HIV-1 gag recruits PACSIN2 to promote virus spreading

Significance

The structural Gag proteins of HIV-1 and other retroviruses contain regions that hijack a cellular membrane fission machinery to facilitate virus release. We now show that these Gag regions additionally recruit a host protein called PACSIN2, which has been shown to remodel cellular membranes and the actin cytoskeleton. Our data indicate that PACSIN2 is recruited indirectly by binding to a small regulatory protein called ubiquitin that becomes attached to Gag. Knockdown and reconstitution experiments revealed a critical role of PACSIN2 in HIV-1 spreading, although it was not required for a single cycle of replication. Overall, our study indicates that PACSIN2 promotes the cell-to-cell spreading of HIV-1, which constitutes the predominant mode of transmission but is poorly understood.

Abstract

The p2b domain of Rous sarcoma virus (RSV) Gag and the p6 domain of HIV-1 Gag contain late assembly (L) domains that engage the ESCRT membrane fission machinery and are essential for virus release. We now show that the PPXY-type RSV L domain specifically recruits the BAR domain protein PACSIN2 into virus-like particles (VLP), in addition to the NEDD4-like ubiquitin ligase ITCH and ESCRT pathway components such as TSG101. PACSIN2, which has been implicated in the remodeling of cellular membranes and the actin cytoskeleton, is also recruited by HIV-1 p6 independent of its ability to engage the ESCRT factors TSG101 or ALIX. Moreover, PACSIN2 is robustly recruited by NEDD4-2s, a NEDD4-like ubiquitin ligase capable of rescuing HIV-1 budding defects. The NEDD4-2s–induced incorporation of PACSIN2 into VLP correlated with the formation of Gag-ubiquitin conjugates, indicating that PACSIN2 binds ubiquitin. Although PACSIN2 was not required for a single cycle of HIV-1 replication after infection with cell-free virus, HIV-1 spreading was nevertheless severely impaired in T cell lines and primary human peripheral blood mononuclear cells depleted of PACSIN2. HIV-1 spreading could be restored by reintroduction of wild-type PACSIN2, but not of a SH3 domain mutant unable to interact with the actin polymerization regulators WASP and N-WASP. Overall, our observations indicate that PACSIN2 promotes the cell-to-cell spreading of HIV-1 by connecting Gag to the actin cytoskeleton.

It is well documented that the replication of HIV-1 and other retroviruses depends on the interaction of Gag with host proteins within virus-producing cells. In particular, retroviral Gag proteins engage the ESCRT membrane fission machinery to promote the release of assembled virions from the cell surface, which requires a membrane fission event to separate the viral lipid envelope from the plasma membrane (1–5). The cellular ESCRT pathway promotes membrane scission from the cytosolic site of narrow membrane necks, such as those formed during cytokinesis, the budding of cellular vesicles into the lumen of late endosomes, or the budding of retroviruses and other enveloped viruses from the plasma membrane (6).

The Gag proteins of HIV-1 and other retroviruses contain so-called “late assembly” (L) domains, which provide docking sites for early-acting components of the ESCRT pathway (1). The primary L domain of HIV-1 maps to a highly conserved PTAP motif within the C-terminal p6 domain of Gag and interacts with TSG101, a component of the ESCRT-I complex (7–10). A secondary YPXL-type L domain within p6 interacts with ALIX, another early-acting ESCRT pathway component (11–13). ESCRT-I and ALIX function in a partially redundant manner to recruit the late-acting ESCRT-III and VPS4 components of the ESCRT pathway to HIV-1 budding sites (14, 15). These late-acting factors are thought to catalyze the membrane fission reaction leading to virus release (1, 6).

A third type of retroviral L domain, first identified within the p2b region of Rous sarcoma virus (RSV) Gag (16), is defined by a PPXY motif that is thought to recruit members of the NEDD4 family of HECT ubiquitin ligases by providing a ligand for their WW domains (17). Like the function of other types of L domains, PPXY motif-dependent retroviral budding depends on the ESCRT membrane fission machinery, but how this pathway is accessed is not fully understood (1, 2).

