NATURE|Vol 438|1 December 2005|doi:10.1038/nature04395 INSIGHT REVIEW
581
Solving the membrane protein
folding problem
James U. Bowie
1
One of the great challenges for molecular biologists is to learn how a protein sequence defines its three-
dimensional structure. For many years, the problem was even more difficult for membrane proteins because
so little was known about what they looked like. The situation has improved markedly in recent years, and we
now know over 90 unique structures. Our enhanced view of the structure universe, combined with an
increasingly quantitative understanding of fold determination, engenders optimism that a solution to the
folding problem for membrane proteins can be achieved.
(ref. 7). This structure illustrates some of the complexities of mem-
brane-protein architecture. Two views of the GlpF monomer (the pro-
tein is a tetramer) are shown in Fig. 2, and I will call them front and
back. The back side presents a simple picture of TM helices packed
together roughly parallel to the membrane normal. Few would have
been surprised by this image, even 30 years ago. The front side pre-
sents a more complex view, however. Most notable is the pair of
helices that penetrate half way into the bilayer, where their amino-ter-
mini meet. Such half TM helices are not very common, but are also
not particularly unusual (about 1 in 20 of all TM helices)
8
. Also visi-
ble in the front view is a common feature of membrane proteins: a
highly distorted TM helix. About 60% of TM helices contain signifi-
cant bends or other distortions
9
. Finally, loop conformations between
In Ilse Aichinger’s famous short story The Bound Man, a man awakes
from a coma to find himself bound by ropes
1
. He learns to move grace-
fully within these constraints and eventually becomes a circus per-
former. Similarly, membrane proteins must perform complex signalling
and transport functions within the strict confines of a lipid bilayer. This
requires elegant structural adaptations to their environment.
Our early views of membrane-protein structure were largely
shaped by the pioneering work of Henderson and colleagues, who
established a modest resolution view of bacteriorhodopsin in 1975
(ref. 2). Confirming the prescient prediction of Lenard and Singer in
1966 (ref. 3), the structure revealed a bundle of long helical rods tra-
versing the membrane. This suggested that membrane proteins could
be largely thought of as an assembly of transmembrane (TM) helices.
This view was reinforced by the stretches of hydrophobic residues
found in membrane-protein sequences that were long enough to form
membrane-spanning helices
4
. But recent structures have challenged
this simple concept of membrane-protein architecture. In this review,
I will briefly illustrate some of the surprising twists and turns that
polypeptide chains make in the bilayer and then discuss our current
understanding how these structures are built. Finally, I will argue that
with concerted effort, practical solutions to the membrane-protein-
folding problem are possible.
Complex architectures
Protein structure within the bilayer can be divided into two general
types: ȋ-barrels and bundles of Ȋ-helices. Because the folding problem
for ȋ-barrels and helix bundles is very different, and because ȋ-barrel
proteins are much less common
5
, I will focus on the helix bundle class.
Figure 1 shows our progress with helix bundle protein-structure deter-
mination since the first high-resolution membrane protein structure
was solved 20 years ago (reviewed in ref. 6). A decade later, the pace of
structure determination quickened. We now know 52 unique helix
bundle structures, and 38 of these were solved in the past 5 years. As a
result, we have a much clearer view of the structural diversity exhib-
ited by membrane proteins. Although long TM helices are still consid-
ered a fundamental building block, polypeptide chains can be
organized into other complex patterns.
Perhaps the first dramatic departure from the standard view was
the structure of the glycerol/water channel GlpF, determined in 2000
1
Department of Chemistry and Biochemistry, UCLA-DOE Center for Genomics and Proteomics, Molecular Biology Institute, Boyer Hall, UCLA, 611 Charles E. Young Drive E, Los Angeles,
California 90095-1570 USA.
Number of Helix bundle structures solved
Year
0
2
4
6
8
10
12
1985
1986
1987
1988
1989
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
1990
Figure 1 | Progress of helix bundle membrane protein structure
determination. Only unique structures are included. The data were
obtained from a website maintained by Stephen White
(http://blanco.biomol.uci.edu/Membrane_Proteins_xtal.html). For a
similar plot including all membrane protein structures, see ref. 6.
Nature Publishing Group ©2005