De novo protein design: what are we learning? William F. DeGrado, Daniel P. Raleigh and Tracey Handel The Du Pont Merck Pharmaceutical Company, Wilmington, Delaware, USA The field of de novo design is examined in the light of current theories of protein folding. The most carefully characterized designed proteins show many of the characteristics of molten globules or protein folding intermediates. Thus, it is unclear whether or not the objective of designing proteins that fully mimic the native states of natural proteins has been achieved. Current Opinion in Structural Biology 1991, 1:984-993 Introduction How do proteins fold into well defined three-dimen- sional structures? How do these structures relate to func- tion? These are intriguing questions, the answers to which lie at the heart of molecular biology and biochem- istry. Traditionally, these questions have been addressed through the systematic characterization of the structures and properties of natural proteins. Site-directed mutage- nesis and protein modification have additionally allowed one to determine which portions of the sequence of a protein are essential for its structure and function. Re- cently, de novo protein design has been introduced as a complementary approach [1,2]. De nora9 design refers to the process of designing a novel protein sequence that is intended to adopt a predetermined three-dimen- sional structure. By definition, the primary sequence of the designed protein should bear no intentional homol- ogy to any natural protein, although the desired overall fold might be precedented in nature. A major objective of de nat)o design is to evaluate the principles believed to influence protein structure and function. One begins with a certain set of assumptions and principles and then attempts to design proteins that should fold and func- tion. The success (or lack thereof) of the ensuing designs helps to evaluate and ultimately refine the principles and hypotheses implicit in the designs. To date, most proteins have been designed simply to fold into a desired struc- ture, whereas a few more ambitious designs have added catalytic functionality. The technique of protein design will yield important new mechanistic information only if the designed proteins are experimentally investigated with the same rigour as the most intensely studied natural proteins. In particu- lar, it is important to compare the physical and thermo- dynamic properties of designed proteins with the known folded states of natural proteins which, in addition to the native state, include the molten-globule state and protein folding intermediates. In recent years, much has been learned about the pathways by which proteins fold, and the intermediates that form in the folding process [3°°,4°,5,6]. These studies have provided attractive tools and conceptual frameworks for the study of designed proteins. Several different proteins have been shown to form compact folding intermediates with native-like secondary structure [7-9,10°o,11]. These intermediates contain hydrophobic binding sites that are presumably formed by clusters of apolar side chains, which are ap- parently not well enough packed to exclude hydropho- bic fluorescent probes such as 8-anilino-l-naphthalene- sulfonic acid (ANS). Also, amide proton-deuteron ex- change studies have shown that compact folding inter- mediates contain a degree of hydrogen bonding and structural stability which is intermediate between the un- folded and fully native states. For instance, when cy- tochrome c is dissolved in D20 , its amide protons ex- change with deuterons at rates 104-109 slower than ex- pected for a random coil -- the corresponding 'slowing factors' for the compact intermediate form of this protein are 101-104 [11]. Because of their transient nature (lifetime in milliseconds to second range), kinetic folding intermediates are dif- ficult to characterize. There is considerable current in- terest, therefore, in the study of 'molten-globule' states, which many investigators believe are good models for compact intermediates [732-16]. Certain combinations of pH, temperature, mutations, ionic strength and or- ganic solvents are known to destabilize the native states of proteins, giving rise to a more mobile but still com- pact form, the molten globule [12,13,16]. Like compact intermediates, molten globules have native-like secondary structure, bind hydrophobic molecules, and show slow- ing factors that are intermediate between those of the na- tive and unfolded form of a given protein. The molten globule also shows a very shallow temperature-depen- dent unfolding transition, in contrast to the sharp co- operative transitions observed for most native proteins [12,13]. Furthermore, NMR studies show that the in- terior side chains in molten globules are relatively mobile and adopt a variety of conformers which interconvert on the millisecond to sub-millisecond time scale [17,18]. AcTyrOEl--acetyl-tyrosine methyl ester; ANS~8-anilino-l-naphthalenesulfonic acid; BPTl~bovine pancreatic trypsin inhibitor; CD~circular dichroism; DDT~ichlorodiphenyltrichloroethanol; NOE--nuclear Overhauser effect. 984 © Current Biology Lid ISSN 0959-440X