*For correspondence:
bjorkman@caltech.edu
Competing interest:
See
page 14
Funding:
See page 14
Received:
29 April 2020
Accepted:
21 July 2020
Published:
22 July 2020
Reviewing editor:
Axel T
Brunger, Stanford University,
United States
Copyright Ladinsky et al. This
article is distributed under the
terms of the
Creative Commons
Attribution License,
which
permits unrestricted use and
redistribution provided that the
original author and source are
credited.
Electron tomography visualization of HIV-
1 fusion with target cells using fusion
inhibitors to trap the pre-hairpin
intermediate
Mark S Ladinsky
1
, Priyanthi NP Gnanapragasam
1
, Zhi Yang
1
, Anthony P West
1
,
Michael S Kay
2
, Pamela J Bjorkman
1
*
1
Division of Biology and Biological Engineering, California Institute of Technology,
Pasadena, United States;
2
Department of Biochemistry, University of Utah School of
Medicine, Salt Lake City, United States
Abstract
Fusion of HIV-1 with the membrane of its target cell, an obligate first step in virus
infectivity, is mediated by binding of the viral envelope (Env) spike protein to its receptors, CD4
and CCR5/CXCR4, on the cell surface. The process of viral fusion appears to be fast compared with
viral egress and has not been visualized by EM. To capture fusion events, the process must be
curtailed by trapping Env-receptor binding at an intermediate stage. We have used fusion
inhibitors to trap HIV-1 virions attached to target cells by Envs in an extended pre-hairpin
intermediate state. Electron tomography revealed HIV-1 virions bound to TZM-bl cells by 2–4
narrow spokes, with slightly more spokes present when evaluated with mutant virions that lacked
the Env cytoplasmic tail. These results represent the first direct visualization of the hypothesized
pre-hairpin intermediate of HIV-1 Env and improve our understanding of Env-mediated HIV-1 fusion
and infection of host cells.
Introduction
The first step of HIV-1 entry into a host target cell, fusion between the viral and target cell mem-
branes, is mediated by the viral envelope spike protein (Env). HIV-1 Env is a trimeric glycoprotein
comprising three gp120 subunits that contain host receptorbinding sites and three gp41 subunits
that include the fusion peptide and membrane-spanning regions. Binding of the primary receptor
CD4 to gp120 triggers conformational changes that expose a binding site for co-receptor (CCR5 or
CXCR4). Coreceptor binding results in further conformational changes within gp41 that promote
release of the hydrophobic fusion peptide, its insertion into the host cell membrane, and subsequent
fusion of the host cell and viral membrane bilayers (
Harrison, 2015
).
Structural studies relevant to understanding Env-mediated membrane fusion include X-ray and
single-particle cryo-EM structures of soluble native-like Env trimers in the closed (pre-fusion) confor-
mation (
Ward and Wilson, 2017
), CD4-bound open trimers in which the co-receptor binding site on
the third hypervariable loop (V3) of gp120 is exposed by V1V2 loop rearrangement
(
Ozorowski et al., 2017
;
Wang et al., 2018
;
Wang et al., 2016
;
Yang et al., 2019
), a gp120 mono-
meric core-CD4-CCR5 complex (
Shaik et al., 2019
), and a post-fusion gp41 six-helical bundle
formed by an
a
-helical trimeric coiled coil from the gp41 N-trimer region surrounded by three heli-
ces from the C-peptide region (
Chan et al., 1997
;
Weissenhorn et al., 1997
;
Figure 1a
). Prior to
membrane fusion and formation of the post-fusion gp41 helical bundle, the viral and host cell mem-
branes are hypothesized to be linked by an extended pre-hairpin intermediate in which insertion of
the gp41 fusion peptide into the host cell membrane exposes the N-trimer (HR1) region of gp41
Ladinsky
etal
. eLife 2020;9:e58411.
DOI: https://doi.org/10.7554/eLife.58411
1 of 18
RESEARCH ARTICLE
(
Chan and Kim, 1998
). Formation of the six-helical bundle and subsequent fusion can be inhibited
by targeting the N-trimer region with C-peptide-based inhibitors; for example the fusion inhibitor
T20 (enfuvirtide [Fuzeon]) (
Kilgore et al., 2003
), T1249, a more potent derivative of T20 (
Eron et al.,
2004
), and a highly potent trimeric D-peptide (CPT31) (
Redman et al., 2018
), or with anti-gp41 anti-
bodies such as D5 (
Miller et al., 2005
;
Figure 1a
).
