Download Precise Assembly of Complex Beta Sheet Topologies from de ... and more Exams Network Design in PDF only on Docsity!
1 Precise Assembly of Complex Beta Sheet Topologies from de novo
2 Designed Building Blocks.
3 Indigo Chris King* †^ , James Gleixner †^ , Lindsey Doyle §, Alexandre Kuzin‡^ , John F. Hunt ‡^ , Rong Xiao x^ , 4 Gaetano T. Montelionex, Barry L. Stoddard§, Frank Dimaio†^ , David Baker†
5 † Institute for Protein Design, University of Washington, Seattle, Washington, United States 6 ‡ Biological Sciences, Northeast Structural Genomics Consortium, Columbia University, New York, New York, United 7 States 8 x^ Center for Advanced Biotechnology and Medicine, Department of Molecular Biology and Biochemistry, Northeast Struc- 9 tural Genomics Consortium, Rutgers, The State University of New Jersey, Piscataway, New Jersey, United States 10 § Basic Sciences, Fred Hutchinson Cancer Research Center, Seattle, Washington, United States 11
12 ABSTRACT
13 Design of complex alpha-beta protein topologies poses a challenge because of the large number of alternative packing
14 arrangements. A similar challenge presumably limited the emergence of large and complex protein topologies in evo-
15 lution. Here we demonstrate that protein topologies with six and seven-stranded beta sheets can be designed by inser-
16 tion of one de novo designed beta sheet containing protein into another such that the two beta sheets are merged to
17 form a single extended sheet, followed by amino acid sequence optimization at the newly formed strand-strand, strand-
18 helix, and helix-helix interfaces. Crystal structures of two such designs closely match the computational design models.
19 Searches for similar structures in the SCOP protein domain database yield only weak matches with different beta sheet
20 connectivities. A similar beta sheet fusion mechanism may have contributed to the emergence of complex beta sheets
21 during natural protein evolution.
22 INTRODUCTION
23 Modular domains constitute the primary structural and functional units of natural proteins. Multi-domain proteins like-
24 ly evolved through simple linear concatenation of successive domains onto the polypeptide chain or through the inser-
25 tion of one or more continuous sequences into the middle of another, now discontinuous domain 1-4^. By analogy, new
26 proteins have been engineered from existing domains by simple linear concatenation or insertion of one domain into
27 another 5-11^. How individual domains evolved, in contrast, is much less clear. Both experimental and computational
28 analyses have suggested that new folds can evolve by insertion of one fold into another 3,12-14^ 15,16, but to our
29 knowledge there is no evidence that complex beta sheet topologies can be formed in this manner. On the protein de-
30 sign front, there has been progress in de novo design of idealized helical bundles 17 and alpha beta protein structures
31 with up to 5 strands 18 , and though new folds have been generated by tandem fusion of natural protein domains fol-
32 lowed by introduction of additional stabilizing mutations19,20^ , assembly of large and complex beta sheets poses a chal-
33 lenge for de novo protein design.
34 One possible route to the large and complex beta sheet topologies found in many native protein domains is recombina-
35 tion of two smaller beta sheet domains. Here we explore the viability of such a mechanism by inserting one de novo
36 designed alpha beta protein into another such that the two beta sheets are combined into one. The backbone geometry
37 at the junctions between the original domains is regularized, and the sequence at the newly formed interface is opti-
38 mized to stabilize the single integrated domain structure. Crystal structures of two such proteins demonstrate that com-
39 plex beta sheet structures can be designed with considerable accuracy using this approach, and provide a proof-of-
40 concept for the hypothesis that complex beta topologies in natural proteins may have evolved from simpler beta sheet
41 structures in a similar manner.
42
43
44 RESULTS
45 A first extended sheet protein was created by inserting a designed ferredoxin domain into a beta turn of the de-
46 signed top7 protein to create a half-barrel structure, with the two sheets fused into a single seven strand sheet flanked
47 by four helices (Figure 1A). The CD spectra show both alpha and beta structure (Figure 2—figure supplement 1). Two
48 crystal structures (NESG target OR327) were solved by molecular replacement and refined to 2.49 Å (PDB entry
49 4KYZ) and 2.96 Å (PDB entry 4KY3) resolutions. Further analysis refers only to the higher resolution structure
50 (4KYZ). The structure shows excellent agreement with the design model (Figure 2A), particularly in low B-factor re-
51 gions, with C-alpha RMSD ranging from 1.76-1.85 Å among the four protomers in the crystal. The relative orientation
52 of the strands packed against the helices is close to that in the design model, and core sidechains at the designed inter-
53 faces are in very similar conformations in the design model and crystal (Figure 2B,2C).
