Endosomal sorting complexes required for transport are cellular proteins that catalyze the fission of membranes and play an important role in biology of diseases, including cancer and infectious virus release [1
]. ESCRTs were discovered in Saccharomyces cerevisiae
when their genetic deletion caused abnormal sorting of cargo in multivesicular bodies [2
]. The functioning units of ESCRTs in multivesicular trafficking were further identified in yeast as the VPS4 protein which is a AAA ATPase [3
] along with multi-protein complexes: ESCRT-I [4
] and ESCRT-II [5
], yeast Bro1 [6
] and ESCRT-III proteins [5
]. Based on identified biochemical interactions, the overall mechanism of ESCRT recruitment was proposed to start with cargo proteins binding to ESCRT-I and ESCRT-II proteins. ESCRT-I and ESCRT-II complexes in turn recruit ESCRT-III proteins. ESCRT-III proteins in turn catalyze the fission of membrane in conjunction with VPS4 recruitment [8
The discovery of the PTAP sequence on the Gag p6 domain of HIV and its importance in infectious virion release [11
] initiated a search for its cellular partners. Binding of the PTAP sequence to ESCRT-1 proteins demonstrated the role of ESCRTs in HIV release [13
]. Further, divergence between early ESCRTs in multivesicular bodies and HIV budding was observed: primarily, in the HIV budding pathway the link to ESCRT-II has been controversial; secondly, there is no direct mammalian homologue within the ESCRT pathway to Bro1 protein identified in Saccharomyces cerevisiae
]. The closest analogue to Bro1 was identified as ALIX which has a Bro domain with similar ESCRT-III interactions as identified in Saccharomyces cerevisiae
. However. mammalian ALIX, in addition to its Bro domain, has a V domain which binds directly to the YP motif, the second HIV late domain located on the HIV Gag-p6 [19
]. While depletion of the ESCRT-I complex completely abrogates infectivity of released HIV virions, abrogating ALIX interactions were shown to only partially affect the infectivity [22
]. In the mammalian system similar to the identified yeast interactions, it was shown that over-expression of ESCRT-III proteins resulted in deformations on the plasma membranes of cells [23
], and helical structures of ESCRT-III proteins were depolymerized by VPS4 [24
]. The overall mechanism of ESCRT recruitment in HIV budding, however, was proposed with similar step-wise recruitment of ESCRT factors—specifically, Gag late domains recruiting ESCRT-I and ALIX who in turn recruit ESCRT-III proteins which polymerize on the neck of the budding virions and catalyze the membrane fission reaction in conjunction with VPS4 [25
ALIX and ESCRT-I were shown to bind Cepp55 at the cleavage furrow during cytokinesis; it was further shown that ESCRT-III proteins and VPS4 are also recruited and play an essential role in cytokinesis [29
]. The functional units of ESCRTs in cytokinesis and HIV budding are similar in that they both require ESCRT-I and both rely on ESCRT-III polymers and VPS4. The difference between the two pathways is firstly their scale and secondly the essential role ALIX plays in cytokinesis compared to its peripheral role in HIV budding [32
In the past decade the number of identified intracellular processes dependent on ESCRTs has increased dramatically and now include exosome release [33
], down regulation of G-protein coupled receptors [35
], plasma membrane repair [36
] and nuclear envelope sealing [37
]. Understandably, major efforts in structural biology have been made to understand the molecular mechanism of ESCRT function. At present, at least some structural information about almost all proteins within the pathway is available [38
]. Reconstructing the ESCRT functions in vitro has also led to identification of ESCRT function in reverse topology membrane fission which has been implicated in endosomal pathway in cells [39
While ESCRTs are implicated in various cellular processes, the fundamental mechanism of ESCRT function is still assumed to be similar to what was proposed in multivesicular body biogenesis, namely, that early ESCRT-I and possibly ESCRT-II proteins bind the cargo and recruit the ESCRT-III and VPS4 proteins to perform the fission reaction. The details of the molecular mechanism of ESCRT function, including how recruitment is orchestrated, how the ESCRT-III proteins work in conjunction with VPS4 to catalyze the fission and how ATP hydrolysis couples to the membrane fission reaction, however, remain obscure.
