ESCRT‐II functions by linking to ESCRT‐I in human immunodeficiency virus‐1 budding

Human immunodeficiency virus (HIV) uses the ESCRT (endosomal sorting complexes required for transport) protein pathway to bud from infected cells. Despite the roles of ESCRT‐I and ‐III in HIV budding being firmly established, participation of ESCRT‐II in this process has been controversial. EAP45 is a critical component of ESCRT‐II. Previously, we utilised a CRISPR‐Cas9 EAP45 knockout cell line to assess the involvement of ESCRT‐II in HIV replication. We demonstrated that the absence of ESCRT‐II impairs HIV budding. Here, we show that virus spread is also defective in physiologically relevant CRISPR/Cas9 EAP45 knockout T cells. We further show reappearance of efficient budding by re‐introduction of EAP45 expression into EAP45 knockout cells. Using expression of selected mutants of EAP45, we dissect the domain requirement responsible for this function. Our data show at the steady state that rescue of budding is only observed in the context of a Gag/Pol, but not a Gag expressor, indicating that the size of cargo determines the usage of ESCRT‐II. EAP45 acts through the YPXL‐ALIX pathway as partial rescue is achieved in a PTAP but not a YPXL mutant virus. Our study clarifies the role of ESCRT‐II in the late stages of HIV replication and reinforces the notion that ESCRT‐II plays an integral part during this process as it does in sorting ubiquitinated cargos and in cytokinesis.

Several viruses have evolved the ability to hijack ESCRT complexes for their own use during productive infection (Votteler & Sundquist, 2013). Human immunodeficiency virus (HIV)-1 is an example of this where ESCRT complexes are utilised at the late stage of its life cycle for budding (Fujii, Hurley, & Freed, 2007;Sundquist & Kräusslich, 2012;Meng & Lever, 2013; Figure 1). The major HIV structural polyprotein Gag forms a multimeric array below the plasma membrane leading to a change of membrane curvature. ESCRT complexes (ESCRT-I and ESCRT-associated protein ALIX) are recruited through the PTAP and YPXL motifs located in the p6 domain of Gag and initiate the budding and scission process Garrus et al., 2001;Strack, Calistri, Craig, Popova, & Gottlinger, 2003;VerPlank et al., 2001;von Schwedler et al., 2003).
The final cleavage involves the recruitment of VPS4 by CHMP4B through CHMP6 or ALIX. Canonically, the formation of ESCRT-II is believed to serve as a part of the curvature sensing supercomplex at the invagination sites leading to nucleation of CHMP4B and recruitment of VPS4 for final scission to occur (Boura et al., 2012;Fyfe, Schuh, Edwardson, & Audhya, 2011;Lee, Kai, Carlson, Groves, & Hurley, 2015). Alternatively, the recruitment of VPS4 can be achieved directly through the ALIX-CHMP4B pathway. In a knockdown study of CHMP6, the budding status seemed to be unperturbed (Morita et al., 2011) providing indirect evidence that there is an alternative route for recruiting CHMP4B that is dependent on ALIX. Indeed, ALIX has been shown to directly interact with CHMP4B (Snf7 in yeast; Kim et al., 2005;McCullough, Fisher, Whitby, Sundquist, & Hill, 