Tanja Kortemme, ^*Marina Ramirez-Alvarado, ^*Luis Serrano

A 20-residue protein (named Betanova) forming a monomeric, three-stranded, antiparallel  sheet was designed using a structuralbackbone template and an iterative hierarchical approach. Structuraland physicochemical characterization show that the -sheet conformation is stabilized by specific tertiary interactions and that the protein exhibits a cooperative two-state folding-unfolding transition, which is a hallmark of natural proteins. The Betanova molecule constitutes a tractable model system to aid in the understandingof -sheet formation, including -sheet aggregation and amyloid fibril formation.

European Molecular Biology Laboratory (EMBL), Meyerhofstrasse 1, Heidelberg D-69117, Germany.
^*   These authors contributed equally to this work.

   To whom correspondence should be addressed. E-mail: Kortemme@EMBL-Heidelberg.DE (T.K.) and Ramirez@EMBL-Heidelberg.DE (M.R.-A.).

A key feature for the successful design of model proteins is to make them simpler than their natural counterparts, yet able to attain a unique conformation in aqueous solution (12). Our goal has therefore been the design of a small, soluble, monomeric, model-sheet protein containing only natural amino acids, which would fold in the absence of disulfide bonds or metal binding sites. We selected a three-stranded antiparallel  sheet composed of four residues per strand and two-residue  turns as our backbone framework (Fig. 1A), which is a minimal unit retaining all the characteristics of larger -sheet proteins. We performed selection and optimization of a sequence compatible with our target backbone structure, considering experimental information on -hairpin stability (13), amino acid -sheet propensities (2, 3), and statisticalpreferences for interstrand residue pairs (14), and evaluatingcombinations of side chain rotamers (side chain rotamer modeling).An iterative design and analysis procedure (Fig. 1B) (15) resultedin the successful design of a 20-amino acid three-stranded  sheet (named Betanova). 

Our initial sequence was a longer version of a designed denovo-hairpin peptide (5) that adopts a folded structure in equilibriumwith random coil conformations (hairpin I). The hairpin structurewas elongated by placement of a Lys⁺-Glu pair at the outer edges of the previous model peptide (hairpin II), leading to an increase in the structured population from about 35% to about 45% (16). The hairpin II sequence was placedon the central ( strand 2) and COOH-terminal ( strand 3)  strands of our framework. For the NH₂-terminal  strand ( strand 1), we searched for the best sequence in terms of statistical analysis of the protein structure database and favorable van der Waals contacts, using rotamer modeling. The turn sequences were selected to be optimal for type I'  turns in hairpins (17). The resulting peptide (sheet I) was soluble and monomeric up to a concentration of 3 mM. NMR analysis showed evidence for structure only in the region corresponding to hairpin II.

In a new set of peptides (sheets II and III), some of the residues in  strand 1 were changed to modify the interstrand packing and to introduce an additional Asp/Glu-Lys⁺ ionic pair. We found that sheets II and III behaved similarly to sheet I. However, the designed sheet I sequence folded into the target structure upon addition of 40% 2,2,2-trifluoroethanol. This indicated that the lack of formation of a well-defined three-stranded sheet in aqueous solution was not the result of side chain-side chain incompatibilities at the interface between the NH₂-terminal and central  strands. A likely explanation is that the amount of hydrophobic surface area buried by the four aliphatic sidechains on  strands 1 and 2 is not sufficient to drive -sheetformation. In contrast, the sequence corresponding to hairpinII buries a large hydrophobic surface area because of the favorablepacking of a Tyr residue ( strand 3) with the side chains ofIle, Val ( strand 2), and Lys ( strand 3). These results prompted us to search for the possibility of introducing an aromatic residue into  strand 1 in order to create extensive contacts with residues on  strand 2. Exploring different rotamers for aromatic residues in  strand 1 showed that an aromatic side chain could pack favorably onto the surface of the -sheet structure only in the absence of -branched residues on  strand 2. A good example of this residue arrangement is found in WW domains (so termed because of the conservation of two tryptophane residues in this protein domain family) (18,19). Considering this information and molecular modeling using the ICM (internal coordinates mechanics) package (20), we modifiedthe sheet I sequence and termed this new sequence Betanova (Fig.1).