Surprisingly, although HIV-1 Gag does not contain PPXY motifs, overexpression of the NEDD4-like ubiquitin ligase NEDD4-2s/NEDD4L efficiently “rescues” the budding defects of HIV-1 L domain mutants unable to engage TSG101 and ALIX (18, 19). NEDD4-2s is a native splice isoform with a truncated C2 domain that is required to rescue HIV-1 release; in contrast, the four WW domains of NEDD4-2s are all dispensable for this activity (18).

More recently, it was shown that HIV-1 Gag can interact with the NEDD4-2s binding partner Angiomotin (AMOT), and that NEDD4-2s and AMOT function together during HIV-1 morphogenesis (20). In cells depleted of AMOT, HIV-1 budding arrested at an earlier stage than in cells depleted of TSG101, suggesting that AMOT is required to complete the immature Gag shell (20). Interestingly, AMOT contains a BAR-like domain, and it has been pointed out that this domain could, in principle, assist Gag in generating membrane curvature during virus assembly (20).

In the present study, we undertook a proteomic analysis to identify endogenous host proteins that are specifically recruited by the intact PPXY-type RSV L domain, but not by an inactive mutant version. Among the host factors that fulfilled these criteria were the NEDD4 family ubiquitin ligase ITCH, several early-acting ESRCT pathway components, and the BAR domain PACSIN2, which has been implicated in the remodeling of membranes and the actin cytoskeleton (21, 22). Our findings indicate that PACSIN2 is also recruited to HIV-1 budding sites by binding to Gag-ubiquitin conjugates induced by HIV-1 p6^gag or by NEDD4-2s. Although PACSIN2 was not required for the completion of a full cycle of HIV-1 replication after infection with cell-free virus, HIV-1 spreading was nevertheless severely impaired in T cells depleted of PACSIN2. Taken together, our results imply that PACSIN2 is important for the cell-to-cell transmission of HIV-1.

Results

A Viral PPXY-Type L Domain Specifically Recruits PACSIN2.

PPXY-type L domains are thought to function as docking sites for the WW domains of NEDD4 family HECT ubiquitin ligases (1). To identify specific host proteins engaged by the PPXY-type RSV L domain, we conducted a proteomic analysis of Optiprep gradient-purified virus-like particles (VLP) produced by the Z_WT-p2b and Z_WT-p2b(Y/G) Gag constructs, which harbor an intact or an inactive RSV L domain, respectively (23). Both Gag constructs are derivatives of Z_WT HIV-1 Gag (Fig. 1A), which has HIV-1 NC and p6 replaced by a leucine zipper (denoted Z in Fig. 1A) and efficiently produces VLP despite the absence of a canonical L domain (23–25). Z_WT-p2b has the p2b peptide of RSV Gag, which contains the PPXY-type L domain, fused to the C terminus of Z_WT (Fig. 1A). Crucially, VLP production by Z_WT-p2b is unaffected by the Y/G mutation, which changes the tyrosine in the PPXY motif to glycine and abrogates the L domain function of RSV p2b (23).

Fig. 1.

RSV p2b and HIV-1 p6 both recruit PACSIN2. (A) Schematic illustration of the HIV-1 Gag constructs used. (B) Western blots showing that the WT but not the inactive Y/G mutant RSV L domain directs the incorporation of PACSIN2 into Z_WT VLP. (C) HIV-1 p6-mediated recruitment of PACSIN2 into Z_WT VLP does not depend on TSG101 binding site in p6. (D) HIV-1 p6-mediated recruitment of PACSIN2 does not depend on ALIX binding site in p6. (E) VLP produced by authentic HIV-1 Gag specifically incorporate PACSIN2.