54
55
56 A second extended sheet protein was created by combining two designed ferredoxin domains via domain insertion
57 to create a half-barrel structure with four alpha helices and six beta strands (Figure 1B). A beta turn segment between
86 We have shown that single designed protein domains can be combined into larger domains with complex beta
87 sheet topologies. This mechanism provides a straightforward route to designing large and complex beta sheet structures
88 capable of scaffolding the pockets and cavities essential for future design of protein functions. Our success in design-
89 ing larger beta sheet domains by recombining smaller independently folded beta sheet proteins suggests a similar
90 mechanism could have played a role in the evolution of naturally occurring complex beta sheet proteins.
92 MATERIALS AND METHODS
93 Our design strategy began with selection of three previously characterized de novo designed protein domains to
94 serve as building blocks for recombination through domain insertion: ferredoxin, rossman 2x2, and top7 18. These
95 three domains were chosen because they were the only Rosetta de novo designed protein domains with both alpha and
96 beta secondary structure for which high resolution experimental structures had been obtained at the time of this work.
97 Each chimeric domain consists of a parent host domain and a parent insert domain. In the insert domain, three residues
98 from from the n-terminus were paired with three residue from the c-terminus to create nine residue pairs. Each residue
99 pair was then aligned against all pairs of residues in the host domain to search for possible insertion points. Insertion
100 points were accepted for residue pair alignment distances of 1 angstrom RMSD or less, replacing host domain seg-
101 ments of less than 5 residues. For every insertion point, a structure is generated by removing the residues between the
102 insertion residues of the host domain and adding linkers between the aligned host and insert domain residues (Figure
103 1). Host and insert were connected by addition of 1-3 residues at the domain junctions using Rosetta Remodel 23 , and
104 12 models in which this junction formed a continuous beta strand were identified. The sequences of these chimeras
105 were optimized using Rosetta Design calculations around the junction regions and the new interface between the for-
106 mer domains. During the design simulation, all amino acid positions within 5 Å of the inter-domain junction interface
107 were redesigned to minimize the predicted free energy of folding with the Rosetta all-atom energy function and a flexi-
108 ble backbone protein design protocol described previously 23. Final designs were selected based on Rosetta energy,
109 packing metrics, and similarity of the junction backbone geometry to local backbone geometry in the PDB. Twelve
110 final domain insertion designs were chosen for expression in E. coli as 6xHis-tag fusions and purified on a Ni-NTA
111 column. Purified proteins were evaluated for the presence of alpha/beta secondary structures via circular dichroism
112 spectroscopy (CD), and three with levels of secondary structure content consistent with the design model were subject-
113 ed to crystallographic analysis. One design based on Rossman 2x2 expressed as soluble protein, but no crystal structure
114 could be obtained. Crystal structures were obtained for two designed proteins: a ferredoxin-top7 chimera and a ferre-
115 doxin-ferredoxin chimera. The design and characterization of these two proteins is described in the Results.
116 Crystal structures were used to search for structural homologs in the SCOP database. First, crystal structures (ferre-
117 doxin-top7: 4KYZ chain A, ferredoxin-ferredoxin: 5CW9 chain A) were used as search queries using TMalign 24. Hits
118 were saved only if the alignment covered 75% or more of the query structure. Results were sorted by TM-score to
119 identify the most similar structures in the SCOP database. Secondary structure topology cartoons were created with the
120 Pro Origami server 25. To map designed protein crystal structures into the protein domains network, the structures were
121 aligned to all domain structures in the protein domains network using the PDBeFold server 26. PDBeFold structural
122 alignment hits were filtered for RMSD less than or equal to 2.5Å and aligned sequence length of greater than or equal
123 to 75 residues. In contrast to the methods of Nepomnyachi et al, sequence similarity thresholds were ignored. Including
124 sequence similarity thresholds eliminates matching hits in the domains network. This is not surprising because the pro-
125 teins were designed de novo and did not evolve from natural proteins. Filtered alignment hits were mapped into the
126 protein domains network using Cytoscape 27. To evaluate neutral drift models of the parent folds, then crystal structures
127 of de novo ferredoxin and Top7 proteins (2KL8 and 1QYS) were obtained and corresponding mutations from the final
128 design proteins were modeled using a flexible backbone protein design algorithm described previously 23. Final Rosetta
129 energies were calculated and subtracted from the Rosetta energies of the original parent protein structures to obtain
130 predictions of the change in free energy of folding.
