Timing of ESCRT-III protein recruitment and membrane scission during HIV-1 assembly

Abstract

The Endosomal Sorting Complexes Required for Transport III (ESCRT-III) proteins are critical for cellular membrane scission processes with topologies inverted relative to clathrin-mediated endocytosis. Some viruses appropriate ESCRT-IIIs for their release. By imaging single assembling viral-like particles of HIV-1, we observed that ESCRT-IIIs and the ATPase VPS4 arrive after most of the virion membrane is bent, linger for tens of seconds, and depart ~20 s before scission. These observations suggest that ESCRT-IIIs are recruited by a combination of membrane curvature and the late domains of the HIV-1 Gag protein. ESCRT-IIIs may pull the neck into a narrower form but must leave to allow scission. If scission does not occur within minutes of ESCRT departure, ESCRT-IIIs and VPS4 are recruited again. This mechanistic insight is likely relevant for other ESCRT-dependent scission processes including cell division, endosome tubulation, multivesicular body and nuclear envelope formation, and secretion of exosomes and ectosomes.

Keywords: ESCRT; HIV-1; cell biology; human; infectious disease; membrane scission; microbiology; viral assembly.

Figure 1.. ESCRT-IIIs and VPS4A transiently recruited prior to scission.

(A) Example trace of Gag-pHluorin assembling into single VLP while the pCO₂ in the imaging media was repeatedly switched between 0% and 10% every 10 s. Moment of scission is indicated by red dashed line. CHMP4B-mCherry was temporarily recruited (indicated by grey zone) to the site of VLP assembly following the loss of pH modulation sensitivity. (B) Histograms of appearance and disappearance of CHMP4B prior to scission. (C-H) Example traces and histograms of appearance and disappearance, relative to scission of the VLP, for mCherry-CHMP2A (C and D), mCherry-CHMP2B (E and F) and mCherry-VPS4A (G and H).

Figure 1—figure supplement 1.. Flow chamber configuration for ESCRT-III assisted membrane scission studies.

Imaging media in reservoirs was preequilibrated with gas containing 0 and 10% CO₂ (balanced with air). During assembly of single HIV particles in cells, the imaging media was modulated between reservoirs, enabling detection of scission of the VLP from the cell.

Figure 1—figure supplement 2.. Example traces of scission relative to recruitment of ESCRT-III or VPS4A.

(A) Gag-pHluorin was observed as pCO₂ was cycled between 0 and 10% every 10 s. VLP scission time (red dashed line) was characterized by half drop in lock-in signal. mCherry-CHMP4B, mCherry-CHMP2A, mCherry-CHMP2B and mCherry-VPS4A recruitment (left to right panels, recruitment highlighted in grey) were simultaneously monitored during Gag assembly. (B) Additional traces of VLP scission during recruitment of mCherry-CHMP4B, mCherry-CHMP2A, mCherry-CHMP2B, and mCherry-VPS4A (left to right columns).

Figure 1—figure supplement 3.. Knockdown of CHMP2A or CHMP2B by siRNA.

HeLa cell lines stably expressing either mEGFP-CHMP2A or mEGFP-CHMP2B were transfected with siRNA to either CHMP2A or CHMP2B or a control siRNA. 48 hr after transfection, presence of tagged ESCRT was significantly reduced and resulted in fewer cells, presumably because of decreased cell division.

Figure 1—figure supplement 4.. CHMP4B was recruited prior to VPS4A.

The times associated with the rising fluorescence edge of VPS4A and CHMP4B were compared relative to each other. CHMP4B appeared on average 5.1 s prior to VPS4A.

Figure 2.. Scission more likely at lower cytoplasmic pH.