As expected (23), the Y/G mutation prevented the appearance of Gag-ubiquitin conjugates in Z_WT-p2b VLP, consistent with the recruitment of a ubiquitin ligase only by the intact RSV L domain (SI Appendix, Fig. S1A). Indeed, microsequencing of gradient fractions that contained comparable amounts of Gag (SI Appendix, Fig. S1A) unequivocally identified the NEDD4 family ubiquitin ligase ITCH in VLP produced by Z_WT-p2b but not by Z_WT-p2b(Y/G) (SI Appendix, Fig. S1B and Dataset S1). Additionally, the ESCRT-0 components HGS, STAM, and STAM2; the ESCRT-I component TSG101; and the ALIX homolog PTPN23 were detected only in Z_WT-p2b VLP (SI Appendix, Fig. S1B and Dataset S1), consistent with ubiquitin-mediated recruitment of the ESCRT pathway (1, 26, 27).

Other host proteins that could only be detected in the presence of the intact RSV L domain included the BAR domain protein PACSIN2 and the Eps15 homology domain-containing proteins EHD1 and EHD4 (SI Appendix, Fig. S1B and Dataset S1), which specifically interact with NPF motifs within PACSIN2 (28). Because the BAR domain protein Angiomotin has been implicated in an early stage of HIV-1 budding (20), we examined the incorporation of HA-tagged PACSIN2 into VLP formed by the Z_WT and Z_WT-p2b Gag constructs. This approach confirmed that the WT but not the inactive Y/G mutant RSV L domain directs the incorporation of PACSIN2 into VLP (Fig. 1B).

PACSIN2 Is Recruited by the p6 Domain of HIV-1 Gag.

HIV-1 Gag lacks a PPXY-type L domain and instead contains a PTAP-type L domain within its C-terminal p6 domain that directly interacts with TSG101. In addition, HIV-1 p6 contains a secondary L domain that interacts with ALIX. Interestingly, like RSV p2b (Fig. 1B), HIV-1 p6 fused the C terminus of the L domain-independent Z_WT Gag construct (Fig. 1A) directed the incorporation of HA-PACSIN2 into VLP (Fig. 1C). In this context, the incorporation of HA-PACSIN2 was unaffected by the T₈I point mutation within the PTAP motif near the N terminus of p6 (Fig. 1C), which has been shown to disrupt both TSG101 binding and HIV-1 release (8, 9). Furthermore, the incorporation of HA-PACSIN2 was only moderately affected by point mutations or premature termination codons within a conserved region near the C terminus of p6 (Fig. 1D), most of which (1-41, L₄₁A, L₄₄P, LF₄₅PS) disrupt ALIX binding (11). Thus, the p6-directed incorporation of HA-PACSIN2 did not depend on the presence of intact binding sites for TSG101 or ALIX. However, we noticed that the 1–46 truncation enhanced both the extent of Gag ubiquitination and the incorporation of HA-PACSIN2 (Fig. 1D). In contrast, mutations that reduced the extent of Gag ubiquitination also moderately reduced the incorporation of HA-PACSIN2 (Fig. 1D). Importantly, HA-PACSIN2 was also specifically incorporated into VLP formed by authentic HIV-1 Gag (Fig. 1E), indicating that p6 interacts with PACSIN2 in its native context.

A NEDD4-Like Ubiquitin Ligase Capable of Rescuing HIV-1 Budding Defects Recruits PACSIN2.

The budding defects of HIV-1 mutants with disrupted TSG101 and ALIX binding sites can be potently rescued through the overexpression of the NEDD4 family HECT ubiquitin ligase NEDD4-2s, a natural splice isoform of NEDD4-2/NEDD4L (18, 19, 29). NEDD4-2s appears to interact with HIV-1 Gag in an unusual manner, since its WW domains are not required for the enhancement of HIV-1 budding, and because HIV-1 Gag lacks PPXY motifs (18). Nevertheless, the ability to associate with HIV-1 Gag is critical for the activity of NEDD4-2s in HIV-1 budding, as is the active-site cysteine (Cys801) within its HECT domain (18).

Because VLP production by Z_WT Gag is largely insensitive to NEDD4-2s, and because Z_WT Gag interacts with NEDD4-2s despite the absence of a canonical L domain (29), we used Z_WT to determine whether NEDD4-2s affects the uptake of PACSIN2 into VLP. As expected (29), the overexpression of WT but not of catalytically inactive C₈₀₁S NEDD4-2s led to the appearance of Gag-ubiquitin conjugates in Z_WT VLP (Fig. 2A). Additionally, WT but not C₈₀₁S NEDD4-2s induced the incorporation of HA-PACSIN2 (Fig. 2A), raising the possibility that PASCIN2 was recruited by Gag-ubiquitin conjugates.

