Distribution of ESCRT Machinery at HIV Assembly Sites Reveals Virus Scaffolding of ESCRT Subunits

See allHide authors and affiliations

Science  07 Feb 2014:
Vol. 343, Issue 6171, pp. 653-656
DOI: 10.1126/science.1247786


The ESCRT (Endosomal Sorting Complex Required for Transport) protein complex plays a role in budding into multivesicular bodies, cytokinesis, and HIV budding, but the details of how the ESCRTs facilitate viral budding are unclear. Now, using high-resolution light and electron microscopical imaging techniques, Van Engelenburg et al. (p. 653, published online 16 January) dissect the role for ESCRT proteins in HIV budding. The findings suggest that the ESCRT machinery required for the scission of HIV particles from infected cells is located within the core of the virus particle and not, as might have been expected based on previous work, on the cellular side of the membrane scission event involved in viral budding.


The human immunodeficiency virus (HIV) hijacks the endosomal sorting complexes required for transport (ESCRT) to mediate virus release from infected cells. The nanoscale organization of ESCRT machinery necessary for mediating viral abscission is unclear. Here, we applied three-dimensional superresolution microscopy and correlative electron microscopy to delineate the organization of ESCRT components at HIV assembly sites. We observed ESCRT subunits localized within the head of budding virions and released particles, with head-localized levels of CHMP2A decreasing relative to Tsg101 and CHMP4B upon virus abscission. Thus, the driving force for HIV release may derive from initial scaffolding of ESCRT subunits within the viral bud interior followed by plasma membrane association and selective remodeling of ESCRT subunits.

One of the key host factors human immunodeficiency virus (HIV) relies on to complete its infection cycle is the endosomal sorting complexes required for transport (ESCRT). By recruiting this machinery, HIV is able to mediate the final step of virus particle fission from the membrane. Otherwise, virus egress is severely inhibited (15). Despite great progress in demonstrating the ESCRT machinery’s role in mediating HIV abscission, the nanoscale organization, and thus function, of ESCRT subcomplexes at native HIV assembly sites remains poorly defined.

ESCRT machinery participates in important cellular membrane remodeling events, such as cytokinesis and multivesicular body biogenesis (MVB), organizing into three general sets of subcomplexes. One set includes ESCRT-0, -I, and -II, which interact directly with cargo or scaffold proteins to direct assembly of downstream ESCRT complexes. Another set includes ESCRT-III subunits, which are thought to polymerize on membranes into a helical architecture to constrict and abscise membrane buds formed by early ESCRT/cargo complexes (6, 7). The final set consists of the AAA+ adenosine triphosphatases (ATPases) Vps4A and Vps4B, which are thought to be necessary for HIV abscission through ESCRT-III filament remodeling and/or recycling (79). Together, the three sets of ESCRT subcomplexes cooperate to drive cellular membrane remodeling. Structural information obtained by means of in vitro assembly (7, 10) and cellular overexpression studies (11) suggest a model of bud formation and abscission in which ESCRT-III filaments encircle and constrict the aperture of a membrane protrusion, acting on the base of the neck, in trans from the protrusion head. On the other hand, when directed by the midbody scaffold during cytokinesis (12, 13) ESCRT-III filaments appear to polymerize from the scaffold and constrict the membrane, acting in cis with respect to the scaffolding structure. Whether ESCRT-III filaments polymerize at HIV bud sites in cis or trans to the scaffolding structure is not known because of the small dimensions of a budding HIV virion (120 to 140 nm diameter). Consequently, it remains unclear how ESCRT machinery acts to mediate viral membrane abscission and propagate HIV infection.

We used interferometric photoactivated-localization microscopy (iPALM) (14) to decipher the three-dimensional (3D) nanoscale organization of ESCRT components at HIV assembly sites and thereby gained insight into the mechanism for viral membrane abscission (supplementary text). ESCRT subcomplexes were selected for our analysis on the basis of their direct interaction with the structural HIV protein Gag [ESCRT-I subunit tumor susceptibility gene 101 (Tsg101)] (15) and their role in HIV membrane abscission (ESCRT-III and Vps4; CHMP2A, CHMP4B, and Vps4A) (16). The ESCRT proteins were modified with either green fluorescent protein (GFP) or photo-switchable cyan fluorescent protein 2 (PSCFP2) and stably expressed in COS7 cells. Stable cell lines expressing fluorescent ESCRT probes were validated for proper function (supplementary text).