The Betanova molecule (21) was soluble in water and monomeric up to a concentration of 2.6 mM, as determined by analytical ultracentrifugation, circular dichroism (CD), and NMR spectroscopy (22). The presence of characteristic long-range backbone-backbone and side chain-side chain nuclear Overhauser effects (NOEs) (23) between residues in adjacent  strands and, more important, the NOEs between aromatic protons of Trp³ and C protons of Thr¹⁷ (Fig. 2) indicated the formation of the desired three-stranded sheet. This was corroborated by the large ³J_NH coupling constant values measured for-strand residues, as well as by the conformational secondary chemical shift profiles of the ¹³C and ¹³C carbon nuclei (24). 

A hallmark of proteins with defined native conformation, when compared to partly folded proteins or peptides, is that theformer exhibit cooperative folding/unfolding transitions. Thedenaturation process of Betanova was monitored by CD and fluorescence spectroscopy (Fig. 3) (25). Betanova displayed cooperative behaviorboth by thermal (Fig. 3A) and chemical (Fig. 3B) denaturation,with a broad transition as expected from its small size (2.3 kD).The experimentally determined dependence of free energy on denaturant concentration (the m value) was 0.4 kcal mol¹ M¹. This value, which is indicative of the amount of surface area exposed upon unfolding, is in very good agreement with the expected values for the calculated total buried surface area (26). The free energy change at 278 K obtained from both thermal and chemical denaturation was similar (around 0.6 to 0.7 kcal mol¹). 

We determined a three-dimensional model structure of Betanova compatible with the NOE and ³J_NH coupling constant data (27). All 45 structures without NOE violations larger than 0.2 A calculated with the dynamicsalgorithm for NMR applications (DYANA) exhibit the expected antiparallel,three-stranded-sheet fold with two-residue turns located at the designed positions and the expected right-handed twist of the  sheet (28) (Fig. 4A). The first and last two residues appear to be disordered, as expected from the design, which includes terminal Arg residues for solubility and Gly residues as flexible linkers to the structured region as described previously (5).Comparison of the averaged minimized structure obtained from theNMR restraints with the target model shows a root-mean-square(rms) deviation of 1.11 A for the backbone, demonstrating thesuccess of our design (Fig. 4B). The packing of aromatic sidechains in sheets appears to contribute significantly to their stability, as well as to establish important conformational constraints for defining a single conformation (Fig. 4C). 

It is remarkable that this designed  sheet has no real hydrophobic core (residues inaccessible to solvent), and although there is a hydrophobic cluster on one face of the  sheet, most of the residues involved in the packing also have polar groups (Fig. 4C). The absence of a hydrophobic core most likely induces a certain flexibility in the  sheet while still allowing definedtertiary interactions.

The simplicity of Betanova, together with the fact that it retains all the structural properties of -sheet proteins, makes it an optimum model to use in refining existing molecular dynamics protocols, as well as to test recent theoretical approaches to protein folding (29). Recently developed computational methods for protein design have shown that it is possible to explore and evaluate all possible amino acid sequences for a given protein framework (10). This raises the question of how many sequences can adopt the same stable fold. Betanova constitutes a suitable system for exploring the sequence space experimentally.

The design of a -sheet protein using information derived from denovodesign and from structural stabilizing motifs demonstratesthat we are starting to understand the principles behind -sheetformation. The small size of the designed protein is critical for rationalizing the rules governing protein architecture and folding, as well as for the development of suitable scaffolds for introducing different functionalities. This knowledge may help to elucidate the mechanisms leading to -sheet aggregation.

