HIV enters cells via endocytosis and dynamin-dependent fusion with endosomes.

Abstract

Enveloped viruses that rely on a low pH-dependent step for entry initiate infection by fusing with acidic endosomes, whereas the entry sites for pH-independent viruses, such as HIV-1, have not been defined. These viruses have long been assumed to fuse directly with the plasma membrane. Here we used population-based measurements of the viral content delivery into the cytosol and time-resolved imaging of single viruses to demonstrate that complete HIV-1 fusion occurred in endosomes. In contrast, viral fusion with the plasma membrane did not progress beyond the lipid mixing step. HIV-1 underwent receptor-mediated internalization long before endosomal fusion, thus minimizing the surface exposure of conserved viral epitopes during fusion and reducing the efficacy of inhibitors targeting these epitopes. We also show that, strikingly, endosomal fusion is sensitive to a dynamin inhibitor, dynasore. These findings imply that HIV-1 infects cells via endocytosis and envelope glycoprotein- and dynamin-dependent fusion with intracellular compartments.

Full text

Page 1

HIV Enters Cells via Endocytosis and
Dynamin-DependentFusionwithEndosomes
Kosuke Miyauchi,1,2Yuri Kim,1,2Olga Latinovic,1Vladimir Morozov,1and Gregory B. Melikyan1,*
1Institute of Human Virology and Department of Microbiology and Immunology, University of Maryland School of Medicine,
725 W. Lombard Street, Baltimore, MD 21201, USA
2These authors contributed equally to this work
*Correspondence: gmelikian@ihv.umaryland.edu
DOI 10.1016/j.cell.2009.02.046
SUMMARY
Enveloped viruses that rely on a low pH-dependent
step for entry initiate infection by fusing with acidic
endosomes, whereas the entry sites for pH-indepen-
dent viruses, such as HIV-1, have not been defined.
These viruses have long been assumed to fuse
directly with the plasma membrane. Here we used
population-based measurements of the viral content
delivery into the cytosol and time-resolved imaging
of single viruses to demonstrate that complete
HIV-1 fusion occurred in endosomes. In contrast,
viral fusion with the plasma membrane did not prog-
ress beyond the lipid mixing step. HIV-1 underwent
receptor-mediated internalization long before endo-
somal fusion, thus minimizing the surface exposure
of conserved viral epitopes during fusion and
reducing the efficacy of inhibitors targeting these
epitopes. We also show that, strikingly, endosomal
fusion is sensitive to a dynamin inhibitor, dynasore.
ThesefindingsimplythatHIV-1infectscellsviaendo-
cytosis and envelope glycoprotein- and dynamin-
dependent fusion with intracellular compartments.
INTRODUCTION
Endocytosis is an obligatory entry step for enveloped viruses
whose fusion proteins are activated by acidic pH (Marsh and
Helenius, 2006). In contrast, viruses that undergo fusion upon
interacting with cognate cellular receptors irrespective of the
pH are thought to fuse directly with a plasma membrane. For
instance, HIV-cell fusion initiated upon sequential interactions
of the envelope (Env) glycoprotein with CD4 and coreceptors
CCR5 or CXCR4 (e.g., Doms and Trono, 2000) has long been
assumed to occur at the cell surface, whereas internalized
virions were thought to be degraded by cells (Maddon et al.,
1988; McClure et al., 1988; Pelchen-Matthews et al., 1995; Stein
et al., 1987). This notion is supported by the fact that HIV can
mediate fusion between adjacent target cells (‘‘fusion from
without’’) and that HIV Env expressed on effector cells promotes
fusion with target cells at neutral pH. In addition, mutations in
CD4 or coreceptors (CR) that impair their ligand-induced inter-
nalization do not block HIV-1 infection (Brandt et al., 2002; Mad-
donetal.,1988).ThefactthatpseudotypingtheHIVcorewiththe
low pH-dependent G glycoprotein of Vesicular Stomatitis Virus
(VSV) eliminates the Nef requirement for optimal infectivity
(Aiken, 1997) is indicative of different entry routes for these and
HIV Env-bearing viruses. In addition, the restriction on HIV-1
infection in resting Tcells imposed by the cortical actinis consis-
tent with fusion at the cell surface (Yoder et al., 2008).
On the other hand, several lines of evidence support the exis-
tence of an alternative endocytic pathway for HIV-1 entry. First,
HIV fusion with endosomes and micropinosomes has been
observed by electron microscopy (Marechal et al., 2001; Pauza
and Price, 1988). Second, blocking the acidification of endoso-
mal compartments can augment HIV infection, apparently by
sparing the virus from degradation in lysosomes (Fredericksen
et al., 2002; Schaeffer et al., 2004; Wei et al., 2005). Third, effi-
cient infection by HIV particles pseudotyped with VSV G (Aiken,
1997) shows that there are no apparent restrictions associated
with the endocytic entry pathway. Finally, inhibition of clathrin-
mediated endocytosis reduces the efficacy of HIV-cell fusion
and infection in HeLa-derived cells (Daecke et al., 2005).
However, this intervention perturbs important cellular functions
and may thus alter the sites of virus entry.
Here, we applied time-resolved single-virus imaging and
a virus population-based fusion assay to delineate the cellular
entry sites of HIV-1. These approaches have revealed that,
surprisingly, complete HIV-1 fusion occurred in endosomal
compartments but not at the plasma membrane of epithelial
and lymphoid cells. We found that endosomal fusion was de-
layed relative to HIV-1 uptake via CD4/CR-dependent endocy-
tosis and thatthe fusion stepwas enhanced bythe largeGTPase
dynamin. Methodologies developed in this work should help
define the entry pathways of other pH-independent viruses.
RESULTS
Toelucidatethesitesofvirusentry,wefirstcomparedtheeffects
of fusion inhibitors blocking surface-accessible viruses and
universalinhibitorsthatblockallvirusesirrespectiveoftheirloca-
tion. If fusion is limited to the cell surface, these interventions
should yield identical results. However, the ability to enter into
and fuse with endosomes would result in the transient appear-
ance of viruses resistant to external inhibitors but sensitive to
inhibitors blocking endosomal fusion. The difference in virus
sensitivity to site-specific and universal inhibitors can thus be
used to deduce the entry sites of pH-independent viruses.
Cell 137, 433–444, May 1, 2009 ª2009 Elsevier Inc. 433

