copy remain associated and sterically inhibit initiation of premature replication rounds (Das and Chattoraj, 2004). The crystal structure of RepA-WH1 identified residues involved in the dimer interface and revealed a hydrophobic core composed of a five-stranded antiparallel b sheet (Giraldo et al., 2003). However, structural information for how RepA molecules might oligomerize remained lacking.
A truncated version of RepA, containing the first WH1 domain (RepA-WH1), forms amyloid fibers in vitro. Although it remains to be seen if these occur in vivo, in Escherichia coli, RepA-WH1 amyloids impact bacterial fitness and shorten cell division cycles (Ferna´ndez-Tresguerres et al., 2010). Many bacteria produce amyloid fibers that share structural and biochemical similarities with pathogenic eukaryotic assemblies. While most bacterial amyloids are secreted from the plasma membrane and are involved in host cell adhesion or biofilm formation (Bednarska et al., 2013), RepA-WH1 fibers are nonfunctional intracellular assemblies. Thus, RepA-WH1 aggregates are an example of an amyloid proteinopathy in bacteria that may serve as a model system for characterizing human disease. In this issue of Structure, Torreira et al. (2015) report the first 3D ultrastructure of a RepA-WH1 assembly. The authors found that soluble RepA-WH1 dimers can be converted to fibers by seeding with amyloid aggregates. Analyzed by both electron microscopy and atomic force microscopy, these fibers were found to be largely heterogeneous and polymorphic. However, by characterizing frayed ends, the authors discovered that fibers were constructed from several filament units. Single-particle image processing methods, applied to isolated filaments, were used to define two filament subtypes. Conformational heterogeneity within each subtype was addressed, and their 3D structures revealed single and double constituent RepA-WH1 helical nanotubes. Segmentation of the reconstructions and fitting of the RepA-WH1 crystal structure showed that both filament subtypes were composed of
10 Structure 23, January 6, 2015 ª2015 Elsevier Ltd All rights reserved
RepA-WH1 monomers. Thus, the structures shown here report a different arrangement of RepA-WH1 than was previously observed in the crystal structure of the dimer. The authors speculate that this dissociation of dimers and the rearrangement of monomeric RepAWH1 may reflect structural transitions of RepA that could occur during initiation of DNA replication. Structural studies of amyloids have demonstrated that both fibers and their protofilament components have a highly variable ultrastructure. Amyloid b protein, associated with Alzheimer’s disease, assembles into nanotubes with a unique triple helix molecular architecture (Nicoll et al., 2013). In contrast, a truncated fragment of transthyretin organizes into three distinct fibril types composed of multiple linear protofilaments (Fitzpatrick et al., 2013). While these studies characterize self-assembly of peptides, Torreira et al., (2015) present the hierarchical organization of a protein domain. All three studies report the variability of amyloids at both the intra- and interprotofilament levels. This polymorphic molecular organization may be reflective of the diversity in clinical subphenotypes for amyloidbased disease. ACKNOWLEDGMENTS I thank S. Matthews for discussions. D.B. is supported by a Career Establishment Award from Cancer Research UK (C26409/A16099). REFERENCES Bednarska, N.G., Schymkowitz, J., Rousseau, F., and Van Eldere, J. (2013). Microbiology 159, 1795–1806. Chiti, F., and Dobson, C.M. (2006). Annu. Rev. Biochem. 75, 333–366.
Structure
Previews Das, N., and Chattoraj, D.K. (2004). Mol. Microbiol. 54, 836–849.
et al. (2013). Proc. Natl. Acad. Sci. USA 110, 5468–5473.
Ferna´ndez-Tresguerres, M.E., de la Espina, S.M., Gasset-Rosa, F., and Giraldo, R. (2010). Mol. Microbiol. 77, 1456–1469.
Giraldo, R., Andreu, J.M., and Dı´az-Orejas, R. (1998). EMBO J. 17, 4511–4526.
Fitzpatrick, A.W., Debelouchina, G.T., Bayro, M.J., Clare, D.K., Caporini, M.A., Bajaj, V.S., Jaroniec, C.P., Wang, L., Ladizhansky, V., Mu¨ller, S.A.,
Giraldo, R., Ferna´ndez-Tornero, C., Evans, P.R., Dı´az-Orejas, R., and Romero, A. (2003). Nat. Struct. Biol. 10, 565–571.
