Structure, Vol. 13, 1521–1531, October, 2005, ©2005 Elsevier Ltd All rights reserved. DOI 10.1016/j.str.2005.07.010
Retroviral Matrix Domains Share Electrostatic
Homology: Models for Membrane Binding Function
throughout the Viral Life Cycle
Paul S. Murray,1 Zhaohui Li,1 Jiyao Wang,1
Chris L. Tang,2 Barry Honig,2 and Diana Murray1,*
1
Department of Microbiology and Immunology and
The Institute for Computational Biomedicine
Weill Medical College of Cornell
New York, New York 10021
2
Howard Hughes Medical Institute
Department of Biochemistry and Molecular Biophysics
Columbia University
New York, New York 10032
Summary
The matrix domain (MA) of Gag polyproteins performs
multiple functions throughout the retroviral life cycle.
MA structures have an electropositive surface patch
that is implicated in membrane association. Here, we
use computational methods to demonstrate that electrostatic control of membrane binding is a central
characteristic of all retroviruses. We are able to explain a wide range of experimental observations and
provide a level of quantitative and molecular detail
that has been inaccessible to experiment. We further
predict that MA may exist in a variety of oligomerization states and propose mechanistic models for the
effects of phosphoinositides and phosphorylation.
The calculations provide a conceptual model for how
non-myristoylated and myristoylated MAs behave similarly in assembly and disassembly. Hence, they provide a unified quantitative picture of the structural
and energetic origins of the entire range of MA function and thus enhance, extend, and integrate previous
observations on individual stages of the process.
Introduction
Retroviruses are enveloped viruses that cause a wide
range of diseases in humans and animals (Goff, 2001;
Vogt, 1997). Newly assembled viruses acquire their lipid
coats by budding through the plasma membrane of
host cells, while mature, infectious viruses shed these
coats upon fusing with the membranes of target cells.
The Gag polyprotein contains the structural proteins of
mature retroviral particles as domains arranged in a linear chain that is cleaved by the viral protease during
budding and maturation (Bukrinskaya, 2004; Freed,
1998; Garnier et al., 1998; Gottlinger, 2001; Scarlata and
Carter, 2003; Weldon and Hunter, 1997). Gag directs the
assembly of new virions and is targeted to the inner
leaflet of the plasma membrane by its matrix domain
(MA) (Zhou et al., 1994). Mature MA forms a proteinaceous shell at the luminal surface of the viral membrane (Kingston et al., 2001; Wilk et al., 2001) and
remains membrane associated until the virus infects a
cell and its components disassemble. The MA may
then participate in nuclear integration, though such a
*Correspondence: [email protected]
role is controversial (Goff, 2001). Therefore, MA/membrane interactions and their regulation play a crucial
role throughout the retroviral life cycle. How, then, does
the MA perform its many roles? In vitro and in vivo
studies provide a great deal of information on the structural features of the MA that are important for function.
However, there have been relatively few studies of the
membrane association of MAs in well-defined systems
(Dalton et al., 2005; Provitera et al., 1999; Zhou et al.,
1994). Here, we use computational approaches to
quantitatively describe the interaction of MAs with phospholipid membranes at different stages of the retroviral
life cycle.
It is unclear how Gag is targeted to the host cell
plasma membrane during viral assembly, but many
studies implicate specific motifs in MA (see e.g., Callahan and Wills, 2000; Dalton et al., 2005; Ono et al.,
2000; Provitera et al., 1999; Soneoka et al., 1997; Zhou
et al., 1994). The MAs of most retroviruses have two
membrane targeting signals: an N-terminal myristate,
which partitions into the membrane hydrocarbon, and
a positively charged surface patch, which is proposed
to interact electrostatically with acidic phospholipid
headgroups (McLaughlin and Aderem, 1995; Resh, 1999).
A number of retroviral MAs are not myristoylated, yet
they still have a basic surface patch, indicating that
electrostatic interactions contribute to membrane association for all MAs (Conte and Matthews, 1998). The
structures of MAs from eight retroviruses have been
solved to date (see Experimental Procedures), and, despite low sequence similarity, all share a similar compact fold that includes a core of four α helices. The
structures of MAs that are not myristoylated were found
to share the same fold as natively myristoylated MAs,
and a recent study directly demonstrates that the structure of the myristoylated HIV-1 MA is very similar to the
structure of the unmyristoylated form (Tang et al., 2004).
Figure 1 illustrates how the basic surface patch “migrates” about the surface of MA. The MAs are depicted
in the same orientation, and these data are obtained
from a multiple structure alignment (Yang and Honig,
2000). Some MAs within the same class share a common basic patch; examples include HIV-1 and SIV MAs
(Figures 1A and 1B) and HTLV-II and BLV MAs (Figures
1C and 1D). However, there is no distinct sequence motif that is predictive of this feature across retroviral
classes (compare any interclass MA pair in Figure 1)
nor even sometimes within classes, e.g., HIV-1 and
EIAV MAs (Figures 1A and 1G). For MAs that are myristoylated (Figures 1A–1F), the basic patch is in close
proximity to the site of myristoylation. Thus, the basic
surface patch is conserved in the absence of sequence
similarity, and, furthermore, there seems to be considerable freedom in the way the MA can interact with
acidic phospholipids.
The membrane association of MA and Gag may be
regulated in a number of ways: (1) oligomerization may
increase the membrane partitioning of Gag (Adamson
and Jones, 2004) by producing a larger, composite basic surface as well as by favoring the exposure of my-
Structure
1522
Figure 1. The Location of the Basic Surface Patch on MA Structures
Is Not Conserved
(A–H) MAs of known structure are represented by their molecular
surfaces and are shown in the same orientation. The electrostatic
potentials were calculated, mapped to the molecular surfaces, visualized in GRASP (Nicholls et al., 1991), and graded continuously
from red (−4 kT/e) to white (0 kT/e) to blue (+4 kT/e). The green
arrows point to the basic patches implicated in membrane binding.
The N terminus of each MA structure is denoted by a small black
arrow, except for EIAV MA, where it is located on the backside of
the protein.
ristate (Tang et al., 2004; Zhou and Resh, 1996); (2) phosphoinositides, which are multivalent acidic lipids, may
stabilize particle structure, promote assembly (Campbell et al., 2001), and determine the site of viral assembly in cells (Ono et al., 2004); (3) the substitution to alanines of serine residues that are phosphorylated in
mature MA impairs infectivity, suggesting that postentry events, including membrane dissociation, are facilitated by MA phosphorylation (Kaushik and Ratner,
2004), though a physiological role for serine phosphorylation has not been clearly established. Each of these
phenomena has the capacity to significantly impact
membrane association by strengthening or weakening
the electrostatic interaction between the MA and acidic
lipids.
The role of electrostatic interactions in the membrane
association of MA and Gag has not been fully characterized. In this study, we computationally assess the
predicted strength of the electrostatic forces involved
by solving the Poisson-Boltzmann equation for atomiclevel models of retroviral MAs and phospholipid bilayers (Honig and Nicholls, 1995). Calculations of this
type have been highly successful in their ability to account for a large number of experimental observations,
involving interactions between charged membranes
and various proteins and peptides (e.g., Ben-Tal et al.,
1996; Murray and Honig, 2002; Diraviyam et al., 2003;
Wang et al., 2004). Most recently, the membrane binding behaviors of Rous sarcoma virus MA as observed
in biochemical assays were accurately predicted, demonstrating the applicability to retroviral matrix domains
(Dalton et al., 2005).