Page 2

HIV-1 Fusion Is Delayed Relative to Its Escape
from a Membrane-Impermeant Fusion Inhibitor
Virus-cell fusion was directly quantified by measuring the
cytosolic activity of viral core-associated b-lactamase (BlaM)
(Cavrois et al., 2002). HIV-1 cores carrying a BlaM-Vpr chimera
were pseudotyped with Env from JRFL (CCR5-tropic) or HXB2
(CXCR4-tropic) HIV-1 strains. We first examined the HIV-1 entry
sites in HeLa-derived indicator cells expressing CD4, CCR5, and
CXCR4 (designated TZM-bl cells; Wei et al., 2002). Viruses were
allowed to bind to cells in the cold, and fusion was initiated by
shifting to 37?C and measured as the extent of cleavage of
a fluorogenic substrate by the cytosolic BlaM-Vpr. To determine
the kinetics of virus-cell fusion, we stopped the reaction after
varied times of incubation at 37?C by adding a recombinant
peptide derived from the C-terminal heptad repeat region of
HIV-1 gp41 (hereafter referred to as C52L; Deng et al., 2007).
C52L and other gp41-derived peptides inhibit fusion by binding
to intermediate gp41 conformations formed upon Env interac-
tions with CD4 and CR and preventing the formation of the final
6-helix bundle structure (reviewed in Eckert and Kim, 2001). The
time of C52L addition experiments revealed that the kinetics of
the JRFL and HXB2 escape from this membrane-impermeant
inhibitor was relatively fast, showing little or no lag and reaching
completion within ?2 hr (Figure 1A).
To block HIV-1 fusion irrespective of its cellular location, we
took advantage of the steep temperature dependence of HIV-1
fusion (Frey et al., 1995; Mkrtchyan et al., 2005). JRFL and
HXB2fusion with TZM-blcellsexhibited awell-defined threshold
at ?22?C (Figure S1A available online). By contrast, the cytosolic
BlaM was active at temperatures that were not permissive for
fusion (data not shown). This allowed kinetic measurements of
virus-cell fusion by quickly reducing the temperature after varied
times of incubation at 37?C, followed by an overnight incubation
at subthreshold temperature to permit substrate cleavage.
Thetemperatureblock(TB)protocolshowedthat,surprisingly,
the cytosolic BlaM delivery was greatly delayed compared to the
virus escape from C52L (Figure 1A). This delayed kinetics can
result from two principal mechanisms. Low temperature can
either block Env-mediated fusion or inhibit post-fusion steps
that may be required for the optimal activity of the viral core-
associated BlaM-Vpr. However, previous work (Cavrois et al.,
2004) and our data (Figure S1B) show that the post-fusion
uncoating step does not enhance the BlaM activity. We found
that this activity was not affected by inhibition of cellular prote-
ases or proteasomes and, importantly, was observed in vitro in
the absence of any cytosolic factors (Figures S1C–S1E). Thus,
the cleavage of BlaM substrate faithfully reports the extent of
virus-cell fusion.
Our experimental strategy to elucidate the sites of virus entry
was further validated using HIV particles pseudotyped with the
low pH-dependent VSV G. As expected for an endocytic entry
pathway, escape from the TB was delayed relative to the virus
uptake measured by the emergence of the BlaM signal resistant
to pronase (Figure 1B). The temperature-dependent steps of
A
C
E
B
D
Figure 1. Dissection of Surface and Endo-
somal HIV-1 Fusion
(A) Virus fusion with TZM-bl cells was stopped by
adding C52L after indicated times of incubation
at 37?C, and incubation was continued up to
90 min, at which point the cells were briefly placed
on ice and loaded with the BlaM substrate. Alter-
natively, fusion was stopped by placing cells on
ice after varied times of incubation at 37?C (TB).
After loading the substrate, cells were incubated
overnight at 13.5?C regardless of the fusion
protocol to allow the substrate cleavage. The red
and blue dashed lines were obtained by subtract-
ing the TB plot from the C52L escape plot for JRFL
and HXB2, respectively. Unless stated otherwise,
data points are means ± standard error of the
mean (SEM) from triplicate measurements.
(B)FusionofVSVGpseudotypes withTZM-blcells
was blocked at indicated times either by treating
cells with 2 mg/ml pronase on ice (10 min), adding
50 mM NH4Cl, or chilling the samples (TB). Cells
were then loaded with CCF2 and incubated over-
night at 12?C.
(CandD)After20minat37?C,virusesremainingat
the surface of TZM-bl cells were rendered nonfu-
sogenic by adding C52L (arrow), and the extent
of fusion over time at 37?C was determined by
chilling cells either immediately or at indicated
time points (green squares).
(E) HXB2 virus escape from C52L and from the TB
inCEMsscells wasmeasured asdescribedabove.
The dashed blue line represents the difference
between the C52L and TB curves. Error bars are
SEM (n = 4).
434 Cell 137, 433–444, May 1, 2009 ª2009 Elsevier Inc.

Page 3

VSV G fusion were completed soon after the completion of low
pH-dependent steps, as measured by the virus, escape from
the block imposed by NH4Cl. The quick appearance of the
TB-resistant BlaM signal after the low pH-induced fusion further
implies that temperature-dependent post-fusion steps are not
required to render BlaM-Vpr active.
HIV-1 Likely Enters Lymphoid Cells through
an Endocytic Pathway
To define the sites of HIV entry in more natural target cells, we
measured the rates of virus escape from C52L and from the
TB in lymphoid CEMss cells expressing CD4 and CXCR4. In
these cells, both rates were considerably faster than in TZM-bl
cells(Figure1EversusFigure1A).However,thelossofsensitivity
to C52L occurred much earlier than the progression beyond
temperature-dependent steps, suggesting that endosomal
fusion is the major HIV-1 entry route in T cells.
HIV-1 Associated with CD4 and Coreceptors Spends
Considerable Time in Endosomes prior to Fusion
The divergent rates of HIV-1 escape from C52L and the
TB demonstrate that the actual fusion is much slower than the
loss of sensitivity to the membrane-impermeant fusion inhibitor,
which has been customarily interpreted as fusion with the
plasma membrane. The difference between the C52L- and
TB-resistant BlaM signals should reflect the fraction of internal-
ized viruses that have not fused at that time point (Figure 1A,
dashed lines). Within the first 20 min of incubation, ?40% of
viruses appeared in C52L-inaccessible compartments, while
only a small fraction acquired resistance to the TB. The high level
ofinternalizedviruseswasmaintainedduringthenext20mindue
tothesimilarratesofescapefromC52L(endocytosis)andtheTB
(fusion). Likewise, nearly half of the viruses were protected from
C52L but did not fuse with CEMss cells within the first 10 min at
37?C (Figure 1E, dashed line). Collectively, these findings show
that HIV-1 fuses primarily, if not exclusively, with endosomes.
In order to separate plasma membrane entry from endosomal
entry,viruseswerepreboundtocellsinthecoldandincubatedat
37?C for 20 min, at which point surface-accessible unfused
viruseswereblocked byC52L.TheBlaMsignal wasthenchased
by dropping the temperature either immediately or after varied
timesofincubationat37?Cinthepresenceoftheinhibitor.Under
these conditions, any increase in the BlaM signal over time
should be exclusively due to viral content release from endo-
somes. The chase experiments revealed that endosomal fusion
progressed slowly (Figures 1C and 1D), reaching completion
within ?1 hr at 37?C. As expected, the regular TB protocol
yielded much greater extents of fusion compared to the chase
protocol (red triangles versus green squares), which was most
likely due to the continued uptake and fusion of surface-
accessible viruses in the absence of the inhibitor. Thus, on
average, HIV-1 spent about 30 min in C52L-inaccessible
compartments prior to releasing its content.
In the absence of surface fusion, protection from C52L should
correspond to productive, CD4/CR-mediated HIV endocytosis.
This notion is supported by the ability of C52L to block fusion
when added at the beginning of incubation, demonstrating that
the gp41 coiled coils are exposed prior to virus uptake; this
exposure is known to occur upon Env binding to CD4 alone or
to CD4 and CR (Eckert and Kim, 2001 and references therein).
Accordingly, HIV-1 acquired resistance to inhibitors blocking
CD4 and CR binding before it escaped from C52L (Figure S1F).
Thus, HIV-1 particles internalized by pathways other than CD4/
CR-mediated endocytosis do not contribute to fusion. These
results and virus imaging data (see below) show that, surpris-
ingly, the major rate-limiting step of HIV-1 fusion occurs after
CR binding and virus endocytosis.
Single-Virus Imaging Distinguishes between Surface
and Endosomal Fusion
To unambiguously identify the sites of HIV-1 entry, we visualized
the fusion of viruses colabeled with the relatively small, diffusible
content marker (NC-GFP, Figures S2B, S2D, and S2E) and the
lipophilic dye DiD (Markosyan et al., 2005 and Experimental
Procedures). Fusion with the plasma membrane should lead to
the disappearance of the viral membrane and content markers
due to their virtually infinite dilution within the plasma membrane
and the cytosol, respectively (Figure 2A). In contrast, virus fusion
with a small intracellular organelle that is not continuous with the
plasmamembraneshouldleadtothelossofviralcontentwithout
the disappearance of a membrane marker. Hence, the fusion
sites can be identified based on the dilution of viral markers.
We validated this strategy by imaging the fusion of pseudovi-
ruses bearing E1/E2 glycoproteins of the low pH-dependent
Semliki Forest Virus (SFV). Normally, SFV fuses with acidic
endosomes, but it can also be forced to fuse with the plasma
membrane by lowering the pH (Marsh and Bron, 1997). As ex-
pected, SFV fusion with endosomes resulted in the disappear-
ance of the viral content while the membrane marker remained
localized within an endosome (Figure S3A and Movie S1). In
contrast, exposure to low pH led to the quick redistribution of
viral lipids, but not of viral content (Figure S3B and Movie S2),
demonstrating the failure of SFV to undergo full fusion with the
plasma membrane.
HIV-1 Fusion with the Plasma Membrane Is Blocked
after the Lipid Mixing Stage
JRFL or HXB2 viruses were prebound to TZM-bl cells in the cold
and triggered to fuse by quickly shifting to 37?C. We observed
three principal outcomes of HIV-cell fusion. First, viruses
released their lipid marker, as seen by the disappearance of
theredsignal,butretainedtheircontentforaslongasweimaged
(Figure2BandMovieS3).Particlesundergoingthistypeoffusion
usually exhibited limited movement before and after the lipid
transfer (Figure 2C). These events were almost assuredly due
to the partial fusion at the cell surface that did not result in the
cytosolic delivery of viral content. Second, following the trans-
port of viruses toward the cell nucleus, typical for endosomal
trafficking (Lakadamyali et al., 2004), the viral content marker
disappeared while the lipid marker continued to move as
a distinct spot (Figures 2D–2G; Movies S4 and S5). These events
observed for both JRFL and HXB2 viruses were interpreted as
the cytosolic release of viral content through fusion with endo-
somes. Third, viral markers were released (disappeared)
sequentially, often exhibiting a considerable delay between lipid
and content transfer (see Figure 5 below). As discussed below,
Cell 137, 433–444, May 1, 2009 ª2009 Elsevier Inc. 435