Note added in proof: Schenck and Gellman reported the characterization of a three-stranded  sheet using D-proline in the turn and an unnatural amino acid in one of the strands (31).

REFERENCES AND NOTES

1. J. W. Kelly, Structure5, 595 (1997)[ISI][Medline];D. R. Booth, et al., Nature 385, 787 (1997)[ISI][Medline].
2. D. L. Minor Jr. and P. S. Kim, Nature 367, 660 (1994)[ISI][Medline];ibid. 371, 264 (1994).
3. C. K. Smith and L. Regan, Science270, 980 (1995)[ISI][Abstract].
4. E. de Alba, M. A. Jiménez, M. Rico, J. L. Nieto, Folding Des. 1, 133 (1996).
5. M. Ramírez-Alvarado, F. J. Blanco, L. Serrano, Nature Struct. Biol. 3, 604 (1996)[Medline].
6. A. Pessi, et al., Nature362, 367 (1993)[ISI][Medline];T. P. Quinn, N. B. Tweedy, R. W. Williams, J. S. Richardson, D. C. Richardson, Proc. Natl. Acad. Sci. U.S.A.91, 8747 (1994)[ISI][Abstract];Y. Yan and B. W. Erickson, Protein Sci.3, 1069 (1994)[Medline].
7. E. Ilyina, V. Roongta, K. Mayo, Biochemistry 36, 5245 (1997)[ISI][Medline].
8. L. Regan and W. F. DeGrado, Science241, 976 (1988)[ISI][Medline];M. H. Hecht, J. S. Richardson, D. C. Richardson, R.C. Ogden, ibid.249, 884 (1990)[Medline];S. Kamtekar, J. M. Schiffer, H. Xiong, J. M. Babik, M. H. Hecht,ibid. 262, 1680 (1993); J. W. Brysonet al., ibid. 270, 935 (1995); C. E. Schafmeister, S.L. LaPorte, L. J. W. Miercke, R. M. Stroud, NatureStruct. Biol. 4, 1039 (1997)[ISI][Medline].
9. K. Goraj, A. Renard, J.A. Martial, Protein Eng. 3, 259 (1990)[ISI][Abstract];M. D. Struthers, R. P. Cheng, B. Imperiali, Science271, 342 (1996)[ISI][Abstract].
10. B. I. Dahiyat and S. L. Mayo, Science278, 82 (1997)[ISI][Abstract/Full Text]; B. I. Dahiyat, C. A. Sarisky, S. L. Mayo, J. Mol. Biol. 273, 789 (1997)[ISI][Medline].
11. D. T. Jones, et al., ProteinsStruct. Funct. Genet. 24, 502 (1996)[ISI];S. Dalal, S. Balasubramanian, L. Regan, Nature Struct. Biol.4, 548 (1997)[ISI][Medline].
12. W. F. DeGrado, Z. R. Wasserman, J. D. Lear, Science 243, 622 (1989)[ISI][Medline].
13. F. J. Blanco, M. Ramírez-Alvarado, L. Serrano, Curr. Opin. Struct. Biol. 8, 107 (1998)[ISI][Medline].
14. M. A. Wouters and P. M. G. Curmi, Proteins Struct. Funct. Genet. 22, 119 (1995)[ISI].
15. All sequence selection steps for -sheet design were based on the use of two peptide backbone templates correspondingto antiparallel  sheets: fragment 57-72 of glyceraldehyde-3-phosphatedehydrogenase from Bacillus stearothermophilus {Protein Data Bank accession code 1gd1 [ T. Skarzynoski, P.C. E. Moody, A. J. Wonacott, J.Mol. Biol. 193, 171 (1987)[ISI][Medline]]}and the WW domain of the mouse formin-binding protein, fragment8-30 (20). The WW domain fragment contained loops between the  strands, which were changed to ideal, two-residue, type I' turns to yield a three-stranded antiparallel  sheet with four residues per strand, connected by two-residue  turns for both backbone templates. The different sequences were evaluated by the lowest van der Waals energies calculated by the ICM package (20). The template structures were prepared for sequence evaluation by a regularization procedure included in the ICM package that undergoes a rotational positioning of methyl groups, an iterative optimization of geometry and energy of the whole structure, and an adjustment of polar hydrogen positions.
16. M. Ramírez-Alvarado, F. J. Blanco, L. Serrano, unpublished data.
17. M. Ramírez-Alvarado, F. J. Blanco, H. Niemann, L. Serrano, J. Mol. Biol. 273, 898 (1997)[Medline].
18. M. J. Macías, et al., Nature 382, 646 (1996)[Medline].
19. M. J. Macías and H. Oschkinat, personal communication.
20. R. A. Abagyan, M. M. Totrov, D. N. Kuznetsov, J. Comp. Chem. 15, 488 (1994).
21. The Betanova polypeptide was synthesized by the laboratory of R. Frank at the Center for Molecular Biology, Heidelberg University (ZmBH), using standard 9-fluorenyl-methoxycarbonyl chemistry. Purification to >95% was achieved by reversed-phase high-performance liquid chromatography. The molecular mass of the peptide was confirmed by mass spectroscopy.
