Energetics of membrane protein folding and stability.

Abstract

The critical role of membrane proteins in a myriad of biological and physiological functions has spawned numerous investigations over the past several decades with the long-term goal of identifying the molecular origins and energetic forces that stabilize these proteins within the membrane. Parallel structural and thermodynamics studies on several systems have provided significant insight regarding the driving forces governing folding, assembly, insertion, and translocation of membrane proteins. The present review surveys families of membrane-associated proteins including alpha-helical and beta-barrel structures, viral surface receptors, and pore-forming toxins, citing representative proteins within each of these classes for further scrutiny in terms of structure-function relationships and global conformational stability. This overview presents seminal findings from pioneering studies on the energetics of membrane protein folding and stability to modern techniques that are exploiting the use of molecular genetics and single molecule studies. An overall consensus regarding the molecular origins of membrane protein stability is that a number of intrinsic properties resemble features of soluble proteins, yet there are distinct energetic differences arising from specific intra- and intermolecular interactions within the membrane. The combined efforts from structural, energetics, and dynamics approaches offer unique insights and improve our fundamental understanding of the driving forces dictating membrane protein folding and stability.

Full text

Page 1

Archives of Biochemistry and Biophysics 453 (2006) 32–53
www.elsevier.com/locate/yabbi
0003-9861/$ - see front matter © 2006 Elsevier Inc. All rights reserved.
doi:10.1016/j.abb.2006.03.023
Review
Energetics of membrane protein folding and stability
Conceição A.S.A. Minetti, David P. Remeta¤
Rutgers—The State University of New Jersey, Department of Chemistry and Chemical Biology, Piscataway, NJ 08854, USA
Received 17 March 2006
Available online 7 April 2006
Abstract
The critical role of membrane proteins in a myriad of biological and physiological functions has spawned numerous investigations
over the past several decades with the long-term goal of identifying the molecular origins and energetic forces that stabilize these proteins
within the membrane. Parallel structural and thermodynamics studies on several systems have provided signiWcant insight regarding the
driving forces governing folding, assembly, insertion, and translocation of membrane proteins. The present review surveys families of
membrane-associated proteins including ?-helical and ?-barrel structures, viral surface receptors, and pore-forming toxins, citing repre-
sentative proteins within each of these classes for further scrutiny in terms of structure–function relationships and global conformational
stability. This overview presents seminal Wndings from pioneering studies on the energetics of membrane protein folding and stability to
modern techniques that are exploiting the use of molecular genetics and single molecule studies. An overall consensus regarding the
molecular origins of membrane protein stability is that a number of intrinsic properties resemble features of soluble proteins, yet there are
distinct energetic diVerences arising from speciWc intra- and intermolecular interactions within the membrane. The combined eVorts from
structural, energetics, and dynamics approaches oVer unique insights and improve our fundamental understanding of the driving forces
dictating membrane protein folding and stability.
© 2006 Elsevier Inc. All rights reserved.
Keywords: Membrane proteins; Thermal and thermodynamic stability; ?-Helical transmembrane proteins; ?-Barrel transmembrane proteins; Porins;
DiVerential scanning calorimetry (DSC); Virus glycoproteins; Pore-forming toxins (PFTs); Enthalpy; Heat capacity; Structure-energetics correlations
Recent advances in elucidating the structural, functional,
and energetic properties of membrane proteins reXect the
increased interest in characterizing this important class of bio-
logical macromolecules. Pioneering structural studies of the
photosynthetic reaction center from the purple bacteria Rho-
dopseudomonas viridis over 20 years ago furnished the inau-
gural high-resolution crystallographic structure of a
membrane protein [1,2]. The unique Wndings of a distinct dis-
tribution of side chains with an unusual predominantly apolar
exterior relative to soluble proteins served as the impetus for a
new era of research studies aimed at unraveling this vastly
unknown world of membrane proteins. Such eVorts have met
with considerable success over the past decade, as an increas-
ing number of high-resolution structures are now available
for proteins associated constitutively with membranes.
The familial origin of membrane proteins is suYciently
broad and encompasses a variety of sources including bacte-
rial, viral, eukaryotic plasma or organelle membranes, as well
as speciWc soluble proteins that undergo membrane insertion
during a deWned stage within the cell cycle. Parallel structural
and functional studies have revealed that membrane protein
families are integrally involved in a multiplicity of specialized
roles spanning the range from enzyme activity [3–8], cell sig-
naling [9,10], receptors [9,11–18], and channels [19–32]. Mem-
brane proteins therefore assume a pivotal role in terms of
regulating speciWc functions within the organelle, cell, or viral
particle. In addition to the extraordinary number of mem-
brane proteins that regulate important functions within the
mammalian cell, both bacterial and viral pathogens have
developed sophisticated strategies to manipulate the host cell
life cycle to their own advantage by utilizing highly specialized
*Corresponding author. Fax: +1 732 445 5312.
E-mail addresses: CMINETTI@RUTCHEM.RUTGERS.EDU (C.A.S.A.
Minetti), DPREMETA@RUTCHEM.RUTGERS.EDU (D.P.Remeta).

Page 2

C.A.S.A. Minetti, D.P. Remeta / Archives of Biochemistry and Biophysics 453 (2006) 32–53
33
molecular mechanisms involving membrane proteins (for a
review refer to [33]).
One of the primary objectives of the present review is to
highlight speciWc advances in structure–function-energetic
relationships amongst membrane proteins, and how these
insights have improved our fundamental understanding of
the driving forces governing the folding, assembly, inser-
tion, translocation, and global stability of these macromol-
ecules. Given the complicating factors associated with
studying this rather heterogeneous class of proteins, a num-
ber of intriguing yet unresolved questions set the frame-
work for future studies. In view of the myriad of functions,
structural features, and energetic properties of membrane
proteins characterized to date, this review is necessarily
restricted to a representative subset of samples that have
been selected speciWcally to highlight some common ubiq-
uitous properties and/or deWne unique features of mem-
brane protein systems relative to soluble proteins.
A retrospective view on membrane protein folding and
energetics
Previous reviews have furnished a comprehensive analy-
sis of the factors that contribute to the stability of a select
group of membrane protein systems [34], based on the
availability of structures characterized at atomic resolution
and how such properties could be related to soluble pro-
teins. The past decade has realized further advances in the
determination of structure, folding and stability of mem-
brane proteins deriving from mitochondria [19,31,35,36],
bacterial outer membranes [32,37], and viral membranes
[17,38–40]. All of these systems exhibit unique features in
terms of their structural and functional properties. SigniW-
cant insight has been gleaned from in vitro folding studies
[41–48], as well as systems that contain a multitude of fold-
ing assisted mechanisms in vivo [35,46,49,50]. In view of
recent advances in the acquisition of high-resolution mem-
brane protein structures and application of state-of-the-art
techniques to characterize in vitro and in vivo folding, sta-
bility, and macromolecular interactions [51,52], there is a
considerable dearth of thermodynamic data describing the
energetics of such complex processes. This review surveys
available thermal and chemical denaturation studies within
the framework of our long-term objective of developing a
structure-energetic database to facilitate eVorts aimed at
elucidating the molecular and thermodynamic driving
forces regulating membrane protein folding and stability.
Basic properties of membrane proteins: structure–function
relationships
A concise deWnition of those protein families that may be
designated as membrane proteins is critical to evaluate the
inherent properties of a particular class as a function of its
partitioning between aqueous and hydrophobic environ-
ments. The relevance of organizing such a vast number of
proteins based upon their functional roles in vivo is evident.
In this particular aspect, the notion of speciWc functional
roles continues to evolve and advance with successive studies.
As a speciWc case in point, the original view of lipoprotein as
a receptor that supplies cells with lipids for energy produc-
tion has been reWned based upon recent evidence which indi-
cates this membrane protein participates in highly specialized
signal transduction processes [18]. A more feasible and prac-
tical approach for organizing this diverse array of membrane
protein families invokes a hierarchal classiWcation scheme. In
this scheme, membrane proteins are categorized according to
their: (1) folding motifs (i.e., helix versus barrel); (2) physio-
logical role within the cell membrane; and (3) intrinsic prop-
erties as constitutive membrane proteins. Constitutive
proteins represent those encoded by the cell genome that are
responsible for a vital function in the cell [53]. Association
processes of proteins with membranes are not limited to con-
stitutive membrane proteins per se, but include those that
translocate or insert into a membrane at a speciWc juncture in
their biological functions. These nonconstitutive forms are
generally soluble proteins that do not fold and assemble via a
constitutive pathway, but develop the ability to insert and/or
translocate membranes under speciWc conditions and/or
when exposed to lipid bilayers.
Transmembrane protein basic folds: ?-helices and ?-barrels
Membrane protein folds may be categorized into two
basic secondary structural motifs, namely ?-helices and ?-
barrels. The primary diVerences in their architecture are
evident in the schematic representation of these two struc-
tural motifs embedded in a hypothetical lipid bilayer as
depicted in Fig. 1. Such folding motifs eVectively circum-
vent the diYculties associated with burying polar peptide
bonds within the non-polar hydrophobic core of lipid
bilayers. The resultant intra- and intermolecular hydrogen
bonds stabilize the membrane protein structure, while con-
straining the folding options for domains located on either
side of the membrane [54].
?-Helical membrane proteins
Resolution of the atomic structures of several ?-helical
membrane proteins has facilitated eVorts to establish spe-
ciWc structural–functional (and limited energetic) relation-
ships for this important class of proteins. In the case of
?-helices, the Wnal structure-determining step in the folding
of membrane proteins involves the coalescence of pre-
formed transmembrane helices to form the native tertiary
structure [55]. The unique helix bundle structures solved to
date have provided insight into the structural diversity of
this type of membrane proteins. Although long transmem-
brane helices are still considered a fundamental building
block, polypeptide chains can be organized into other com-
plex patterns (refer to [56]). Amongst the ?-helical trans-
membrane proteins, the channel-forming proteins and
pumps are probably one of the most well characterized.
Recent reviews furnish insight into the properties of this