2008;Wemmer et al., 2011), which may provide an alternative activation route for CHMP4B in sorting ubiquitinated cargo (Tang et al., 2016) and cytokinesis (Christ et al., 2016). During or soon after budding, virus maturation occurs in which a series of proteinase cleavages transform the spherical array of Gag and Gag/Pol in the immature virion into mature viral products (Bell & Lever, 2013;Sundquist & Kräusslich, 2012). The last stage of Gag cleavage where p24 is cleaved from its precursor p24-p2 is the slowest (Pettit et al., 1994), delay of which is a hallmark for impairment of budding (Dussupt et al., 2011;Strack et al., 2003;von Schwedler et al., 2003). The roles of ESCRT-I and -III in HIV budding are established (Garrus et al., 2001;Morita et al., 2011). However, it was not until recently that it was demonstrated that ESCRT-II, which is known to play an integral part in remodelling membranes (Boura et al., 2012;Fyfe et al., 2011;Lee et al., 2015), is also required for efficient budding of HIV (Meng, Ip, Prestwood, Abbink, & Lever, 2015).
ESCRT-II is composed of two copies of EAP20 and one copy each of EAP30 and EAP45 arranged in a Y-shaped conformation (Teo, Perisic, Gonzalez, & Williams, 2004). Initial studies using siRNA knock down suggested that ESCRT-II was not required for HIV budding in 293 T cells (Langelier et al., 2006), but evidence from in vitro studies has shown otherwise (Carlson & Hurley, 2012 (Im & Hurley, 2008;Langelier et al., 2006). In yeast, this domain contains two Npl4-type zinc finger domains, NZF1 and NZF2; the former has been demonstrated to interact with VPS28 of ESCRT-I (Teo et al., 2004).
The interaction between ESCRT-I and -II plays an important role in the endosomal recruitment of ESCRT-II (Mageswaran, Johnson, Odorizzi, & Babst, 2015 (Carlton & Martin-Serrano, 2007). The role of ESCRT-II in this process was also controversial with an early report suggesting it is not recruited to the midbody (Carlton & Martin-Serrano, 2007) or that depletion resulted in only a minimal defect in cytokinesis (Morita et al., 2007;Morita et al., 2010). However, recent studies reported that exogenously expressed EAP45 is recruited to the midbody, and later, endogenous ESCRT-II was also found to be localised to the same site as ALIX during this process (Christ et al., 2016;Goliand, Nachmias, Gershony, & Elia, 2014). Hence, it appears that ESCRT-II is important in fulfilling the requirement to recruit the ESCRT-III complex and that two pathways involved in recruiting ESCRT-III converge. Both pathways are important in recruiting CHMP4B and VPS4 for membrane scission to occur as is also described in sorting ubiquitinated cargos into MVB (Tang et al., 2016).
Previously, we utilised a CRISPR-Cas9 EAP45 knock out (KO) HAP1 cell line to eliminate the possibility of any residual ESCRT-II remaining after knockdown confounding the results and to fully address the involvement of ESCRT-II in HIV budding (Meng et al., 2015). We demonstrated that ESCRT-II is important in efficient HIV budding but also identified additional post-transcriptional effects of