22. Sedimentation equilibrium studies were performed at 278 K at a peptide concentration of 1.3 mM in aqueous solution (pH 5.0), using a Beckman XL-A ultracentrifuge equipped with a An-50 Ti rotor at a speed of 50,000 rpm. Fitting of the equilibrium radial concentration distribution to an ideal single-component model using a partial specific volume of 0.6825 × 10³ m³ kg¹, calculated from the amino acid sequence, resulted in a molecular mass of 2259 daltons, which is in excellent agreement with the expected molecular mass for a monomer of 2257 daltons. The monomeric state of Betanova was additionally confirmed by the concentration independence of the far-ultraviolet (UV) CD spectra between 5 µM and 1 mM in 5 mM sodium acetate (pH 5.0) and of the NMR line widths and chemical shift values at peptide concentrations between 20 µM and 2.6 mM, in 90% (v/v) ¹H₂O/10% ²H₂O (pH 5.0).
23. NMR measurements were performed at temperatures of 273, 278, and 280 K in 90% (v/v) ¹H₂O/10% ²H₂O (pH 5.0) at a peptide concentration of 2.6 mM, or in 99.8% ²H₂O (pH 5.0) (uncorrected for deuterium isotope effects) at apeptide concentration of 2.2 mM. Spectra were acquired on Bruker DRX-500 or DRX-600 spectrometers, operating at 500.13 and 600.13 MHz, respectively. Samples contained sodium 3-trimethylsilyl (2,2,3,3-²H₄) propionate as an internal reference. double-quantum filter correlated spectroscopy (DQFCOSY) [ U. Piantini, O. W. Sørensen, R. R. Ernst, J. Am. Chem. Soc. 104, 6800 (1982)[ISI]],NOE spectroscopy (NOESY) [ J. Jeener, B. H. Meier, P. Bachmann, R. R. Ernst, J. Chem. Phys. 71, 4546 (1979)[ISI];S. Macura and R. R. Ernst, Mol. Phys.41, 95 (1980)[ISI];mixing time (_m) = 100, 130, and 200 ms], and rotating frame overhauser effect spectroscopy (ROESY) [ A. A. Bothner-By, R. L. Stephens, J.M. Lee, C. D. Warren, R. W. Jeanloz, J. Am. Chem. Soc. 106, 811 (1984)[ISI]; _m = 100, 130, and 200 ms] spectra were acquired for resonance assignment using standard procedures [K. Wüthrich, NMR of Proteins and Nucleic Acids (Wiley, New York, 1986)]. NOESY and ROESY spectrawere jointly analyzed to discard artifactual NOEs as those arisingfrom spin diffusion. Total correlated spectroscopy (TOCSY) [ A. Bax and D. G. Davies, J. Magn. Reson. 65, 355 (1985)[ISI]; _m = 80 ms] spectra were acquired using the standard MLEV-17 spin lock sequence. Water suppression was achieved by selective presaturationduring the relaxation delay (1.2 s) or field-gradient pulses [ M. Piotto, V. Saudek, V. Sklenar, J. Biomol. NMR 2, 661 (1992)[ISI][Medline];V. Sklenar, M. Piotto, R. Leppik, V. Saudek, J.Magn. Reson. A102, 241 (1993)[ISI]].The spin-spin coupling constants ³J_HN were measured from 2D TOCSY and NOESY spectra using the method of Stonehouse and Keeler as implemented in the program MEDEA [ J. Stonehouse and J. Keeler, J. Magn. Reson. A112, 43 (1995)[ISI]]. ¹³C and ¹³C chemical shifts were obtained from natural abundance ¹³C heteronuclear multiple quantum coherence (HMQC) spectra [ A. Bax, R. H. Griffey, B. L. Hawkins, ibid.55, 301 (1983)].
24. T. Kortemme, M. Ramírez-Alvarado, L. Serrano, data not shown. The conformational shifts of the ¹³C and ¹³C nuclei are available as supplementary material at Science Onlineat http://www.sciencemag.org/feature/data/980831.shl.
25. CD measurements were recorded on a JASCO-710 spectropolarimeter equipped with a Peltier-type temperature control system. Thermal denaturation was monitored at 217 nm in a 2-mm stoppered cuvette, with a temperature slope of 50°C hour¹, at a peptide concentration of 100 µM in 5 mm sodium acetate (pH 5.0) containing 10% (v/v) glycerol. The glycerol was present to allow measurements at temperatures down to 268 K. Measurements between 273 and 368 K gave identical traces in the presence and absence of glycerol. Reversibility was demonstrated by cooling the solution and repeating the thermal denaturation, resulting in a profile identical to the initial measurement. Van't Hoff analysis of the