Fig. 2.

PACSIN2 recruitment triggered by NEDD4-2s correlates with Gag ubiquitination. (A) Western blots showing that WT but not catalytically inactive NEDD4-2s triggers the uptake of PACSIN2 into Z_WT VLP. (B) Schematic illustration of NEDD4-2s mutants used. (C) Effects of WT and mutant versions of NEDD4-2s and of WWP1s on the appearance of Gag-ubiquitin conjugates and on the uptake of PACSIN2 into VLP. (D) Effects on the uptake of ALIX.

We therefore compared the effects of NEDD4-2s on the incorporation of PACSIN2 and ALIX, which can bind to K63-linked ubiquitin chains (30, 31), the chain type preferentially synthesized by at least some NEDD4-type ubiquitin ligases (32). In addition to WT and catalytically inactive C₈₀₁S NEDD4-2s, we used WW1-4m NEDD4-2s (Fig. 2B), which lacks intact WW domains but nevertheless interacts with HIV-1 Gag and induces Gag-ubiquitin conjugates (18). In contrast, Δ1–31 NEDD4-2s (Fig. 2B), which lacks the residual C2 domain of NEDD4-2s, and WWP1s, a version of the NEDD4-type ubiquitin ligase WWP1 that lacks the exact portion of the C2 domain that is naturally absent from NEDD4-2s, fail to associate with HIV-1 Gag and do not induce Gag ubiquitination (18, 29). We observed that WT and WW1-4m NEDD4-2s, which induced the ubiquitination of Z_WT Gag as expected, robustly induced the incorporation both of PACSIN2 (Fig. 2C) and ALIX (Fig. 2D) into Z_WT VLP. In marked contrast, Δ1–31 NEDD4-2s, C₈₀₁S NEDD4-2s, and WWP1s lacked all of these activities (Fig. 2 C and D). Thus, the ability of the ubiquitin ligases to induce Gag ubiquitination correlated precisely with their ability to induce the incorporation of both PACSIN2 and ALIX (Fig. 2 C and D). This finding supports the notion that PACSIN2, like ALIX, can bind to Gag-ubiquitin conjugates.

PACSIN2 Enhances HIV-1 Spreading.

Pacsins belong to the F-BAR domain family of proteins, which can remodel both membranes and the actin cytoskeleton (33). To determine whether PACSIN2 plays a role in the HIV-1 life cycle, we stably depleted endogenous PACSIN2 in MOLT3 cells by lentiviral delivery of a short-hairpin RNA (shRNA) (Fig. 3A). MOLT3 cells depleted of PACSIN2 expressed normal surface levels of the viral receptors CD4 and CXCR4, as well as the adhesion molecules LFA-1, ICAM-1, ICAM-2, and ICAM-3, which have been implicated in the spreading of HIV-1 between T cells (34) (SI Appendix, Fig. S2). Nevertheless, these cells were poorly permissive for HIV-1_NL43 replication, as monitored by examining Gag protein expression levels in the infected cells by Western blotting (Fig. 3B), or alternatively by measuring the release of p24 antigen over time (Fig. 3C). Similar results were obtained with more permissive CD4^high versions of the same cells (Fig. 3D), which were generated by transduction with a retroviral vector encoding a truncated CD4 without a cytoplasmic domain to limit its endocytosis.

Fig. 3.

PACSIN2 enhances HIV-1 spreading in infected cultures. (A) Western blots showing depletion of endogenous PACSIN2 in MOLT3 cells stably expressing a specific shRNA. (B) HIV-1_NL4-3 replication in MOLT3 cells depleted of PACSIN2 after infection with 2 ng/mL p24. Gag protein expression at day 8 and day 16 after infection (pI) was examined by Western blotting as a measure of virus replication. (C) Virus replication in the same cells examined by monitoring p24 accumulation in the culture media over time. (D) HIV-1_NL4-3 replication in CD4^high MOLT3 cells depleted of PACSIN2. (E) Depletion of endogenous PACSIN2 in primary human PBMC by the indicated shRNAs. (F) HIV-1_NL4-3 replication in PHA-stimulated PBMC depleted of PACSIN2 after infection with 1 ng/mL p24.