Page 3

34
C.A.S.A. Minetti, D.P. Remeta / Archives of Biochemistry and Biophysics 453 (2006) 32–53
vast family of proteins [57], and how these highly special-
ized proteins selectively transport ions.
The thermal and thermodynamic stability of several repre-
sentative ?-helical transmembrane proteins are of particular
interest and will be discussed in subsequent sections of this
review. Fig. 2 presents the molecular structure of the light-
driven proton pump bacteriorhodopsin which converts solar
energy into proton electrochemical potential and is one of the
most widely studied membrane proteins. While speciWc struc-
tural–functional relationships have been established for bacte-
riorhodopsin, the origins of its high stability to a wide range
of harsh environmental conditions remains a matter of
inquiry. The continued interest in unraveling such molecular
origins and driving forces has been extended to in vitro studies
on folding, mutagenesis, and energetics of bacteriorhodopsin,
the latter employing spectroscopic and calorimetric tech-
niques (as reviewed in [34] and [58]). Cytochrome-c oxidase, a
multisubunit enzyme that functions as the terminal oxidase of
the respiratory chain, is a member of a large superfamily of
related enzymes originating from numerous sources (e.g.,
microorganisms to mitochondria). The thermal stability of
several ?-helical transmembrane proteins within this impor-
tant class of enzymes has been investigated using calorimetric
approaches (as reviewed in [34]). Contrary to the mammalian
enzyme or the yeast enzyme, both of which are composed of
13 subunits, the bacterial enzyme (e.g., Paracoccus denitriW-
cans) contains only three or four subunits, thereby providing a
unique opportunity to examine the magnitude of the forces
that stabilize this enzyme and derive accurate structural-ener-
getic correlations [59]. Photosystem II (PS II)1 is another ?-
helical membrane protein that catalyzes the water-splitting
reaction and associated electron transfer reactions in oxygenic
photosynthesis (as reviewed in [34]). The overall complexicity
of these multidomain proteins represents a formidable chal-
lenge to ongoing eVorts aimed at understanding the forces
that govern folding and stability within membranes and
enable these proteins to perform their highly specialized tasks
in the cell.
1Abbreviations used: PS II, Photosystem II; HA, hemagglutinin; SIV, sim-
ian immunodeWciency virus; ?-HL, ?-hemolysin; DSC, diVerential scanning
calorimetry; LPS, lipopolysaccharides; FRET, Xuorescence resonance ener-
gy transfer; CD, Circular dichroism; FTIR, fourier transform infrared;
DMPC, dimyristoyl phosphatidylcholine; PFT, pore-forming toxins.
Fig. 1. Basic transmembrane folds embedded within a hypothetical lipid bilayer. Ribbon diagram representations of an ?-helical bundle (rhodopsin fold)
and a ?-barrel (folding motif of a theoretical porin monomer isolated from its neighboring subunits within the trimer). The molecular structures derive
from PDB accession numbers 2OMF [181] and 2BRD [182] with the ribbon diagram prepared using Chimera [183].
Fig. 2. Molecular structure of bacteriorhodopsin. The molecular structure
derives from PDB accession number 2BRD [182] with the ribbon diagram
prepared using Chimera [183].

Page 4

C.A.S.A. Minetti, D.P. Remeta / Archives of Biochemistry and Biophysics 453 (2006) 32–53
35
?-barrel membrane proteins
The “constitutive” transmembrane ?-barrel fold motif
(as reviewed in [60] and [37]) is a characteristic structural
feature of both bacterial outer membrane proteins (e.g.,
porins) and the mitochondrial porin (i.e., VDAC). While
protecting gram-negative bacteria against a harsh environ-
ment, the outer membrane accommodates embedded pro-
teins that fulWll a number of crucial functions within the
bacterial cell, including solute and protein translocation
and signal transduction. In recent years, the atomic struc-
tures of six families of outer membrane proteins have been
determined. These include the OmpA membrane domain,
the OmpX protein, phospholipase A, general porins
(OmpF, PhoE), substrate-speciWc porins (LamB, ScrY), and
the TonB dependent iron siderophore transporters FhuA
and FepA (as reviewed in [61]). Fig. 3 presents the molecu-
lar structure of Escherichia coli OmpF in which the trimeric
arrangement of the ?-barrels are viewed sideways and per-
pendicular to the plane of the membrane. An overview of
energetic studies is presented for outer membrane proteins
employing porins as the prototypes. The availability of
such thermodynamic data contributes to our understand-
ing of how these ?-barrel proteins retain their structural
features and functional roles within the protective environ-
ment of the outer membrane.
Membrane proteins comprising large extracellular domains:
viral protein receptor ectodomains
A unique class of proteins associated with membranes is
the virus surface receptors of which inXuenza virus hemag-
glutinin (HA) represents one of the most extensively stud-
ied viral receptor proteins. HA is in fact primarily exposed
to solvent as an ectodomain, with a short C-terminal pep-
tide region inserted into the membrane [17,38]. In the sec-
tions that follow, the energetics of unfolding HA exposed
to varying conditions of pH is described in detail. A pri-
mary function of this protein during viral infection and rep-
lication is to participate in the low pH-mediated fusogenic
activities that occur between the viral and lysosomal mem-
branes within the mammalian cell. Another important class
of virus surface receptors of particular interest is the immu-
nodeWciency virus gp41. Infection by simian immunodeW-
ciency virus (SIV) and its human HIV counterpart involves
fusion between virus and target cell membranes, which is
mediated by the trimeric envelope glycoprotein gp41 (a
product of the precursor gp160). This protein has been the
focus of considerable interest as a potential target in inter-
ventions aimed at eradication of such viruses. Elucidation
of the structural and energetic properties of these viral pro-
teins are critical determinants for understanding the driving
forces that modulate folding, stability, and conformational
changes associated with membrane fusion and infection
preceding viral replication in the host cell. This review
examines the results of energetics studies on the folding and
stability of viral membrane proteins (e.g., HA and gp41),
which provides insights and assists in the overall elucida-
tion of structure–function-energetic relationships.
Nonconstitutive membrane-associated proteins: soluble
proteins that undergo membrane insertion and translocation
Soluble proteins that exert a biological function by
undergoing transient bilayer insertion may be classiWed as
“nonconstitutive” proteins (refer to [53]). A number of
these proteins adopt a ?-barrel fold in the latter stages of
their functional pathways, including some of the pore-
forming toxins (PFTs) such as ?-hemolysin (?-HL). This
bacterial cytotoxin assembles from a water-soluble mono-
mer to form a membrane-bound heptameric ?-barrel [62],
which subsequently perforates the cell surface resulting in
cell lysis. The E. coli ?-HL inserts irreversibly into lipid
bilayers via a mechanism that is presumably Ca+2-depen-
Fig. 3. Molecular structure of E. coli OmpF. The molecular structure derives from PDB accession number 2OMF [181] with the ribbon diagram prepared
using Chimera [183]. The trimeric arrangement of the ?-barrels in E. coli OmpF are viewed sideways (left) and perpendicular (right) to the plane of the
membrane.

