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

Reviewer #2:

[…] Overall, this is a very interesting manuscript whose key strength is imaging of ESCRT function in relation to precise timing of scission in live cells using inventive and complex methodology. Despite this the take home message is simple and important. I have some comments and queries below that the authors should consider addressing:

1) It is not clear from the manuscript how many waves of ESCRT-III recruitment happen for the average budding event. Do most scission events require only one? How many fail to progress to scission over the time frame despite multiple waves of ESCRT-III recruitment?

A majority of our scission events occurred following the first recruitment of ESCRTs, but multiple rounds of ESCRT recruitment were also observed. In this work approximately 20% of our scission events consisted of more than one round of ESCRT recruitment. This is in rough agreement with our previous work in which we quantified the number of rounds of waves of recruitment (Jouvenet et al., 2011). In the previous work we did not simultaneously study scission, but in the current manuscript we demonstrate that scission only occurs following the last ESCRT recruitment. We have added a statement to the main text to clarify how often we observed multiple recruitment events. We rarely observed the recruitment of ESCRT-III without subsequent scission (~5%). However, we imaged cells for only a finite time and for those few events where we did not observe scission, we believe it occurred after we stopped imaging.

2) It is well established in artificial systems that Gag can induce membrane curvature, but the key novelty here is measuring it in real time in live cells. However, to definitively demonstrate ESCRT-independence, it would seem important to me to examine the membrane curvature of assembly of a Gag lacking the late domain in comparison – one should expect the same profile but subtle differences in timing might be interesting.

To address this question, we have repeated the Gag assembly experiments with either the intact Gag or with a deletion of the late domain (p6) of Gag, which contains sites for ESCRT recruitment. These were imaged and quantified using the same polarization-based excitation technique to monitor the orientation of the fluorophore (new Figure 4A). By measuring relative emission of GFP when excited with an electric field parallel (S-polarized) or perpendicular (P-polarized) to the glass we observed a similar change in the emission ratio (P-polarized/S-polarized) in the native Gag and Gag deleted of p6. Like wild type Gag assembly, the Gag missing p6 also had an initial P/S ratio of ~2 which then dropped to ~1.4 during the early parts of VLP formation (example trace added to Figure 4A; and additional examples shown in a new figure, Figure 4—figure supplement 2). We attribute this change to the VLP assembling into a spherical shell throughout assembly.

These results demonstrate that eliminating the late domain and the recruitment of early acting ESCRTs (like TSG101 or ALIX), have no effect that can be detected on the timing or extent of membrane bending. Thus, they are not required for VLPs to form the spherical shell structure, consistent with previous electron microscopy results in which Gag missing the late domain still formed spherical shells (Martin-Serrano and Bieniasz, 2003).

3) The authors should include representative videos of the events described.

We have now included representative movies for the following scenarios:

Video 1: Gag-pHluorin assembly (left) in whole cells with CO₂ switching every 10s along with simultaneous recruitment of mCherry-CHMP4B (right).

Video 2: Example of individual puncta of Gag-pHluorin assembly with CO₂ switching (left, every 10s) with simultaneous recruitment of mCherry-CHMP4B (right). CHMP4B was visible starting at 13:29 min.

Video 3: Gag-pHluorin assembly (top) in whole cells with CO₂ switching every 10s along with simultaneous imaging of mCherry-VPS4A recruitment (right).

Video 4: Example of individual puncta of Gag-pHluorin assembly with CO₂ switching (left, every 10s) with simultaneous recruitment of mCherry-VPS4A (right). VPS4A was visible starting at 20:49 min.

Video 5: Gag-pHluorin assembly (left) in whole cells with CO₂ switching every 120s along with simultaneous imaging of mCherry-VPS4A recruitment (right).

Video 6: Example of individual puncta of Gag-pHluorin assembly with CO₂ switching (left, every 120s) with simultaneous recruitment of mCherry-VPS4A (right). VPS4A was visible starting at 14:33 min.

Video 7: Gag-mEGFP assembly in whole cells with ratio image of perpendicular and parallel excitation (P/S) and total emission (P+2S).

Video 8: Example of individual puncta of Gag-mEGFP with P/S and P+2S images

We have added references and legends for these videos to the text.

4) Can the authors measure the relative concentration to carbonic anhydrase in purified virions vs the cytoplasm?