In another experiment, we used three different shRNAs that efficiently knocked down PACSIN2 in MOLT3 cells (SI Appendix, Fig. S3A) and observed that in all cases the amount of virus released into the supernatant by day 9 after infection was severely reduced (SI Appendix, Fig. S3B). In addition, we knocked down PACSIN2 in MOLT4 cl. 8 cells, which are highly susceptible to X4-tropic HIV-1 (35). In MOLT4 cl. 8 cells stably depleted of PACSIN2 (SI Appendix, Fig. S4A), Gag protein expression as determined by Western blotting 5 d after infection was markedly reduced (SI Appendix, Fig. S4B), consistent with a defect in virus spreading. Importantly, the replication of HIV-1_NL43 was also markedly attenuated in primary PHA-stimulated human peripheral blood mononuclear cells (PBMC) depleted of PACSIN2 (Fig. 3 E and F).

To exclude an off-target effect, MOLT3 CD4^high cells stably depleted of PACSIN2 were transduced with a retroviral vector encoding shRNA-resistant HA-PACSIN2 (denoted HA-P2*). Western blotting confirmed that endogenous PACSIN2 was replaced by slower migrating HA-P2* in these cells (Fig. 4A). Crucially, HA-P2* fully restored HIV-1_NL43 replication in cells depleted of endogenous PACSIN2 (Fig. 4B).

Fig. 4.

Exogenous PACSIN2 restores HIV-1 replication in MOLT3 cells depleted of endogenous PACSIN2. (A) Western blots showing expression of shRNA-resistant HA-PACSIN2 (HA-P2*) in CD4^high MOLT3 cells stably expressing a nontargeting control shRNA (sh-Ctrl) or a shRNA targeting PACSIN2 (sh_P2_1). (B) Expression of HA-P2* in CD4^high MOLT3 cells depleted of endogenous PACSIN2 rescues HIV-1_NL4-3 replication monitored by p24 ELISA. The cultures were infected with 1.6 ng/mL p24.

An Intact SH3 Domain Is Required for the Effect of PACSIN2 on HIV-1 Replication.

The SH3 domain at the C terminus of PACSIN2 interacts with the cytoskeletal regulators WASP and N-WASP (36–38). In vitro, the SH3 domain of PACSIN2 activates N-WASP and triggers ARP2/3 complex-dependent actin filament formation (39). In vivo, PACSIN1 and PACSIN2 can induce filopodia formation when overexpressed, and in the case of PACSIN1, this activity is disrupted by a point mutation in the SH3 domain that abolishes the interaction with N-WASP (40). To examine the role of the SH3 domain in HIV-1 replication, we transduced MOLT3 cells stably depleted of PACSIN2 with retroviral vectors encoding wild-type or P478L mutant versions of HA-P2*. Although both versions were expressed at comparable levels (Fig. 5A), only wild-type HA-P2* rescued HIV-1_NL43 replication (Fig. 5B). These observations support the notion that the actin remodeling activity of PACSIN2 is involved in HIV-1 replication, since the P478L mutation disrupts the interaction between the SH3 domain of PACSIN2 and WASP/N-WASP (37).

Fig. 5.

The SH3 domain of PACSIN2 is essential for its ability to promote HIV-1 replication. (A) Western blots showing expression of WT and mutant versions of shRNA-resistant HA-PACSIN2 (HA-P2*) in MOLT3 cells depleted of endogenous PACSIN2. (B) Replication curves showing that WT but not P478L HA-P2* rescues HIV-1_NL4-3 spreading in MOLT3 cells depleted of endogenous PACSIN2. The cultures were infected with 1 ng/mL p24.

PACSIN2 Is Not Necessary Following Infection with Cell-Free Virus.