data yielded values of the enthalpy change at 278 K (H_(278K) = 3.3 kcal mol¹), of the entropy change at 278 K (S_(278K) = 0.01 kcal mol¹ K¹), and of the free energy change at 278 K (G_(278K) = 0.6 kcalmol¹), indicating the presence of 80 to 90% folded structure under these conditions. The energy values have to be considered approximate because of the difficulty in defining the upper baseline of the transition. The value for the heat capacity at constant pressure (C_P) used in the analysis was estimated from the change in accessible polar and nonpolar surface area upon unfolding, according to (26)or as described by K. P. Murphy and E. Freire [Adv. Prot. Chem. 43, 313 (1992)]. In both cases, the value forC_P was around 0.14 kcal mol¹ K¹. Chemical denaturation by urea was followed using fluorescence spectroscopy. Urea concentrations were determined refractometrically.Fluorescence emission at 352 nm (excitation, 290 nm) was measuredat 278 K at a peptide concentration of 6 µM using an Aminco Bowman Series 2 luminescence spectrometer. The denaturation profile was fitted to F = {F_F + A × [urea] + ((F_U + B × [urea]) × exp ((G_H2O+ (m[urea]))/RT))}/{1 + exp ((G_H2O + (m[urea]))/RT)}, where A and B are the slopes of the urea dependence of the fluorescence of the folded and unfolded states, respectively. F, F_F, and F_U are the observed fluorescence values at a particular urea concentration and the fluorescence of the folded state and that of the unfolded state, respectively.G_H2O is the free energy of unfolding in the absence of urea, and m is the dependence of G_H2O on denaturant concentration. Fitting yielded an m value of 0.4 kcal mol¹ M¹ and aG_H2O of 0.7 kcal mol¹. The obtained m value is in excellent agreement with the one expected (0.44 kcal mol¹ M¹) for a protein of this size, using the empirical correlation described in (26). However, the experimental values of m andG_H2Ohave to be taken with caution because of the small change in fluorescencebetween folded and unfolded states and the large error in the estimation of the slopes of the urea dependence of the fluorescence of the folded and unfolded states, arising from the solvent exposure of the Trp side chain.
26. J. K. Myers, C. N. Pace, J. M. Scholtz, Protein Sci. 4, 2138 (1995)[ISI][Medline].
27. Cross-peak intensities were obtained from 130-ms NOESY spectra collected at 273 and 280 K in 90% (v/v) ¹H₂O/10% ²H₂O (pH 5.0) or 99.8% ²_H2O (pH 5.0). Each NOE was classified as strong, medium, weak, or very weak by visual inspection of the contour levels or intensityintegration of the cross-peaks in NOESY spectra and was assignedto upper-limit distance restraints as follows: strong (2.5 Å),medium (3.5 Å), weak (5.0 Å), and very weak (5.5 Å). Dihedral restraints were derived from the ³J_HN coupling constants. With a set of 40 sequential, 48 nonsequential,and 11 dihedral restraints, a family of 45 structures withoutNOE violations larger than 0.2 Å were generated using DYANA [ P. Güntert, C. Mumenthaler, K. Wüthrich, J. Mol. Biol. 273, 283 (1997)[Medline]].The rms deviation from the average minimized structure in the well-ordered region of residues 3-18 was 0.96 Å for the backbone atoms and 1.93 Å for all heavy atoms.
28. C. Chothia, J. Mol. Biol.75, 295 (1973)[ISI][Medline].
29. T. Lazaridis and M. Karplus, Science278, 1928 (1997)[ISI][Abstract/Full Text].
30. J. Prieto, M. Wilmans, M.A. Jiménez, M. Rico, L. Serrano, J. Mol. Biol.268, 760 (1997)[ISI][Medline].
31. H. L. Schenck and S. H. Gellman, J.Am. Chem. Soc. 120, 4869 (1998)[ISI].
32. We thank M. J. Macías and H. Oschkinat for communicating results before publication; H. van der Zandt for help with the sedimentation equilibrium measurements; T. Creighton and M. Saraste for critical reading of the manuscript; M. Sattler for discussions concerning NMR experiments; and M. Petukhov, E. Lacroix, and J. Martínez for helpful discussions. This work was partly supported by a European Union Biotechnology grant (BIO4-CT97-2086).