Superresolution image analysis of HIV Gag demonstrated the ability of the iPALM method to sensitively locate and resolve subviral details of membrane-enveloped particles emanating from the plasma membrane of expressing cells (Fig. 1 and supplementary text). With the capacity to differentiate subviral protein organization within HIV bud structures, we next performed two-channel iPALM imaging of both HIV Gag and ESCRT subcomplexes. Imaging of HIV Gag and ESCRT-I (PSCFP2-Tsg101) revealed a small cytosolic pool of Tsg101 protein punctuated by clustering at Gag assembly sites (Fig. 2A and fig. S6A). PSCFP2-Tsg101 was localized within the interior of the Gag lattice of assembling particles (Fig. 2D), which is consistent with the direct interaction between the p6 domain of HIV Gag and the UEV domain of Tsg101 (3). PSCFP2-CHMP2A and -CHMP4B (ESCRT-III) cell lines expressing HIV Gag showed high cytosolic pools of PSCFP2 followed by membrane-localized co-clustering with Gag clusters (Fig. 2, A to D, and fig. S6B). When co-clustering was observed, the 3D distribution of ESCRT-III and Gag at assembly sites was similar and distinctly separable from the cytosolic pool of PSCFP2-CHMP2A or -CHMP4B (Fig. 2, A, B, and D). To correlate the apparent observation of ESCRT-III signal residing beyond the plane of the plasma membrane, we visualized the plasma membrane using a lipidated fluorescent probe. PSCFP2-CHMP2A signal was observed in discrete distributions extending away from the cytosol and beyond the membrane plane in Gag-expressing cells (Fig. 2C). Co-clusters of HIV Gag and PSCFP2-CHMP2A were confirmed with scanning electron microscopy (SEM) to be membrane-budding structures resembling virus-like particles (Fig. 2E). We further confirmed our observations of ESCRT-III localization to the HIV Gag interior using immuno-gold transmission electron microscopy (fig. S7).

Fig. 1 Interferometric superresolution microscopy resolves the 3D organization of cell-associated HIV particles.

(A) Diffraction-limited (left) and iPALM-based superresolution (right) images of COS7 cells expressing HIV Gag-FLAG labeled with AlexaFluor 647 (AF647)–conjugated antibody (Life Technologies, Grand Island, NY). Gag clusters from different surfaces of the membrane are clearly resolved (pseudocolored for height; white arrowheads). Scale bars, 10 μm (full image); 100 nm (right). (B) Correlative iPALM (left) and SEM (middle) demonstrate HIV Gag-FLAG clusters to be membrane-enveloped particles (overlay, right). Scale bars, 250 nm. (C) Probability density of aligned clusters from HIV Gag-internal-FLAG (top, n = 1181 clusters) and HIV Gag-FLAG (bottom, n = 1398 clusters) highlight the shell-like lattice (isosurface threshold = 0.5 localizations × nm−3). (D) Probability slices (10 nm) display a ring-like distribution for Gag-internal-FLAG clusters (top middle) compared with the more compact Gag-FLAG clusters (bottom middle). Distance between crosshatch is 100 nm (right). Radial averages highlight the density differences owing to probe position within the HIV Gag molecule (inset model).

Fig. 2 Two-color 3D superresolution imaging localizes ESCRT-I and ESCRT-III subunits within the HIV Gag lattice of budding particles.

(A and B) HIV Gag fused internally to mEOS2 protein (Gag-imEOS2) detects assembly sites containing ESCRT-I subunits Tsg101 [(A), top] and ESCRT-III subunits CHMP4B [(A), bottom] and CHMP2A (B) fused to PSCFP2 protein. (B) (Top) iPALM measurements resolve the cytoplasmic pool of PSCFP2-CHMP2A (top arrowhead), peripheral membrane-associated Gag-imEOS2 (middle arrowhead), and densely coclustered Gag and CHMP2A structures extending from the membrane (bottom arrowhead). Scale bars, 100 nm (A); 500 nm (B). (C) PSCFP2-CHMP2A coexpressed with unlabeled HIV Gag and membrane marker tdEOS-farnesyl (red) show displacement of a CHMP2A cluster (green) extending beyond the plane of the plasma membrane. Scale bar, 500 nm. (D) Single clusters representative of AF647-labeled Gag-FLAG and PSCFP2-ESCRT probes. Axial (top) and lateral (bottom) views show localization of PSCFP2-ESCRT probes within the HIV Gag lattice. Scale bars, 50 nm. (E) Correlative two-color iPALM and SEM of two virus-like particles (white arrowheads) emanating from a cell expressing Gag-FLAG (red) and PSCFP2-CHMP2A (green). The left membrane bud shows confinement of PSCFP2-CHMP2A to the head of the particle. The particle on the right is at an earlier stage of budding, apparent from the extended lateral fluorescence distribution (right arrowhead). Scale bar, 250 nm.