To determine whether PACSIN2 is important for early stages of the HIV-1 replication cycle, PACSIN2 knockdown cells reconstituted or not with wild-type HA-P2* were infected at a high multiplicity of infection with Env-deficient HIV-1_NL43 trans-complemented with Env_NL43. Control infections were carried out in the presence of reverse transcriptase (RT) inhibitors to help distinguish de novo Gag protein expression from Gag derived from input virus. After overnight incubation, the cells were extensively washed to remove input virus. Western blotting after another 3 d of culture revealed that the reintroduction of PACSIN2 into MOLT3 cells depleted of endogenous PACSIN2 did not increase de novo Gag protein expression after infection with single-cycle HIV-1 pseudovirions (SI Appendix, Fig. S5A). This result implied that PACSIN2 is not critical for HIV-1 Env-mediated entry, reverse transcription, or integration.

To examine whether PACSIN2 is critical for later stages of the replication cycle, MOLT3 cells depleted of endogenous PACSIN2 and reconstituted or not with wild-type HA-P2* were incubated with VSVG-pseudotyped HIV-1_NL43 to increase the efficiency of infection. Virus-containing supernatants were either subjected to ultracentrifugation to concentrate progeny virions for Western blotting or used directly to infect TZM-bl indicator cells. We found that the reintroduction of PACSIN2 did not affect the production of progeny virions, the processing of virion-associated Gag, or the levels of virion-associated gp120 and gp41 (SI Appendix, Fig. S5B), and had only a moderate effect on the infectivity of progeny virions (SI Appendix, Fig. S5C). Of note, contamination with input virus was minimal, as determined by incubating cells with VSVG-pseudotyped virus in the presence of RT inhibitors (SI Appendix, Fig. S5B). Together, these observations imply that PACSIN2 is not crucial for the completion of a single cycle of replication after infection with cell-free virus.

Impaired HIV-1 Transmission from MOLT3 Cells Depleted of PACSIN2.

We next used an assay designed to quantify productive virus transmission from an infected donor population to a cocultured target population. The donor cells used in this assay (MOLT3/RFP cells) stably express uniformly high levels of the far-red fluorescent protein RFP657 (SI Appendix, Fig. S6A), and either a control shRNA or a shRNA that knocks down PACSIN2 (SI Appendix, Fig. S6B). The target cells (MOLT3/ZsGreen cells) contain a reporter gene encoding ZsGreen-NLS that is trans-activated by Tat upon infection with HIV-1. Infected donor cells are washed to remove cell-free virus, mixed with reporter target cells, and virus transmission is blocked at defined time points with an entry inhibitor (AMD3100). After further incubation to allow ZsGreen expression, productive virus transmission is quantified by flow cytometry based on the percentage of cells expressing ZsGreen but not RFP657. Of note, HIV-1 Env-mediated fusion of donor and target cells is expected to result in syncytia that express ZsGreen and RFP657.

After infection with replication-competent HIV-1_NL4-3 under conditions that limited virus replication to a single cycle, MOLT3/RFP cells depleted or not of PACSIN2 expressed similar amounts of HIV-1 Gag proteins (SI Appendix, Fig. S6C). Nevertheless, when these cells were cocultured with MOLT3/ZsGreen reporter cells and virus transmission was allowed to proceed for 3 h, productive virus transmission from the donor cells depleted of PACSIN2 was reduced about 15-fold (SI Appendix, Fig. S6D). Importantly, background ZsGreen fluorescence was negligible when AMD3100 was added when the cocultures were initiated (SI Appendix, Fig. S6D). We infer that PACSIN2 is crucial for HIV-1 transmission under conditions where cell-to-cell spreading can occur.