Abstract of this Article Reprint (PDF) Version of this Article Similar articles found in:  SCIENCE Online ISI Web of Science PubMed PubMed Citation This Article has been cited by:  Search Medline for articles by: Kortemme, T. || Serrano, L. Search for citing articles in:  ISI Web of Science (20) Alert me when:  new articles cite this article Download to Citation Manager

Collections under which this article appears: Biochemistry

This article has been cited by other articles:

• DeGrado, W. F., Summa, C. M., Pavone, V., Nastri, F., Lombardi, A. (1999). DENOVODESIGN AND STRUCTURAL CHARACTERIZATION OF PROTEINS AND METALLOPROTEINS. Annu. Rev. Biochem. 68: 779-819 [Abstract][Full Text]
• Sung, S.-S. (1999). Monte Carlo Simulations of beta -Hairpin Folding at Constant Temperature. Biophys J 76: 164-175 [Abstract][Full Text]
• Ibragimova, G. T., Wade, R. C. (1999). Stability of the beta -Sheet of the WW Domain: A Molecular Dynamics Simulation Study. Biophys J 77: 2191-2198 [Abstract][Full Text]
• Walsh, S. T. R., Cheng, H., Bryson, J. W., Roder, H., DeGrado, W. F. (1999). Solution structure and dynamics of a de novodesigned three-helix bundle protein. Proc. Natl. Acad. Sci. U. S. A. 96: 5486-5491 [Abstract][Full Text]

Volume 281, Number 5374 Issue of 10 Jul 1998, pp. 253 - 256 
©1998 by The American Association for the Advancement of Science.