To facilitate a robust statistical analysis of ESCRT organization at HIV assembly sites, we performed two-color single-particle averaging. The resulting probability distributions of PSCFP2-ESCRT probes and HIV Gag revealed significant signal from each ESCRT subunit within the interior of the viral lattice (Fig. 3, A to C; fig. S8; and supplementary text). Radially averaged probability density plots of HIV Gag and ESCRT probes revealed 92% of PSCFP2-Tsg101, 61% of PSCFP2-CHMP4B, 80% of PSCFP2-CHMP2A, and 78% of PSCFP2-Vps4A integrated probability residing within the estimated radius of the HIV Gag shell (Fig. 3E). This analysis highlighted a pool of PSCFP2-CHMP4B residing beyond the HIV Gag lattice (39% remaining probability), potentially indicating a signature of ESCRT-III polymer extending toward the neck of the budding structure. As a control, we observed an order of magnitude less PSCFP2-CHMP4B signal residing within the probability shell of mutant Gag particles, unable to recruit ESCRT-I (GagΔPTAP) (1, 2), as compared with wild-type Gag (Fig. 3, D and E). This suggests that the Gag PTAP motif is required for localizing ESCRT-III subunits to the particle interior. These results support our single-particle observations and highlight a distinct scaffolding mechanism of ESCRT-III and Vps4A subunits within the interior of the Gag lattice.

Fig. 3 Single-cluster averaging demonstrates ESCRT subunits within the interior of the HIV Gag lattice.

(A to D) Cell-associated HIV Gag-FLAG clusters from PSCFP2-ESCRT–expressing cells were subjected to 3D single-cluster averaging [(A) PSCFP2-Tsg101, n = 313 clusters; (B) PSCFP2-CHMP2A, n = 460 clusters; (C) PSCFP2-CHMP4B, n = 509 clusters; and (D) PSCFP2-CHMP4B/HIV Gag ΔPTAP, n = 278 clusters]. Two-color isosurfaced probability densities show a high-probability core of PSCFP2-Tsg101 (A) and ESCRT-III subunits PSCFP2-CHMP2A/CHMP4B [(B) and (C)] (green) with respect to the HIV Gag lattice (red) (isosurface threshold, HIV Gag-FLAG = 2.5 × 10−4 nm−3 and PSCFP2-ESCRTs = 2.5 × 10−5 nm−3). Right insets [(A) to (D)] depict 40-nm sections along the z axis [top HIV Gag (red), bottom PSCFP2-ESCRT (green)]. (D) A marked reduction in the probability density of PSCFP2-CHMP4B signal was observed upon expression of release-defective HIV Gag-FLAG ΔPTAP. (E) Radial average plots of sections from (A) to (D) define the highest-probability densities of ESCRT (green) subunits residing within the HIV Gag probability shell (red). Approximately 90% of PSCFP2-Tsg101 (top left), 80% of PSCFP2-CHMP2A (top right), and 61% of PSCFP2-CHMP4B (bottom right) integrated probability reside within the half-maximum Gag probability shell, whereas 10% of the integrated probability for PSCFP2-CHMP4B remains when ESCRT-I recruitment is inhibited (Gag-FLAG ΔPTAP; bottom right). Line thickness represents SEM and is <1 nm.