Page 5

36
C.A.S.A. Minetti, D.P. Remeta / Archives of Biochemistry and Biophysics 453 (2006) 32–53
dent and triggered by increased surface hydrophobicity
[63]. Another class of toxic blood cell lytic proteins is the
cytoplasmic family of cytolysins (e.g., pneumolysin). There
is currently intense interest in this class of proteins consid-
ering their integral role in pathophysiological processes and
potential use as a target in therapeutic vaccine interven-
tions [64]. These cholesterol-dependent cytolysins are
believed to interact with cholesterol-rich membranes of the
host cell and oligomerize to form sizable (>150Å) pores (as
reviewed in [65]).
The aforementioned examples illustrate that association
processes of proteins with membranes are not strictly lim-
ited to membrane proteins per se, or proteins that are inher-
ently functional within membranes, but include those that
undergo insertion, translocation, and/or pore formation in
a critical step of their biological processes. Considering the
tendency for most of these proteins to form large aggre-
gates, particularly when conformational changes are trig-
gered in aqueous solution in the absence of the proper lipid
environment, assessment of conformational stability repre-
sents a signiWcant experimental challenge that often limits
such measurements to monitoring the onset of aggregation
as an indication of thermal stability. There are nevertheless
some examples of PFTs for which a more comprehensive
thermodynamic analysis has been conducted. Examples
include the pore-forming equinatoxin from sea anemones
[66–68] and the colicin family of plasmid encoded cytotox-
ins produced by E. coli and responsible for channel forming
toxic activity against E. coli susceptible strains [69–72].
In typical processes involving membrane insertion, spe-
ciWc classes of proteins may undergo conformational changes
prior to or during the insertion process. In fact, one intui-
tively expects a water-soluble protein to undergo structural
rearrangements to adopt the requisite surface properties by
exposing high aYnity groups to interact with membranes.
The resultant conformers are inherently more hydrophobic,
and consequently, more prone to aggregation. Since such
membrane-induced structural rearrangements may lead to
less compact states similar to those triggered by alterations in
solution conditions (e.g., pH and temperature), there is con-
siderable speculation regarding the exact nature of the con-
formational changes accompanying water to lipid insertion
and whether these constitute compact intermediate or molten
globule states. Inasmuch as conformational changes are evi-
dent in a number of systems [66,69,72,73], these Wndings sug-
gest that folding intermediates (including the molten globule
state) may indeed mediate membrane interaction processes
[74] and thereby warrants further investigation.
ClassiWcation of those protein families that comprise the
vast group of membrane proteins is absolutely essential to
accurately evaluate the inherent properties of such proteins
as a function of their partitioning between aqueous and
hydrophobic environments. Since there are multiple situa-
tions in which a protein associates with a membrane regard-
less of whether the protein is strictly deWned as a membrane
protein per se, Papahadjopoulos et al. [75] have attempted to
organize these proteins into speciWc classes according to
their eVects on gel/liquid crystalline phase transitions of
phospholipids as observed by diVerential scanning calorime-
try (DSC), vesicle permeability, and monolayer expansion.
The resultant classes are grouped into three categories with
respect to the nature of their interactions and overall impact
on the lipid transition temperature (Tc) and enthalpy (?H).
Although this classiWcation scheme cannot be applied
unequivocally to each membrane protein, it provides some
basis for understanding the origins of such unique protein–
membrane interactions.
Experimental strategies to study membrane protein folding
and stability
Model systems for in vitro studies
The isolation and characterization of cell membrane
proteins usually requires the solubilization of the mem-
brane and a method of separating the various membrane
proteins and glycoproteins (as reviewed in [76]). A range of
surfactants is available for membrane solubilization that
requires knowledge of the mode(s) by which these deter-
gents interact with membranes. An equally important con-
sideration is the potential requirement for additives that are
normally present in vivo. These include macromolecules
such as lipopolysaccharides (LPS) [77], which assume an
integral role in the formation of an assembly-competent
intermediate of the outer-membrane protein E. coli PhoE
[77]. Additional evidence suggests that LPS is implicated in
the folding and function of some outer membrane proteins
such as porin from Yersinia pseudotuberculosis [78].
A signiWcant obstacle that initially impeded progress in
characterizing membrane proteins involved the challenge of
expressing suYciently high quantities of these proteins to
conduct structural and thermodynamic studies. Such limita-
tions have been satisfactorily overcome by developing novel
methods of expression and/or puriWcation that provide yields
comparable to those of soluble proteins [79–83]. A particu-
larly eYcient method entails expression of membrane pro-
teins as inclusion bodies from which the native folded
functional state is attained via a process that involves unfold-
ing and refolding from denaturants to amphiphilic detergents
([80] and references therein). The ability to produce suYcient
quantities of membrane proteins eVectively allows one to sys-
tematically evaluate solution conditions that appropriately
mimic the native environment.
An important caveat that must be addressed when con-
sidering methods available for characterizing protein fold-
ing and stability is how one adapts these protocols to study
membrane or outer membrane proteins. SpeciWcally, the
majority of these protein systems are insoluble in simple
aqueous solutions, thereby requiring the presence of co-sol-
vents such as detergents, sub-denaturing concentrations of
chaotropes, etc. One must recognize that model systems for
studying membrane protein folding and stability require
the proper conditions in which to suspend these proteins.
Ideally, lipid bilayers are preferred to provide the requisite

Page 6

C.A.S.A. Minetti, D.P. Remeta / Archives of Biochemistry and Biophysics 453 (2006) 32–53
37
hydrophobic environment, although a number of deter-
gents have proven useful as co-solvents for ensuring the sol-
ubility of integral membrane proteins and retention of the
native folded state for conducting biophysical studies [84].
Recent Wndings have demonstrated that the nature and the
ratio of the detergent are often critical for both proper
refolding and structural stability of membrane proteins
[34,84].
Critical information on the folding and stability of mem-
brane proteins has been derived from mutagenesis-based
studies. Protein engineering of designed hydrophobic and
amphipathic polypeptides have assisted overall eVorts
aimed at elucidating the thermodynamic driving forces of
folding and orientation of such peptides within their bio-
logical membranes [85]. The speciWcity and thermodynamic
stability of transmembrane helix–helix interactions or the
assembly of ?-sheets using small peptides have furnished
quantitative data on the interactions that drive membrane
protein folding (as reviewed in [55]).
Integral membrane proteins: in vitro folding
Given the advances and improvement of technology
dedicated to the successful production of high yields of
puriWed functional membrane and outer membrane pro-
teins [45,80–83], in vitro folding studies in this relatively new
and challenging Weld have provided compelling results, as
illustrated by a signiWcant number of recent studies [41–
48,74,86–90]. It is evident that the choice of solubilizing
detergents and lipids can ultimately determine the success
of folding membrane proteins, as it appears that particular
lipid properties can be used to control the rate and
eYciency of folding. Mutagenesis and fragment studies on
membrane proteins have also proven pivotal in terms of
advancing novel concepts related to the folding problem.
For ?-helical proteins such as bacteriorhodopsin, analysis
of individual fragments indicate that whereas most of these
components may fold spontaneously [91], the preference
for a non-native conformation exhibited by some of the
peptides suggests the involvement of external constraints
that include cellular “chaperones” or translocases. SpeciWc
model systems have been studied such as diacylglycerol
kinase for which folding and misfolding data have been
obtained [92]. In vitro folding of transmembrane ?-helices
has also been performed recently [41,93].
Although a considerable body of evidence suggests that
in vivo assisted folding is required for a large number of
membrane proteins, many of these spontaneously refold
in vitro without the presence of “chaperones.” As a speciWc
example, OmpA can successfully fold in vitro into a range
of model membranes of diVerent phospholipid composi-
tions (i.e., into bilayers of lipids of diVerent headgroup,
structures, and hydrophobic chain lengths) [48]. Other
examples include ?-barrel membrane proteins such as
porins that successfully refold in vitro upon isolation and
extraction from outer membranes [94] or upon overexpres-
sion and refolding from inclusion bodies [95]. Following
puriWcation to homogeneity, the resultant porin prepara-
tions are fully active [82] and adopt a native-like conforma-
tion [80,81].
Folding and unfolding pathways
Although signiWcant eVorts have been dedicated towards
elucidating the driving forces of membrane protein folding
as well as associated mechanisms and pathways, the pro-
posal of a two-stage model by Popot (as reviewed in [96,97])
still represents a useful concept in terms of its basic tenets.
In this model, helical membrane protein folding and oligo-
merization can be conceptualized as involving two energeti-
cally distinct stages, namely the formation and subsequent
side-to-side association of independently stable transbilayer
helices. The two-stage model therefore represents a practi-
cal approach to simplifying discussions of stability within
the framework of membrane protein folding [96]. This is
particularly evident when surveying the results of in vitro
studies that oftentimes raise additional inquiries regarding
the feasibility of identifying speciWc forces that drive the
folding of integral membrane proteins [98]. In fact, a com-
prehensive analysis of the structural segments that stabilize
such intricate systems supports the recently articulated
three-stage model of membrane protein folding. These
studies and previous single-molecule force spectroscopy
experiments reveal that secondary structures can establish
suYcient molecular interactions to act independently as
stable units [97]. SigniWcant insights have derived from
studies that present evidence for a third stage in protein
folding [99], in which two diVerent structural elements
together establish molecular interactions that stabilize these
grouped structures.
Amongst ?-barrel proteins studied to date, three mem-
brane-bound folding intermediates have been identiWed in
time-resolved Xuorescence quenching experiments of
OmpA associated with dioleoylphosphatidylcholine bilay-
ers [48]. These folding intermediates are structurally distin-
guished by the relative positions of the Wve Trp residues of
OmpA in projection to the membrane normal. FomA rep-
resents another major outer membrane protein of Fusobac-
terium nucleatum that folds and inserts into lipid bilayers
via parallel folding pathways [100]. Unfolding pathways
have also been proposed for bacteriorhodopsins [101],
some of which are temperature-dependent [102]. The
unfolding and in vitro refolding of a tetrameric ?-helical
membrane protein, the prokaryotic potassium channel
KcsA, has also been assessed [44]. The controlled unfolding
and refolding of a single sodium-proton antiporter has
been monitored using atomic force microscopy [103].
Energetic landscapes in membrane protein folding
The protein-folding problem has elicited the interest of
numerous investigators for the past Wve decades as succes-
sive studies have generated novel concepts that seek to
identify and characterize the driving forces governing