| EAP45 acts at a post-integration stage in HIV replication
In our previous study, either control or EAP45 KO cells were transfected with HIV WT proviral DNA, and viral protein expression was quantified by western blotting (Meng et al., 2015). The data showed that Gag expression is reduced in the EAP45 KO cells. It was concluded that this reduction is most likely due to a transcriptional or posttranscriptional defect. However, it remained possible that EAP45 was also involved in the pre-integration stage(s). To formally address this, we infected either control or EAP45 KO cells with VSV-G pseudotyped HIV. At 48 hr post infection, the total cellular DNA was extracted and examined by quantitative real-time polymerase chain reaction (PCR) analysis using a probe specifically targeting the integrated provirus (Brussel & Sonigo, 2003). The data show that the amount of integrated provirus in EAP45 KO cells is comparable with that in control cells ( Figure S1), with no significant difference by one sample t-test (p = .583) suggesting EAP45 is not involved at pre-integration stages and that its function is restricted to post-integration events.

| EAP45 knockout reduces viral replication and secondary spread in a T cell line
We were interested to examine if the phenotype from HAP1 EAP45 KO cells was recapitulated in a disease-relevant cell type. For this, two copies of gRNA expressors targeting exon 4 and exon 9 of the EAP45 gene together with a hygromycin fused Cas9 expressor were con-

| Functional analysis of EAP45 in virus rescue
We previously demonstrated that there is a decrease in virus production upon depletion of EAP45. To ascertain that this effect is specifically due to the absence of EAP45, we exogenously expressed C terminally tagged HA-EAP45 (Meng et al., 2015) in HAP1 EAP45 KO cells together with the WT provirus. The production of EAP45 protein and mRNA was confirmed by immunoblotting ( Figure 3a) and reverse transcription polymerase chain reaction (RT-PCR; Figure 3b). With the increasing amount of the transfected EAP45 expressor, we observed an increase in EAP45 production both at the protein and mRNA levels that coincided with a two-to threefold increase in virus release ( Figure 3c). This EAP45-mediated rescue effect is specific to EAP45 KO cells because expressing EAP45 in the parental cells failed to produce any increase in virus production ( Figure S2). In our previous study, we observed an accompanying defect at the last step of Gag cleavage that is classically associated with defects in viral budding (Meng et al., 2015). Interestingly, densitometric analysis of the traces of p24 and its precursor p24-p2 revealed that this defect in cleavage was also reversed upon exogenous expression of EAP45 ( Figure 3d).
Our data clearly demonstrate that EAP45 is involved in efficient virus budding and that the phenotype produced by defective expression can be rescued by expressing EAP45 in trans.
Next, we investigated the domain requirement of EAP45 region when the Glue domain is removed. This is plausible as the level of protein is restored when H0 is also removed in ΔGΔH. Upon transfecting EAP45 variants with WT provirus into EAP45 KO cells, we only observed rescue in virus release from EAP45 mutants still containing the H0 region (FL and ΔG, Figure 4c). G domain on its own and ΔGΔH failed to rescue the budding defect even though intracellular viral protein levels were unaffected ( Figure 4d). Of note, ΔG could still rescue the defect efficiently despite its comparatively lower level of expression. This suggests that binding to ubiquitin is not essential in EAP45-mediated virus release (Slagsvold et al 2005). The H0 linker region is known to be important for the ability of the Glue domain to recruit ESCRT-I complex (Im & Hurley, 2008;Langelier et al., 2006).
To investigate the functionality of this region in our virus rescue assay, we used a construct (H0M) where the linker region is specifically mutated to destabilise the interaction with ESCRT-I (Im & Hurley, 2008). Mutating this region alone abolished the rescue effect completely (Figure 4c), reinforcing the notion that the H0 domain of EAP45 is critical for successful virus rescue.
Densitometric analysis of the traces of p24 and its precursor p24-p2 showed restoration to that of WT Gag cleavage upon expressing the FL EAP45 and ΔG, but not with the G only or ΔGΔH comparatively low level of ΔG expression indicates that this mutant is particularly effective at enabling particle release. Taken together, our data show that the H0 linker region is crucial in regulating the EAP45-mediated budding process through an evolutionarily conserved route by linking to the ESCRT-I complex.

| Electron microscopic examination of virus rescue
We used electron microscopy (EM) to examine the morphology of the budding virions. In HAP1 EAP45 KO cells, we observed a large number of budding viruses arrested at the plasma membrane, with stalks tethering the particles to it (Figure 5a,b, and d).
Interestingly, quantitative analysis of the fold change in virus release revealed that export of virus particles increased up to threefold in the presence of exogenously expressed EAP45 ( Figure 5c,f ). Virions at the termini of filopodia risk being counted as "released" if the latter is not visible, so we were careful only to include virions at a distance from the plasma membrane in the released category to minimise this skewing the data. The increase in virus production is consistent with the biochemical analysis of the released virions shown by immunoblotting (Figure 4b). In a parallel experiment, we examined the morphology of released virions from ΔGΔH expressing cells, which fail to rescue the budding defect in EAP45 KO cells (Figure 5d). The arrested particles were readily seen at the plasma membrane, and the number of released particles was comparable with that in EAP45 KO cells that are not expressing exogenous EAP45 (Figure 5f). Taken together, the EM results are consistent with the data shown in Figure 4, reinforcing the notion that exogenously expressing EAP45 in trans rescues the budding defect and that this effect is dependent on an intact H0 linker region.