The presence of ESCRT-III subunits within the Gag lattice suggests that a portion of this pool may become trapped within the viral particle while mediating bud neck constriction and abscission. To test this possibility, we purified virus-like particles (VLPs) from GFP-ESCRT–expressing cells and performed quantitative fluorescence microscopy (Fig. 4A). Using HIV Gag-GFP as a molecular standard, we estimated that on average hundreds of copies of GFP-ESCRT probes reside within released particles. Qualitative coincidence detection between Gag-mCherry and GFP-ESCRT proteins showed that GFP-ESCRT probes were detected in the majority of released particles (50 to 90%) (fig. S9). Correlative two-color iPALM and SEM imaging confirmed the presence of ESCRT probes within released VLPs (Fig. 4B). Western blot analysis showed that GFP-Tsg101 and GFP-CHMP4B were readily detected in purified VLP fractions (Fig. 4C). Levels of GFP-CHMP2A, but not GFP-CHMP4B or GFP-Tsg101, were reduced in released virus particles relative to membrane-associated particles (Fig. 3, A to C). This suggested that CHMP2A subunits may undergo selective remodeling upon virus abscission. This may be mediated by the ATPase activity of Vps4, given the established interaction between Vps4 and CHMP2A (17).

Fig. 4 Released VLPs contain ESCRT-I and ESCRT-III subunits, but reduced levels of CHMP2A relative to CHMP4B.

(A) Quantitative fluorescence microscopy analysis of GFP-ESCRT and Gag-GFP containing single particles. The fluorescence intensity of Gag-GFP is predicted to arise from ~600 to 1000 molecules (assuming 3000 to 5000 Gag molecules per particle), suggesting that signal from GFP-ESCRT particles corresponds to hundreds of molecules per particle (data represent ≥25,000 particles from ≥2 independent samples for each probe). (B) Correlative two-color iPALM and scanning electron micrographs of a single released (top, red) Gag- and (top, green) CHMP4B-positive VLP. Image registration residual uncertainty of 11 nm. Scale bar, 50 nm. (C) Western blot analysis of purified VLP fraction from GFP-ESCRT stable cell lines. GFP-Tsg101 and GFP-CHMP4B are detected in released particles (top); however, GFP-CHMP2A levels are dramatically reduced from VLPs upon particle release (bottom; asterisk indicates proteolysis product). Reduction in the levels of GFP-CHMP2A in released particles correlates between the biochemical and fluorescence data. (D) Proposed virus-scaffolding model for ESCRT-mediated HIV particle release from host cells. A pre-scission pool of ESCRT-I subunit Tsg101 (gray) interacts with HIV Gag (red), leading to assembly of ESCRT-III subunits CHMP4B and CHMP2A (green). Hypothetical ESCRT-III polymerization beyond the Gag lattice and Vps4 (orange) remodeling of CHMP2A subunits leads to membrane aperture constriction and particle abscission (middle and right).

Current models have been unable to distinguish whether ESCRTs localize to the base of the neck of a virion (in trans) or to the head of the virion (in cis) in order to constrict and release the viral particle from the plasma membrane. Our results show that ESCRT-III and Vps4A probes concentrate on the interior of the HIV Gag lattice and occupy a similar volume to that of ESCRT-I. This architecture is consistent with previous cytokinesis observations that ESCRT-III subunits initiate proximal to the midbody scaffold protein CEP55 and ESCRT-I (12). Indeed, HIV Gag and CEP55 share features, such as dual Tsg101 and ALIX recruitment domains (1820), as well as an apparent independence of ESCRT-II for mediating membrane abscission (1921). Collectively, these observations suggest that structural proteins, such as HIV Gag and CEP55, act to nucleate structures composed of ESCRT-I, ESCRT-III, and Vps4 subunits (supplementary text). These subunits then serve as distinct templates (acting in cis relative to the structural scaffold) for ESCRT-III polymerization and Vps4-mediated disassembly/remodeling of CHMP2A, ultimately leading to membrane abscission (Fig. 4C). Further studies will be required to dissect the interplay between scaffolding and polymerizing/depolymerizing pools of ESCRT-III and the role this dynamism plays in mediating viral membrane constriction and abscission.

Supplementary Materials

Materials and Methods

Figs. S1 to S10

References (2233)

References and Notes

  1. Acknowledgments: The authors are thankful for insight and helpful discussions from N. Elia, B. Kopek, and B. Lorenz. The authors are grateful to R. Villasmil for assistance with flow cytometry and N. Tsai for DNA sequencing. The authors also thank P. Bieniasz and M. Davidson for providing vectors.
View Abstract

Stay Connected to Science

Navigate This Article