Page 4

A
D
F
B
C
E
G
Figure 2. Identification of HIV-1 Fusion Sites by Single-Virus Imaging
(A) Schematic presentation of redistribution of viral lipid and content markers upon fusion with a plasma membrane (left) and with an endosome (right). Viruses
colabeled with membrane (red) and content (green) markers are pseudocolored yellow.
436 Cell 137, 433–444, May 1, 2009 ª2009 Elsevier Inc.

Page 5

these events (hereafter dubbed the two-step fusion) likely reflect
the full fusion proceeding in two distinct temporally and, in most
cases, spatially separated steps.
JRFL and HXB2 pseudoviruses exhibited distinct fusion
phenotypes. The majority of JRFL particles exchanged lipids
withtheplasmamembrane,whilethecontentreleasefromendo-
someswaslessfrequent(Figure3A).Bycomparison,mostHXB2
particles underwent endosomal fusion, very few released the
lipid marker, and none exhibited the two-step phenotype seen
for JRFL viruses. The overall low probability of HIV-cell fusion
is in agreement with our previous data (Markosyan et al.,
2005). In control experiments, viral lipid and content transfer
was inhibited by a high concentration of C52L (Figure 3A),
demonstrating that the overwhelming majority of viral lipid and
content transfer events were mediated by HIV-1 Env. Viral
content release was not detected between JRFL pseudoviruses
and HeLa cells expressing CD4 but not CCR5 (Figure 3A). Thus,
under our experimental conditions, the deterioration of the GFP
signal caused by low pH in late endosomes/lysosomes was
negligible (see also Figure S2C).
The restriction on virus fusion at the cell surface was not
limited to HIV-1 and SFV pseudoviruses. We found that the
Env glycoprotein of pH-independent amphotropic Murine
Leukemia Virus (aMLV) also mediated virus fusion with endo-
somes. Out of 14 detected events, 11 released their content
from endosomes (Figures S3C and S3D), 2 transferred only the
lipid marker at the cell surface, and 1 exhibited sequential
(two-step) lipid and content release.
Endosomal Fusion Can Lead to Infection
To relate single-virus fusion to infectivity, we evaluated the frac-
tion of cells for which at least one content transfer event was
detected by imaging. Endosomal fusion was observed for
8.0% and 10.5% of cells incubated with JRFL and HXB2 parti-
cles, respectively (Figure 3B). Under identical conditions, 8.6%
of cells were infected by JRFL and 5.7% by HXB2 (i.e., multi-
plicity of infection [moi] was ?0.1). The fraction of fusion-sup-
porting cells was clearly underestimated due to the missed
events. The relatively low fraction of double-labeled particles
produced by the labeling protocol and the relatively short
imaging time limited our ability to track all fusion events. None-
theless,comparableefficaciesofviralcontentdeliveryandinfec-
tion indicate that a significant fraction of endosomal fusion
established productive infection.
Infectious HIV-1 Fuses with an Endosome
but Not with the Plasma Membrane
To take advantage of the diffusible NC-GFP marker, our initial
imaging experiments employed MLV-based pseudoviruses. To
ensure adequate incorporation of HIV-1 Env into these particles,
we used the gp41 construct lacking the cytoplasmic domain
(Figures S2A and S2B and Markosyan et al., 2005). To rule out
the possibility that deletion of the cytoplasmic domain or pseu-
dotyping with the MLV core alters the virus entry sites, we
labeled infectious viruses by cotransfecting the cells with the
proviral HIV-1 R9 clone encoding a full-length CXCR4-tropic
Env (Gallay et al., 1997) and a new vector expressing GFP-
tagged HIV Gag. In this construct (referred to as MA-GFP-CA),
the EGFP coding sequence (flanked by the viral protease
cleavage sites)was inserted between the MA and CA sequences
of Gag polyprotein (see Supplemental Experimental Proce-
dures). Upon virus maturation, MA-GFP-CA is cleaved by viral
protease, yielding a free GFP (Figures S2F–S2I). These
viruses were colabeled with DiD and allowed to fuse with
TZM-bl cells.
Similar to pseudovirus fusion, infectious HIV-1 exhibited lipid
mixing at the cell surface and content transfer from endosomes
(Figures 3A and 3C–3F and Movie S6), whereas the sequential
lipid and content release (two-step events, see Figures 5E and
5F) was less frequent. All fusion-related activities were abro-
gated in the presence of C52L (n = 972). We observed the
same fusion phenotype for particles produced by pseudotyping
the HIV-1 core with the full-length Env (Figures 3A, S3E, and
S3F). Together, these results imply that, irrespective of the origin
of the viral core or the presence of the cytoplasmic domain,
HIV-1 Env-mediated content delivery into HeLa-derived target
cells occurs through fusion with endosomes. Moreover, in
CEMss cells, infectious HIV-1 also underwent partial fusion (lipid
transfer) with the plasma membrane and complete fusion with
endosomes (Figure S4).
Endosomal Fusion Is Delayed Relative to Lipid Transfer
at the Cell Surface
The kinetics of single HIV-1 pseudovirus fusion with the plasma
membrane and with endosomes was determined by measuring
the waiting time from raising the temperature to each lipid or
content transfer event, respectively (Figure 4A). Content release
from endosomes started after a considerable delay (?10 min),
whereas lipid transfer proceeded without an apparent lag. The
rates of surface and endosomal fusion differed markedly regard-
less of the virus tropism (JRFL versus HXB2) and regardless of
whether MLV core-based pseudoviruses or infectious HIV-1
viruses (Figure 4C) were imaged. The delayed release of the viral
content marker is consistent with the lag in the cytosolic BlaM
delivery measured by the TB protocol (Figure 1A). In fact, after
renormalization to correct for the shorter imaging time, the rates
of viral content delivery measured by single-virus and BlaM
assays were indistinguishable (Figure S5). This finding validates
the usage of the TB protocol for measuring the rate of formation
of relatively small fusion pores and implies that complete pore
dilation is not required for detecting the BlaM signal in the
cytosol.
(B and C) Partial fusion of JRFL with the plasma membrane of TZM-bl cells. The time from the beginning of imaging is shown. The two-dimensional projection of
the particle’s trajectory (cyan) is overlaid on the last image. Changes in fluorescence intensities (in arbitraryunits) of membrane (red) and content (green) markers,
as well as the instantaneous velocity (blue trace) of the particle, are shown.
(D–G) Complete fusion of JRFL (D and E) and HXB2 (F and G) viruses following the fast retrograde movement from the cell periphery (cyan traces on last images).
Fusion is evident from the disappearance of GFP signal. Graphs (E and G) show changes in fluorescence of membrane and content markers (smoothed for visual
clarity) and 3D trajectories for the viruses marked by arrows in (D) and (F). See Movies S3–S5.
Cell 137, 433–444, May 1, 2009 ª2009 Elsevier Inc. 437