| EAP45 functions through the ALIX pathway
It is generally believed that of the two pathways leading to activation of ESCRT-III in HIV budding, the PTAP pathway (ESCRT-I-II-III axis) is dominant for virus egress; however, cell type specificity is also documented . In HAP1 cells, we found that the YPXL pathway (ALIX axis) is preferentially used by the virus for budding ( Figure S3). We reproducibly observed an even more  KO cells when transfected with the PTAP mutant. We previously observed an unexpected additional reduction (Meng et al., 2015). This suggests instead that the functions of ESCRT-I and -II in this process may not completely overlap. To delineate the relationship between these two pathways, we investigated the effect of EAP45 expressed in trans together with either PTAP or YPXL mutated HIV.
On providing EAP45 exogenously, the release of the PTAP mutant, strikingly, is increased, and this effect is accompanied by an increase in efficiency of the final cleavage of Gag ( Figure 6). This was not observed from the YPXL mutant, suggesting that the increase of virus production and enhancement of Gag cleavage efficiency are both dependent on accessibility to ALIX. This effect of EAP45 in rescuing the budding defect of a PTAP mutant is highly reminiscent of what is seen when overexpression of ALIX rescues release of a PTAP mutant (Fisher et al., 2007;Usami, Popov, & Gottlinger, 2007). This strongly implies that

| Release of virus-like particles in EAP45 KO cells is independent of EAP45
VLPs (virus-like particles) are produced from Gag only expressors lacking late domains, and the level of their release is comparable with that of a WT Gag expressor at the steady state (Bendjennat & Saffarian, 2016). Although this may suggest that ESCRT is not required in VLP production, the release of such VLPs is in fact dependent on ESCRT complexes as overexpression of the dominant negative form of VPS4 abolished their release (Bendjennat & Saffarian, 2016 Figure 9). This suggests that a functional viral protease is not a prerequisite for the enhanced budding mediated by EAP45.

| The final cleavage of Gag correlates with virus release
Failure of the cleavage of p24 from p24-p2 by the viral protease is regarded as the signature defect associated with impairment of ESCRT mediated viral budding (Demirov, Ono, Orenstein, & Freed, 2002;Strack et al., 2003;von Schwedler et al., 2003) Figure 10). This is indicative that the two processes are strongly linked.