Visualizing the pre-hairpin intermediate that joins the host and viral membranes has not been
straightforward. Despite 3-D imaging by electron tomography (ET) of HIV-1 infection of cultured
cells (
Carlson et al., 2008
;
Carlson et al., 2010
;
Do et al., 2014
;
Earl et al., 2013
) and tissues
(
Kieffer et al., 2017a
;
Kieffer et al., 2017b
;
Ladinsky et al., 2019
;
Ladinsky et al., 2014
), viruses
caught in the act of fusion have not been unambiguously found. In our ET imaging of HIV-1–infected
humanized mouse tissues, we have identified hundreds of budding virions at various stages of egress
and thousands of free mature and immature virions (
Kieffer et al., 2017a
;
Kieffer et al., 2017b
;
Ladinsky et al., 2019
;
Ladinsky et al., 2014
), but not a single example of a virus attached to a host
cell via a pre-hairpin intermediate or in the process of fusing its membrane with the target cell mem-
brane. The absence of observed viral fusion events might be explained if fusion is a fast process
compared with viral budding; thus when cells or tissues are immobilized for EM or ET, the relatively
slow process of viral budding would be more easily captured compared with the faster process of
fusion. We assume that fusion could theoretically be observed if a virus were caught at exactly the
right time, but this might require examining thousands or millions of images.
Here, we report visualizing the pre-hairpin intermediate by ET after treatment of HIV-1–exposed
target cells with inhibitors of six-helix bundle formation that bind the N-trimer region of gp41 that is
exposed during the fusion process. Using optimally preserved samples for ET with a nominal
resolution
~
7 nm, we found >100 examples of HIV-1 virions linked to TZM-bl target cells by 2–4 nar-
row rods of density (spokes) in inhibitor-treated samples, but none in untreated or control-treated
samples. The approximate dimensions of the majority of the spokes (
Ausubel et al., 1989
) matched
models of gp41-only pre-hairpin intermediates in which the Env gp120 subunit had been shed. The
average number of observed spokes connecting a virion to a target cell increased slightly when
using a virus containing an Env with a cytoplasmic tail deletion, suggesting that the increased lateral
mobility of cytoplasmic tail-deleted Envs in the viral membrane (
Crooks et al., 2008
;
Roy et al.,
2013
;
Pezeshkian et al., 2019
) allowed more Envs to join the interaction with the target cell. We
discuss the implications of these studies for understanding HIV-1 Env-mediated membrane fusion
and how these results differ from a previous ET study of the ‘entry claw’ that is formed upon HIV-1
or SIV interactions with target cells (
Sougrat et al., 2007
).
Results
Experimental design
A previous study used ET to visualize HIV-1 and SIV virions in contact with target cells after promot-
ing a temperature-arrested state (
Mkrtchyan et al., 2005
) in which viruses can remain attached to
cells prior to fusion (
Sougrat et al., 2007
). For that study, target cells were incubated with virus at 4 ̊
C to allow binding but not fusion, warmed to 37 ̊C, and then fixed after incubations ranging from 15
min to 3 hr (
Sougrat et al., 2007
). At all time points after warming, viruses were found attached to
target cells by a cluster of 5–7 ‘rods,’ each
~
100 A
̊
long and
~
100 A
̊
wide. The fact that the attach-
ment structure was not found when the viruses and target cells were incubated in the presence of
C34, a gp41 N-trimer–targeting C-peptide inhibitor related to T20 (
Sougrat et al., 2007
), suggests
that the rod structure that was trapped during the temperature-arrested state did not involve the
pre-hairpin intermediate.
We hypothesized that addition of an HIV-1 fusion inhibitor that binds to the exposed gp41 N-tri-
mer after host cell receptor and coreceptor binding would slow or stop virus-host cell membrane
fusion such that we could visualize pre-hairpin intermediate structures by ET (
Figure 1b
). We charac-
terized three fusion inhibitors of different sizes and potencies that target the exposed gp41 N-trimer
region of the pre-hairpin intermediate for attempts to visualize the pre-hairpin intermediate: T1249-
Fc, a C-peptide–based inhibitor that we linked to human Fc (MW = 65 kDa), D5 IgG (MW = 150
kDa) (
Miller et al., 2005
), and CPT31, a high-affinity D-peptide inhibitor linked to cholesterol
(MW = 9 kDa) (
Redman et al., 2018
;
Welch et al., 2010
;
Figure 1a
;
Figure 1—figure supplement
Ladinsky
etal
. eLife 2020;9:e58411.
DOI: https://doi.org/10.7554/eLife.58411
2 of 18
Research article
Cell Biology
Microbiology and Infectious Disease
1a
). We measured their neutralization potencies using in vitro HIV-1 pseudovirus neutralization
assays (
Montefiori, 2009
) against the SC4226618 and 6535 viral strains. We found potencies rang-
ing from 50% inhibitory concentration (IC
50
) values of
~
0.13 ng/mL for CPT31 to
40
m
g/mL for D5
Figure 1.