Here, we examine a broader range of MA functions
in a unified way and consider many retroviral MAs in
addition to HIV-1 MA. We provide mechanistic molecular models for (1) the membrane-associated forms of
MAs, (2) the effects of experimentally characterized
mutations in the MA on membrane association, (3) how
MAs interact with phosphoinositides, (4) how phosphorylation of the MA weakens its membrane interaction, and (5) the role of oligomerization in enhancing the
membrane association of both myristoylated (myr) and
nonmyristoylated (non-myr) MAs. By examining the
biophysical characteristics of MA structures and predictive models for MAs from all of the retroviral classes,
we find that MAs share a striking electrostatic homology in the absence of strong sequence homology. This
feature indicates that the basic surface patch and electrostatic interactions with membrane surfaces play an
important functional role throughout the retroviral life
cycle. Our computational analysis supports this hypothesis by providing quantitative models for each of
the major functions of MA.
Results
Electrostatic Interactions Are Predicted
to Significantly Enhance the Membrane
Association of MAs of Known Structure
In many cell types, retroviruses assemble at and bud
from the plasma membrane. The inner leaflet of the
plasma membrane of a typical mammalian cell is relatively enriched in phosphatidylserine (PS) with respect
to other intracellular membranes and contains w30
mole percent PS (Buckland and Wilton, 2000). Hence,
we use 2:1 PC:PS (PC, phosphatidylcholine) bilayers in
our calculations and 0.1 M KCl to mimic physiological
ionic strength. Throughout, ⌬Gel represents the minimum electrostatic free energy of interaction between
the MA and the membrane (see Experimental Procedures).
Figure 2 illustrates our model for HIV-1 MA in its minimum electrostatic free energy orientation at the membrane surface. Far from the MA, the negative contour
of the membrane is flat, but in the vicinity of the MA, it
becomes dramatically distorted, illustrating qualitatively the strong favorable electrostatic interaction,
⌬Gel w−5 kcal/mol (Table 1, row 2). (This electrostatic
Electrostatic Homology of Retroviral MAs
1523
Figure 2. HIV-1 MA Interacts Favorably with a Negatively Charged
Membrane
HIV-1 MA (from 1hiw) is depicted adsorbed to the surface of a 2:1
PC:PS lipid bilayer in 0.1 M KCl. The +1 kT/e (blue) and −1 kT/e
(red) electrostatic equipotential profiles were calculated with Delphi
(Gallagher and Sharp, 1998) and imaged with GRASP (Nicholls et
al., 1991). HIV-1 MA is depicted as a molecular surface (white), and
the lipid bilayer is depicted in atomic detail (carbon and hydrogen,
white; oxygen, red; nitrogen, blue; and phosphate, yellow). The arrow points to the N terminus of the protein where the myristate
would be attached.
interaction is not expected to change due to the existence of the C-terminal extension of the Gag polyprotein.) In the following argument, we use ⌬Gel as an approximation to the electrostatic contribution to the
absolute binding (Ben-Tal et al., 1996). Assuming a lipid
concentration of w10−3 M at the surface of a spherical
cell of radius 10 ␮m (McLaughlin and Aderem, 1995), a
membrane binding free energy of −5 kcal/mol (or a molar partition coefficient of w103 M−1) is not sufficient
to localize more than half of the protein at the plasma
membrane. However, if an additional driving force is
available, binding can be significantly enhanced. The
myristate on HIV-1 MA (Figure 2, arrow) is well positioned to insert into the membrane and can also contribute w103 M−1 to membrane partitioning (Peitzsch
and McLaughlin, 1993). Because of their close proximity in space, the energetic contributions of the basic
residues and the myristate can be additive and the
binding constants can multiply; based on these properties a total membrane partition coefficient of w106 M−1
results. Thus, the electrostatic and nonpolar components
together can produce stable membrane association.
However, while this characterization of MA/membrane
association is very likely relevant from a biochemical
perspective, the contributions of the two membrane
binding motifs under physiological conditions is not as
easily discernable. Effects related to the sequestration/
exposure of the myristate and oligomerization are addressed below.
Table 1 lists our ⌬Gel predictions for a variety of MAs.
It is important to note that ⌬Gel is, at best, a relative
measure of the electrostatic contribution to membrane
association. As described in Experimental Procedures,
calculating the absolute binding constant is beyond the
scope of this paper. The ⌬Gel values for the MAs of
known structure range from −3 kcal/mol (for MMLV MA)
to −5 kcal/mol (for HIV-1 MA), and, hence, our results
indicate that the basic patches contribute significantly
to membrane association. Indeed, our prediction for
RSV MA is in good agreement with experimental observations (Dalton et al., 2005). Furthermore, the favorable
electrostatic interactions are not a trivial consequence
of net charge. For example, EIAV MA, which has a net
charge of 0, has ⌬Gel = −4.6 kcal/mol. Importantly, our
calculations account for a wide range of observations
described in the literature concerning relative changes
in membrane binding upon residue substitutions (Table
1, column 4). The calculations have been shown to be
highly reliable for such predictions in other protein/
membrane systems (Murray and Honig, 2002). However, some of the experimental readouts listed in Table
1, e.g., “no particle production,” are not necessarily indicative of decreased membrane association and may
be indicative of some other mechanism. Examples of
agreement between our calculations and experimental
data include: (1) HIV-1 MA (Ono et al., 2000): mutating
basic residues in the vicinity of the basic patch to glutamates (K29E/K31E; Table 1, row 6) was shown to decrease membrane binding, and our calculations correctly predict a significant increase in ⌬Gel; (2) MMLV
MA (Soneoka et al., 1997): a series of mutations that
decrease the number of basic residues in the surface
patch either decreased or abolished membrane binding
of Gag. In agreement with experimental data (Table 1,
rows 15–19), our calculations predict that (a) ⌬Gel increases but is still favorable for the two cases in which
membrane binding is observed merely to decrease, and
(b) ⌬Gel increases and becomes positive (representing
a repulsive electrostatic interaction with the membrane)
for the three cases in which membrane binding was
observed to be abolished.
Homology Models of MAs of Unknown Structure
Suggest that the Basic Surface Patch Is a
Conserved and Functionally Important Feature
To test the generality of these results, we built homology models for 40 sequences of MAs from the alpha-,
beta-, gamma-, delta-, and epsilon-retrovirus and the
lentivirus classes (see Experimental Procedures and our
website: http://maat.med.cornell.edu/MA.html). We were
able to obtain reliable models for some, but not all, epsilon-retroviral MAs.
All of our models except for one (avian spleen necrosis virus MA) have a significant basic surface patch.