131 The ferredoxin – TOP7 protein (NESF ID OR327) was expressed, and purified following standard protocols devel-
132 oped by the NESG for production of selenomethionine labeled protein samples 28. Briefly, Escherichia coli BL
133 (DE3) pMGK cells, a rare-codon enhanced strain, were transformed with the DNA sequence-verified OR327-21.
134 plasmid. A single isolate was cultured in MJ9 minimal media supplemented with selenomethionine, lysine, phenylala-
135 nine, threonine, isoleucine, leucine, and valine for the production of selenomethionine-labeled OR327. Initial growth
136 was carried out at 37 °C until the OD600 of the culture reached 0.8 units. The incubation temperature was then de-
137 creased to 17 °C, and protein expression was induced by the addition of isopropyl-β-D-thiogalactopyranoside (IPTG)
138 at a final concentration of 1 mM. Following overnight incubation at 17 °C, the cells were harvested by centrifugation
139 and resuspended in Lysis Buffer [50 mM Tris, pH 7.5, 500 mM NaCl, 1 mM tris (2-carboxyethyl)phosphine, 40 mM
140 imidazole]. After sonication, the supernatant was collected by centrifugation for 40 min at 30,000 × g. The supernatant
141 was loaded first onto a Ni affinity column (HisTrap HP; GE Healthcare) and the eluate loaded into a gel filtration col-
142 umn (Superdex 75 26/60; GE Healthcare). Yields were 60-90 mg / L. The purified 6His-OR327 construct in buffer
170 2. Berrondo, M., Ostermeier, M. & Gray, J.J. Structure prediction of domain insertion proteins 171 from structures of individual domains. Structure 16 , 513-527 (2008). 172 3. Lupas, A.N., Ponting, C.P. & Russell, R.B. On the evolution of protein folds: Are similar motifs 173 in different protein folds the result of convergence, insertion, or relics of an ancient peptide 174 world? Journal of Structural Biology 134 , 191-203 (2001). 175 4. Pandya, C. et al. Consequences of domain insertion on sequence-structure divergence in a 176 superfold. Proceedings of the National Academy of Sciences of the United States of America 110 , 177 E3381-E3387 (2013). 178 5. Ay, J., Gotz, F., Borriss, R. & Heinemann, U. Structure and function of the Bacillus hybrid 179 enzyme GluXyn-1: native-like jellyroll fold preserved after insertion of autonomous globular 180 domain. Proc Natl Acad Sci U S A 95 , 6613-8 (1998). 181 6. Collinet, B. et al. Functionally accepted insertions of proteins within protein domains. Journal of 182 Biological Chemistry 275 , 17428-17433 (2000). 183 7. Cutler, T.A., Mills, B.M., Lubin, D.J., Chong, L.T. & Loh, S.N. Effect of Interdomain Linker 184 Length on an Antagonistic Folding-Unfolding Equilibrium between Two Protein Domains. 185 Journal of Molecular Biology 386 , 854-868 (2009). 186 8. Doi, N. & Yanagawa, H. Design of generic biosensors based on green fluorescent proteins with 187 allosteric sites by directed evolution. Febs Letters 453 , 305-307 (1999). 188 9. Edwards, W.R., Busse, K., Allemann, R.K. & Jones, D.D. Linking the functions of unrelated 189 proteins using a novel directed evolution domain insertion method. Nucleic acids research 190 36 (2008). 191 10. Guntas, G. & Ostermeier, M. Creation of an allosteric enzyme by domain insertion. Journal of 192 Molecular Biology 336 , 263-273 (2004). 193 11. Ostermeier, M. Engineering allosteric protein switches by domain insertion. Protein Engineering 194 Design & Selection 18 , 359-364 (2005). 195 12. Grishin, N.V. Fold change in evolution of protein structures. Journal of Structural Biology 134 , 196 167-185 (2001). 197 13. Soding, J. & Lupas, A.N. More than the sum of their parts: on the evolution of proteins from 198 peptides. Bioessays 25 , 837-846 (2003). 