Nicoll, A.J., Panico, S., Freir, D.B., Wright, D., Terry, C., Risse, E., Herron, C.E., O’Malley, T., Wadsworth, J.D., Farrow, M.A., et al. (2013). Nat. Commun. 4, 2416. Torreira, E., Moreno-del A´lamo, M., FuentesPerez, M.E., Ferna´ndez, C., Martı´n-Benito, J., Moreno-Herrero, F., Giraldo, R., and Llorca, O. (2015). Structure 23, this issue, 183–189.
A New Test of Computational Protein Design: Predicting Posttranslational Modification Specificity for the Enzyme SMYD2 Kimberly A. Reynolds1,* 1Green Center for Systems Biology, University of Texas Southwestern Medical Center, 6001 Forest Park Road, ND11.136, Dallas, TX 75390-8597, USA *Correspondence: [email protected] http://dx.doi.org/10.1016/j.str.2014.12.004
In this issue of Structure, Lanouette and colleagues use a combination of computation and experiment to define a specificity motif for the lysine methyltransferase SMYD2. Using this motif, they predict and experimentally verify four new SMYD2 substrates. Protein lysine methyltransferases (PKMTs) such as SMYD2 transfer one, two, or three methyl groups to the ε-amine of a lysine side chain. This protein family methylates histones, executing a critical step in the epigenetic regulation of chromatin structure and gene expression. PKMTs also modify several nonhistone substrates (p53, Retinoblastoma protein, and estrogen receptor a) and are frequently associated with aspects of cell cycle progression and chromosome segregation. Despite the central importance of PKMTs to cell biology, the comprehensive identification of PKMT targets is currently limited by technical obstacles. In particular, the small change in molecular weight and lack of charge associated with modification by methylation makes detection by mass spectrometry nontrivial and the generation of specific antibodies difficult (Lanouette et al., 2014). In this issue of Structure, Lanouette et al. (2015) address this challenge by taking an alternate approach to the identification of SMYD2 substrates. The idea is simple: to conduct comprehensive saturation point mutagenesis for the peptide
substrate in silico and use the results to define a specificity motif for methyltransferase activity. To do this, they begin with a crystal structure of SMYD2 in complex with a p53 substrate peptide. The peptide residues 1, +1, and +2 relative to the methylated lysine are then individually mutated to all other amino acid residues with the exception of proline, and the new point mutant peptide/SMYD2 complex is energy minimized and scored using a physics-based potential energy function. The energetic scores (referred to in the article as substitution fitness values) are then used to select tolerated or beneficial mutations in the peptide, resulting in a calculated binding motif for SMYD2. For these calculations, the authors make a key strategic choice: to represent the complex between SMYD2 and peptide ligand as a conformational ensemble. This approach, often referred to as ‘‘multistate design,’’ attempts to capture the flexibility of natural proteins by calculating energy over a conformational ensemble (typically on the order of 10–102 structures) instead of for a fixed single-structure template (Friedland and
Kortemme, 2010). Indeed, prior work has shown that multistate design can improve the de novo design of small proteins (Allen et al., 2010) and the accuracy of peptide binding specificity predictions (Smith and Kortemme, 2010). In this article, Lanouette et al. (2015) directly compare the predictive capacity of multistate and single-state computational design protocols to experimental methyltransferase data. Using peptide SPOT arrays, SMYD2 methyltransferase activity is tested for a collection of p53 substrate peptides where the 1, +1, and +2 positions are individually mutated to all other amino acid residues. The results are striking: the multistate design calculations largely reproduce the experimentally determined specificity motif (Figure 1). In contrast, single-state design predicts unfavorable energies for several amino acids well tolerated by SMYD2 experimentally. The observation that multistate design calculations result in fewer false negatives for this system is consistent with the well-known observation that fixed backbone calculations often induce steric clashes that would be
Structure 23, January 6, 2015 ª2015 Elsevier Ltd All rights reserved 11