Representative examples are provided in Figure 3. As
in Figure 1, the MAs are depicted in the same orientation, as obtained from a multiple structure alignment,
and are annotated by their electrostatic surface potentials. Importantly, to our knowledge, our work provides
Structure
1524
Table 1. Electrostatic Free Energies of Interaction for MA Structures, Models, and Mutants
Matrix Domain: Structure, Mutant, or Homology Model
⌬Gel (kcal/mol)
Net Charge
HIV-1, 1tam
HIV-1, 1hiw: A
HIV-1, 2hmx: A
HIV-1, 1uph:A
HIV-1, 1tam, membrane contains 1% PIP2
HIV-1, 1tam, K29E/K31E
HIV-1, 1tam, pS8/pS66/pS71/pS76
SIV, 1ecw
HTLV-2, 1jvr
RSV, 1a6s
MPMV, 1bax
EIAV, 1hek
BLV
MMLV, 1mn8
MMLV, K21N
MMLV, K32N/R34N
MMLV, K31N/K32N/R33N/R34N
MMLV, K31E/K32N/R33E/R34N
MMLV, K17N/R21Q/K31N/K32N/R33N/R34N
HTLV-1
HTLV-1, K47I/K48I/K51I
FIV
FIV, K26E/K28E/K29E
FIV, K26N/K28N/K29N
−5.2
−4.7
−4.7
−5.2
−7.7, −6.3
−0.8
−3.0
−4.2
−4.2
−4.0
−2.9
−4.6
−3.5
−2.7
−1.5
−0.3
+1.1
+3.7
+1.6
−4.4
−1.8
−2.5
+1.7
+0.9
+6
+3
+1
+1
+6
+2
−2
+2
+6
+3
+2
0
+2
+2
+1
0
−2
−4
−4
+6
+3
+4
−2
+1
Mutation Phenotype
Decreased membrane binding
Decreased infectivity
Decreased membrane binding
Decreased membrane binding
Abolishes membrane binding
Abolishes membrane binding
Abolishes membrane binding
No particle production
No particle release
No particle release
The electrostatic free energy of interaction (⌬Gel) was calculated for the MA structures and mutants for which there exists experimental data.
The calculations were performed with 2:1 PC:PS lipid bilayers and 0.1 M KCl as described in Experimental Procedures. Results related to
MAs of known structure are highlighted in bold, those related to mutants of known structure are shown in regular text, and those related to
homology models and their mutants are italicized.
the first structural models for MAs in the epsilon-retrovirus class (Figure 3E). In each of the three epsilonretrovirus models, the large basic patch is adjacent to
the N-terminal glycine, which is believed to be myristoylated.
Using these models, we predicted the effects of mutations in HTLV-1 and FIV MAs, which were examined
in the literature (Le Blanc et al., 1999; Manrique et al.,
2001). Our results suggest that impaired MA/membrane
association may account for the mutant phenotypes.
(1) HTLV-1 MA: a Gag mutant in which three of the basic
residues proposed to contribute to the basic surface
patch were mutated to isoleucines resulted in no particle production. Our prediction that ⌬Gel increases by
w2.5 kcal/mol is consistent with this observation (Table
1, rows 21–22). (2) FIV MA: charge reversal mutations
in the proposed basic surface patch resulted in no particle production, and our calculations predict that ⌬Gel
for these mutants is positive (repulsive; Table 1, rows
23–25).
Phosphoinositides Strengthen the Electrostatic
Interaction of HIV-1 MA with Membranes
Phosphatidylinositol 4,5-bisphosphate (PI(4,5)P2) is primarily localized to the inner leaflet of the plasma membrane (McLaughlin et al., 2002). It is present at about 1
mole percent and is the predominant polyvalent phosphoinositide for a cell in its resting state. Several pieces
of evidence suggest that phosphoinositides and, in
particular, PI(4,5)P2 play important roles in retroviral assembly: (1) PI(4,5)P2 is the ligand for a number of host
factors involved in membrane trafficking and viral bud
fission (Martin, 2001); (2) Campbell et al. (2001) showed
that inositol derivatives induce the formation of viral-
like particles in vitro; and (3) Ono et al. (2004) showed
that PI(4,5)P2 plays a critical role in directing the subcellular localization of HIV-1 assembly. Our calculations
predict that the minimum electrostatic free energy for
HIV-1 MA with a membrane containing 1 mole percent
PI(4,5)P2 (i.e., 67:32:1 PC:PS:PIP2) is −7.5 kcal/mol; this
increase in membrane association (compare rows 1 and
5 in Table 1) is due solely to nonspecific electrostatic
interactions, i.e., a decrease in ⌬Gel from −5 to −7.5
kcal/mol. Thus, our calculations predict that enhanced
membrane association is mediated by an interaction
between the basic patch on MA and PI(4,5)P2, and that
there need not be a specific coordination between particular residues on MA and the head group of PI(4,5)P2.
However, the calculations do not address the possibility of specific binding of inositol polyphosphates, as
has been suggested by the results of Campbell et al.
(2001).
The ⌬Gel value of −7.5 kcal/mol is a lower limit on the
effect, as there is, on average, at least 1 PI(4,5)P2 for
every 100 lipids, and the imprint of MA on the membrane encompasses only about 30 lipids; thus, there
must be significant demixing or lateral reorganization
of PI(4,5)P2 for an interaction with the MA to occur. In
addition, if the MA is already adsorbed to the membrane, it may laterally sequester PI(4,5)P2 through nonspecific electrostatic interactions as suggested by Figure 1: the strong positive potential at the membrane
surface surrounding HIV-1 MA may serve as a basin of
attraction for polyanionic phospholipids. Such a phenomenon has been observed for a basic peptide that
mimics the effector domain of the protein MARCKS
(Rauch et al., 2002; Wang et al., 2004). Our calculations
predict that retroviral MAs may similarly sequester phos-
Electrostatic Homology of Retroviral MAs
1525
tegration complex, a role that requires postentry dissociation from the plasma membrane of an infected cell.
Recent studies have suggested that phosphorylation of
serines within the MA is required for viral infectivity and
replication. In particular, Kaushik and Ratner (2004) demonstrate that an MA mutant, in which serines at positions 9, 67, 72, and 77 are changed to alanines, greatly
impairs infectivity with respect to wild-type. Our calculations predict that phosphorylation of serines at these
positions significantly weakens the electrostatic interaction between HIV-1 MA and the plasma membrane
surface, i.e., ⌬Gel increases from −5 to −3 kcal/mol (Table 1, row 7). This suggests that phosphorylation may
play a role in favoring the membrane-dissociated state
of MA, thus facilitating the nuclear targeting function
of MA.
Figure 3. Structural Models of MAs of Unknown Structure Suggest
that All Retroviral MAs Share Electrostatic Homology
(A–F) Homology models were constructed as described in Experimental Procedures and are shown in the same orientation, with the
exception of WEHV-2 MA, which is rotated 180° about the vertical
axis. The visualization protocol is the same as that in Figure 1. The
percent sequence identity with the structural template is given.
phoinositides in the plane of the plasma membrane: a
single MA is predicted to sequester a PI(4,5)P2 lipid
with a Boltzmann-averaged electrostatic free energy of
−1.7 kcal/mol. However, the entropic cost for localizing
a PI(4,5)P2 lipid to a region corresponding to the imprint
of MA is w+0.7 kcal/mol (see Experimental Procedures
for details). Hence, at most, an upper limit to the net
free energy gain due either to MA adsorbing to a PIP2containing membrane or to the sequestration of a
PI(4,5)P2 by an isolated, membrane-associated MA is
−1.0 kcal/mol. However, an array of membrane-associated MAs formed during assembly is expected to laterally accumulate many PI(4,5)P2 lipids (see Figure 5).
Our calculations support two mutually consistent
models: PI(4,5)P2 may direct Gag to the plasma membrane by enhancing membrane association, and/or
membrane bound MA may laterally sequester PI(4,5)P2
in the plasma membrane, either to stabilize membrane
association or to facilitate the targeting of cellular factors required for budding and fission.