Page 6

HIV-1 Fusion May Proceed through a Stable
Hemifusion-like Intermediate
Even though the loss of a lipid marker during the two-step
fusion precludes unambiguous determination of the site of
subsequent content release, the latter step appears to occur
in endosomes. First, the rates of sequential lipid and content
transfer for the two-step fusion were statistically indistinguish-
able (p > 0.180 and p > 0.594, respectively) from the respective
rates of separate surface and endosomal fusion events
(Figure 4B). Hence, by analogy to endosomal fusion, the
content release through the two-step events likely occurs in en-
dosomes. This result also implies that the two-step events are
a subset of ‘‘regular’’ fusion events, in which lipid transfer
occurred prior to virus uptake. Second, the pronounced delay
between lipid and content transfer during the two-step events
(half-time of about 10 min, Figure 4B) was sufficiently long to
permit virus endocytosis (t1/2= 13.5 min, Figure 1A) prior to
content release.
A
CD
EF
B
Figure 3. Fusion of HIV Core-Based Pseudoviruses and of Infectious HIV-1 with TZM-bl Cells
(A) The efficiency of lipid mixing with the plasma membrane, the viral content release from endosomes, and the sequential two-step fusion events mediated by
JRFL Env (pseudotyped with MLV or HIV core), HXB2 Env, and infectious R9 viruses. The first row is the negative control for JRFL fusion using HeLa-CD4 cells
lacking CCR5. The number of respective fusion-related events was normalized to the total number of cell-associated double-labeled virions at the beginning of
theexperiment.Theextentsofvirallipidandcontentmixinginthepresenceof4mMC52Lareshowninparentheses.Therarefalse-positiveeventsinthepresence
ofC52Lwereusuallydelayedrelative tothoseintheabsenceoftheinhibitor(data notshown) andwereminimizedbylimiting thedurationofimagingexperiments.
ND, not determined.
(B) The fraction of cells supporting viral content release was measured by imaging (orange bars), and the fraction of infected cells in the same sample was deter-
minedby ab-gal assay(blackbars) after anadditional48hrcultivationinthepresenceofC52L.Thesomewhatlarger fractionof cellssupporting HXB2fusionwas
due to a more efficient binding of HXB2 to target cells compared to JRFL viruses (data not shown). Error bars are SEM (n = 3).
(C and D) Lipid transfer initiated by the infectious R9 HIV-1 labeled with DiD and MA-GFP-CA at the surface of the TZM-bl cell.
(E and F) Complete endosomal fusion of the infectious R9 particle. See Movie S6.
438 Cell 137, 433–444, May 1, 2009 ª2009 Elsevier Inc.

Page 7

Third, by tracking the two-step fusion events, we found that
viruses tended to accelerate (>0.2 mm/s) prior to or at the time
of content release (Figures 5A and 5B). These particles were
thus judged to have entered an endocytic pathway and fused
with endosomes. Only 1 out of 22 particles released approxi-
mately half of its content (Figures 5C and 5D, arrow) without
a significant prior displacement. However, the incomplete
release of viral content shows that, even if this fusion pore was
formedatthecellsurface,itclosedsoonafteropening(schemat-
ically shown by the thick line above Figure 5D). Importantly, viral
content release did not resume until after the onset of fast move-
ment (double arrow and Movie S7) associated with virus uptake.
We also found that the content release during the two-step
events exhibited by infectious HIV-1 usually coincided with the
fast particle movement (Figures 5E and 5F).
These data imply that, even when HIV-1 establishes lipid
continuity with the plasma membrane, it fails to form a fusion
pore that permits the transfer of a content marker. This fusion
phenotype is operationally defined as membrane hemifusion
(Chernomordik and Kozlov, 2005). The temporal separation of
lipid and content transfer events suggests that the two-step
fusion proceeds through a remarkably long-lived hemifusion
intermediate. We cannot rule out the possibility that the lipid
mixing at the cell surface represents a nonproductive pathway,
in which case, distinct Env trimers would be responsible for
subsequent endosomal fusion. However, given the paucity of
Env trimers in HIV-1 particles (Zhu et al., 2006), formation of
more than one fusion complex per virion appears unlikely,
suggesting that hemifusion is a bona fide intermediate of HIV
entry.
Clathrin- and Dynamin-Dependent Endocytosis
Is a Prerequisite for HIV-1 Fusion
To obtain further evidence that HIV-1 is internalized prior to
fusion, we blocked clathrin- and caveolin-mediated endocytosis
by pretreating the TZM-bl cells with dynasore, a small-molecule
inhibitor of the dynamin GTPase activity that prevents the scis-
sion of clathrin-coated pits from the plasma membrane (Macia
et al., 2006). At a concentration that blocked transferrin uptake
(Figure S7A), dynasore diminished virus internalization and
strongly inhibited HIV-1 infection and fusion with TZM-bl and
CEMss cells (Figures 6A and 6B, respectively). As expected for
viruses entering cells via a clathrin-dependent pathway (Sun
et al., 2005), fusion and infection by VSV G pseudoviruses
were suppressed by the drug. The diminished HIV-1 fusion
was not caused bythe downregulation of CD4 or CR expression,
reduction of virus binding, or the compromised ability of Env to
promote fusion in the presence of dynasore (Figures S7B–S7D).
To control against possible adverse effects of dynasore, we
assessed the effect of MiTMAB, a surface-active inhibitor that
blocks dynamin’s interactions with phospholipids (Quan et al.,
2007). The diminished HIV-1 fusion in cells pretreated with
MiTMAB (Figure 6C) supported the essential role of dynamin in
virus entry. By comparison, the small-molecule inhibitor of
Cdc42 GTPase, secramine A (Pelish et al., 2006), augmented
HIV-1 fusion. The specificity of small-molecule dynamin inhibi-
tors was further verified by showing that HIV-1 fusion was sup-
pressed in cells overexpressing the dominant-negative K44A
mutant of dynamin (Damke et al., 1994) (Figure S8). These
results, along with the inhibition of HIV-1 fusion by a hypertonic
medium (Figure 6C) known to inhibit clathrin-mediated endocy-
tosis, suggest that HIV-1 is internalized via a clathrin-dependent
pathway prior to undergoing fusion.
To determine which step of HIV fusion is blocked by dynasore,
we performed single-virus imaging experiments. Viruses bound
to cells pretreated with the drug exhibited highly restricted
mobility compared to control experiments (data not shown),
consistent with inhibited virus endocytosis. Most importantly,
dynasore abolished the viral content release but permitted the
lipid transfer to the plasma membrane (Figure 6G). In the
absence of virus uptake, partial fusion at the cell surface led to
A
B
C
Figure 4. Lipid Mixing at the Cell Surface Precedes the Content
Release from Endosomes
(A) The kinetics of JRFL (circles) and HXB2 (triangles) fusion with TZM-bl cells.
Waitingtimesfromtheshiftto37?Ctothepointoflipid(redsymbols)orcontent
(green symbols) transfer were measured, rank ordered, and plotted as cumu-
lative distributions of the fraction of fused viruses over time.
(B) Comparative kinetics of partial fusion at the cell surface (red circles), endo-
somal fusion (green circles), and sequentiallipid and contenttransferexhibited
during the two-step events (red and green triangles, respectively). The time
intervals between sequential lipid (TL) and content (TC) transfer were ranked
and plotted as cumulative distribution (crosses).
(C) The kinetics of lipid and content mixing during fusion of infectious R9 HIV-1
viruses with TZM-bl cells.
Cell 137, 433–444, May 1, 2009 ª2009 Elsevier Inc. 439

