Design and Synthesis of a Globin Fold ² Yasuhiro Isogai,* ,‡ Motonori Ota, § Tetsuro Fujisawa, | Hiroyuki Izuno, ⊥ Masahiro Mukai, ‡ Hiro Nakamura, ‡ Tetsutaro Iizuka, ‡ and Ken Nishikawa § The Institute of Physical and Chemical Research (RIKEN), 2-1 Hirosawa, Wako, Saitama 351-0198, Japan, National Institute of Genetics, Yata, Mishima, Shizuoka 411-8540, Japan, The Institute of Physical and Chemical Research (RIKEN), RIKEN Harima Institute, Mikazuki-cho, Sayo, Hyogo 679-5143, Japan, and Department of Physics, Faculty of Science, Gakushuin UniVersity, Mejiro, Toshima-ku, Tokyo 170-0031, Japan ReceiVed December 21, 1998; ReVised Manuscript ReceiVed April 6, 1999 ABSTRACT: We propose a simple method to find an amino acid sequence that is foldable into a globular protein with a desired structure based on a knowledge-based 3D-1D compatibility function. An asymmetric R-helical single-domain structure of sperm whale myoglobin consisting of 153 amino acid residues was chosen for the design target. The optimal sequence to fit the main-chain framework has been searched by recursive generation of the protein 3D profile. The heme-binding site was designed by fixing His64 and His93 at the distal and proximal positions, respectively, and by penalizing residues that protrude into the space with a repulsive function. The apparent bumps among side chains in the computer model of the converged, self-consistent sequence were removed by replacing some of the bumping residues with smaller ones according to the final 3D profile. The finally obtained sequence shares 26% of sequence with the natural myoglobin. The designed globin-1 (DG1) with the artificial sequence was obtained by expression of the synthetic gene in Escherichia coli. Analyses using size-exclusion chromatography, circular dichroism spectroscopy, and solution X-ray scattering showed that DG1 folds into a monomeric, compact, highly helical, and globular form with an overall molecular shape similar to the target structure in an aqueous solution. Furthermore, it binds a single heme per protein molecule, which exhibited well-defined spectroscopic properties. The radius of gyration of DG1 was determined to be 20.6 Å, slightly larger than that of natural apoMb, and decreased to 19.5 Å upon heme binding based on X-ray scattering analysis. However, the heme-bound DG1 did not stably bind molecular oxygen as natural globins do, possibly due to high conformational diversity of side-chain structures observed in the NMR and denaturation experiments. These results give insight into the relationship between the sequence selection and the structural uniqueness of natural proteins to achieve biological functions. De novo protein design is now thought to be an essential approach to elucidate the principles of protein architecture and has potential applications to yield novel molecules for medical and industrial aims such as drug discovery (for recent reviews, see refs 1-3). The novel paradigm to design artificial sequences that fold into a desired three-dimensional (3D) 1 structure may originate from the conceptual proposal in the early 1980s, at nearly the same time that the term ‘protein engineering’ began to be used (4, 5). Since then, many efforts have been made to determine the structural factors that govern the folding, stability, and functions of artificial and native proteins using both experimental and theoretical approaches. The first successful design of a well- defined global fold was achieved on a four-helix-bundle motif which is widely found in native protein structures (6- 8). The methodology was based on manual model building using the simple binary patterns of polar and nonpolar residues along the sequences for the helices. The sequence selection for this fold has been made more sophisticated to increase the local structural specificity, and a native-like, highly ordered structure has been obtained (9-11). Besides symmetrical helical structures, successful, or at least partially successful, designs of -sheets containing folds have been achieved by considering the geometry and statistical prefer- ences of amino acids for the secondary structure (12-14). On the other hand, systematic and quantitative methods using computer algorithms have been developed to establish a general strategy for designing a desired, more challenging 3D structure. These design algorithms, which mainly origi- nate from 3D structure prediction, have been tested and advanced in redesigning separate parts of native proteins ² This work was supported in part by the Biodesign Research Program and MR Science Program of RIKEN (to Y.I., H.N., and T.I.), by the Special Postdoctoral Researchers’ Program of RIKEN (to M.M.), and by Grant-in-Aids for Scientific Research from the Ministry of Education, Science, Culture, and Sports of Japan (to Y.I. and to K.N.). * To whom correspondence should be addressed. Fax: 81-48-462- 4660. E-mail: yisogai@postman.riken.go.jp. ‡ The Institute of Physical and Chemical Research (RIKEN). § National Institute of Genetics. | RIKEN Harima Institute. ⊥ Gakushuin University. 1 Abbreviations: CD, circular dichroism; Cm, midpoint denaturant concentration; 1D, one-dimensional; 3D, three-dimensional; DG, designed globin; Gd-HCl, guanidine hydrochloride; HPLC, high- performance liquid chromatography; K d, dissociation constant; Mb, myoglobin; Mr, relative molecular mass; NMR, nuclear magnetic resonance; PAGE, polyacrylamide gel electrophoresis; pI, isoelectric point; Rg, radius of gyration; SCS, self-consistent sequence; SDS, sodium dodecyl sulfate; TFA, trifluoroacetic acid; UV, ultraviolet. 7431 Biochemistry 1999, 38, 7431-7443 10.1021/bi983006y CCC: $18.00 © 1999 American Chemical Society Published on Web 05/15/1999