Phosphorylation of HIV-1 MA Significantly Weakens
the Electrostatic Interaction with Membranes
The MA is also implicated in steps that occur early in
the viral life cycle, after infection. Mature MA is thought
to aid in the nuclear targeting and import of the prein-
Calculations with Trimer and Dimer Models of MAs
Support Plasticity in Gag Oligomerization
Retroviruses are enveloped viruses, and, although various oligomeric structures of Gag and Gag domains have
been isolated from virions or assembled in vitro, the entire
viral particle requires the lipid coat for its structural integrity (Adamson and Jones, 2004). Early electron microscopy studies assumed that MAs were arranged icosahedrally at the viral envelope (Wilk and Fuller, 1999); the
crystal structures of the SIV and HIV-1 MA trimers (Hill
et al., 1996; Rao et al., 1995) are consistent with these
observations. However, more recent studies reveal that
there is no long-range symmetry among HIV-1 MA in particles, confounding the determination of overall particle
structure. In addition, local 6-fold as well as 3-fold and 2fold symmetry was discerned, suggesting the existence
of both trimers and dimers of Gag and, perhaps, MA as
well (Wilk and Fuller, 1999). Furthermore, retroviral capsid (CA) and nucleocapsid (NC) domains, which are
thought to drive Gag oligomerization, have been shown
to interact in different oligomeric combinations, most
notably dimers (Adamson and Jones, 2004). Hence, it
is likely that MA in the context of Gag in the immature
particle and as the cleaved protein in the mature particle may adopt a variety of intermolecular arrangements.
The structure of HIV-1 MA has been determined by
NMR in a monomeric state both in its myr (Tang et al.,
2004) and un-myr forms (Massiah et al., 1994; Matthews et al., 1995) and by X-ray crystallography in a
trimeric state (Hill et al., 1996). As described in Hill et
al. (1996) and depicted in Figure 4A, the trimer structure
of HIV-1 MA is well suited for productive membrane association: all three monomers contribute to a contiguous basic surface patch that may interact with acidic
lipids in the membrane, while the three myristates (one
from each monomer) may simultaneously partition into
the membrane interior. In addition, the C termini project
away from the membrane, suggesting that the rest of
Gag may point radially toward the center of the virion,
as observed microscopically; i.e., the trimer is “assembly competent.” More recently, the MAs of EIAV and
MMLV have been crystallized, and the fundamental
units of these structures are dimers (Hatanaka et al.,
2002; Riffel et al., 2002). Our calculations predict that
the EIAV and MMLV MA dimers, like the HIV-1 and SIV
MA trimers, associate strongly with membrane surfaces
Structure
1526
Figure 4. Oligomerization of MA Increases
the Effective Size of the Basic Surface:
Structures and Models
(A–F) Table 2 provides the quantitative results for the MA/membrane interactions illustrated here. Monomeric subunits in the
oligomeric structures (or models) are differentially colored for clarity, and the membrane surface is represented schematically
as a black line. The sites of myristoylation
on the MAs that are myristylated (HIV-1 and
MMLV MA) face the membrane surface in all
cases and are denoted by black arrows. The
+1 kT/e (blue) and −1 kT/e (red) equipotential
electrostatic profiles were calculated and
imaged with GRASP (Nicholls et al., 1991).
in assembly-competent orientations (Figures 4D and
4F). It is worth pointing out that the interfaces observed
among MAs are expected to interact weakly (Hill et al.,
1996), which suggests that MA/MA interactions should
occur only under conditions of high MA concentrations,
i.e., during assembly and in the viral particle. Here, we
compare quantitatively how the MAs of HIV-1, EIAV, and
MMLV interact electrostatically with membranes as dimers and trimers.
In Figures 2 and 4A, the HIV-1 MA monomer and trimer, respectively, are depicted in their minimum free
energy orientations: ⌬Gel(monomer) = −5 kcal/mol and
⌬Gel(trimer) = −9 kcal/mol. We calculated ⌬Gel for each
of the monomers in their trimer configuration; the sum
of the individual electrostatic free energies is equal to
⌬Gel(trimer) so that the contribution of each monomer
(w−3 kcal/mol) to the trimer is additive. As seen by
comparing Figures 2 and 4A, the monomers in the context of the trimer are not in their minimum free energy
orientations. Therefore, to adopt the trimer configuration, each monomer gives up favorable protein/membrane interaction free energy to gain favorable protein/
protein interaction free energy. Assuming that the myristate remains partitioned into the membrane interior
for some time after a membrane bound trimer of MA
breaks up, our calculations predict that it is energetically more favorable for three MAs to be monomeric
rather than to retain the trimer form at the membrane
surface: 3 × ⌬Gel(monomer) = 3 × (−5 kcal/mol) = −15
kcal/mol < ⌬Gel(trimer) = −9 kcal/mol. This is important
in the context of postentry events (see Discussion).
However, either in the context of Gag, where downstream domains drive Gag-Gag interactions, or at the
luminal surface of the virion membrane, where MA concentration is high, MA/MA interactions may be stabilized.
It was suggested (Hatanaka et al., 2002) that EIAV MA
may adopt a similar trimeric arrangement, as observed
in the crystal structures of HIV-1 MA. A model for trimeric EIAV MA, based on the HIV trimer, is depicted in
Figure 4C; this arrangement brings the negatively
charged lobe of EIAV MA (top of Figure 1G) in proximity
to the membrane surface. Whereas ⌬Gel for the EIAV
MA monomer is −4.6 kcal/mol (Table 2, row 5), the electrostatic free energy for the trimer model is −1.7 kcal/
mol (Table 2, row 6). Since EIAV MA is not myristoylated
and does not have any prominent surface hydrophobic
features, electrostatic interactions, mediated by its
basic surface patch, are likely critical to viral assembly.
Hence, the trimer model, at least that based on the
HIV-1 MA trimer structure, does not appear to be a viable configuration for EIAV assembly. However, as stated
above, EIAV MA crystallized as a dimer. This structure
is shown in its minimum free energy orientation (Figure
Electrostatic Homology of Retroviral MAs
1527
Table 2. Oligomerization of MA Increases the Electrostatic Free
Energy of Interaction
Matrix Protein Oligomeric State:
Structure or Model
⌬Gel, kcal/mol
HIV-1 MA monomer, 1hiw:A
HIV-1 MA trimer, 1hiw:A,B,C
HIV-1 MA dimer, model (1hek)
HIV-1 MA dimer, model (1mn8)
EIAV MA monomer, 1hek:A
EIAV MA trimer, model (1hiw)
EIAV MA dimer, 1hek:A,B
MMLV MA monomer, 1mn8:A
MMLV MA trimer, model (1hiw)
MMLV MA dimer, 1mn8:A,B
MMLV MA dimer, 1mn8:C,D
−5.1
−9.1
−8.3
−5.5
−4.6
−1.7
−8.0
−2.3
−6.9
−5.4
−5.6
The electrostatic free energy of interaction (⌬Gel) was calculated
for the structures and models of the oligomeric MAs. The
calculations were performed with 2:1 PC:PS lipid bilayers and 0.1
M KCl as described in Experimental Procedures. Results related to
experimentally determined structures are highlighted in bold, and
those related to models constructed with structure superposition
are shown in regular text.
4D), for which ⌬Gel = −8.0 kcal/mol (Table 2, row 7).
Similar to HIV-1 MAs, we predict that the monomer is
the more energetically favorable state for mature EIAV
MA: 2 × ⌬Gel(monomer) = 2 × (−4.6 kcal/mol) = −9.2
kcal/mol < ⌬Gel(dimer) = −8 kcal/mol.
Conversely, it is interesting to consider a dimer model
for HIV-1 MA based on the EIAV MA dimer structure. As
shown in Figure 4B, this model has properties similar
to those of the trimer structure (Figure 4A) and is assembly competent as well (⌬Gel = −8.3; Table 2, row 3).