Page 7

38
C.A.S.A. Minetti, D.P. Remeta / Archives of Biochemistry and Biophysics 453 (2006) 32–53
protein folding. Such investigations have expanded to a
rather challenging class of proteins that are associated with
a lipid bilayer in vivo, the latter represented by either a
plasma membrane, viral membrane, organelle membrane,
or outer membrane as in the case of gram-negative bacteria.
These membranes share a common property of comprising
an organized hydrophobic environment in which speciWc
proteins may insert and be embedded and associated with
the lipid components. Similar to the soluble protein-folding
problem, the folding of membrane proteins presumably
proceeds via a funnel-shaped energy landscape to an energy
minimum. Consistent with such a folding funnel view is the
observation of multiple pathways for membrane protein
folding (for a recent review refer to [56]). One major diVer-
ence between soluble protein folding and membrane pro-
tein folding is that the starting point is much more
constrained, since the overall secondary structure and
topology may be determined by the insertion process,
thereby suggesting that the unfolded protein resides further
within the folding funnel and closer to the folded state
when compared to soluble proteins. Consequently, the fold-
ing-energy landscape may be deWned by a complex inter-
play between various forces, including polypeptide
partitioning in the bilayer, interactions between lipid and
protein, and interactions within the protein itself. An inter-
esting study has recently probed the energy landscape of
the membrane protein bacteriorhodopsin [104].
The protein folding problem: employing soluble proteins as
references to elucidate the forces governing membrane
protein stability
The literature contains an extensive number of studies
on the energetic and thermodynamic properties of soluble
proteins in terms of folding, stability, and assembly (for
speciWc reviews refer to [105–108]). The notion of protein
folding is increasingly discussed within the context of popu-
lating stable intermediates, the latter surmised from ther-
mal denaturation studies. Unlike chaotrope-induced
denaturation, a thermally unfolded protein often presents
characteristics of a compact denatured state that retains
speciWc secondary structural elements. The problem in
identifying the thermal denatured state of a number of pro-
teins is how to characterize such a state in terms of its
apparent molar heat capacity. Using amino acid mixtures
resembling the completely unfolded state, one may estimate
this quantity by direct summation of the partial molar heat
capacity contributions of the amino acid residues constitut-
ing the chain [109,110]. Comparison of the latter with the
empirically measured molar heat capacity reveals that the
majority of proteins undergoing thermal denaturation are
only partially unfolded in contrast with the chaotrope-
induced random coil conWguration. The heat capacity eVect
monitored in thermal unfolding experiments is propor-
tional to the presence and exposure of non-polar residues,
and is therefore proportional to the hydrophobicity and
compressibility of a protein (as reviewed in [106] and [107]).
Thermodynamic characterization of protein folding/
unfolding yields accurate estimates of the enthalpy/entropy
and resultant free energy for the process, although it does
not permit speciWc resolution of the internal bonds dis-
rupted versus those that are newly formed. In summary, the
heat capacity eVect is proportional to the presence and
exposure of non-polar residues and therefore a measure of
the overall hydrophobicity of a protein (as reviewed in
[106]).
The magniWcent body of studies describing the forces
driving protein folding and stability had its glorious origins
over three decades ago [111]. Aside from the pioneering
studies of Sturtevant and colleagues on bacteriorhodopsin
[112], only recently has this vast reservoir of knowledge
been exploited to understand protein folding and stability
within membranes. The realization that such systems are
often characterized by irreversible unfolding processes and
therefore require more rigorous experimental approaches
has prompted the development of methods that address
these deWciencies with the proper due diligence [113]. A
number of recent studies have conWrmed the challenging
nature of such complex systems and incorporated novel
experimental strategies to overcome these limitations
[59,69,114–117]. This review presents a retrospective survey
of thermodynamic data gleaned from studies on represen-
tative systems within each class of membrane proteins
including discussion of novel methodologies and strategies
employed in such investigations (see for instance [116]).
This overview and analysis of experimental data facilitates
derivation of structural–energetic correlations that deWne
this unique class of proteins and provides insight regarding
the molecular origins of folding and stability in membrane
proteins. Recent developments are discussed including use
of a new generation of model systems [118] incorporating
small peptides that adopt secondary structures typical of a
transmembrane protein motif to address the membrane
protein folding problem via a “minimalist approach” [56].
In parallel to improving experimental approaches for char-
acterizing membrane proteins, structure-based models and
methods that enhance overall predictive capabilities regard-
ing speciWc structure–function-energetic properties are
Wnally beginning to emerge [119].
Extending studies to membrane proteins
Membrane proteins have been designed to exist in an
environment from which water is speciWcally excluded [34].
Since most forces driving protein folding are coupled with
solvent properties (e.g., the “hydrophobic eVect,” hydrogen
bonding and electrostatic interactions), it is not surprising
that the driving forces stabilizing membrane protein struc-
ture diVer signiWcantly from those of soluble proteins (as
reviewed in [120]). Consequently, the impact of unfolding a
polypeptide within a lipid bilayer on the resultant enthalpy
(?H) and heat capacity changes (?Cp) is not suYciently
understood. Recent studies using model peptides that
exhibit a remarkable ability to undergo ?-sheet formation

Page 8

C.A.S.A. Minetti, D.P. Remeta / Archives of Biochemistry and Biophysics 453 (2006) 32–53
39
within membranes [118] have provided some limited quan-
titative estimates of these thermodynamic parameters. The
increasing number of systems characterized to date and
their diversity in terms of solvent exposed surface areas (i.e.,
?ASA), protein–protein interface contacts, and relative
membrane embedded areas, reveals that a number of vari-
ables with respect to protein–solvent, protein–protein, and
protein–lipid interactions must be considered when evalu-
ating the molecular origins of energetic and thermody-
namic folding/unfolding
contribution of each of these variables to global protein
stability provides a complete characterization of a protein
membrane system in terms of its energetic and structural
parameters. The resultant data are also inXuenced by a
number of experimental conditions including the choice of
surfactant (i.e., lipid bilayers and/or detergents), the pres-
ence of ligands (as deemed appropriate), and solution con-
ditions (e.g., ionic strength and pH).
parameters. The relative
Methods to assess the conformational stability of membrane
proteins
A number of techniques have been employed in the
study of protein folding and stability that may be applied to
membrane proteins. Important information regarding pro-
tein conformation and interactions has been gleaned from
Xuorescence approaches [9,48,121,122] including intrinsic
Trp Xuorescence [67,68,80,117] and Xuorescence resonance
energy transfer (FRET) techniques [55,121,123]. The oligo-
meric state of membrane proteins has been elucidated from
gel electrophoresis [55,80,81] and analytical ultracentrifuga-
tion [55,117]. Newer analytical techniques including single-
molecule force spectroscopy [51,52] have been employed to
characterize speciWc molecular interactions in diVerent bac-
teriorhodopsin assemblies (as reviewed in [55]). The ener-
getics of membrane protein folding may be assessed using a
combination of spectroscopic and calorimetric techniques.
Circular dichroism (CD), fourier transform infrared
(FTIR), and Xuorescence spectroscopy may be employed to
monitor the structural integrity and conformational stabil-
ity of membrane proteins. The far UV CD and/or FTIR are
diagnostic probes of secondary structure, whereas the near
UV CD and/or intrinsic Xuorescence are sensitive to pertur-
bations of tertiary structure [7,80,81,117,124]. Use of acid,
denaturants, and/or temperature permits characterization
of the unfolding process, thereby providing important
information on the conformational stability of membrane
proteins. Due to inherent complications associated with the
membrane environment, it is often advisable to study the
stability of such proteins in the presence of amphiphilic
detergents, since the latter approximate the native bilayer
environment while eliminating artifacts arising from lipid
transitions. SpeciWc methods have been designed to assess
membrane protein structural stability by thermally unfold-
ing membrane proteins embedded in mixed micelles (i.e.,
non-ionic/ionic detergent mixtures) [125,126]. This quanti-
tative assessment of membrane protein stability employs a
method originally proposed by Lau and Bowie [125] and
subsequently modiWed by Sehgal et al. [126] to characterize
the impact of the micelle environment on the overall
thermal stability.
A number of precautions must be considered when con-
ducting thermal unfolding studies of membrane proteins.
Several fundamental concerns are directly related to solu-
tion characteristics, as one must ensure the overall compat-
ibility between the solvent (i.e., buVer) and co-solvent (i.e.,
detergent) as well as co-solvent tolerance to the experimen-
tal conditions employed. Reversibility is often observed
over a limited temperature range above which the protein–
detergent interactions essential for maintaining the protein
in its solubilized form are disrupted. Folding and oligomer-
ization are often coupled events [55] that can be monitored
via analytical techniques that permit discrimination
between monomers and multimers. Such methods are use-
ful for characterizing coupled processes involving either
folding/oligomerization or unfolding/dissociation. Native
polyacrylamide gel electrophoresis has been used success-
fully to study a number of membrane proteins [55].
SDS–PAGE analysis may also furnish qualitative informa-
tion regarding oligomeric stability in that most membrane
proteins do not dissociate in the presence of SDS unless
heated to high temperatures. Numerous membrane pro-
teins including the porins are SDS-resistant oligomeric
structures [80,127] with a few notable exceptions such as
Neisserial meningitidis PorB class 3 protein [81].
Ideally, one should assess the oligomeric state of a protein
via quantitative techniques such as analytical ultracentrifu-
gation. This method has proven useful for characterizing the
oligomeric states of a number of detergent solubilized mem-
brane proteins including inXuenza virus hemagglutinin (HA)
[117], thereby conWrming that this protein adopts an
expected trimeric conformation in the presence of octylglu-
coside. Such viral membrane proteins are primarily exposed
to the solvent milieu with only a small portion of the C-ter-
minus embedded into the membrane. Although detergent is
imperative to ensure the mono-trimeric form of HA, the
ectodomain remains soluble in the absence of detergent,
albeit adopting a “rosette” structure that results from associ-
ation of 5–6 trimers [117]. The multitrimeric solution struc-
ture does not appear to interfere in the overall analysis of the
unfolding process, as it yields energetic parameters that are
nearly identical to those obtained for the mono-trimers. One
of the unique features of this class of proteins is the fact that
its membrane-associated trimeric form may either be isolated
from viral membranes and thereby retain the transmembrane
C-terminal region, or it may be speciWcally cleaved by prote-
ases (e.g., bromelain) and subsequently release the ectodo-
main to solution. As noted previously, the ectodomain
adopts a soluble macromolecular assembly of a “rosette”
structure as deduced from cryo-electron microscopy and
analytical ultracentrifugation studies in aqueous solution
[117,128].
The stability of a protein is determined by group–group
interactions as well as group interactions with solvent,