Page 8

AC
BD
EF
Figure 5. Sequential Lipid and Content Transfer Events Exhibited by HIV-1
(A and B) A two-step fusion event exhibiting a short delay between lipid and content transfer.
(C and D) A rare two-step event characterized by stepwise release of viral content. Complete content discharge (double arrow) occurs after the onset of quick
movement (see Movie S7).The predicted dynamics of afusion pore (D) is shown by a thick line above the panel. The initial and final coordinates of particles on 3D
plots are marked by pink crosses and stars, respectively.
(E and F) The two-step fusion of infectious R9 HIV-1 colabeled with DiD and MA-GFP-CA. Changes in fluorescence intensities of viral lipid and content markers
upon incubation with TZM-bl cells at 37?C are smoothed for visual clarity.
440 Cell 137, 433–444, May 1, 2009 ª2009 Elsevier Inc.

Page 9

HIV inactivation, as evidenced by the lack of recovery of the
BlaM signal after the removal of dynasore (Figure S7E). These
results confirm that HIV-1 is unable to fuse at the cell surface
and that the fusion block occurs after the lipid mixing stage.
Endocytosis Reduces the Window of Opportunity
for the Inhibitory Peptide to Bind to Intermediate
Conformations of gp41
Clearance from the cell surface prior to the completion of fusion
can protectHIV-1 against antibodies and inhibitors targeting Env
epitopes that are exposed during the slow fusion reaction. To
test this notion, we sought to reversibly slow down virus uptake
while permitting its interactions with CD4 and CR. After unsuc-
cessful attempts to reversibly arrest virus uptake using different
intervention strategies, we chose to create a temperature-ar-
rested stage (TAS) that has been described before (Henderson
and Hope, 2006; Mkrtchyan et al., 2005). After binding to cells
in the cold, the viruses/cells were incubated for several hours
at a temperature that minimized productive endocytosis and
virus-cell fusion (for details, see the legend to Figure S6). This
intermediate stage was reversible, as fusion quickly ensued
upon raising the temperature (Figures S6A and S6C). At this
A
D
G
EF
BC
Figure
Inhibits HIV-1 Uptake, Fusion, and Infection
(A) TZM-bl cells were pretreated with 80 mM dynasore
and allowed to bind viruses in the cold. Virus uptake
after a 1 hr incubation at 37?C was measured by the
accumulation of intracellular p24 (black bars, n = 6),
asdescribedintheSupplementalExperimentalProce-
dures. The extent of fusion was quantified by the BlaM
assay(darkcyanbars,n=4),andsingle-cycleinfection
was measured by a b-gal assay (orange bars, n = 6).
(B) Inhibition of HXB2 fusion with CEMss cells pre-
treated with dynasore.
(C) Inhibition of HXB2 fusion in TZM-bl cells pretreated
with 0.45 M sucrose, 80 mM MiTMAB, and 15 mM secr-
amine A.
(D) The kinetics of HXB2 escape from 80 mM dynasore
added at indicated times of incubation at 37?C. The
background fusion (?20% of total) in the presence
of dynasore was subtracted from the dynasore curve
to ease the comparison with the C52L and TB curves.
(E) Fusion of wild-type (WT) and Nef-deficient (DNef)
viruses bearing HXB2 Env with TZM-bl cells.
(F) Kinetics of fusion mediated by the wild-type and
cytoplasmic tail-deleted (DCT) JRFL Env.
(G) Pretreatment of TZM-bl cells with dynasore blocks
the HXB2 content (NC-GFP) transfer but permits lipid
transfer (disappearance of DiD) to the plasma
membrane. The loss of the red signal was due in
part (?50%) to dynasore directly quenching the DiD
fluorescence (data not shown).
Unlessindicatedotherwise,errorbarsareSEM(nR3).
6. Blocking the DynaminFunction
stage, a fraction of fusion events became
resistant to CD4- and CR-binding inhibitors,
demonstrating the formation of ternary
complexes, which in turn resulted in the
exposure of the gp41 coiled-coil regions tar-
geted by inhibitory peptides, such as T20
and C34 (Eckert and Kim, 2001). We were thus able to control
the exposure of pretriggered gp41 to C34 by varying the interval
between adding this peptide at the TAS and inducing endocy-
tosis (and fusion) by shifting to 37?C. The longer exposure to
C34 enhanced its inhibitory activity (Figures S6B and S6D),
consistent with the notion that endocytic entry of HIV-1 might
attenuate the potency of this class of fusion inhibitors.
HIV-Endosome Fusion Does Not Rely on an Intact
Cytoskeleton but Depends on Dynamin Activity
The reliance of HIV-1 fusion on endosomal pathways prompted
us to examine the effects of actin- and microtubule-disrupting
agents also known to interfere with endosomal trafficking and
maturation (Bayer et al., 1998). Pretreatment of cells with latrun-
culin A or nocodazole led to a modest reduction of the extent but
not therate of HIV-1 fusion (Figure S9). This finding indicates that
the BlaM signal is not critically affected by actin- and microtu-
bule-dependent processes, in agreement with the previous
work (Campbell et al., 2004).
Dynasore’s ability to quickly block endocytosis (Macia et al.,
2006 and Figure S7F) permitted us to perform time-of-addition
experiments and delineate the step(s) of HIV-1 fusion sensitive
Cell 137, 433–444, May 1, 2009 ª2009 Elsevier Inc. 441