This is a good question, which has previously been addressed. Mass spectrometry studies of purified virions of HIV-1 that have probed for cellular proteins could not detect the presence of carbonic anhydrase (Ott, 2008). This is consistent with our observation that in isolated VLPs, modulating the pCO₂ does not affect the pH in the VLPs. If there was significant carbonic anhydrase in isolated VLPs then CO₂ modulation would result in pH (and thus fluorescence) changes – which we don’t observe (this work and Jouvenet, 2008).

Thus, the change in pH reported by the Gag-pHluorin in the forming bud is not the consequence of carbonic anhydrase in the bud. It is the result of a diffusive pathway in the neck that is sufficient size to allow the transit of protons. After scission, the connecting pathway is lost and protons in the cell cannot enter the nascent VLP. We have expanded the text to clarify our hypothesis on carbonic anhydrase in the VLPs.

Reviewer #3:

The authors have used a number of interesting biophysical techniques to analyse the later stages of HIV budding and the temporal dynamics of recruitment of ESCRT-III components and membrane scission. Oscillating fluorescence of a pH dependent fluorophore was used to determine the point of membrane scission. I am curious to know whether they are actually measuring completion of the Gag sphere or actual membrane pinching off. In the original work of Gottlinger many of the particles that remained adherent to the cell appeared to have completed Gag sphere formation. Would it be useful to test out their oscillating pH system using such a budding mutant to determine whether they see slowing of oscillation caused by Capsid shell completion without particle release?

As described above, we believe that we are measuring scission of the membrane as opposed to completion of the Gag shell. Electron microscopy studies (Briggs et al., 2009; Carlson et al., 2008; Woodward et al., 2015; Wright et al., 2007) show that the immature Gag lattice is generally incomplete with openings in the Gag lattice at the junction with the mother cell, thus continued movement of protons back and forth. The rapid transition to a state where protons cannot move between the mother and bud indicates a loss of continuity thus scission. Based on our results carbonic anhydrase is not in the nascent VLP. Additionally, mass spectrometry of nascent virions looking at cellular proteins did not detect carbonic anhydrase in the virions (Ott, 2008). Thus, any CO₂ that passes the membrane of the VLP can only be very slowly converted to bicarbonate and protons. We have previously demonstrated that with a late domain mutant, the Gag inside the nascent virion remains sensitive to the cytosolic pH (Jouvenet, 2008, Figure 3F).

They previously showed ESCRT-III recruitment kinetics using a fluorophore based super resolution microscopy approach so some of the data here is more confirmatory rather than novel. The timing of CHMP4 recruitment and departure appears to be shorter than in their previously published work.

First, one of the key novelties in this work is imaging the recruitment kinetics of ESCRT-III simultaneously with membrane bending or simultaneously with scission. This has not been done before by us, or anyone else.

Second, there is a difference in relative timing between recruitment of CHMP4B and VPS4A, but it is a difference not between previous work and this work, but a subtle difference that varies with the cell type. In both this work, which mainly examined HEK293T cells, and in our previous work on HeLa cells (Bleck et al., 2014), we monitored the time course of recruitment of CHMP4B simultaneously with recruitment of VPS4A. In both cell types, after ESCRT-III appeared (see red tracing in Figure 3 of the current manuscript), we detected VPS4A (see the green tracing). The ESCRT-III fluorescence increased, with the VPS4A following. The CHMP4B reached a maximum shortly before VPS4A, and then they both decreased.

In the HEK293T cells we observed that the lag to VPS4A reached its ½ max 5.1 seconds after CHMP4B and in the HeLa cells the lag was 9 seconds (Author response image 1 and Figure 1E from our previous paper, Bleck et al., 2014). From re-examining the histograms we are uncertain what, if any, the significance is of that delay. For this work we performed the majority of measurements in HEK293T cells because scission was observed to occur much more frequently, allowing us to study many more events. The difference in the lag between the HeLa and HEK293T is consistent between our previous work and this work (here we examined both). However, we did not examine the difference in the lag closely and therefore we are reticent to make any statements about its biochemical significance. For this publication we focused on the relative timing of the ESCRTs and VPS4 to the scission – where there is a much greater and consistent difference.

Author response image 1

Download asset Open asset

Acidification of the cytosol accelerates scission. My understanding is that the inner surface of the plasma membrane carries a generalized negative charge, so I am not sure how important the PIP₂ components are in affecting scission. PIP₂ acts as an important bridge binding the Matrix region of Gag to the inner plasma membrane but this occurs with all the Gag polyproteins from the first stages of assembly even before membrane bending and is not specific to those interactions involved at the budding neck. It would be of interest to see whether formation of vesicles triggered by ESCRT, such as in MVB formation, is similarly affected by pH changes to define how specific this is (or isn't) for virion budding and whether PIP₂ is of specific importance.