MMLV MA is also capable of forming dimeric (Figure 4E)
and trimeric (Figure 4F) structures that interact favorably
with negatively charged membranes: ⌬Gel(dimer) = −5.4
kcal/mol (Table 2, row 10) and ⌬Gel(trimer) = −6.9 kcal/
mol (Table 2, row 9).
Overall, our calculations and models suggest that
there is significant plasticity in the ways in which MAs
may interact with each other in order to accommodate
different Gag and MA oligomerization schemes (Adamson and Jones, 2004). An important feature of our models is that they predict that myr and non-myr MA oligomers will dissociate at the membrane surface; this will
ultimately lead to membrane dissociation of MA monomers after viral entry as described below. Note that this
is not true for MA in the context of Gag because downstream protein-protein interactions, as well as NC/RNA
interactions, are potent mediators of Gag oligomerization, and the free energy gained by these interactions
is very likely more favorable than free energy gained
due to enhanced MA/membrane interactions upon dissociation into monomers.
Discussion
The Role of MA/Membrane Interactions
throughout the Retroviral Life Cycle
Our computational modeling highlights the importance
of electrostatic interactions in regulating the membrane
association of MA during assembly, in the virion and
during postentry events. Our predictions complement
experimental observations and provide a unifying picture of the role of MA/membrane interactions in the retroviral life cycle. Figure 5 synthesizes information from
the literature (cited above) with our computational results for both myr and non-myr MAs by using HIV-1
and EIAV, respectively, as specific examples. Figure 5
strikingly emphasizes both the central role of electrostatics and the similar manner in which myr and nonmyr MAs behave.
Membrane Targeting and Assembly
Myristate is expected to be sequestered by MA on monomeric HIV-1 Gag (Tang et al., 2004; Zhou and Resh,
1996). As suggested by Figure 5A, the membrane may
occasionally compete with MA for the myristate, but
the overall change in free energy will be small because
the hydrophobic component should be approximately
equivalent whether the myristate is bound to MA or partitioned into the membrane. Hence, the membrane
binding free energies for both monomeric HIV-1 Gag
and EIAV Gag are weak (w−5 kcal/mol) but sufficient
to direct a significant amount of protein to the plasma
membrane surface, which is preferentially enriched in
acidic phospholipids with respect to other intracellular
membranes (Pomorski et al., 2001). PIP2 (red circles)
serves to favor the membrane-associated state of both
Gags by at least 1 kcal/mol (Table 1). Membrane association, as well as the directionality it confers on Gag,
significantly increases the probability that Gag monomers and oligomers will find and efficiently bind each
other (Figures 5A and 5B). Gag oligomerization is due
mainly to interactions mediated by CA and NC. Oligomerization can occur either in solution or at the membrane surface and will result in strong membrane association for both HIV-1 and EIAV Gags (Figure 5B): (1)
Oligomerization is proposed to favor the myristateexposed state for HIV-1 Gag (Tang et al., 2004; Zhou
and Resh, 1996). Either Gag dimers or trimers will be
strongly anchored at the membrane surface. For example, a trimer of HIV-1 Gags is predicted to have a membrane binding free energy of ⌬G w−9 kcal/mol (electrostatics) + −12 kcal/mol (myristates) w−21 kcal/mol. (2)
Oligomerization similarly enhances the membrane association of EIAV Gag dimers: ⌬G w−8 kcal/mol (electrostatics). Since, Gag-Gag interactions are facilitated
by membrane association, such oligomeric complexes
of Gag molecules may act as seeds for the formation
of extended patches, which are visualized microscopically as electron-dense regions at the cell surface (Figure 5C; [Gelderblom, 1991]). Most likely, the major difference between myr and non-myr Gags lies in the
specific region of the plasma membrane that is targeted: Myr Gag is thought to be localized to putative
lipid rafts because of the saturated nature of the myristate chain (Lindwasser and Resh, 2002; Briggs et al.,
2003).
Budding and Maturation
When w1500–3000 Gag molecules assemble at the
plasma membrane, Gag-Gag interactions collaborate
with nucleocapsid/RNA interactions and host factors
(including PIP2, red circles) to produce a viral bud and,
ultimately, an immature particle. During this process,
the viral protease cleaves Gag so that its domains become independent proteins, and the particle subsequently assumes its mature, infectious morphology
Structure
1528
Figure 5. MA/Membrane Interactions Are Differentially Regulated throughout the Virus
Life Cycle
This figure synthesizes our computational results with what is known about the function
of MA from experimental studies. Our results
provide quantitative measures that provide
molecular insight into the roles of MA (see
text).
(Figure 5D). In the mature virion, the MA to lipid concentration ratio is very high; thus, MA will remain in the
oligomer form. Our calculations predict that all of the
MA protein will remain associated with the luminal surface of the viral membrane, i.e, membrane binding is
not saturated under these conditions For example,
using the EAIV MA dimer structure to create an extended array of membrane-associated dimers, we predict that each monomer, in its dimer orientation, has
the same membrane binding free energy, both in the
absence and presence of other membrane-adsorbed
MAs (data not shown). This is in agreement with similar
earlier studies on basic peptides and “discreteness of
charge” effects (Murray et al., 1999). The shell of matrix
proteins depicted in Figure 5D may help stabilize the
virion structure as well as provide interaction sites for
the luminal component of the viral coat proteins (Goff,
2001).
Entry and Infection
Upon gaining entry into a new cell, the components of
the virion must disassemble in order for the virus to be
infective. Upon fusion of the virus with the host cell, the
inner leaflets of the two membranes become contiguous. This dramatically increases the membrane surface
accessible to the MAs (Figure 5E). The MAs then diffuse
away from each other, because, in the absence of other
forces favoring oligomerization, it is energetically more
favorable for the MAs to be in a monomer state (see
Results). Once monomeric, the MA may compete with
the membrane and eventually sequester the myristate
(see, e.g., Resh, 2004). In both cases (myr and nonmyr), monomeric MA/membrane association is weak
(w−5 kcal/mol), and a fraction of the protein will dissociate from the membrane. This may be functionally important, as HIV-1 MA is thought to play a role in directing the preintegration complex to the nucleus. Kaushik
and Ratner (2004) suggest that phosphorylation of
HIV-1 MA is crucial for infectivity. Our calculations (Table 1) predict that phosphorylation of the MA further
weakens its interaction with the membrane (Figure 5F),
thus providing a larger pool of cytosolic MA.
Summary
In essence, MAs appear to have been designed to associate and dissociate on cue: (1) during assembly,
downstream domains drive Gag oligomerization, which
enhances Gag/membrane association through the membrane binding properties of the MA domain; (2) in the
virion, the high concentration of MA induces MA/MA
interactions, which are responsible for the formation of
Electrostatic Homology of Retroviral MAs
1529
a shell of MA on the luminal surface of the virion, but
which force each MA to adopt suboptimal orientations
with respect to membrane binding; (3) postentry, the
MA concentration decreases, which favors MA/MA dissociation, as each individual MA adopts its minimum
free energy orientation with respect to the membrane.
In the monomeric form, each MA is now relatively weakly
bound to the plasma membrane of the newly infected
cell. Our study provides numerical estimates for the
strength of MA/membrane interactions at each of the
stages of the retroviral life cycle for both the myr and
non-myr MA. As such, it provides a quantitative picture
of the structural and energetic origins of the diverse
functions that may be mediated by MA and thus enhances, extends, and integrates previous suggestions
as to the role of different factors in the individual stages
of the process. The work presented here is generalizable, in part, to matrix domains/proteins from other enveloped RNA viruses, e.g., influenza virus, rhabdoviruses, and filoviruses (data not shown). Although the
true physiological picture is surely more complex than
that suggested here, as well as that described in the
literature, our calculations provide a foundation upon
which to construct more detailed models of MA function.