Page 10

to this compound. Dynasore was added after varied times of
virus-cellincubation,and the
compared to that obtained by the C52L and TB protocols. The
loss of sensitivity to inhibitors of virus endocytosis is expected
to occur at the time of the virus’ escape from C52L. Remarkably,
however, HIV-1 escape from dynasore was markedly delayed
and was indistinguishable from its escape from the TB
(Figure 6D). This finding indicates that dynamin plays a role
both in HIV-1 uptake and in virus-endosome fusion.
Next, we asked whether dynamin-2 could augment Env-
mediated fusion with endosomes through its specific binding
to the accessory HIV-1 Nef protein (Pizzato et al., 2007) exposed
to the cytosol as a result of fusion. However, the identical rates
(Figure 6E) and extents (data not shown) of fusion of wild-type
and Nef-deficient viruses did not support this possibility. Thus
dynamin appears to promote HIV-1 fusion indirectly, perhaps
by interacting with effector protein(s) involved in a variety of
cellular processes, such as cytokinesis, membrane trafficking,
cell migration, and adhesion (Kim and Chang, 2006; Kruchten
and McNiven, 2006; Peters et al., 2004). We also tested whether
the unusually long cytoplasmic domain of gp41 is involved in
pore formation or dilation either directly or by interacting with
its cellular partners. Deletion of the cytoplasmic domain did not
considerably alterthekinetics ofviralescapefromtheTBrelative
to its escape from C52L (Figure 6F), implying that this domain is
not essential for HIV-1 fusion.
drug-resistantfusion was
DISCUSSION
Time-resolved imaging of single viruses and differential blocking
of fusion by site-specific and universal inhibitors revealed that
HIV-1 co-opts the endocytic machinery to enter into and fuse
with target cells. By contrast, fusion with the plasma membrane
did not progress beyond the lipid mixing step, suggesting that
endosomal entry is the pathway that leads to productive infec-
tion. Endocytic entry offers several advantages, including the
sheltering of HIV-1 from antibodies and inhibitors targeting inter-
mediate conformations of Env during the unusually slow fusion
reaction. Indeed we found that the delayed virus uptake
increased the potency of the inhibitory C34 peptide. Thus, in
order to efficiently block intracellular fusion events, the next
generation of HIV entry inhibitors must be able to permeate the
cell membrane.
The failure of HIV to fuse with the plasma membrane is in stark
contrast to cell-cell fusion mediated by Env glycoproteins of this
and other pH-independent viruses. While the basis for this
discrepancy is unclear, the much larger number of Env involved
in cell-cell contact compared to a few Env responsible for virus
entry could increase the likelihood of fusion at the cell surface.
Another manifestation of differences between these experi-
mentalsystemsistheprolonged lagbetweenCD4/CR-mediated
HIV-1 uptake and endosomal fusion. By comparison, the forma-
tion of ternary Env-CD4-CR complexes abrogated the lag before
cell-cell fusion(Melikyanetal.,2000;Mkrtchyanetal.,2005).The
delayed endosomal fusion of HIV-1 is indicative of a rate-limiting
step downstream of coreceptor-dependent steps and down-
streamofahemifusion-like intermediate. Slowporeenlargement
is unlikely to contribute to this delay because both single-virus
imaging and the BlaM assay appear to detect relatively small
pores. It is thus possible that the formation of higher-order Env
oligomers and/or gp41 folding into the 6-helix bundle are rate-
limiting for HIV-cell fusion. Alternatively, the lag before fusion
could reflect the time required for HIV-1 delivery into permissive
intracellular compartments, such as late endosomes.
Accumulating evidence suggests that entry of viruses other
than HIV-1 by direct fusion with the plasma membrane is also
disfavored. For instance, pH-independent MLV and respiratory
syncytialvirusappeartoutilizeanendocyticpathwayforinfection
(Beer et al., 2005; Katen et al., 2001; Kolokoltsov et al., 2007).
Infection by several low pH-dependent viruses can be hindered
when fusion with a plasma membrane is forced by acidic pH
(Marsh and Bron, 1997; Matlin et al., 1982; Mothes et al., 2000).
The failure of HIV-1, aMLV, and SFV to progress beyond the lipid
mixing step at the cell surface (Figures 2, 3, and S3) shows that
the block for pH-dependent and pH-independent infection is at
the stage of formation and/or dilation of a fusion pore. The lack
of complete fusion at the cell surface could be due to restrictions
imposedbythecorticalactinorotherfactorspresentinoraround
the plasma membrane. However, the modest effect of actin
depolymerization on viral content delivery (Figure S9) does not
support this notion. An alternative possibility discussed below
is that viruses rely on yet unidentified endosomal factors to
promote complete fusion.
Our data revealed a novel role for dynamins in HIV-1 content
release from endosomes. Why would HIV-1 need cellular factors
to promote fusion? Several lines of evidence suggest that the
formation and enlargement of a fusion pore are the most
energy-intensive steps (reviewed in Melikyan, 2008) that require
the concerted action of several viral proteins. Considering the
low number of Env per virion (Zhu et al., 2006), HIV-1 may not
be able to sustain a fusion pore on its own without cellular
partners. The ability of dynamin to regulate actin remodeling
and/or to associate with membrane-bending proteins (Kruchten
and McNiven, 2006) could provide an additional driving force to
expand pores and permit the release of the HIV-1 core. It is thus
possible that cellular factors involved in membrane trafficking
are responsible for the virus’ strong preference for endocytic
entry.
EXPERIMENTAL PROCEDURES
BlaM Assay for Virus-Cell Fusion
TZM-bl cells (4$104cells/well) were grown overnight in 96-well plates in a
phenol red-free growth medium. Viruses were added to cells at moi 0.7–1.0
and centrifuged at 2095 g, 4?C for 30 min. The cells were washed with
a cold medium to remove free viruses, and fusion was initiated by shifting to
37?C. After an indicated time at 37?C, fusion was stopped by adding C52L
(1mM)or other fusion inhibitors. All samples weremaintained at37?C foratotal
of 90 min (unless indicated otherwise), chilled by briefly placing on ice, loaded
with the CCF2-AM substrate (GeneBLAzer in vivo Detection Kit, Invitrogen),
and incubated overnight at 13.5?C (or as indicated). In the temperature block
protocol, cells were placed on ice until the end of the experiment and then
loaded with the BlaM substrate. Fusion of HIV-1 pseudotyped with VSV G
was carried out as described above but stopped at indicated times by either
treating cells with pronase, placing cells on ice (TB), or adding 50 mM
NH4Cl. The temperature chase experiments were carried out by introducing
C52L peptide after 20 min of incubation, as in the standard fusion protocol.
The BlaM signal was then chased by placing cells on ice either immediately
442 Cell 137, 433–444, May 1, 2009 ª2009 Elsevier Inc.