Page 9

40
C.A.S.A. Minetti, D.P. Remeta / Archives of Biochemistry and Biophysics 453 (2006) 32–53
thereby representing a balance between these two types of
association processes [107]. DiVerential scanning calorimetry
allows evaluation of model-independent unfolding enthalpies
(?Hcal) and associated heat capacity changes (?Cp) accom-
panying thermal denaturation. Whereas the former provides
information on the extent and/or magnitude of intramolecu-
lar interactions, the latter reXects the temperature-dependent
exposure of polar and apolar residues to solvent. A rigorous
description of the unfolding process requires knowledge of
the heat capacity change accompanying the unfolding transi-
tion [116]. Multiple approaches are available to determine
?Cp including direct analysis of the resultant DSC endo-
therms, as well as KirchhoV plots in which the temperature-
dependence of the unfolding enthalpy is evaluated (i.e.,
?CpD??Hm/?Tm). An important caveat to ensure accurate
determination of the requisite thermodynamic parameters
from analysis of DSC endotherms is unequivocal evidence
that the unfolding process is indeed reversible (i.e., scan-rate
independent). Furthermore, derivation of ?Cp from Kirch-
hoV plots is only feasible when such studies are performed
under various solution conditions incorporating an agent
(e.g., pH, chaotropes, etc.) that thermally destabilizes the
native conformation. The relative paucity of thermodynamic
data in the literature reXects the overall challenges and diY-
culties associated with obtaining accurate estimates of the
energetic parameters characterizing the thermal unfolding of
membrane proteins.
A survey of representative examples for each class of
membrane associated proteins
A review published over one decade ago highlighted
some of the important distinctions between typical water-
soluble and membrane proteins based on a comprehensive
analysis of temperature-dependent spectroscopic and diVer-
ential scanning calorimetric studies ([34] and references
therein). The survey focused on those membrane protein
systems that had been characterized from both a structural
and thermodynamic perspective. The resultant compilation
of unfolding data revealed a fundamental diVerence
between soluble and membrane proteins, namely the intra-
bilayer secondary structure elements of membrane proteins
are highly stable and resist thermal unfolding, whereas their
extramembraneous regions behave much like soluble pro-
teins. The authors postulated that the reduced denaturation
enthalpies measured for membrane relative to soluble pro-
teins is indicative of incomplete unfolding [34]. In the inter-
vening decade, signiWcant advances in protein expression
methodologies and improvements in analytical instrumen-
tation have spawned numerous studies on a diverse array of
membrane proteins. The present review seeks to survey rep-
resentative examples of membrane-associated protein sys-
tems for which there are available structural and
thermodynamic data. This retrospective is not intended to
be all inclusive considering the vast number of studies pub-
lished to date, but oVer a glimpse into the broad Weld of
membrane protein structure and conformational stability.
The membrane proteins selected for further scrutiny are
intended to provide the reader with an appreciation of the
overall complexity of such systems and how parallel struc-
tural/functional/energetic studies may furnish additional
insight into the nature of the driving forces stabilizing
membrane-associated proteins.
Structural stability of transmembrane ?-helices
A prototypical ?-helical membrane protein is bacteriorho-
dopsin (Mr 26kDa), which is active as a monomer but assem-
bles as a trimeric structure. The conformational stability of
bacteriorhodopsin has been assessed in both membranes and
detergent–lipid micelles [112,129]. The resultant DSC endo-
therm is characterized by a relatively high transition temper-
ature (Tm»100°C) [112] and low unfolding enthalpy
(?H»110kcalmol¡1 or 3.7calg¡1) compared to the average
value of 12calg¡1 measured for soluble proteins [34,130]. The
cooperative ratio of ?HVH/?Hcal»3 [112] suggests that the
protein dissociates/denatures cooperatively in the membrane
as a trimer. Subsequent studies questioned the validity of this
analysis based on the irreversibility of the unfolding process
as veriWed by the scan-rate dependence of the DSC thermo-
grams [113,131]. Although the estimates of ?Cp are compro-
mised by the irreversibility of thermal unfolding, the linear
increase in ?H above 80°C yields a tentative value of
?Cp»1.2kcalK¡1 mol¡1 or 0.046calK¡1g¡1. SigniWcantly,
the magnitude of ?Cp is threefold lower than that estimated
for soluble proteins such as myoglobin (i.e., ?Cp»0.15
calK¡1g¡1 [34]). This Wnding suggests that with the excep-
tion of the extracellular loop regions in bacteriorhodopsin,
minimal hydrophobic surface area is exposed to solvent upon
thermal unfolding. Furthermore, the reduced enthalpy [112]
conWrmed by subsequent studies on cleaved bacteriorhodop-
sin systems [132,133] signiWes that bacteriorhodopsin under-
goes incomplete unfolding upon thermal denaturation.
Cytochrome-c oxidase is an ?-helical protein that retains
as much as 45% of its native intramembraneous structure
upon thermal denaturation [59]. Calorimetric studies to
assess the structural stability of cytochrome-c oxidase (as
reviewed in [34]) have been performed on a number of sys-
tems and under a variety of experimental conditions. The
presence of endogenous lipids shifts the transition tempera-
ture by 5°C when compared to delipidated protein. More-
over, the magnitude of its unfolding enthalpy depends
signiWcantly on the in vitro reconstitution protocol, which
suggests a linkage between protein denaturation and gel/
liquid phase transition of the surrounding lipids. Considering
the complexity of studying these membrane protein systems
from the standpoint of irreversible unfolding, the presence of
multiple subunit domains, and the implications of protein–
lipid linking events (as reviewed in [34]), a wealth of informa-
tion has been derived with respect to the structural stability
of these macromolecules. A speciWc case in point is beef
oxidase reconstituted with dimyristoyl phosphatidylcholine
(DMPC) in which the resultant DSC endotherm exhibits a
biphasic proWle with transitions centered at temperatures of