| DISCUSSION
During or soon after HIV buds from the cell, autocleavage of Gag/Pol occurs to produce activated viral protease, which leads to sequential proteolytic cleavage of Gag and Gag/Pol (Bell & Lever, 2013;Sundquist & Kräusslich, 2012 (Bendjennat & Saffarian, 2016;Chamontin et al., 2015). This loss of protease is thought to be instrumental in producing the maturation defect observed in budding defective viruses.
Depletion of EAP45 gives rise to viruses in which the last step of Gag cleavage is also defective, which is the first indication that EAP45 plays a role at the late stage of the HIV life cycle (Meng et al., 2015). This defect is accompanied by a defect in virus release, and we show that it can be successfully restored when EAP45 is expressed exogenously ( Figure 3). The increase in virion release triggered by EAP45 rescue is, however, not dependent on the presence of an active protease ( Figure 9), suggesting the rescue effect of EAP45 in EAP45 KO cells is independent of virion maturation, and that budding neck closure precedes the activation of the viral protease. Our data are consistent with the notion that in EAP45 KO cells, recruitment of EAP45 is required to restore defective budding kinetics and to facilitate the sealing of the budding neck so that the activated protease is confined within the budding virions to allow Gag maturation to proceed (Bendjennat & Saffarian, 2016;Chamontin et al., 2015). The significant correlation between the final cleavage of Gag and virion release also supports this notion ( Figure 10). Interestingly, this rescue of budding is only observed when Gag/Pol is present (Figure 8). This suggests that EAP45 (and possibly the entire ESCRT-II complex) acts like TSG101 and ALIX at the early stage of the budding process and is more effective at packaging a large cargo. This necessity to package Gag/Pol in ESCRT-II-mediated budding provides an alternative explanation for an earlier study where VLP release was found to be unaltered when ESCRT-II was depleted (Pincetic, Medina, Carter, & Leis, 2008). Our data suggest that the virus uses an additional safeguard to favour production of infectious virions, regulated by early ESCRT proteins facilitating virion maturation and acting as a checkpoint to ensure Gag/Pol inclusion into the virus; failure of autoprocessing of Gag and Gag/Pol and consequential absence of protease would otherwise render the virus noninfectious.
The lipid targeting ability of ESCRT-II at the plasma membrane is not fully understood. However, it is possible that the ESCRT-II-CHMP6 complex senses the negative membrane curvature as Gag multimerises at the plasma membrane in the presence of EAP45. The accumulation of such complexes could lead to further nucleation of CHMP4B for final scission to occur. However, the membrane curvature sensing ability of these complexes is necessary but not sufficient for a successful budding event because our evidence shows that the H0 linker region in EAP45, which is known to bind ESCRT-I, is also required; only mutants containing H0 rescue the budding defect in EAP45 KO cells (Figure 3). The H0 linker region both recruits the ESCRT-I complex and stabilises the ESCRT-II complex for recruiting F I G U R E 7 EAP45 functions through the ALIX pathway. The HAP1 EAP45 KO cells were transfected with sicontrol or siALIX followed by cotransfection of PTAP deleted proviral DNA and an EAP45 expressor. Western blots of total cell lysates and purified virions using antibodies specific to viral p24 and host proteins ALIX and GAPDH are shown. The blot is representative of three independent experiments CHMP6 for viral budding. The evidence thus points to ESCRT-II bridging ESCRT-I and -III complexes in this system as it does in others; mutating the H0 linker region alone completely abolishes the effect in rescue, and Gag cleavage remains defective (Figure 3). Our quantitative EM analysis also shows that the budding particles remain tethered at the plasma membrane when H0 is absent, whereas the particle production is dramatically increased in the presence of intact EAP45 (Figure 5f). Intriguingly, the Glue domain therefore seems to play an inhibitory role for the activation of ESCRT-II as upon removal of this region; the virus rescue is almost as efficient as that achieved by the FL protein ( Figure 4). An autoinhibitory role within members of ESCRT is not uncommon (McCullough et al., 2008) and has been proposed also for ESCRT-II for its role in MVB sorting in yeast (Mageswaran et al., 2015). However, activation of ESCRT-II in MVB sorting is slightly different from that in HIV budding, as the removal of the Glue domain is not sufficient to rescue the cargo sorting defective phenotype; a mutation is additionally required at its C terminus.
We observed that expression of EAP45 rescues virus release of a PTAP mutant, accompanied by an increased efficiency of the final cleavage step of Gag ( Figure 6). This mimics the phenotype of overexpression of ALIX in the presence of a PTAP mutant (Fisher et al., 2007;Usami et al., 2007). This increase in virus release fails to occur when ALIX is also knocked down in EAP45 KO cells (Figure 7), suggesting EAP45 rescues virus release through the ALIX pathway. Consistent with this model, overexpression of EAP45 fails to increase viral release of the YPXL mutant, suggesting ESCRT-II is dispensable in this regard possibly due to a direct interaction between VPS28 of ESCRT-I and CHMP6 of ESCRT-III providing an alternative bypass (Pineda-Molina et al., 2006). ESCRT-II may, however, function through ALIX, mediated by the interaction between ALIX and TSG101 (Langelier et al., 2006;Strack et al., 2003;von Schwedler et al., 2003). Binding of YPXL to the central V domain of ALIX releases the sequestered C-terminal domain of ALIX and transforms ALIX from a closed to an open position (Zhou et al., 2008). This conformational change provides an additional binding site for TSG101. The recruitment of TSG101 in this process may further recruit downstream CHMP4B through binding to EAP45 of the ESCRT-II complex. This logically explains why rescue can only be observed by expression of EAP45 with the PTAP mutant but not with the YPXL mutant ( Figure 9). The same network of interactions has recently been clarified in cytokinesis whereby the role of ESCRT-II is to aid recruitment of CHMP4B for abscission (Christ et al., 2016). However, an alternative possibility, which is not mutually exclusive, is that overexpression of EAP45 enhances the interaction of Gag and ALIX probably through the NC domain, and this enhanced interaction leads to an acceleration in CHMP4B polymerisation (Sette et al., 2016). If so, it is tempting to speculate that the accessibility of ALIX is important in nucleating CHMP4B at the budding site to lower the threshold that is needed for