Page 11

or at the indicated times of incubation at 37?C in the presence of C52L. The
BlaM activity was quantified using Synergy HT fluorescence plate reader
(Bio-Tek Instr., Germany). The extent of virus-cell fusion was determined
from the ratio of blue (440–480 nm) and green (518–538 nm) emission upon
exciting the cells at 405–415 nm. HIV-1 fusion with CEMss cells was carried
out by resuspending the cells in media containing HXB2 pseudoviruses (moi
1–1.5)andcentrifugingat2095g,4?Cfor30min.Thefreeviruseswerewashed
off, and fusion was initiated by shifting to 37?C and stopped after indicated
times either by adding C52L peptide or by placing cells on ice. The cells
were then loaded with the CCF2-AM substrate and transferred into a 96-well
plate (1$105cells/well), and the BlaM activity was determined as described
above.
Single-Particle Imaging and Analysis
Viruses were centrifuged onto TZM-bl cells cultured on a cover glass (2095 g
for 1 hr at 12?C). The cells were washed to remove unbound viruses and trans-
ferred into an imaging chamber. Virus-cell fusion was triggered by quickly and
locally raising the temperature to 37?C using a home-built temperature-jump
setup (Melikyan et al., 2000) and visualized using a Zeiss LSM 510 Meta
confocal microscope. Unless noted otherwise, samples were simultaneously
excited at 488 and 633 nm, and the emitted light was collected by a C-Apo
403/1.2 water immersion or a Neofluar 403/1.3 oil immersion objective, split,
and passed through 505–550 nm band-pass and 650 nm long-pass filters. To
minimize photobleaching, only 3–4 Z stacks spaced by 2.5 mm were acquired
every6–7sfor 35–40min.Single-particletrackingwasperformedusingVoloc-
ity image analysis software (Improvision, Perkin Elmer). Briefly, the total
number of cell-associated viruses containing detectable amounts of DiD
and GFP-based content marker was determined by identifying contiguous
pixels with fluorescence intensity at least 3-fold greater than background.
The waiting times for fusion were estimated as the time interval from raising
thetemperature to37?Ctothepoint whenthesignal fromeitherthemembrane
or content marker dropped to the background level. Three-dimensional
tracking of particles over time was performed by adjusting the intensity
threshold and the maximal particle displacement between consecutive
frames.
SUPPLEMENTAL DATA
Supplemental Data include Supplemental Experimental Procedures, nine
figures, and seven movies and can be found with this article online at http://
www.cell.com/supplemental/S0092-8674(09)00268-2.
ACKNOWLEDGMENTS
The authors are grateful to Dr. T. Kirchhausen for dynasore and to Drs. C.
Aiken, M. Alizon, J. Binley, J. Cunningham, M. Kielian, G. Lewis, M. Lu, W.
Mothes, J. Strizki, L. Wang, and J. Young, as well as the NIH AIDS Research
and Reference Reagent Program for reagents, expression vectors, and cell
lines.WethankDr.M.Reitz forhis help indesigning theMA-GFP-CAconstruct
and Drs. L. Chernomordik and R. Dutch for stimulating discussions. This work
was supported by NIH R01 GM054787 and AI053668 grants to G.B.M. Y.K.
was partially supported by the SDG from the American Heart Association.
Received: October 17, 2008
Revised: January 3, 2009
Accepted: February 25, 2009
Published: April 30, 2009
REFERENCES
Aiken, C. (1997). Pseudotyping human immunodeficiency virus type 1 (HIV-1)
by the glycoprotein of vesicular stomatitis virus targets HIV-1 entry to an endo-
cytic pathway and suppresses both the requirement for Nef and the sensitivity
to cyclosporin A. J. Virol. 71, 5871–5877.
Bayer, N., Schober, D., Prchla, E., Murphy, R.F., Blaas, D., and Fuchs, R.
(1998). Effect of bafilomycin A1 and nocodazole on endocytic transport in
HeLa cells: implications for viral uncoating and infection. J. Virol. 72,
9645–9655.
Beer, C., Andersen, D.S., Rojek, A., and Pedersen, L. (2005). Caveola-depen-
dent endocytic entry of amphotropic murine leukemia virus. J. Virol. 79,
10776–10787.
Brandt, S.M., Mariani, R., Holland, A.U., Hope, T.J., and Landau, N.R. (2002).
Association of chemokine-mediated block to HIV entry with coreceptor inter-
nalization. J. Biol. Chem. 277, 17291–17299.
Campbell, E.M., Nunez, R., and Hope, T.J. (2004). Disruption of the actin
cytoskeleton can complement the ability of Nef to enhance human immunode-
ficiency virus type 1 infectivity. J. Virol. 78, 5745–5755.
Cavrois,M.,DeNoronha, C., and Greene,W.C. (2002).A sensitive and specific
enzyme-based assay detecting HIV-1 virion fusion in primary T lymphocytes.
Nat. Biotechnol. 20, 1151–1154.
Cavrois, M., Neidleman, J., Yonemoto, W., Fenard, D., and Greene, W.C.
(2004). HIV-1 virion fusion assay: uncoating not required and no effect of Nef
on fusion. Virology 328, 36–44.
Chernomordik,L.V.,andKozlov,M.M.(2005).Membranehemifusion:crossing
a chasm in two leaps. Cell 123, 375–382.
Daecke, J., Fackler, O.T., Dittmar, M.T., and Krausslich, H.G. (2005). Involve-
ment of clathrin-mediated endocytosis in human immunodeficiency virus type
1 entry. J. Virol. 79, 1581–1594.
Damke, H., Baba, T., Warnock, D.E., and Schmid, S.L. (1994). Induction
of mutant dynamin specifically blocks endocytic coated vesicle formation.
J. Cell Biol. 127, 915–934.
Deng, Y.,Zheng, Q.,Ketas, T.J., Moore, J.P., and Lu, M. (2007). Protein design
of a bacterially expressed HIV-1 gp41 fusion inhibitor. Biochemistry 46,
4360–4369.
Doms, R.W.,and Trono, D. (2000). The plasma membrane as acombat zone in
the HIV battlefield. Genes Dev. 14, 2677–2688.
Eckert, D.M., and Kim, P.S. (2001). Mechanisms of viral membrane fusion and
its inhibition. Annu. Rev. Biochem. 70, 777–810.
Fredericksen, B.L., Wei, B.L., Yao, J., Luo, T., and Garcia, J.V. (2002). Inhibi-
tion of endosomal/lysosomal degradation increases the infectivity of human
immunodeficiency virus. J. Virol. 76, 11440–11446.
Frey,S.,Marsh,M.,Gunther,S.,Pelchen-Matthews,A.,Stephens,P.,Ortlepp,S.,
andStegmann,T.(1995).Temperaturedependenceofcell-cellfusioninducedby
the envelope glycoprotein of human immunodeficiency virus type 1. J. Virol. 69,
1462–1472.
Gallay, P., Hope, T., Chin, D., and Trono, D. (1997). HIV-1 infection of nondi-
viding cells through the recognition of integrase by the importin/karyopherin
pathway. Proc. Natl. Acad. Sci. USA 94, 9825–9830.
Henderson, H.I., and Hope,T.J. (2006). The temperature arrestedintermediate
of virus-cell fusion is a functional step in HIV infection. Virol. J. 3, 36.
Katen,L.J.,Januszeski,M.M.,Anderson,W.F.,Hasenkrug,K.J.,andEvans,L.H.
(2001). Infectious entry by amphotropic as well as ecotropic murine leukemia
viruses occurs through an endocytic pathway. J. Virol. 75, 5018–5026.
Kim, Y.,andChang,S.(2006). Ever-expanding networkof dynamin-interacting
proteins. Mol. Neurobiol. 34, 129–136.
Kolokoltsov,A.A.,Deniger,D.,Fleming,E.H.,Roberts,N.J.,Jr.,Karpilow,J.M.,
and Davey, R.A. (2007). Small interfering RNA profiling reveals key role of
clathrin-mediated endocytosis and early endosome formation for infection
by respiratory syncytial virus. J. Virol. 81, 7786–7800.
Kruchten, A.E., and McNiven, M.A. (2006). Dynamin as a mover and pincher
during cell migration and invasion. J. Cell Sci. 119, 1683–1690.
Lakadamyali, M., Rust, M.J., and Zhuang, X. (2004). Endocytosis of influenza
viruses. Microbes Infect. 6, 929–936.
Macia,E.,Ehrlich,M.,Massol,R.,Boucrot,E.,Brunner,C.,andKirchhausen,T.
(2006).Dynasore,acell-permeableinhibitorofdynamin.Dev.Cell10,839–850.
Maddon, P.J., McDougal, J.S., Clapham, P.R., Dalgleish, A.G., Jamal, S.,
Weiss, R.A., and Axel, R. (1988). HIV infection does not require endocytosis
of its receptor, CD4. Cell 54, 865–874.
Cell 137, 433–444, May 1, 2009 ª2009 Elsevier Inc. 443