The results from live imaging observations of ESCRTs have added to the mystery of ESCRT function. During imaging of multivesicular body biogenesis in yeast cells, ESCRT-III and VPS4 proteins were observed polymerizing and depolymerizing on the membrane before catalyzing the fission reaction, which was surprisingly found to be ATP-independent [40
]. During imaging of HIV virus-like particle (VLP) budding from mammalian cells in culture, membrane fission was detected to occur up to a minute after all the ESCRTs had been released back into the cytosol [41
Aside from live imaging, when kinetic release experiments were performed on release of infectious HIV particles, it was found that abrogating ESCRT interactions did not result in a full blockage of the release of virions as previously assumed. Instead it was shown that virions which had abrogated interactions with early ESCRTs eventually managed to release with a considerable delay. It was further shown that this delay led to an untimely activation of the HIV protease and release of non-infectious virions [42
Using live imaging we have visualized the recruitment of ALIX, CHMP4 and VPS4 during budding of HIV with abrogated Gag-ALIX interactions. ALIX interacts directly with HIV Gag through the YPXL late domain motif on Gag p6 [22
]. In Gag (YP−
) we abrogated this interaction by incorporating (36
) in place of (36
), as previously characterized [20
]. Under these conditions, based on the canonical view, we were expecting to find reduced recruitment of ALIX into HIV Gag VLPs. Instead, we report observing multiple rounds of transient recruitment of ALIX, CHMP4 and VPS4 after completion of Gag assembly during virion budding. We further show that during each transient recruitment, the stoichiometry of all ESCRT components remained the same when compared to WT condition. We also show that the timely recruitment of ALIX to the budding VLP is dependent on the intact PTAP domain. Our results demonstrate that recruitment of ESCRTs is driven by a robust network of interactions resulting in an “on/off” switch behavior, and ALIX’s interactions with late domains of HIV Gag play a crucial role during final the final stages after assembly of the full ESCRT machinery.
2. Materials and Methods
2.1. Cell Culture and Transfection
HEK 293T, HeLa and U2OS cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM; Invitrogen, Carlsbad, CA) supplemented with fetal calf serum (10%), sodium pyruvate (1 mM) and L-glutamine (2 mM). HeLa cell lines stably expressing eGFP-tagged ALIX were maintained in the same medium supplemented with G-418 (0.5 mg/mL) for selection. For TIRF experiments, cells were incubated in CO2-independent medium (LifeTechnologies, Grand Island, NY, USA).
Cells were seeded 18 h before transfection on sterile 4 chamber dishes at 60% confluence. Transfection was carried out using Lipofectamine2000 (Invitrogen, Carlsbad, CA, USA) and DNA plasmid at the ratio of 3:1 in HeLa and 2:1 for 293T cells and U2OS cells with total DNA of 1500 ng for imaging and 2000 ng for western blot. For stable cell line, the sample was supplemented in CO2-independent medium and moved to the microscope for imaging 4 h post transfection. In the case of transient co-transfection of DNA plasmids in normal HeLa, cells were used for imaging 6–7 h after transfection. The cells were kept at 37 °C during the imaging.
2.2. siRNA Transfections
HeLa cells were seeded at 40% confluence and were transfected 24 h later with siRNA targeting luciferase (CUGCCUGCGUGAGAUUCUCdTdT) or Alix (GAAGGAUGCUUUCGAUAAAUU) using Lipofectamine-2000. After 72 h, cells were re-transfected with siRNA. Again, after another 48 h cells were transfected with siRNA along with the desired plasmid. Cells were imaged 7 h later.
Live images were acquired using iMIC Digital Microscope made by TILL photonics controlled by TILL’s Live Acquisition imaging software as previously described [44
] using Andor iXon camera. Two wavelengths of laser, 488 nm diode laser (Coherent, Saphire 488, Santa Clara, CA, USA) and 561 nm diode-pumped solid state (DPSS) laser (Cobolt Jive, 561 nm Jive High Power, San Jose, CA, USA), were used to excite eGFP and mCherry, respectively. 60× objective was used for the experiments. Laser beams passed through an AOTF (acousto-optical tunable filter) and focused into a fiber which delivers the light to TILL Yanus digital scan head and then Polytrope II optical mode switch. Polytrope hosts a quadrant photodiode used for TIRF penetration depth calibration, which was set to 150 nm for the experiments in this manuscript. Once the penetration depths for the experiments are set at the beginning of acquisition, a feedback loop keeps the focus of the objective on the sample by constantly monitoring the position of the back reflected beam with respect to the original beam.