Page 10

C.A.S.A. Minetti, D.P. Remeta / Archives of Biochemistry and Biophysics 453 (2006) 32–53
41
52 and 64°C. The lower temperature endotherm corresponds
to the unfolding of subunit III and is characterized by an
enthalpy of 4.8calg¡1, whereas subunits I and II unfold via a
complex multiphasic process at the higher temperature.
Upon reconstitution in detergent, the biphasic proWle simpli-
Wes to a single broad endotherm with Tm»56°C and
?H»2.7calg¡1, the latter signiWcantly lower than the
?H»6.9calg¡1 expected for soluble proteins at this temper-
ature [34]. A lipid-reconstituted preparation of yeast cyto-
chrome-c oxidase (Mr 175kDa) denatures with an enthalpy
of 2.4calg¡1 (Freire and colleagues as reviewed in [34]).
Calorimetric and spectroscopic approaches have been
employed to characterize the structural stability of cyto-
chrome-c oxidase from P. denitriWcans [59]. The resultant
proWles reveal that subunits I and II form a highly coopera-
tive complex that denatures as a single cooperative unit at
67°C, while subunit III denatures at »47°C. The magni-
tude of the unfolding parameters and the infrared spectro-
scopic experiments corroborate the notion that the enzyme
does not unfold completely upon thermal denaturation,
experiencing a reduction of ?-helical content from 45 to 30
percent. Presumably, most of the extramembraneous struc-
tural elements are denatured while the intramembranous
secondary structure is maintained. It is interesting to note
that the unfolding enthalpies (normalized to weight) of 2.4–
2.9calg¡1 calculated for the cytochrome-c oxidase family of
proteins are in the overall range of that determined for bac-
teriorhodopsin, yet signiWcantly less than the unfolding
enthalpies measured for soluble proteins [108,130]. Another
observation that emerges from these studies is the nature of
the partially unfolded structure, which resembles the com-
pact denatured state observed for some soluble proteins
under speciWc solution conditions [134].
Photosystem II (PS II) is a multidomain ?-helical mem-
brane protein (Mr 600kDa) that unfolds with an enthalpy
of 3300kcalmol¡1 or 5–6calg¡1 (i.e., ?HD0.55kcalmol-
residue¡1) [135]. The resultant DSC proWles are extremely
sensitive to the concentrations of detergent utilized in the
PS II preparations, as noted by substantial peak broaden-
ing in the presence of higher Triton X-100 concentrations
(i.e., 0.1% as opposed to 0.01%). This Wnding has been cor-
roborated by a recent study that suggests both the nature
and the ratio of surfactants assume a critical role in the
folding and stability of membrane proteins [48]. An overall
observation inferred from the aforementioned calorimetric
and temperature-dependent spectroscopic studies is that
the resultant enthalpies are characteristic of incomplete
unfolding for most of the ?-helical membrane proteins
studied to date, and the lower than expected heat capacities
reXect minimal apolar surface exposure to solvent (refer to
Fig. 10 in [34]). Recent studies aimed at elucidating the
driving forces stabilizing transmembrane ?-helical proteins
suggest that the majority of such systems are compromised
by the poor repeatability (and in most cases) irreversibility
of unfolding transitions [131,133,136,137], which is readily
observed from the scan-rate dependence of the transition
temperature in the respective endotherms [131,136]. Never-
theless, characterization of the thermal stability of mem-
brane proteins may still provide some useful information in
terms of assessing the impact of disrupting speciWc molecu-
lar interactions such as the temperature-induced perturba-
tion of secondary and/or tertiary structure.
Structural stability of transmembrane ?-barrels
Porins
One of the most extensively studied classes of ?-barrel
proteins are the porins that are present in the mitochondria
outer membrane as monomeric structures in mammalian
cells (e.g., VDAC), and as integral trimeric protein structures
within the outer membrane of gram-negative bacteria (for
recent reviews refer to [32,37,138]). The thermal stability of a
number of ?-barrels has been assessed by spectroscopic and
calorimetric techniques. Similar to thermal stability studies of
?-helical membrane proteins, ?-barrels often undergo exten-
sive aggregation thereby impeding rigorous thermodynamic
analysis [138]. Nevertheless, the porin family is extraordi-
narily thermally resistant and exhibits transition tempera-
tures well above 85°C that are characterized by partial
disruption of secondary structure [138], undergoing revers-
ible unfolding under certain conditions [139]. Denaturation
of neisserial porins [80,81] corroborates the unusual thermal
stability of this class of ?-barrel transmembrane proteins [34].
Fig. 4 presents temperature-dependent circular dichroism
spectra of PorB class 2 protein that are characterized by a
reversible unfolding transition with a midpoint of 88°C.
Inspection of the respective CD melting proWles monitored at
290nm (Fig. 4A) and 217nm (Fig. 4B) reveals essentially
complete disruption of tertiary structure that is accompanied
by only partial disruption of secondary structure [80]. Res-
cans monitoring recovery of the molar residue ellipticity in
the far and near UV regions upon cooling to 20°C demon-
strates that the trimeric assembly is regained with greater
than 95% reversibility [80].
The conformational stability of native and recombinant
sources of porins from N. meningitidis have been studied
extensively, revealing striking results that reXect their
intrinsic property as either SDS-resistant (e.g., class 2) [80]
or SDS-susceptible (e.g., class 3) [81] trimeric structures.
The latter property as veriWed by SDS–PAGE analysis has
been exploited for the identiWcation and characterization of
the various meningococcal serotypes [127,140]. Given the
SDS-susceptibility of N. meningitidis PorB class 3 relative
to its serotype Por B class 2, the conformational as well as
trimeric stability of this porin has been evaluated in mixed
micelles by increasing the mole fraction of SDS and moni-
toring the retention of secondary structure via circular
dichroism spectroscopy and trimeric assembly by native
PAGE under isothermal conditions (i.e., 25°C) [81]. Fig. 5
presents a family of far UV CD proWles for PorB class 3 pro-
tein monitoring the SDS-induced transition from ?-sheet to
a less ordered ?-helical-like conformation with an SDS0.5
»0.35%. SDS-induced denaturation is accompanied by

Page 11

42
C.A.S.A. Minetti, D.P. Remeta / Archives of Biochemistry and Biophysics 453 (2006) 32–53
increased protease susceptibility and trimer dissociation,
the latter monitored by gel analysis of the oligomeric struc-
ture as depicted in the inset of Fig. 5 [81]. These studies
reveal that intrinsic properties distinguish individual pro-
teins within the family of PorB neisserial porins. SpeciW-
cally, SDS-induced partial unfolding and dissociation of
class 3 protein are linked processes that are abrogated in its
SDS-resistant class 2 serotype which neither unfolds nor
dissociates in the presence of high anionic detergent con-
centrations [80]. The topology model of PorB class 3 pro-
tein presented in Fig. 6 may reveal the structural origin of
such distinct diVerences in conformational stability within
this class of neisserial porins. In this comparative represen-
tation of the PorB class 3 relative to class 2 proteins, residue
deletions are noted by the magenta dots that occur princi-
pally in the putative loop regions (particularly loops III–V).
SigniWcantly, these deletions represent the primary struc-
tural distinction between the two proteins and these regions
are likely to be part of the trimeric interface. In view of the
topology model, the SDS-susceptibility of class 3 relative to
class 2 is consistent with the proposal that the high thermal
and thermodynamic stability of such ?-barrels is further
enhanced by the extramembraneous loop regions. Addi-
tional support of this hypothesis arises from studies that
demonstrate the latching loop 2 contributes to the stability
of the E. coli OmpF trimer [139]. This calorimetric investi-
gation reveals that E. coli OmpF porin exhibits unfolding
enthalpies on the order of »430kcalmol¡1 (i.e.,
?H»0.41kcalmol-residue¡1), which are reduced signiW-
cantly upon loop 2 deletion. In fact, mutagenesis of speciWc
residues located in the extracellular loop 2 region (e.g.,
E71Q) destabilizes the porin with a resultant unfolding
enthalpy of »201kcalmol¡1 (i.e., ?H»0.19kcalmol-resi-
due¡1). These ?-barrels partially unfold with ?H values
ranging from 2.0 to 4.3calg¡1, which is signiWcantly lower
than the unfolding enthalpies typical for soluble proteins
[108].
Model peptides to study the driving forces that stabilize
?-sheet transmembrane proteins
Due to the complexity in elucidating the forces that sta-
bilize transmembrane proteins, the use of model peptides
designed to study the thermodynamics of protein folding in
membranes becomes increasingly imperative. Wimley and
White [118] have studied the hexapeptide acetyl-Trp-Leu5
(AcWL5), which has the ability to assemble reversibly and
spontaneously into ?-sheets on lipid membranes as a conse-
quence of monomer partitioning followed by cooperative
assembly. This calorimetric investigation represents one of
the few examples of reversible thermal unfolding of pep-
tides in membranes and may thereby facilitate eVorts to
deWne the contribution of hydrogen bonding to the extreme
thermal stability of membrane proteins. Whereas ITC anal-
ysis reveals that the enthalpy for partitioning of monomeric
unstructured AcWL5 from water into membranes approxi-
mates zero (within experimental error) over the tempera-
ture range of 5–75°C, the corresponding DSC experiments
reveal that the ?-sheet aggregates undergo reversible
unfolding, as characterized by an asymmetric concentra-
tion-dependent endotherm with transition temperatures of
60–80°C and unfolding/disaggregation enthalpy of
»8kcalmol¡1 (i.e., ?H»1.3kcalmol-residue¡1). This value
is consistent with a numerical nucleation model of oligo-
meric ?-sheets proposed earlier to describe ?-sheet forma-
tion in membranes [53]. The ?-sheet self-association
promoted by membranes has been discussed in terms of
possible biological implications regarding membrane-
induced ?-amyloid peptide self-association [53].
Structural stability of glycoprotein viral receptors
InXuenza virus hemagglutinin
InXuenza virus hemagglutinin (HA) is a receptor-bind-
ing and membrane fusion glycoprotein as well as a target
for infectivity-neutralizing antibodies [17]. HA has been
characterized by X-ray crystallography to exist in three
Fig. 4. Thermal stability of PorB class 2 protein. Temperature-dependent
circular dichroism proWles monitoring the thermal unfolding of class 2
protein tertiary structure at 290nm (A) and secondary structure at 217nm
(B). The insets of (A) and (B) present the respective isothermal near and
far UV CD spectra recorded at 20.0°C (blue), and 95.0°C (magenta). Res-
cans of renatured protein following cooling reveal greater than 95%
recovery of both tertiary and secondary structure (CD spectra not shown)
as reported in [80].
20 30 4050 60708090100
0
2
4
6
8
10
(θ)290 (10·deg·cm2·dmol-1)
Temperature (ºC)
2030 4050 6070 8090 100
-5
-4
-3
-2
-1
0
(θ)217 (103·deg·cm2·dmol-1)
Temperature (ºC)
240 260 280
Wavelength (nm)
300320340
-3
0
3
6
9
12

(θ) (10·deg·cm2·dmol-1)
180190200210220 230 240250
-5
0
5
10
15
20

Wavelength (nm)