Page 12

Marechal, V.,Prevost,M.C., Petit,C., Perret, E.,Heard,J.M.,and Schwartz, O.
(2001). Human immunodeficiency virus type 1 entry into macrophages medi-
ated by macropinocytosis. J. Virol. 75, 11166–11177.
Markosyan, R.M., Cohen, F.S., and Melikyan, G.B. (2005). Time-resolved
imaging of HIV-1 Env-mediated lipid and content mixing between a single
virion and cell membrane. Mol. Biol. Cell 16, 5502–5513.
Marsh, M., and Bron, R. (1997). SFV infection in CHO cells: cell-type specific
restrictionstoproductivevirusentryatthecellsurface.J.CellSci.110,95–103.
Marsh, M., and Helenius, A. (2006). Virus entry: open sesame. Cell 124,
729–740.
Matlin, K.S., Reggio, H., Helenius, A., and Simons, K. (1982). Pathway of vesi-
cular stomatitis virus entry leading to infection. J. Mol. Biol. 156, 609–631.
McClure, M.O., Marsh, M., and Weiss, R.A. (1988). Human immunodeficiency
virus infection of CD4-bearing cells occurs by a pH-independent mechanism.
EMBO J. 7, 513–518.
Melikyan, G.B. (2008). Common principles and intermediates of viral protein-
mediated fusion: the HIV-1 paradigm. Retrovirology 5, 111.
Melikyan,G.B.,Markosyan,R.M.,Hemmati,H.,Delmedico,M.K.,Lambert,D.M.,
andCohen,F.S.(2000).EvidencethatthetransitionofHIV-1gp41intoasix-helix
bundle,notthebundleconfiguration,inducesmembranefusion.J.CellBiol.151,
413–424.
Mkrtchyan, S.R., Markosyan, R.M., Eadon, M.T., Moore, J.P., Melikyan, G.B.,
and Cohen, F.S. (2005). Ternary complex formation of human immunodefi-
ciencyvirustype1Env,CD4,andchemokine receptor capturedasaninterme-
diate of membrane fusion. J. Virol. 79, 11161–11169.
Mothes, W., Boerger, A.L., Narayan, S., Cunningham, J.M., and Young, J.A.T.
(2000). Retroviral entry mediated by receptor priming and low pH triggering of
an envelope glycoprotein. Cell 103, 679–689.
Pauza, C.D., and Price, T.M. (1988). Human immunodeficiency virus infection
of T cells and monocytes proceeds via receptor-mediated endocytosis. J. Cell
Biol. 107, 959–968.
Pelchen-Matthews, A., Clapham, P., and Marsh, M. (1995). Role of CD4 endo-
cytosis in human immunodeficiency virus infection. J. Virol. 69, 8164–8168.
Pelish,H.E.,Peterson,J.R.,Salvarezza,S.B.,Rodriguez-Boulan,E.,Chen,J.L.,
Stamnes,M.,Macia,E.,Feng,Y.,Shair,M.D.,andKirchhausen,T.(2006).Secr-
amine inhibits Cdc42-dependent functions in cells and Cdc42 activation
in vitro. Nat. Chem. Biol. 2, 39–46.
Peters, C., Baars, T.L., Buhler, S., and Mayer, A. (2004). Mutual control of
membrane fission and fusion proteins. Cell 119, 667–678.
Pizzato, M., Helander, A., Popova, E., Calistri, A., Zamborlini, A., Palu, G., and
Gottlinger, H.G. (2007). Dynamin 2 is required for the enhancement of HIV-1
infectivity by Nef. Proc. Natl. Acad. Sci. USA 104, 6812–6817.
Quan, A., McGeachie, A.B., Keating, D.J., van Dam, E.M., Rusak, J., Chau, N.,
Malladi, C.S., Chen, C., McCluskey, A., Cousin, M.A., and Robinson, P.J.
(2007). Myristyl trimethyl ammonium bromide and octadecyl trimethyl
ammonium bromide are surface-active small molecule dynamin inhibitors
that block endocytosis mediated by dynamin I or dynamin II. Mol. Pharmacol.
72, 1425–1439.
Schaeffer, E., Soros, V.B., and Greene, W.C. (2004). Compensatory link
between fusion and endocytosis of human immunodeficiency virus type 1 in
human CD4 T lymphocytes. J. Virol. 78, 1375–1383.
Stein, B.S., Gowda, S.D., Lifson, J.D., Penhallow, R.C., Bensch, K.G., and
Engleman, E.G. (1987). pH-independent HIV entry into CD4-positive T cells
via virus envelope fusion to the plasma membrane. Cell 49, 659–668.
Sun, X., Yau, V.K., Briggs, B.J., and Whittaker, G.R. (2005). Role of clathrin-
mediated endocytosis during vesicular stomatitis virus entry into host cells.
Virology 338, 53–60.
Wei, B.L., Denton, P.W., O’Neill, E., Luo, T., Foster, J.L., and Garcia, J.V.
(2005). Inhibition of lysosome and proteasome function enhances human
immunodeficiency virus type 1 infection. J. Virol. 79, 5705–5712.
Wei, X., Decker, J.M., Liu, H., Zhang, Z., Arani, R.B., Kilby, J.M., Saag, M.S.,
Wu, X., Shaw, G.M., and Kappes, J.C. (2002). Emergence of resistant human
immunodeficiency virus type 1 in patients receiving fusion inhibitor (T-20)
monotherapy. Antimicrob. Agents Chemother. 46, 1896–1905.
Yoder, A., Yu, D., Dong, L., Iyer, S.R., Xu, X., Kelly, J., Liu, J., Wang, W.,
Vorster, P.J., Agulto, L., et al. (2008). HIV envelope-CXCR4 signaling activates
cofilin to overcome cortical actin restriction in resting CD4 T cells. Cell 134,
782–792.
Zhu, P., Liu, J., Bess, J., Jr., Chertova, E., Lifson, J.D., Grise, H., Ofek, G.A.,
Taylor, K.A., and Roux, K.H. (2006). Distribution and three-dimensional struc-
ture of AIDS virus envelope spikes. Nature 441, 847–852.
444 Cell 137, 433–444, May 1, 2009 ª2009 Elsevier Inc.