2.4. Microscopy Data Analysis
Images from the microscope were stored as TIFF files and analyzed using Matlab software (Mathworks, Natick, MA, USA) as described previously [44
]. The intensity of the fluorescent signal collected from each diffraction limited spot is proportional to the number of molecules within that position; however, the intensity is also proportional to the laser intensity, position of molecules with respect to glass during TIRF and substitution level of WT versus fluorescent molecules in each particular cell. To compare intensities of the ESCRT recruitments in between various cells and experimental conditions, average intensity (considering 25 VLPS from each cell) of the HIV unaffiliated ESCRT recruitments at the plasma membrane were used to normalize the fluorescent intensities in between cells. The intensity plots of VLPs are fitted using Boltzman growth equation. The timings of the recruitments were measured from the start of the Gag stationary phase to the intensity rise of the fluorescent ESCRTs. The later spikes are accounted in the histograms after adding them to the times of the first recruitments.
2.5. Cell Detachment Experiments
U2OS cells were transfected with Gag–eGFP or Gag–eGFP (YP−) and observed by TIRF imaging. At 12 h post-transfection, cells with VLPs were first imaged using TIRF and then cells were gently detached using TryplE (LifeTechnologies). Detachment was achieved by removing the medium and washing once with PBS; a thin layer of TryplE was added to cover cells to allow cell to detach. After a few minutes, the glass was again imaged with released VLPs left on the glass support.
2.6. Western Blot Analysis
Virion and cell lysates were separated on 4–15% polyacrylamide gels and transferred to Immobilon-FL membranes. Anti-p24 (183-H12-5C, NIH AIDS Reagent Program), anti–eGFP (Santa Cruz) and infrared dye coupled secondary antibodies (LI-COR) were used for immunoprobing. Scanning was performed with the Odyssey infrared imaging system (LI-COR) in accordance with the manufacturer’s instructions at 700 or 800 nm, accordingly.
2.7. Infectivity Assay
HEK 293T cells (60% confluent in 4 cm plates) were transfected using lipofectamine-2000 with NL4.3 vector alone or along with ΔCMV–eGFP–flex–CHMP4b plasmid. The supernatant was harvested 48 h later. Infectivity was measured by adding the supernatant to TZM-B1 cells (80% confluent); 48 h later cells were lysed using britelite plus Reporter Gene Assay (Perkin Elmer, Waltham, MA, USA) and luminosity was measured using a Cytation 5 microscope, experiments were carried out in triplicate.
All conditions tested contained 20 analyzed virus-like particles (supplementary Figures S1, S6 and S7
) or 40 virus-like particles (main Figures 1–4) analyzed from 4–5 cells. The experiments were performed at least 3 times. There was no data selection applied to the sample, therefore all relevant data collected from the microscopy were analyzed and plotted in the figures.
2.9. Availability of Data
All data and reagents used in this study are available upon request; that includes the ΔCMV–eGFP–flex–CHMP4B plasmid characterized in this study and its sequence which is available upon request and also includes the Matlab code used for analysis of the imaging data.
2.10. Cell Lines
The HeLa, 293T and U2OS cells were obtained from ATCC. TZM-b1 cells were acquired from NIH AIDS Reagent Program.
The recruitment of ALIX, CHMP4b and VPS4 with almost the same number of molecules under various late domain mutations argues that recruitment of ESCRTs is driven by a cooperative network which can be triggered through multiple entry points with identical net resulting recruitment (Figures S2, S4 and S5
). More recently, binding of ubiquitin to Gag through ubiquitin ligases and membrane curvature were shown to play roles in recruitment of ESCRTs [50
]; however, how these events are choreographed on the plasma membrane remains to be explored.
HIV-1 mainly infects CD4+
helper T cells and macrophages in vivo [56
]. Here, we have mostly used HIV Gag constructs and performed the experiments in HeLa cells because of their ideal membrane configuration for imaging purposes and the mostly static phenotype of Gag VLPs building on their membrane. This system has allowed easy tracking and observation of multiple rounds of ESCRT recruitment in the same VLP. Observing multiple rounds of assembly during HIV budding in T-cells remains technically out of reach, due to the limited membrane and cellular movements of T-cells [58
Multiple VPS4 recruitments are rarely observed after completion of Gag WT assembly, and even when they are observed, no more than two recruitments are observed on the same VLP [41
] and Figure 1
, Figure 2
, Figure 3
and Figure 4
. The stuttering recruitment of the full ESCRT machinery which is characterized by more than three recruitments of the ESCRT machinery in the same YP−
VLPs is therefore significant and novel and suggests that ALIX plays a major role during end stages of ESCRT function. We hypothesize that the catastrophic disassembly of all of the ESCRT machinery during stuttering on the YP- VLPs indicates that a failure of ALIX to connect properly with Gag results in a catastrophic disassembly of all ESCRT components. This disassembly is then followed by re-assembly of the full ESCRT machinery, resulting in stuttering recruitment of ESCRTs. The molecular mechanism and the ultra-structural localizations of all the ESCRT machinery during membrane fission are unclear and would require significant new investigations.