HIV-1 Env-mediated fusion between viral and host cell membranes. (
a
) Top: Schematics of host receptors, HIV-1 Env trimer, pre-hairpin
intermediate, and fusion inhibitors. Bottom: steps in fusion: (i) Closed, prefusion structure of HIV-1 Env trimer in which the V1V2 loops (orange) occlude
the coreceptor-binding site on V3 (blue) (e.g., PDB code 5CEZ). The Env trimer is embedded in the viral membrane, while the host receptor (CD4) and
coreceptor (CCR5) are embedded in the target cell membrane. (ii) CD4-bound open HIV-1 Env trimer in which V1V2 loops have been displaced to
expose the coreceptor-binding site on V3 (e.g. PDB 6U0L). (iii) Hypothetical CD4- and CCR5-bound open Env trimer with rearrangements of gp41
N-trimer/HR1 to form a pre-hairpin intermediate structure that is linked to the target cell membrane by the gp41 fusion peptide (red). (iv) Hypothetical
pre-hairpin intermediate formed by gp41 trimer after shedding of gp120s. (v–vi) Formation of the post-fusion gp41 six-helical bundle (e.g. PDB 1GZL)
that juxtaposes the host cell and viral membranes (step v) for subsequent membrane fusion (step vi). (
b
) Approximate binding sites (red circles) for
fusion inhibitors shown on schematics of steps iii and iv (panel a). Entry inhibitor binding sites might be partially sterically occluded for binding to the
T1249-Fc or D5 fusion inhibitors. Schematics shown above as models are from PDB codes 6U0L and 1AIK with approximate dimensions indicated. (
c
)
Schematic illustrating why fewer HIV-1 Envs might be involved in attaching to a target cell when the attachment site is flat versus a concave surface.
Top: attachment site (described here) formed during a 37 ̊C incubation of virions, target cells, and a fusion inhibitor. Bottom: attachment site (described
in
Sougrat et al., 2007
) formed in the absence of a fusion inhibitor when virions and target cells were incubated in a temperature jump protocol (4 ̊C
incubation followed by warming to 37 ̊C).
The online version of this article includes the following figure supplement(s) for figure 1:
Figure supplement 1.
Characterization of fusion inhibitors and viral infectivity.
Ladinsky
etal
. eLife 2020;9:e58411.
DOI: https://doi.org/10.7554/eLife.58411
3 of 18
Research article
Cell Biology
Microbiology and Infectious Disease
IgG (
Figure 1—figure supplement 1a
). T1249-Fc exhibited intermediate potencies (IC
50
s = 0.99
m
g/
mL; 17
m
g/mL) (
Figure 1—figure supplement 1a
), higher than IC
50
s measured for T1249 peptide
alone, consistent with limited steric accessibility resulting in decreased potencies for larger fusion
inhibitors (
Hamburger et al., 2005
).
Incubating with a fusion inhibitor at 37 ̊C obviated the need for a 4 ̊C incubation of virus and tar-
get cells, which we reasoned was desirable since low temperatures alter membrane fluidity
Figure 2.
Identification of attached HIV-1 virions. (
a
) Montaged projection overview of a field of cultured TZM-bl
cells from a 400 nm section. Note extensive blebbing and surface projections that are typical of the cell type.
Inset: Projection detail of a HIV-1 virion adjacent to TZM-bl cell surface. (
b
) Slice (5.6 nm) from a tomographic
reconstruction of the virion shown in the inset of panel a (from a dataset collected with the T1249-Fc inhibitor). The
bullet-shaped core identifies the particle as mature HIV-1 (see also
Figure 3—figure supplement 1
). Two pre-
hairpin intermediate ‘spokes’ (red arrowheads) attach the virion to the cell surface. (
c
) 3-D isosurface rendering of
the spokes shown in panel b. (
d
) Examples of extra densities observed in some data sets collected using the D5
IgG inhibitor. These appear as ‘hook-like’ structures projecting from the sides of spokes, adjacent to the virion
surface, which are visible in two sequential tomographic slices (small and large black arrows). Extra densities may
represent portions of D5 IgGs attached to the prehairpin intermediate. Similar densities were not seen in
experiments with the T1249-Fc or CPT31 inhibitors.
The online version of this article includes the following figure supplement(s) for figure 2:
Figure supplement 1.
Confirmation of pseudovirions in tomograms and experimental controls.
Ladinsky
etal
. eLife 2020;9:e58411.