Experimental Procedures
Structures and Sequences
Structures of MAs were obtained from the PDB, except where
noted. HIV-1: 1tam (Matthews et al., 1995), 2hmx (Massiah et al.,
1994), 1hiw (Hill et al., 1996), 1uph (Tang et al., 2004); SIV: 1ecw
(Rao et al., 1995); HTLV-2: 1jvr (Christensen et al., 1996); BLV: coordinates kindly supplied by S.J. Matthews (Matthews et al., 1996);
MMLV: 1mn8 (Riffel et al., 2002); MPMV: 1bax (Conte et al., 1997);
EIAV: 1hek (Hatanaka et al., 2002); RSV: 1a6s (McDonnell et al.,
1998). Multiple structure alignments were created with PrISM (Yang
and Honig, 2000), and models of MA oligomers were built by using
structural superposition (Shindyalov and Bourne, 1998). The sequences modeled were taken from the ICTV database: http://
www.ictvdb.rothamsted.ac.uk/Ictv/fs_retro.htm. Of the many HIV
and SIV Gag sequences available, we manually chose a representative from each clade in order to avoid modeling highly redundant
sequences. Databases of MA structures and sequences were created in PrISM for further analysis.
Modeling Protocol
The following steps were followed in generating homology models:
(1) each of the target sequences was scanned against the structure
database in PrISM. If an alignment had a high normalized SmithWaterman score (>30), and the subsequent Needleman-Wunsch
alignment produced high target coverage (>90%), this latter alignment and PrISM’s tools were used to construct the homology
model. PSI-BLAST searches gave similar alignments for these targets. These models scored well according to the structure evaluation tools Verify3D (Luthy et al., 1992), ProsaII (Sippl, 1993), and pG,
which provides a normalized ProsaII z score (Sanchez and Sali,
1998). (2) For the remaining sequences, a combination of profile
and fold recognition methods, i.e., 3D-PSSM (Kelley et al., 2000),
PSIPRED (McGuffin et al., 2000), FUGUE (Shi et al., 2001), and
HMAP (Tang et al., 2003), were used to identify suitable structural
templates and initial alignments for modeling. PrISM was used to
construct models for each sequence based on alternative templates and alignments. These models were evaluated, and, if necessary, the alignments were manually edited by using secondary
structure information as a guide. This process was iterated either
until a well-evaluated model was obtained or the search was abandoned. Models that passed our screening criteria were viewed as
reliable. This was further confirmed by sensitivity analysis in which
known MA structures were used as templates for modeling the se-
quences of other MAs of known structure. In each case, even for
distantly related MAs, i.e., across retroviral classes, the overall
electrostatic properties of the structures were faithfully reproduced
by these “control models.”
Electrostatic Calculations
The electrostatic free energy component of the membrane interaction of MAs was obtained from a modified version of the DelPhi
program (Gallagher and Sharp, 1998) that solves the nonlinear
Poisson Boltzmann equation for protein/membrane systems (BenTal et al., 1996). The application of this method has been highly
successful for a wide range of peptide/membrane and protein/
membrane systems, as judged by theory/experiment comparisons
(see http://maat.med.cornell.edu/refs.html for additional references
that could not be included here due to space limitations). DelPhi
produces finite difference solutions to the PB equation (the FDPB
method) for a system in which the solvent is described in terms of
a bulk dielectric constant and concentrations of mobile ions, while
solutes (here, MAs and membranes) are described in terms of the
coordinates of the individual atoms. Hydrogen atoms were added
to the heavy atoms in the MA structures with CHARMM (Brooks et
al., 1983). Phospholipid bilayers were built, and all atomic charges
and radii in the system were assigned as described previously
(Ben-Tal et al., 1996). The protein/membrane/solvent system was
mapped onto a three-dimensional lattice of l3 points. Regions inside the molecular surfaces of the protein and membrane are assigned a dielectric constant of 2 to account for electronic polarizability, and those outside are assigned a dielectric constant of 80.
An ion exclusion layer of 2 Å is added to the solutes.
The potential is considered converged when its value changes
less than 10−4 kT/e between successive iterations. Electrostatic
free energies are obtained from the potentials, and the electrostatic
free energy of interaction is determined as the difference between
the electrostatic free energy of MA docked at the membrane surface, Gel(P,M), and the electrostatic free energies of the MA, Gel(P),
and the membrane, Gel(M), taken separately:
⌬Gel = Gel(P · M) − [Gel(P) + Gel(M)]
A sequence of focusing runs of increasing resolution (0.375, 0.75,
1.5, 3.0 grid/Å) with lattice sizes of 2733–3533 was employed. In the
initial calculation, the MA/membrane model encompassed a small
percentage of the lattice (w10%), and the potentials at the boundary points are approximately zero. The precision in the electrostatic
free energies of interaction is <0.3 kcal/mol for all calculations, but
is typically <0.1 kcal/mol. We did not calculate the full membrane
partition coefficient. Rather, we calculated the orientation of minimum electrostatic free energy. Previous work established that this
orientation describes well both the relative binding free energies
and the electrostatic contribution to the absolute binding free energies (e.g., Ben-Tal et al., 1996; Murray and Honig, 2002).
Calculations with PIP2
We followed the theoretical scheme outlined in previous work
(Wang et al., 2004). The initial and final states of the system are
composed of two equal-sized regions of membrane containing the
same mole percent monovalent acidic lipid: in the initial state, one
has MA adsorbed to its surface, Gel(P·M), and one has a single PIP2
lipid, Gel(PIP2·M); in the final state, one has PIP2 in the vicinity of
the membrane-adsorbed MA, Gel(PIP2·P·M), and one has no PIP2,
Gel(M). The electrostatic free energy for PIP2 sequestration was determined as:
⌬Gel(PIP2) = (Gel(PIP2 · P · M) + Gel(M)) − (Gel(P · M) + Gel(PIP2 · M))
⌬Gel(PIP2) was calculated for the w30 lipid positions within the imprint of the MA on the membrane surface. The average electrostatic free energy of sequestration and the average entropic cost of
PIP2 demixing were calculated as Boltzmann-weighted averages:
< ⌬Gel(PIP2) > = Σ⌬Gj(exp( − ⌬Gj / kBT) / Q)
S = − kBΣ[exp( − ⌬Gj / kBT) / Q]ln(exp( − ⌬Gj / kBT) / Q),
where Q = Σ (exp(−⌬Gi/kBT) is the partition function, kB is the Boltzmann constant, T is the absolute temperature, ⌬Gj is the electro-
Structure
1530
static free energy associated with PIP2 moving to position j in the
membrane, and the sum is over all positions, j. Since the membrane contains 1 mole percent PIP2, we summed over 100 positions. The energetic and entropic costs for PIP2 sequestration by a
membrane-adsorbed MA are given by ⌬Gel = <⌬Gel(PIP2)>2 −
<⌬Gel(PIP2)>1, and ⌬GS = −T (S2 − S1), where <⌬Gel(PIP2)>1 and S1
are the averages in the absence of the MA, and <⌬Gel(PIP2)>2 and
S2 are the averages in the presence of MA. We used the “bootstrap” resampling method to determine the statistical significance
of these values (Efron and Tibshirani, 1993), which are precise to
within 0.05 kcal/mol.