Thank you for the emphasizing the presence of PIP₂ throughout the Gag bud, and pointing out that PIP₂ might not be a critical lipid for the scission. Based on the pHluorin fluorescence change we observed when switching between 0% and 10% CO₂ we believe the cytosolic pH is modulated between ~7.5 and ~6.5. Phosphatidic acid (PA), phosphatidylethanolamine (PE), phosphatidylcholine (PC), phosphatidylglycerol (PG), and phosphatidylserine (PS) all have pK_a values that are away from this range. The two phosphates on the inositol of PIP₂ have a pK_a of 6.5 and 7.7, which are in the range of sensitivity of our measurements and therefore we highlighted the possibility of this lipid being sensitive to charge change in the pH range for our experiments. However, we acknowledge that this is just one possible component that is sensitive to charge modulation in this pH range, and we have modified the discussion to more clearly state this possibility.

MVB formation would be an excellent system to also study membrane scission and although outside the scope of this work, we hope our research will provide inspiration for a similar study. In particular, due to differences in lipid composition in MVBs, it would be interesting to determine how scission in MVBs are sensitive to pH.

The demonstration of repeated recruitment and loss of VPS4A and CHMP4B is interesting and novel. It begs the question as to what is the critical trigger for scission as it would appear that although they are important for the process they themselves are necessary but not sufficient. A significant caveat is that the experiments as far as I can tell are performed with Gag alone and there is evidence that the Gag/Pol polyprotein may influence budding significantly with a suggestion that cargo size affects ESCRT function (Bendjennat et al., 2016). Similarly, there is no genomic RNA and this may affect virion assembly. Either of these may be involved in effecting the scission pathway and might be part of the definitive trigger, which would render repeated VPS4A/CHMP4B recruitment unnecessary.

We have previously observed the repeated recruitment and loss of VPS4A and ESCRT components (Jouvenet et al., 2011). We had struggled with the meaning of this until this current work. The results now demonstrate that the durations of ESCRT recruitment and loss are fairly stereotypic in duration. At the end of each round, there is either scission, or not. If there is scission, there are no further rounds of ESCRT recruitment. If there is not scission, then there is another round. We think that the ESCRTs are pulling the neck just an extra bit closer to facilitate scission, but then have to move out of the way. This is repeated until scission occurs. This suggests that it may be possible for scission to occur even in the absence of ESCRTs, albeit at a reduced rate. Thus, we don’t believe there is a definitive trigger for ESCRT-III recruitment. Instead, we hypothesize that the presence of the late domains helps to initiate recruitment of ESCRT-IIIs in a probabilistic fashion and may also be governed by the current neck structure. This is supported by Bendjennat and Saffarian (2016) which showed that the absence of the late domain (deletion of PTAP) delayed particle release by 10 hours but did not prevent viral particle release. Thus, without the late domains, scission events might still be observed, though would be rarer. We have updated the text to include information that scission may occur in the absence of late domain. However, when we previously monitored for scission in the presence of a late domain mutant (Jouvenet et al., 2008), we did not see scission during the 30-60 minute imaging periods we used.

We have previously examined if this assembly is affected by whether Gag or Gag and Gag-Pol was recruited, and at the level of the single virion, the assembly kinetics were the same for Gag alone and for the provirus Gag/Gag-Pol (Jouvenet et al., 2008). We also could not distinguish any differences in assembly rate in the presence of the HIV-1 genome (Jouvenet et al., 2009). The referee points out that in ensemble measurements larger cargo sizes appear to slow down rate of viral particle release (Bendjennat and Saffarian, 2016). We think that this may be an indirect effect seen at the macroscopic level not seen at the level of single virions. Larger cargos take longer to synthesize and slow the overall rate of synthesis in the cell. We have previously demonstrated that rate of viral assembly is very sensitive to concentration of Gag. Gag-Pol takes longer to synthesize than just Gag and thus, it would take longer to get to a critical concentration to initiate assembly and longer to assemble.

I note that nowhere is the role of ESCRT-II mentioned. There is now clear biochemical and EM evidence that it is involved in HIV budding (Meng et al., 2015).

Just as viral particle release rate may depend on presence of ESCRT-I/TSG101 and particle size (Bendjennat and Saffarian, 2016), the presence of ESCRT-II, which works in concert with ESCRT-I, is also expected to affect the rate of viral particle release (Meng et al., 2015) which we have now added to the manuscript.