| Quantitative RT-PCR for quantitation of integrated proviruses
A total of 293 T cells were transfected with pBH10ΔBglII-WT and pCMV-VSV-G expressing plasmids using TransIT LT1 (Mirus) as described previously (Meng et al., 2015). At 2-days post transfection, the supernatant was harvested, and virus infectivity was measured in a TZmbl infectivity assay. The same amount of pseudotyped of either control or EAP45 KO cells. After the 2-day incubation at 37 C, the supernatant was removed before the total DNA was extracted using a Qiagen blood and tissue extraction kit (Qiagen).
The integrated proviral DNA was quantified by Alu PCR based on a published protocol (Brussel & Sonigo, 2003). In a typical quantitative RT-PCR analysis, the Ct value from non-Alu primers reaction was subtracted from the Ct value of a reaction where Alu primers were added. Relative quantitation (ΔΔCt) of the amount of integrated provirus in EAP45 KO cells compared with that of control cells was measured.

| Construction of SupT1 EAP45 KO cells
The lentivirus that packages two gRNA targeting the EAP45 gene

| Viral protease inhibition assay
Either control or EAP45 KO cells in a 24-well plate were transfected with a total amount of 200 ng of pBH10ΔBglII-WT and pCMV-VSV-G expressing plasmid in the absence or presence of Tipranavir (200-or 500-nM final concentration; NIH AIDS reagent). In some transfections, an EAP45-HA expressor or EV was also cotransfected to examine the rescue effect of EAP45 during TPV treatment.

| Immunoblotting
Typically, 2 days post transfection, supernatants were harvested for virion purification by Optiprep ultracentrifugation, and monolayers were lysed in cell culture lysis buffer for immunoblotting. Cell lysates and purified virions were electrophoresed by SDS-PAGE followed by transfer to a nitrocellulose membrane. Blots were incubated with the primary antibody followed by incubating with secondary antibody before detection with either ECL (Promega) or ECL prime (GE healthcare). The primary antibodies used in this study are anti-p24 (NIBSC), anti-GAPDH (Abcam), anti-HA (Thermofisher), anti-ALIX (Cell signalling), and anti-TSG101 (Abcam), and secondary antibodies used were goat anti-mouse IgG-HRP (cell signalling) and goat anti-rabbit IgG-HRP (Thermofisher). Densitometric analysis of the target bands was done by ImageJ (NIH) with traces of p24 and p24-p2 highlighted, where needed, to show the differences in the final cleavage product.

| Transmission electron microscopy of transfected cells
Transmission electron microscopy was performed as described previously with a slight modification (Meng et al., 2015). 4.10 | siRNA knock down siRNA knock down was performed essentially as described previously (Meng et al., 2015). Briefly, 2 × 10 5 EAP45 KO cells were transfected by TransIT-TKO (Mirus) with one dose of siRNA (siControl and siALIX; Thermofisher) on Day 1. A second dose of siRNA was transfected together with the total amount of 200 ng of pBH10ΔBglII-PTAP and pCMV-VSV-G expressing plasmid with either 45 ng of EAP45-HA expressor or EV. On Day 4 post transfection, the supernatant was harvested and virions purified as described above. The monolayer was lysed with 1× cell culture lysis buffer before immunoblotting for viral or host proteins.