Acknowledgments
This work was supported by National Institutes of Health grant
AI54167 and National Science Foundation (NSF) grant MCB030028
for advanced computational resources at the Pittsburgh Supercomputing Center (D.M.). Support from NSF grant MCB-0416708 is
also acknowledged (B.H.). The authors are grateful to Stephen J.
Matthews for the coordinates of BLV MA and to Stephen Goff,
Volker Vogt, Amanda Dalton, Marilyn Resh, and Michael Summers
for stimulating discussions.
Received: July 1, 2005
Revised: July 1, 2005
Accepted: July 9, 2005
Published: October 11, 2005
Diraviyam, K., Stahelin, R., Cho, W., and Murray, D. (2003). Computer modeling of the membrane interaction of FYVE domains. J.
Mol. Biol. 328, 721–736.
Efron, B., and Tibshirani, R. (1993). An Introduction to the Bootstrap
(New York: Chapman and Hall).
Freed, E.O. (1998). HIV-1 gag proteins: diverse functions in the virus
life cycle. Virology 251, 1–15.
Gallagher, K., and Sharp, K.A. (1998). Electrostatic contributions
to heat capacity changes of DNA-ligand binding. Biophys. J. 75,
769–776.
Garnier, L., Bowzard, B., and Wills, J.W. (1998). Recent advances
and remaining problems in HIV assembly. AIDS 12, S5–S16.
Gelderblom, H.R. (1991). Assembly and morphology of HIV: potential effect of the structure on virus function. AIDS 5, 617–638.
Goff, S.P. (2001). Retroviridae: The retroviruses and their replication. In Fundamental Virology, D.M. Knipe and P.M. Howley, eds.
(Philadelphia, PA: Lippincott Williams and Wilkins), pp. 843–912.
Gottlinger, H.G. (2001). The HIV-1 assembly machine. AIDS 15,
S13–S20.
Hatanaka, H., Iourin, O., Rao, Z., Fry, E., Kingsman, A., and Stuart,
D.I. (2002). Structure of equine infectious anemia virus matrix protein. J. Virol. 76, 1876–1883.
Hill, C.P., Bancroft, D.P., Christensen, A.M., and Sundquist, W.I.
(1996). Crystal structures of the trimeric human immunodeficiency
virus type I matrix protein: implications for membrane binding and
assembly. Proc. Natl. Acad. Sci. USA 93, 3099–3104.
References
Honig, B.H., and Nicholls, A. (1995). Classical electrostatics in biology and chemistry. Science 268, 1144–1149.
Adamson, C.S., and Jones, I.M. (2004). The molecular basis of HIV
capsid assembly-five years of progress. Rev. Med. Virol. 14, 107–121.
Kaushik, R., and Ratner, L. (2004). Role of human immunodeficiency virus type 1 matrix phosphorylation in an early postentry
step of virus replication. J. Virol. 78, 2319–2326.
Ben-Tal, N., Honig, B., Peitzsch, R.M., Denisov, G., and McLaughlin,
S. (1996). Binding of small basic peptides to membranes containing
acidic lipids: theoretical models and experimental results. Biophys.
J. 71, 561–575.
Kelley, L.A., MacCallum, R.M., and Sternberg, M.J.E. (2000). Enhanced genome annotation using structural profiles in the program
3D-PSSM. J. Mol. Biol. 299, 499–520.
Briggs, J.A., Wilk, T., and Fuller, S.D. (2003). Do lipid rafts mediate
virus assembly and pseudotyping? Curr. Opin. Struct. Biol. 84,
757–768.
Kingston, R.L., Olson, N.H., and Vogt, V.M. (2001). The organization
of mature Rous sarcoma virus as studies by cryoelectron microscopy. J. Struct. Biol. 136, 67–80.
Brooks, B.R., Bruccoleri, R.E., Olafson, B.D., States, D.J., Swaminathan, S., and Karplus, M. (1983). CHARMM: a program for
macromolecular energy, minimization, and dynamics calculations.
J. Comp. Chem. 4, 187–217.
Le Blanc, I., Rosenberg, A.R., and Dokhelar, M. (1999). Multiple
functions for the basic amino acids of the human T-cell leukemia
virus type 1 matrix protein in viral transmission. J. Virol. 73, 1860–
1867.
Buckland, A.G., and Wilton, D.C. (2000). Anionic phospholipids, interfacial binding and the regulation of cell functions. Biochim. Biophys. Acta 1483, 199–216.
Lindwasser, O.W., and Resh, M.D. (2002). Myristoylation as a target
for inhibiting HIV assembly: unsaturated fatty acids block viral budding. Proc. Natl. Acad. Sci. USA 99, 13037–13042.
Bukrinskaya, A.G. (2004). HIV-1 assembly and maturation. Arch.
Virol. 149, 1067–1082.
Luthy, R., Bowie, J.U., and Eisenberg, D. (1992). Assessment of
protein models with three-dimensional profiles. Nature 356, 83–85.
Callahan, E.M., and Wills, J.W. (2000). Repositioning basic residues
in the M domain of the Rous sarcoma virus Gag protein. J. Virol.
74, 11222–11229.
Campbell, S., Fisher, R.J., Towler, E.M., Fox, S., Issaq, H.J., Wolfe,
T., Phillips, L.R., and Rein, A. (2001). Modulation of HIV-like particle
assembly in vitro by inositol phosphates. Proc. Natl. Acad. Sci. USA
98, 10875–10879.
Christensen, A.M., Massiah, M.A., Turner, B.G., Sundquist, W.I., and
Summers, M.F. (1996). Three-dimensional structure of the HTLV-II
matrix protein and comparative analysis of matrix proteins from the
different classes of pathogenic human retroviruses. J. Mol. Biol.
264, 1117–1131.
Conte, M.R., and Matthews, S. (1998). Retroviral matrix proteins: a
structural perspective. Virology 246, 191–198.
Conte, M.R., Klikova, M., Hunter, E., Ruml, T., and Matthews, S.
(1997). The three-dimensional solution structure of the matrix protein from the type D retrovirus, the Mason-Pfizer monkey virus, and
implications for the morphology of retroviral assembly. EMBO J.
16, 5819–5827.
Dalton, A.K., Murray, P.S., Murray, D., and Vogt, V.M. (2005). Biochemical characterization of Rous sarcoma virus MA protein interaction with membranes. J. Virol. 79, 6227–6238.
Manrique, M.L., Celma, C.C.P., Gonzalez, S.A., and Affranchino, J.L.
(2001). Mutational analysis of the feline immunodefficiency virus
matrix protein. Virus Res. 76, 103–113.
Martin, T.F. (2001). PI(4,5)P2 regulation of surface membrane traffic.
Curr. Opin. Cell Biol. 13, 493–499.
Massiah, M.A., Starich, M.R., Paschall, C., Summers, M.F., Christensen, A.M., and Sundquist, W.I. (1994). Three-dimensional structure of the human immunodeficiency virus type 1 matrix protein. J.
Mol. Biol. 244, 198–223.
Matthews, S., Barlow, P., Clark, N., Kingsman, S., Kingsman, A.,
and Campbell, I. (1995). Refined solution structure of p17, the HIV
matrix protein. Biochem. Soc. Trans. 23, 725–738.
Matthews, S., Mikhailov, M., Burny, A., and Roy, P. (1996). The solution structure of the bovine leukaemia virus matrix protein and similarity with lentiviral matrix proteins. EMBO J. 15, 3267–3274.