Discussion

Our findings reveal that RSV p2b and HIV-1 p6, which are both involved in the recruitment of the ESCRT pathway, additionally recruit PACSIN2. HIV-1 replication was markedly impaired in T cell lines and in primary PBMC depleted of PACSIN2, and rescue experiments confirmed that PACSIN2 supports HIV-1 replication. PACSIN2, which unlike other PACSINs is ubiquitously expressed, possesses an N-terminal F-BAR domain that binds membranes and can induce curvature, as indicated by its ability to induce the in vitro tubulation of liposomes (41). Interestingly, the cellular protein Angiomotin, which has been proposed to contain a region that functions as a BAR domain (42), has recently been shown to help in the formation of Gag spheres during early stages of HIV-1 budding (20). However, our results indicate that PACSIN2 does not function in HIV-1 budding, and that it is indeed dispensable for the production of infectious progeny virions after infection with cell-free virus. Rather, our results point to a role for PACSIN2 in the cell-to-cell transmission of HIV-1, which constitutes the predominant mode of HIV-1 transfer between T cells (43, 44).

Cryoelectron tomograms of intact human cells expressing HIV-1 indicate that HIV-1 buds are often found adjacent to filamentous actin (45). Indeed, half of the buds analyzed in that study were on the sides or tips of filopodia-like strucures (45). Also, live imaging of infected dendritic cells has revealed an abundance of HIV-1 on the tips of filopodia (46). Such HIV-tipped filopodia are mobile and appear to promote viral replication by scanning the surface of target cells, thereby creating multiple opportunities for the cell-to-cell transmission of HIV-1 (46). Interestingly, PACSINs 1 and 2 can induce filopodia upon overexpression, and PACSIN1 has been shown to localize to the very tips of such filopodia (40). The induction of filopodia by PACSINs depends on their C-terminal SH3 domains (40), which interact with the cytoskeletal effectors N-WASP and WASP (36, 37). Indeed, the SH3 domain of PACSIN1 is sufficient to activate N-WASP and to stimulate ARP2/3 complex-mediated actin polymerization downstream of N-WASP (39). In support of the possibility that the role of PACSIN2 in HIV-1 spreading depends on its actin remodeling activity, we have observed that the ability of exogenous PACSIN2 to rescue HIV-1 replication in PACSIN2-depleted cells is abrogated by a point mutation (P478L) that disrupts the interaction with N-WASP/WASP. However, the P478L mutation also disrupts the ability of PACSIN2 to interact with other potential effectors (37).

A small percentage of HIV-1 Gag is ubiquitinated within the p6 domain (47), and our data suggest that PACSIN2 can interact with such Gag-ubiquitin conjugates. However, the mechanism of p6 domain ubiquitination remains unknown. In the present study, we observed a robust incorporation of PACSIN2 even in the absence of p6 when the ubiquitin ligase NEDD4-2s was coexpressed. We and others have shown that ectopic NEDD4-2s is unique among NEDD4 family ubiquitin ligases in its ability to potently enhance HIV-1 budding in the absence of all known L domains, and that this activity correlates with the induction of Gag-ubiquitin conjugates (18, 19, 29). Similarly, the NEDD4-2s–induced incorporation of PACSIN2 correlated with the appearance of additional Gag species that represent Gag-ubiquitin conjugates, as we have previously shown (18).

Interestingly, ectopic NEDD4-2s also induced the robust incorporation of the ESCRT factor ALIX into Z_WT VLP, although Z_WT Gag lacks NC and p6, the HIV-1 Gag domains known to engage ALIX (48, 49). As in the case of PACSIN2, the NEDD4-2s–triggered incorporation of ALIX depended on the ubiquitin ligase activity of NEDD4-2s and on its truncated C2 domain, which interacts with Gag and is required for the induction of Gag-ubiquitin conjugates (29). In contrast, the WW domains of NEDD4-2s, which function in the recognition of cellular substrates but are dispensable for the induction of Gag-ubiquitin conjugates (18), were dispensable for the NEDD4-2s–triggered incorporation of PACSIN2 or ALIX. Several NEDD4 family members have been shown to preferentially synthetize K63-linked ubiquitin chains, and the ability to promote VLP budding correlates with the ability to induce the formation of K63-linked ubiquitin chains on Gag (29). Furthermore, ALIX is a K63-specific diubiquitin and polyubiquitin binding protein (30, 31). We thus hypothesize that NEDD4-2s can promote HIV-1 budding even in the absence of TSG101 and ALIX binding sites by recruiting ALIX and other ubiquitin-binding ESCRT pathway components indirectly via the formation of K63-linked ubiquitin chains on Gag. Furthermore, the striking parallels between the effects of wild-type and mutant versions of NEDD4-2s on the incorporation of PACSIN2 and ALIX suggest that PACSIN2 also binds to ubiquitin chains.