199 14. Krishna, S.S. & Grishin, N.V. Structural drift: a possible path to protein fold change. 200 Bioinformatics 21 , 1308-1310 (2005). 201 15. Friedberg, I. & Godzik, A. Connecting the protein structure universe by using sparse recurring 202 fragments. Structure 13 , 1213-1224 (2005). 203 16. Ben-Tal, N. & Kolodny, R. Representation of the Protein Universe using Classifications, Maps, 204 and Networks. Israel Journal of Chemistry 54 , 1286-1292 (2014). 205 17. Park, K. et al. Control of repeat-protein curvature by computational protein design. Nat Struct 206 Mol Biol 22 , 167-74 (2015). 207 18. Koga, N. et al. Principles for designing ideal protein structures. Nature 491 , 222-7 (2012). 208 19. Hocker, B., Claren, J. & Sterner, R. Mimicking enzyme evolution by generating new (beta 209 alpha)(8)-barrels from (beta alpha)(4)-half-barrels. Proceedings of the National Academy of 210 Sciences of the United States of America 101 , 16448-16453 (2004). 211 20. Shanmugaratnam, S., Eisenbeis, S. & Hocker, B. A highly stable protein chimera built from 212 fragments of different folds. Protein Engineering Design & Selection 25 , 699-703 (2012). 213 21. DiMaio, F. et al. Improved low-resolution crystallographic refinement with Phenix and Rosetta. 214 Nature Methods 10 , 1102-1104 (2013). 215 22. Murzin, A.G., Brenner, S.E., Hubbard, T. & Chothia, C. Scop - a Structural Classification of 216 Proteins Database for the Investigation of Sequences and Structures. Journal of Molecular 217 Biology 247 , 536-540 (1995). 218 23. Huang, P.S. et al. RosettaRemodel: A Generalized Framework for Flexible Backbone Protein 219 Design. PloS one 6 (2011).
220 24. Zhang, Y. & Skolnick, J. TM-align: a protein structure alignment algorithm based on the TM- 221 score. Nucleic acids research 33 , 2302-2309 (2005). 222 25. Stivala, A., Wybrow, M., Wirth, A., Whisstock, J.C. & Stuckey, P.J. Automatic generation of 223 protein structure cartoons with Pro-origami. Bioinformatics 27 , 3315-3316 (2011). 224 26. Krissinel, E. & Henrick, K. Secondary-structure matching (SSM), a new tool for fast protein 225 structure alignment in three dimensions. Acta Crystallogr D Biol Crystallogr 60 , 2256- 226 (2004). 227 27. Shannon, P. et al. Cytoscape: a software environment for integrated models of biomolecular 228 interaction networks. Genome Res 13 , 2498-504 (2003). 229 28. Xiao, R. et al. The high-throughput protein sample production platform of the Northeast 230 Structural Genomics Consortium. Journal of Structural Biology 172 , 21-33 (2010). 231 232
233 FIGURE SUPPLEMENT TITLES/CAPTIONS
234 Figure 1. Domain insertion strategy for combining ferredoxin-top7 (A) and ferredoxin-ferredoxin (B).
235 Two beta strands from each partner (red and purple) are concatenated to form the central strand pair of
236 the fusion protein (pink).
237
238 Figure 2. Crystal structure of ferredoxin-top7 (4KYZ, chain A) aligned with design model (A) showing
239 core packing of the insert (B) and host (C) domains. Crystal structure colored by B-factor. Design model
240 in gray.
241
242 Figure 2—figure supplement 1. Circular dichroism spectra showing alpha and beta structure at 25°C for
243 ferredoxin-top7.
244
245 Figure 3. Crystal structure of ferredoxin-ferredoxin (5CW9) aligned with design model showing overall
246 alignment of helices (A) and the fused beta sheet (B). Crystal structure colored by B-factor. Design
247 model in gray.
248
249 Figure 3—figure supplement 1. Circular dichroism spectra showing alpha and beta structure at 25°C for 250 ferredoxin-ferredoxin. 251 252 253 Figure 3—figure supplement 2. Ferredoxin-Ferredoxin 2Fo-Fc omit map superimposed with crystal
254 structure shows core packing of host (A) and insert (B) domains.