McDonnell, J.M., Fushman, D., Cahill, S.M., Zhou, W., Wolven, A.,
Wilson, C.B., Nelle, T.D., Resh, M.D., Wills, J.W., and Cowburn, D.
(1998). Solution structure and dynamics of the bioactive retroviral
M domain from Rous sarcoma virus. J. Mol. Biol. 279, 921–928.
McGuffin, L.J., Bryson, K., and Jones, D.T. (2000). The PSIPRED
protein structure prediction server. Bioinformatics 16, 404–405.
McLaughlin, S., and Aderem, A. (1995). The myristoyl-electrostatic
Electrostatic Homology of Retroviral MAs
1531
switch: a modulator of reversible protein-membrane interactions.
Trends Biochem. Sci. 20, 272–276.
McLaughlin, S., Wang, J., Gambhir, A., and Murray, D. (2002). PIP2
and proteins: interactions, organization and information flow. Annu.
Rev. Biophys. Biomol. Struct. 31, 151–175.
Murray, D., and Honig, B. (2002). Electrostatic control of the membrane targeting of C2 domains. Mol. Cell 9, 145–154.
Murray, D., Arbuzova, A., Mihaly, G., Ghambir, A., Ben-Tal, N.,
Honig, B., and McLaughlin, S. (1999). Electrostatic properties of
membranes containing acidic lipids and adsorbed basic peptides:
theory and experiment. Biophys. J. 77, 3176–3188.
Nicholls, A., Sharp, K.A., and Honig, B. (1991). Protein folding and
association: insights from the interfacial and thermodynamic properties of hydrocarbon. Proteins 11, 281–296.
Ono, A., Orenstein, J.M., and Freed, E.O. (2000). Role of the Gag
matrix domain in targeting human immunodeficiency virus type 1
assembly. J. Virol. 74, 2855–2866.
Ono, A., Ablan, S.D., Lockett, S.J., Nagashima, K., and Freed, E.O.
(2004). Phosphatidylinositol (4,5) bisphosphate regulates HIV-1 Gag
targeting to the plasma membrane. Proc. Natl. Acad. Sci. USA 101,
14889–14894.
Tang, C., Loeliger, P., Liuncsford, P., Kinde, I., Beckett, D., and Summers, M.F. (2004). Entropic switch regulates myristate exposure in
the HIV-1 matrix protein. Proc. Natl. Acad. Sci. USA 101, 517–522.
Vogt, V.M. (1997). Retroviral virons and genomes. In Retroviruses,
J.M. Coffin, S.H. Hughes, and H.E. Varmus, eds. (Cold Spring Harbor, NY: Cold Spring Harbor Laboratory Press), pp. 22–70.
Wang, J., Gambhir, A., McLaughlin, S., and Murray, D. (2004).
A computational model for the electrostatic sequestration of
PI(4,5)P2 by membrane-adsorbed basic peptides. Biophys. J. 86,
1969–1986.
Weldon, R.A., and Hunter, E. (1997). Molecular requirements for retrovirus assembly. In Structural Biology of Viruses, W. Chiu, ed.
(New York, NY: Oxford University Press), pp. 381–403.
Wilk, T., and Fuller, S.D. (1999). Towards the structure of the human
immunodeficiency virus: divide and conquer? Curr. Opin. Struct.
Biol. 9, 231–243.
Wilk, T., Gross, I., Gowen, B.E., Rutten, T., De haas, F., Welker, R.,
Krausslich, H., Boulanger, P., and Fuller, S.D. (2001). Organization
of immature human immunodeficiency virus type 1. J. Virol. 75,
759–771.
Peitzsch, R.M., and McLaughlin, S. (1993). Binding of acylated peptides and fatty acids to phospholipid vesicles: pertinence to myristoylated proteins. Biochemistry 32, 10436–10443.
Yang, A.S., and Honig, B. (2000). An integrated approach to the
analysis and modeling of protein sequences and structures. I. Protein structural alignment and a quantitative measure for protein
structural distance. J. Mol. Biol. 301, 665–678.
Pomorski, T., Hrafnsdottir, S., Devaux, P.F., and van Meer, G. (2001).
Lipid distribution and transport across cellular membranes. Semin.
Cell Dev. Biol. 12, 139–148.
Zhou, W., and Resh, M.D. (1996). Differential membrane binding of
the human immunodeficiency virus type 1 matrix protein. J. Virol.
70, 8540–8548.
Provitera, P., Bouamr, F., Murray, D., Carter, C., and Scarlata, S.
(1999). Binding of equine infectious anemia virus matrix proteins to
membrane bilayers involves multiple interactions. J. Mol. Biol. 296,
887–898.
Zhou, W., Parent, L.J., Wills, J.W., and Resh, M.D. (1994). Identification of a membrane-binding domain within the amino-terminal region of human immunodeficiency virus type 1 Gag protein which
interacts with acidic phospholipids. J. Virol. 68, 2556–2569.
Rao, Z., Belyaev, A.S., Fry, E., Roy, P., Jones, I.M., and Stuart, D.I.
(1995). Crystal structure of SIV matrix antigen and implications for
virus assembly. Nature 378, 743–747.
Rauch, M.E., Ferguson, C.G., Prestwich, G.D., and Cafiso, D.
(2002). Myristoylated alanine-rich C kinase substrate (MARCKS) sequesters spin-labeled phosphatidylinositol-4,5-bisphosphate in
lipid bilayers. J. Biol. Chem. 277, 14068–14076.
Resh, M.D. (1999). Fatty acylation of proteins: new insights into
membrane targeting of myristoylated and palmitoylated proteins.
Biochim. Biophys. Acta 1451, 1–16.
Resh, M.D. (2004). A myristoyl switch regulates membrane binding
of HIV-1 Gag. Proc. Natl. Acad. Sci. USA 101, 417–418.
Riffel, N., Harlos, K., Iourin, O., Rao, Z., Kingsman, A., Stuart, D.,
and Fry, E. (2002). Atomic resolution structure of the Moloney murine leukemia virus matrix protein and its relationship to other retroviral matrix proteins. Structure 10, 1627–1636.
Sanchez, R., and Sali, A. (1998). Large-scale protein structure modeling of the Sacharomyces cerevisiae genome. Proc. Natl. Acad.
Sci. USA 95, 13597–13602.
Scarlata, S., and Carter, C. (2003). Role of HIV-1 Gag domains in
viral assembly. Biochim. Biophys. Acta 1614, 62–72.
Shi, J., Blundell, T.J., and Mizuguchi, K. (2001). FUGUE: sequencestructure homology recognition using environment-specific substitution tables and structure-dependent gap penalties. J. Mol. Biol.
310, 243–257.
Shindyalov, I.N., and Bourne, P.E. (1998). Protein structure alignment by incremental combinatorial extension (CE) of the optimal
path. Protein Eng. 11, 739–747.
Sippl, M.J. (1993). Recognition of errors in three-dimensional structures of proteins. Proteins 17, 355–362.
Soneoka, Y., Kingsman, S.M., and Kingsman, A. (1997). Mutagenesis analysis of the murine leukemia virus matrix protein: identification of regions important for membrane localization and intracellular transport. J. Virol. 71, 5549–5559.
Tang, C.L., Xie, L., Koh, I.Y., Posy, S., Alexov, E., and Honig, B.
(2003). On the role of structural information in remote homology
detection and sequence alignment: new methods using hybrid sequence profiles. J. Mol. Biol. 334, 1043–1062.