(θ) (103·deg·cm2·dmol-1)
A
B

Page 12

C.A.S.A. Minetti, D.P. Remeta / Archives of Biochemistry and Biophysics 453 (2006) 32–53
43
distinct states, namely as the single-chain precursor HA0,
the native neutral-pH conformation, and a low pH-induced
conformation, the latter lacking the HA1-head domains.
These structures have provided a framework for under-
standing HA receptor recognition, membrane fusion dur-
ing viral infection, and the mechanisms associated with
Fig. 5. Conformational stability of PorB class 3 protein. Family of far UV circular dichroism proWles depicting the susceptibility of class 3 protein second-
ary structure to increasing concentrations of the anionic detergent SDS (0–1.0%). The inset presents a gel analysis of oligomeric stability by monitoring the
SDS-induced dissociation of class 3 trimeric structure (in accordance with [81]).
180190200210220230240250
-8
-4
0
4
8
12
16
Wavelength (nm)
(θ) (103·deg·cm2·dmol-1)
Fig. 6. Topology model of PorB class 3 relative to Por B class 2. The topology of PorB class 3 protein is represented as a 16-strand antiparallel ?-barrel
transmembrane domain with eight extracellular loops designated I–VIII. Conserved amino acid residues appear in blue, replacements are in red, and dele-
tions in class 3 relative to class 2 are depicted by magenta dots (adapted from [81]).
••
•
G
•••
S
K
D
Y
L
D
W
S
V
N
G
N
D
A
D
T
G
I
•
D
A
E
G
A
N
D
N
E
Y
VQ
D
D
I
•
K
G
N
K
F
V
Q
E
T
A
G
E
ALVLVQQKE
A
RDQK
VG
AW
VG
A
Q
IY
SYL
K
S
Q
D
A
SGVKA
Q
HAYQILFELK
VA
G
FTVRGHV
S
VIVGI
SAG
L
ALG
A
S
V
S
G
Q
S
E
VEH
A
VVYGGRNL
W
I
KG
LYAYSSQF
S
YLKIY
DYRAGVNLDRGAGL
FSFYYFYGSGGKFS
S
V
G
LD
F
KAP
V
FLK
V
D
A
I
S
G
W
G
A
S
R
N
G
T
L
N
•
•••
R
H
RN
T
P
V
G
L
R
H
A
KS
F
T
D
E
T
Q
S
H
•••
S
N
S
A
N
A
K
L
R
H
H
Q
V
Q
E
G
L
••
•
N
•
•
•
••
•
•
•
I
Y
K
E
V
E
T
S
R
S
••
V
T
QV
A
T
T
I
G
Q
N
H
E
V
N
P
I
K
QN
G
GG
L
E
A
D
NG
G
N
K
R
E
F
VIII
VIIVI
V
IV
IIIIII
T
D
S
N
P
E
A
•
T
••
GF
G
KL
R
D

Page 13

44
C.A.S.A. Minetti, D.P. Remeta / Archives of Biochemistry and Biophysics 453 (2006) 32–53
re-emerging epidemics [17]. The conformational stability of
HA has been assessed by a combination of spectroscopic
and calorimetric techniques as a function of both pH and
temperature [117]. It is important to note that the majority
of the polypeptide chain is located in the extramembrane
region as an ectodomain, which in detergent solutions
retains a trimeric assembly that undergoes thermal unfold-
ing with a Tm centered at »66°C and an unfolding ?H of
»800kcalmol¡1 per trimer [117]. Although intrinsic trypto-
phan Xuorescence reveals exposure of subunit interfaces
upon thermal unfolding, sedimentation equilibrium studies
indicate that the trimeric assembly is not disrupted. In spite
of the fact that the ectodomain HA1 “heads” dissociate
upon thermal unfolding, the trimer remains assembled via
intermolecular contacts amongst the HA2 domains.
It has been proposed that the “low pH structure” exists
at a lower energy than the neutral pH structure and that the
refolding reaction is essentially irreversible. The neutral pH
structure is thus postulated to be a kinetically trapped
metastable conformation that can release stored energy
only upon reducing the activation energy associated with
the structural transition. SpeciWc environmental conditions
that lower the activation energy include exposing the mole-
cule to low pH, high temperature, or denaturants [141]. The
resultant membrane fusion mechanism has been termed
“spring-loaded” since release of the “clamp” that retains
the neutral pH structure in the metastable state results in
complete inversion of the helices and reorientation of the
N- and C-terminal ends of HA2 ([142] and reviewed in
[40]). Although numerous studies proposing alternative
models have appeared in attempts to explain such a com-
plex process, controversial opinions have arisen from such
studies [17,136]. Evaluating the calorimetric and spectro-
scopic studies of full length puriWed HA [117], it is not evi-
dent how these data may assist in current interpretations
regarding the acid-induced fusogenic activity of HA. Nev-
ertheless, a number of subsequent studies and reviews have
interpreted this information within the context of driving
forces governing the conformational changes that lead to
fusogenic activity either by supporting the concept of meta-
stability [143,144] or refuting such a hypothesis
[136,145,146]. Clearly, the energetic data provides impor-
tant insights regarding the overall conformational stability
of the ensemble of multiple states populating the protein
unfolding pathway(s), particularly the interesting Wnding
that the resultant thermally denatured state is comprised of
species exhibiting a high content of secondary structure yet
lacking speciWc tertiary structural elements. In fact, the
unfolding enthalpy at neutral pH corresponds to 3.6calg¡1,
which suggests that the protein undergoes only partial
unfolding when compared to soluble proteins. An impor-
tant caveat is that such Wndings neither address the inquiry
of whether the neutral pH structure is indeed in a metasta-
ble state, nor whether the fusogenic state is achieved by a
“spring loaded” [142] or “spring-loaded boomerang” mech-
anism [40], thereby yielding the more stable low pH fuso-
genic state. The current hypothesis regarding the
mechanism of HA “metastability” therefore remains a mat-
ter of inquiry that cannot be resolved based on critical anal-
ysis of available studies on puriWed intact HA.
Irrespective of the controversy surrounding the alleged
“metastability” of the neutral pH structure, a number of
unequivocal observations remain. The overall thermody-
namic stability of the global HA structure is suYciently
high and thermal unfolding is characterized by an enthalpy
of 800kcalmol¡1 per trimer. Interestingly, acidiWcation of
HA to pH 5.0 results in a proton-induced loss of coopera-
tivity (i.e., ?Hcal/?HVH»1.2) relative to that measured for
the neutral pH state (i.e., ?Hcal/?HVH»3.0) [117]. These
data are interpreted to reXect the absence of HA1–HA1
interface contacts within the trimeric conformation at low
pH in which the three subunits remain associated via the
HA2–HA2 contacts within the coiled-coil assembly. The
low pH HA structure exhibits properties of a “compact
denatured state” as observed for several other proteins
when exposed to low pH, as gleaned from the retention of
signiWcant secondary structure (i.e., far UV CD spectrum)
and the absence of noticeable tertiary structural elements
(i.e., near UV CD and intrinsic Trp Xuorescence) [117].
Limited information derived from DSC studies on crude
preparations of membranes [147] or whole viruses [145]
does not permit a meaningful elucidation of thermody-
namic parameters due to the complexity of studying such
systems. Nevertheless, studies on the thermal stability of
intact inXuenza virus have generally corroborated the
results on the puriWed protein [117], and this method has
recently gained popularity as a means of obtaining quality
control parameters in virus preparations aimed at immuni-
zation protocols [148]. SigniWcantly, the major endotherm
observed in DSC proWles of the intact virus is completely
abrogated in samples pre-treated with bromelain [145], rep-
resenting further evidence that the cooperative transition in
the intact inXuenza virus arises almost exclusively from
thermally induced unfolding of HA.
Simian immunodeWciency virus gp41
Infection by primate (i.e., human and simian) immuno-
deWciency virus involves fusion between the virus and tar-
get cell membranes, which is mediated by the envelope
glycoprotein gp41 that is a product of the precursor gp160.
By analogy to the inXuenza virus HA, the gp41 trimer
undergoes conformational rearrangements from a postu-
lated metastable native structure [141,142] to a more stable
fusion-active conformation. Thermodynamic analysis of
this “trimer of hairpins” has been hampered by its tendency
towards irreversible denaturation/aggregation [116]. How-
ever, Jelesarov and Lu [116] have designed a construct {i.e.,
N34(L6)C28 in lieu of the original N36(L6)C34} which
exhibits a similar structure to the wild type primate gp41,
yet unfolds via a reversible single two-state transition. The
estimated free energy of 88.4calmol¡1-residue¡1 (i.e.,
?GD370Jmol¡1-residue¡1) at 25°C is within the values
observed for other proteins (i.e., ?G»200–600Jmol¡1- res-

