Protein Folding and Macromolecular 
Recognition 
Peter S. Kim, Ph.D. — Assistant Investigator 
Dr. Kim is also Member of the Whitehead Institute for Biomedical Research, Associate Professor of Biology 
at the Massachusetts Institute of Technology, and Assistant Molecular Biologist at the Massachusetts 
General Hospital, Boston. His undergraduate degree in chemistry was obtained at Cornell University, 
where he studied with George Hess. After receiving the Ph.D. degree in biochemistry from Stanford 
University, where he studied with Robert Baldwin, Dr. Kim moved to the Whitehead Institute for 
Biomedical Research as a Whitehead Fellow. 
INFORMATION transfer in biology generally 
proceeds from DNA to RNA (transcription) and 
then from RNA to protein (translation) . The lin- 
ear, unfolded protein chains made during transla- 
tion must fold into a three-dimensional shape to 
be functional. Although the basic mechanisms of 
transcription and translation are understood, at 
least in outline, the transfer of information from 
one to three dimensions — i.e., protein folding — 
remains a major unsolved problem in molecular 
biology. To understand protein folding is a prime 
objective of this laboratory. 
A second effort is aimed at understanding the 
principles of macromolecular recognition: spe- 
cific protein-protein interactions and interac- 
tions between protein molecules and DNA. These 
interactions are central to much of molecular 
physiology and developmental biology. We have 
focused on a structural motif called the leucine 
zipper, which occurs in several different DNA- 
binding proteins, including the products of some 
nuclear oncogenes. 
A third and new effort is the de novo design of 
peptides and proteins. 
Protein Folding 
Much of our work in this area is centered on 
bovine pancreatic trypsin inhibitor (BPTI) , argu- 
ably the protein most thoroughly characterized 
in biophysical terms. It is difficult to determine 
the structures of protein-folding intermediates, 
because protein folding is a cooperative process. 
Trapped disulfide-bonded intermediates, such as 
those identified in the early folding steps of BPTI, 
are often rather insoluble; this hinders detailed 
structural characterization by nuclear magnetic 
resonance (NMR) . We have developed a peptide 
model approach that circumvents the problem of 
cooperativity and improves solubility, so that the 
structures contained within protein-folding in- 
termediates can be characterized in detail. 
Peptide models that simulate two crucial early 
intermediates in the folding of BPTI have been 
designed and synthesized chemically. By using 
two-dimensional NMR, we find that the struc- 
tures contained within these peptide models are 
remarkably native- like, corresponding to subdo- 
mains of BPTI. These results suggest that a large 
part of the protein-folding problem can be re- 
duced to identifying and understanding subdo- 
mains of native proteins. 
Earlier work by others, however, concluded 
that there are well-populated, nonnative states in 
the oxidative folding of BPTI. This conclusion 
complicates efforts to understand protein fold- 
ing. Recently we reexamined the spectrum and 
population of intermediates present during the 
folding of BPTI, taking advantage of improve- 
ments that have been made in separation technol- 
ogies in the years since the original BPTI-folding 
experiments. In contrast to earlier studies, we 
find that all of the well-populated intermediates 
in the folding of BPTI contain only native disul- 
fide bonds and that the essential features of the 
BPTI-folding reaction are determined in large 
part by native structure. These results emphasize 
the importance of native protein structure for un- 
derstanding protein folding. 
A recombinant model for an early predominant 
intermediate, containing a single disulfide bond 
between residues 5 and 55, has been made by 
replacing the cysteines not involved in the disul- 
fide bond with alanine. Remarkably, this model 
folds essentially into the same conformation as 
native BPTI, as judged by two-dimensional NMR, 
and it inhibits trypsin. These findings provide an 
explanation for the properties of this interme- 
diate in the folding of BPTI and demonstrate that 
the native fold of BPTI can be obtained without 
the assistance of nonnative disulfide species. The 
recombinant model also provides an attractive 
model system for studies of protein folding. This 
work is supported by a grant from the National 
Institutes of Health. 
Other efforts are directed at evaluating electro- 
static fields at the ends of a-helices and develop- 
ing a model system to evaluate the /3-sheet pro- 
pensities of different amino acid residues. 
Macromolecular Recognition 
In this area, we have focused on the leucine 
zipper class of DNA-binding transcriptional acti- 
225 