(A) Example traces of Gag-pHluorin assembly while pCO₂ was switched every 120 s between 0% (red dashed line, greater fluorescence emission) and 10% (black dashed line, lower fluorescence emission). In these traces the fluorescence intensity became fixed in the low pH state (10% pCO₂) after reaching an assembly plateau. VPS4A appeared and disappeared during the first low pH state. (B) Examples in which fluorescence intensity became fixed in the high pH state (0% pCO₂). VPS4A appeared and disappeared during the first trapped high pH state. (C) Example in which fluorescence became fixed in low pH state. VPS4A disappeared during the first trapped low pH state, but appeared during the previous high pH state. (D) Example in which fluorescence became fixed in high pH state. VPS4A disappeared during the first trapped high pH state, but appeared during the previous low pH state. (E) Example trace in which fluorescence intensity because fixed in an intermediate state. (F) Bar graph of cytoplasmic pH condition in which scission occurs (N = 45). Scission is ~3-fold more likely at low pH (10% pCO₂) compared to high pH (0% pCO₂) condition. A small percentage of VLPs were trapped in an intermediate state.

Figure 3.. A new round of ESCRT-III recruitment required following failed scission event.

(A) Example traces of multiple waves of VPS4A and CHMP4B recruited following cessation of Gag accumulation. (B) Example trace of multiple recruitments of VPS4A prior to scission (red dashed line). (C) Example traces of multiple recruitments of CHMP4B prior to scission (red dashed line).

Figure 4.. Structural changes in VLPs throughout Gag accumulation.

(A) Example traces of wild-type Gag-GFP (black line, sf3 in Matrix of Gag) and Gag-∆p6 (grey line, missing ESCRT-I recruiting p6 domain) assembling into single VLPs. Images were collected every 30 s with excitation illumination polarized either perpendicular (${\displaystyle \hat{\mathbf{p}}}$) or parallel (${\displaystyle \hat{\mathbf{s}}}$) to the glass surface. Total Gag characterized by P+2S (top) and relative average dipole orientation by P/S (bottom). P+2S from wild-type Gag was fit to an exponential and used to predict an expected P/S (blue dashed line) assuming membrane bending throughout assembly. (B) Comparison of average P/S from all traces before VLP assembly (membrane background) and after VLP assembly (plateau region) for three different tagged versions of wild-type Gag. p6-GFP (N = 8), MA-sf3 (N = 9), and MA-sf11 (N = 7). Error bars represent s.d. (C) To compare the evolution of VLP structure to the assembly of Gag the time for each assembly trace was normalized from 0 (beginning of Gag assembly) to 1 (end of Gag assembly). A normalized time difference for each trace between Gag half assembly [½ (P+2S)_max] and the dipole half drop [½ (P/S)_max] was found and all normalized differences were compiled into a histogram. (D) Illustration of sphere budding from flat membrane. (E) Illustration of coordinate system with θ and ϕ representing position on the sphere and β and ψ representing orientation of excitation dipole. (F) Predicted P/S when β = 45°, background intensity is 45% of full assembly intensity, and evanescent field penetration depth is the same as the radius of the VLP.

Figure 4—figure supplement 1.. Example traces of Gag accumulation (quantified as P+2S) and fluorophore polarization (quantified as P/S) during VLP assembly.

Three different tagging schemes were used: labeled with mEGFP after p6 domain at carboxy terminal of Gag (A), labeled with circularly permutated GFP variant three in matrix of Gag (B), labeled with circularly permutated GFP variant 11 in matrix of Gag (Pédelacq et al., 2006) (C).

Figure 4—figure supplement 2.. Example traces of Gag-∆p6-mEGFP accumulation (quantified as P+2S) and fluorophore polarization (quantified as P/S) during VLP assembly.

Figure 5.. Proposed temporal model of ESCRT-III-mediated scission of HIV from cell plasma membrane.

(I) The viral particle structure changes throughout accumulation of Gag until (II) a spherical topology prevents incorporation of additional Gags. (III) ESCRT-IIIs (examples: CHMP2, CHMP4) are recruited to the neck and polymerize, with (IV) removal by VPS4 resulting in constriction of the neck. (V) After removal of all ESCRT-IIIs the narrow neck undergoes spontaneous fission, (VI) freeing the virus. If membrane fission does not occur a new round of ESCRT-III recruitment is required (V→ II).

Author response image 1.. Measurement of the lag in the rise to ½ max between CHMPB and VPS4A in HEK293T cells.