DOI: https://doi.org/10.7554/eLife.58411
4 of 18
Research article
Cell Biology
Microbiology and Infectious Disease
(
Avery et al., 1995
;
Simons and Vaz, 2004
;
Quinn, 1988
), which could affect one or more steps in
membrane fusion. Since target cells for HIV-1 are several microns in height, much thicker than the
0.5–1
m
m limit for cryo-ET (
Beck and Baumeister, 2016
), we used stained, plastic-embedded sam-
ples that could be cut into 300–400 nm sections using a microtome, and then examined the samples
in 3-D using ET. Although ET of stained, plastic-embedded sections results in lower resolution than
cryo-ET, the minimal effects of radiation damage in plastic sections (
Glaeser, 2016
) was an advan-
tage for locating rare attached virions. Thus, more cells could be assayed in plastic sections than in
samples prepared by cryo-ET methods (e.g. by examining thin leading edges of cells or using
focused-ion-beam milling [
Villa et al., 2013
] to prepare a sufficiently thin sample), therefore allowing
for statistically significant observations of virion attachment events. We prepared samples by light
fixation followed by high-pressure freezing/freeze substitution fixation (HPF-FSF) instead of the tradi-
tional chemical fixation protocol used previously (
Sougrat et al., 2007
) because HPF vitrifies cells
at
~
10,000 ̊/s, stopping all cellular movement within ms and allowing optimal preservation of ultra-
structural features (
Kellenberger, 1991
;
McIntosh et al., 2005
;
Sartori et al., 1993
;
Dahl and Stae-
helin, 1989
). By contrast, chemical fixation immobilizes elements in the cell at different rates, and
movement and rearrangement of transmembrane proteins may continue even in the presence of
aldehyde fixatives (
Brock et al., 1999
;
Stanly et al., 2016
;
Tanaka et al., 2010
). Following HPF-FSF,
samples were plastic embedded and stained with uranyl acetate and lead citrate as described in our
previous ET studies of HIV-1 in infected tissues (
Kieffer et al., 2017b
;
Ladinsky et al., 2019
;
Ladinsky et al., 2014
). Since biosafety requirements for the current study necessitated the use of
HIV-1 pseudoviruses instead of infectious HIV-1, we verified that the ultrastructure of HIV-1 pseudo-
viruses, including approximate numbers and dimensions of Env trimer spikes and the presence of
collapsed (in mature virions) versus C-shaped (in immature virions) cores (
Carlson et al., 2008
;
Benjamin et al., 2005
;
Ganser, 1999
;
Wright et al., 2007
), was preserved during the fixation,
embedding, and staining procedures (
Figure 2—figure supplement 1a
) consistent with our previous
publications involving ET of infectious HIV-1 in tissue samples (
Kieffer et al., 2017b
;
Ladinsky et al.,
2019
;
Ladinsky et al., 2014
). These results are also consistent with previous direct comparisons of
tomograms of stained and plastic-embedded versus unstained and cryopreserved SIV virions
(
Sougrat et al., 2007
).
We conducted ET experiments by first incubating TZM-bl cells, a HeLa cell line that stably
expresses high levels of human CD4 and coreceptors CCR5 and CXCR4 (
Platt et al., 1998
), with 130
m
g/mL of inhibitor (either T1249-Fc, D5, or CPT31) and
~
5000 TCID
50
/mL of HIV-1 pseudovirus at 37
̊C for 2, 4, or 48 hr, followed by HPF, FSF, plastic embedding, sectioning, and visualization by ET. In
order to verify that results were not dependent upon a particular viral strain, we used pseudoviruses
derived from two primary isolate HIV-1 strains: SC4226618 (Tier 2) and 6535 (Tier 1B) (
Li et al.,
2005
), chosen for their sensitivity to the fusion inhibitors and because we had both wild-type and
Env cytoplasmic tail-deleted forms of the 6535 pseudovirus (
Figure 1—figure supplement 1a
).
TZM-bl cells are contaminated with ecotropic murine leukemia virus (
Takeuchi et al., 2008
), which
does not affect their use for HIV-1 in vitro neutralization assays (
Platt et al., 2009
). In our surveys of
TZM-bl cells incubated in the presence or absence of fusion inhibitors, we occasionally observed
budding MLV virions (
Figure 2—figure supplement 1b
). As MLV serves as a control for non-specific
inhibition in TZM-bl–based HIV-1 in vitro neutralization assays (
Montefiori, 2005
), the fusion inhibi-
tors used in our experiments are known to have no effect on MLV fusion, thus contaminating MLV
virions were not captured during fusion in our experiments.
To identify attached virions by EM (
Figure 2
;
Figure 2—figure supplement 1
), the peripheries of
TZM-bl cells were scanned to locate roughly spherical objects with diameters
~
100 nm that were
near a cell surface. Regions of interest were then examined at higher magnification and at tilts of 0 ̊,
35 ̊ and