We also show that PTAP mutation delays but does not stop the recruitment of ALIX. Therefore, we argue that the prevalent linear biochemical interaction map between ESCRTs may unnaturally simplify the in vivo function of these interactions. There are up to five different ALIX interactions functioning at late stages of VLP assembly: (i) a direct ALIX–Gag interaction through the YPXL late domain motif on Gag p6 [22
], (ii) a direct ALIX–Gag interaction through a binding site on Gag NC [61
], (iii) ALIX interactions with ubiquitin [64
], (iv) ALIX–TSG101 interactions [20
] and (v) interactions with ALIX itself, including relief of ALIX autoinhibition [67
],opening of the V domain [67
] and possibly ALIX dimerization [69
]. Our results suggest that the exact choreography of these interactions and what role they play during the function of the full ESCRT machinery is not simply recruitment and remains to be visualized in vivo.
We have previously shown that mutations within the late domain of HIV result in a delayed release of the virus, which in turn results in budding of non-infectious virions due to premature protease activation [42
]. These kinetic biochemical assays showed an approximate delay of 20 min for HIV Gag (YP−
) in U2OS cells and 70 min for fully infectious virions with YP−
]. This delay is consistent with the time between the first recruitment of ESCRTs and the last recruitment of ESCRTs on individual VLPs visualized in this study. Our measurements show an average time between these recruitments as 10 ± 8 min for HIV Gag (YP−
) in HeLa cells, 18 ± 13 min for HIV Gag (YP−
) in U2OS cells and 37 ± 45 min for NL4.3 (iGFP)(ΔENV)(YP−
) in HeLa cells. Recent studies have shown recruitment of ESCRTs is followed by release of the virion within a 20 s time window from the last ESCRT recruitment [41
]. Therefore, we suggest that in our experiments, virion release happened after the last recruitment of ESCRTs after the stuttering events on individual virions. During imaging of HIV virus-like particle (VLP) budding from mammalian cells in culture, membrane fission was detected to occur up to a minute after all the ESCRTs had been released back into the cytosol [41
]. In our study we did not directly measure viral release; we only report a constant time delay in release of Gag VLPs with the delay associated with stuttering. We therefore cannot report on the exact moment of virion release with respect to the last ESCRT recruitment in our study. Experiments to visualize both the fluorescence recruitment and virion release are, however, the focus of future work.
The ALIX homologue Bro1 in yeast is proposed to be recruited through interactions with Snf7, a yeast homologue of CHMP4 [6
]. ALIX has a Bro1 domain analogous to the yeast Bro1, along with a V domain and a PRR which does not exist in the yeast homologue Bro1 [7
]. The binding of late domain YPXL has been mapped to the ALIX V domain [22
] and Cepp55 binds the PRR [30
]. Such apparent diversity had suggested that the recruitment and possibly function of ALIX is evolutionarily separate from the yeast homologue Bro1. In contrast to this view, our observations showing that the late domain does not play a role in recruitment of ALIX is more in agreement with the findings in yeast where Bro1 was shown to regulate the function of ESCRT-III protein Snf7 during membrane scission [70
]. While ALIX has been shown to be important in function of ESCRTs in all membrane scission reactions, a unified understanding of its function has been lacking. Based on all above data and available literature we suggest that ALIX plays a critical role during the final stages of membrane fission along with ESCRT-III and VPS4 proteins.
Live imaging of HIV and MVB budding has been previously used for visualizing recruitment of ESCRTs during membrane scission events [40
]. Our study shows how disturbing previously characterized biochemical interactions can result in surprising recruitment profiles of ESCRTs observed in live cells and therefore underscores the usefulness of the imaging methods for further characterizing these interactions in vivo.