Page 14

C.A.S.A. Minetti, D.P. Remeta / Archives of Biochemistry and Biophysics 453 (2006) 32–53
45
idue¡1). Interestingly, the free energy of the wild type con-
struct (i.e., ?GD18.2 kcalmol¡1) at pH 7.0 and 25°C is
further increased by single point mutations (i.e., T582I and
T586I) in the central coiled-coil core (i.e., ??G»7.0–
8.4kcalmol¡1), indicating that polar residues lead to signiW-
cant stabilization of the core structure. The Wnding of
buried polar residue mutations that stabilize the core has
been speculated as a possibility that such mutations may
serve as the “tip of balance” between the metastable and
stable conformations [116]. Thermal denaturation studies
of the trimeric N34(L6)C28 protein reveals an unfolding
enthalpy of 42.3kcalmol¡1 and corresponding heat capac-
ity of 2.15kcalK¡1mol¡1 (i.e., ?CpD10.8calK¡1mol-resi-
due¡1) [116]. This value of the heat capacity is suYciently
lower than that estimated from the ?ASA-based model (i.e.,
?CpD2.9–3.8kcalK¡1mol¡1), which is expected for com-
plete unfolding and hydration of the polypeptide chain.
The aforementioned results have signiWcant implications in
terms of the proposition that this envelope glycoprotein
undergoes incomplete unfolding, resulting in a compact
partially solvated-denatured state. Analysis of the resultant
thermodynamic data supports such a proposal in that the
unfolding enthalpy of 0.55kcalmol-residue¡1
?HD2.3kJmol-residue¡1) measured at 60°C, is suYciently
lower than the average unfolding enthalpy of 0.84kcalmol-
residue¡1 (i.e., ?HD3.5kJmol-residue¡1) estimated for
globular proteins at this temperature.
(i.e.,
Structural stability of the membrane-associated
nonconstitutive protein colicin E1
Colicin is a plasmid encoded protein family that is pro-
duced by E. coli and responsible for channel forming toxic
activity against susceptible strains of E. coli [69,70,72]. Coli-
cin E1 is a three domain protein in which each domain
assumes a diVerent functional role, namely as channel
forming (C-terminus or C domain), receptor binding (cen-
tral R domain), and translocation (N-terminus or T
domain) from the outer to the inner membrane. Deconvo-
lution of the DSC proWles for colicin E1 allows assignment
of diVerent thermal transitions to each of these structural
domains [69]. Two well-deWned transitions are resolved at
T1D51°C and T2D66°C, the Wrst representing a non two-
state transition (i.e., ?Hcal/?HVHD1.9) characteristic of the
presence of two independent folding domains (assigned as
R and T), while the second more stable species is character-
ized by a single two-state transition (i.e., ?Hcal/?HVH»1.0)
that can be assigned to unfolding of the C domain. An inde-
pendent analysis of the corresponding truncated forms of
colicin E1 has facilitated conWrmation of these assignments.
Interestingly, the presence of the N-terminus T domain
thermally stabilizes the central R domain, whereas the C
domain is thermally destabilized by the presence of T–R
domains as evidenced by the 2.0°C increase in transition
temperature of the isolated C domain. The presence of
three domains in the DSC proWle of intact colicin E1 is in
accordance with the structural description of this class of
proteins. One of the striking features deriving from these
calorimetric studies of colicin E1 is the distinction between
thermal and thermodynamic stability. SpeciWcally, the R
domain unfolds at a lower temperature than the T and C
domains, yet is actually more stable as deduced from the
temperature-dependence of the free energies for these three
domains (refer to Fig. 10 in [69]). The unusually large
unfolding enthalpy of
?HD8.13calg¡1) measured for colicin E1 at TmD52°C
may be interpreted to reXect disruption of the extended
interacting helices (e.g., a coiled-coil) [69] and is consistent
with enthalpies measured for unfolding ?-helices (i.e.,
?HD1.2–1.67kcalmol-residue¡1 [149]). During membrane
binding, the coiled-coil R domain is proposed to “uncoil”
resulting in binding of the T domain to TolC and transloca-
tation through the TolC channel [69] and [150]. These stud-
ies demonstrate the utility of correlating structure,
energetics, and function, the integration of which provides
signiWcant insights regarding the properties of such an
important family of toxins.
136.2kcalmol¡1 (i.e.,
Thermal stability of membrane proteins
Inspection of published data on the thermal stability of
membrane proteins reveals that there are only a handful of
cases in which complete reversibility (and repeatability) is
observed. In most studies, there is a clear scan-rate depen-
dence of thermal unfolding [115,131,133] that in the most
severe situations manifests itself in the form of post-transi-
tion aggregation at high temperatures. The resultant
unfolding enthalpies and heat capacities are generally lower
than what is expected for water-soluble proteins
[4,6,112,117,151–153], an indication that membrane pro-
teins often undergo partial unfolding. Fig. 7 presents a
comparative analysis of normalized unfolding enthalpies
(on a per residue basis) for several representative mem-
brane protein systems. In all cases, the values of ?H for
membrane proteins (i.e., 0.1–0.6kcalmol-residue¡1) are sig-
niWcantly less than the average unfolding enthalpy deter-
mined for soluble proteins (i.e., 0.84kcalmol-residue¡1),
poly-alanine ?-helices in water (i.e., 1.5kcalmol-residue¡1),
and ?-sheet hexapeptide in lipids (i.e., 1.3kcalmol-resi-
due¡1). An alternative interpretation for the reduced
unfolding enthalpies is that the magnitude of ?H is com-
mensurate with unfolding extramembraneous domains as
opposed to the entire structure. In fact, analysis based
solely on the soluble portion of these proteins provides bet-
ter agreement with the magnitudes of ?H expected for solu-
ble proteins. Additional structural-energetic studies are
therefore required to accurately account for the transmem-
brane region in terms of the forces stabilizing such protein
domains. The latter necessarily requires the application of
protein engineering, mutagenesis, and minimization
approaches employing model peptides to conduct experi-
mental and theoretical studies devoted to map out micro-
scopic intra- and intermolecular interactions. It is worth
noting that some model-based systems have circumvented

Page 15

46
C.A.S.A. Minetti, D.P. Remeta / Archives of Biochemistry and Biophysics 453 (2006) 32–53
such deWciencies through the use of genetically modiWed
proteins that are amenable to rigorous thermodynamic
analysis (i.e., two-state reversible unfolding versus non two-
state irreversible aggregation/precipitation) [116]. Applica-
tion of novel methodologies to search for small peptides
that reversibly fold into transmembrane secondary struc-
tures promises to provide signiWcant insight towards global
eVorts to elucidate and quantify the driving forces of mem-
brane protein stability [118].
Membrane proteins within the context of intra- and
intermolecular interactions: impact on overall membrane
protein stability
Intramolecular interdomain interactions
Thermal denaturation studies have provided valuable
information regarding the molecular origins of the stability
of rather complex macromolecular systems. A previous sec-
tion described the biphasic DSC proWle of intact colicin E1
in which its three structural domains exhibit distinct diVer-
ences in terms of their respective thermal stabilities relative
to the isolated fragments. Thermal denaturation of the
membrane protein phenylalanine oxidase has been assessed
by infrared studies [124] delineating interactions between
the N- and C-terminal domains [124]. Such representative
examples illustrate the signiWcant insight that thermal
unfolding studies aVord in terms of elucidating complex
intramolecular and interdomain interactions and their
impact on the overall thermal and thermodynamic stability
of membrane proteins.
Oligomeric state: origins of intermolecular interactions and
impact on stability
A number of studies have proposed that the folding of
porins is thermodynamically linked with oligomerization
[80,81,154], a Wnding that has also been reported for homo-
oligomeric ?-helical proteins [55]. In contrast, cytolysins are
soluble monomeric proteins that undergo conformational
changes prior to insertion/oligomerization within blood cell
membranes, thereby eliciting pore formation and cell lysis
(as reviewed in [65]). Mutagenesis studies reveal that
replacement of functionally important residues often ren-
ders these pore forming toxins (e.g., pneumolysin) inactive.
SigniWcantly, a recent investigation suggests that replace-
ment of speciWc residues which are neither functionally crit-
ical nor impart folding/stability actually abrogates cell lysis
[64], conceivably by inhibiting the conformational changes
facilitating membrane insertion and oligomerization within
the erythrocyte membrane.
The impact of multimerization on the conformational sta-
bility of hemagglutinin has been investigated by comparing
the detergent suspended homotrimers with the multimeric
forms, the latter derived via extensive dialysis to remove octyl-
glucoside. It is worth noting that the protein remains soluble
in aqueous solution in the absence of detergent, albeit adopt-
ing a “rosette” structure comprised of 5–6 trimers. Compari-
son of the energetic parameters indicates that inter-trimeric
associations within HA rosettes provides a relatively modest
contribution to the overall trimer stability at neutral pH (i.e.,
?TmD1.0°C; ??HcalD 100kcalmol¡1) [117].
The membrane protein bacteriorhodopsin exists as both
trimeric and monomeric structures [34]. Comparative studies
reveal that the latter undergoes unfolding with a transition
temperature that is approximately 20°C lower than its tri-
meric counterpart, although the unfolding enthalpies of both
structures are comparable (i.e., ?H»100kcalmol¡1). These
data suggest that the trimeric assembly is stabilized entropi-
cally relative to the monomer, contributing on the order of
5kcalmol¡1 to the overall free energy. The calculated cooper-
ative ratios of 1 and 3 reXect cooperative unfolding and sta-
bilization by intersubunit interactions in the monomer and
trimer, respectively ([34] and references therein).
Impact of ligands and/or receptors on the structural stability
of membrane proteins
In view of the substantial body of literature dedicated to
membrane protein structure and function, there are rela-
tively few studies devoted to folding and energetics, partic-
ularly the impact of receptor binding on membrane protein
conformation. One such investigation involves interaction
Fig. 7. Normalized unfolding enthalpies for representative membrane pro-
tein systems compared to soluble proteins. The unfolding enthalpies are
expressed on a per residue basis in kcalmol-residue¡1 for proteins: (1)
bacteriorhodopsin [112]; (2) and (3) are yeast cytochrome-c in lipids and
in detergent (Tween 80), respectively [34]; (4) and (5) are wild type and
minus subunit III-P. denitriWcans cytochrome-c, respectively [34]; (6) P.
denitriWcans cytochrome-c subunit III [34]; (7) Beef cytochrome-c subunit
III [34]; (8) Photosystem II [135]; (9), (10), and (11) are hemagglutinin at
pH 7.0 in 1% octylglucoside (trimer), aqueous solution (rosette), and pH
5.0 (rosette), respectively [117]; (12) and (13) are wild type and mutant
(E71Q) OmpF [139], respectively; (14) soluble proteins (§0.8kcalmol-resi-
due¡1) [130]; (15) ?-helical polyalanine in water [149]; and (16) ?-sheet
hexapeptide in lipids [118]. In the inset, the unfolding enthalpies (normal-
ized to molecular weight) for several membrane protein systems are com-
pared with the temperature-dependence of the average enthalpies
measured for unfolding soluble proteins (dashed line).
123456789 10111213 14 1516
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
Enthalpy (kcal·mol-residue -1)
Protein/Peptide
20406080100
3
6
9
12


Enthalpy (cal·g -1)
Temperature (ºC)