Consistent with this scenario, we observed that the PPXY-type RSV L domain specifically recruited the NEDD4 family ubiquitin ligase ITCH, which catalyzes K63-linked chains (32), along with PACSIN2 and ESCRT pathway components such as TSG101 and the ALIX homolog HD-PTP. Moreover, while ALIX could be detected even when the L domain was mutated, ALIX nevertheless appeared enriched in the presence of the wild-type PPXY L domain (Dataset S1). Taken together, our observations indicate that PPXY-type L domains promote viral budding by recruiting TSG101 and ALIX in an indirect manner. Furthermore, the observation that the RSV L domain specifically recruits PACSIN2 suggests that retroviruses other than HIV-1 may also depend on this host factor for their efficient transmission.

Materials and Methods

Please see SI Appendix, SI Materials and Methods for information regarding plasmids and retroviral vectors used in this study, and for a description of the analysis of VLP-associated proteins, protein identification, single-cycle replication studies, and the quantification of virus transmission to cocultured reporter cells.

Depletion and Reconstitution of PACSIN2.

MOLT3, CD4^high MOLT3, and MOLT4 cl. 8 cells were transduced with pLKO.1-based lentiviral vectors encoding shRNAs as previously described (50), followed by selection with 1 μg/mL puromycin (Sigma). CD4^high MOLT3 cells were obtained by retroviral transduction with pCXbsrCD4_ΔCT and selection with blasticidin. PBMC were isolated from the blood of healthy donors by Ficoll-Hypaque density gradient centrifugation and immediately transduced with pLKO.1-based lentiviral vectors in the presence of 2.5 μg/mL phytohemagglutinin (Sigma). After 36 h, the culture medium was replaced with medium containing 20 U/mL interleukin 2 (Roche Applied Science) and 2 μg/mL puromycin. Transduced cells were maintained in medium containing puromycin until no viable cells remained in parallel cultures of nontransduced cells that had also been kept in puromycin-containing medium. The pLKO.1-based lentiviral vectors targeting PACSIN2 included clones TRCN0000037980 (here denoted sh_P2_1) and TRCN0000037982 (denoted sh_P2_4), which were purchased from Dharmacon. Additional pLKO.1-based vectors encoding shRNAs targeting PACSIN2 were obtained by inserting annealed oligonucleotides into pLKO.1. The sites targeted by these shRNAs are AGGCAGATGAGCTGGTCATTT (sh-P2-2) and AGACGCAGAACAACAGAAATA (sh_P2_3). In the same manner, pLKO.1-based vectors encoding shRNAs targeting GFP or firefly luciferase were made, which were used as controls. Ectopic HA-PACSIN2 expression cassettes were introduced into MOLT3 cells stably expressing a control shRNA or sh_P2_1 by retroviral transduction with MSCVhygHA-P2* or pCXbsrHA-P2*, followed by selection with hygromycin (Invitrogen) or blasticidin (Sigma). PACSIN2 expression was examined by Western blotting with a rabbit anti-PACSIN2 antibody (GTX104204; GeneTex). Protein loading was assessed with anti-actin antibody AC-40 (Sigma).

Virus Replication Studies.

Replication-competent HIV-1 was produced by transfecting 293T cells with the prototypic infectious molecular clone pNL4-3 (51). Additionally, the nef-deficient variant NL4-3/nef⁻ (52) was used in the experiment shown in SI Appendix, Fig. S4B. Virus-containing supernatants were passed through 0.45-μm filters, normalized for p24 antigen with a HIV-1 p24 ELISA kit (PerkinElmer), and used to infect target cells in T25 flasks at a p24 concentration of 1–2 ng/mL. Virus replication was monitored by comparing Gag protein levels in the infected cells by Western blotting using anti-CA antibody 183-H12-5C and by measuring p24 antigen in the culture supernatants by ELISA.