In this study, we used CD4-Gag chimeras to demonstrate that HIV-1 Gag contains a dileucine-like endocytosis signal that promotes rapid internalization. There is now abundant evidence in the literature to indicate that retroviral Gag proteins interact with proteins involved in various stages of endocytosis (16
). For example, HIV-1 Gag binds to Tsg101, RSV Gag binds to Nedd4, and EIAV as well as HIV-1 Gag proteins bind to AIP1/ALIX. Moreover, Gag proteins have been shown to localize to and/or bud from endocytic compartments. Moloney murine leukemia virus Gag is present in lysosomes and recycling endosomes, and HIV-1 is localized to and buds from MVBs in macrophages (5
). The mechanism whereby Gag becomes localized to endocytic compartments is not known. One possibility is that Gag may first bind to the plasma membrane and then localize to endocytic compartments as a consequence of internalization. Our finding that a mutated Gag protein that lacks a functional endocytosis signal, Gag(IL-333,334AA), does not localize to MVBs supports the latter hypothesis. We propose that the internalization of Gag from the plasma membrane via an endocytic pathway could account for the presence of Gag in endocytic compartments and could facilitate the interactions of Gag proteins with endocytic proteins. Alternatively, newly synthesized Gag may first bind directly to MVBs and then later traffic to the plasma membrane, as has recently been proposed (43
). In this case, the IL motif would be a signal for MVB targeting. It is also possible that the effect of the IL mutation on VLP production is distinct from its effect on Gag trafficking.
The region of Gag that promoted rapid internalization of a CD4-Gag chimera contained the C-terminal domain of CA, p2, and NC. This same region also promotes the multimerization of Gag. Experiments performed with CD4T(−) fused to DsRed, a protein with a well-documented ability to form tetramers (4
), suggest that multimerization alone is not sufficient to drive rapid internalization of CD4 chimeras. As depicted in Fig. , despite the presence of large aggregates of CD4T(−)DsRed constructs on the cell surface, increased rates of internalization for these chimeras were not observed. It does, however, remain possible that the CD4-DsRed constructs are not themselves multimerized but, rather, are associated with large cellular aggregates. Thus, the possibility that, within the context of Gag, multimerization may play a role in internalization remains.
Gag containing the LL321,322AA mutation was defective for membrane binding, which may in part be due to decreased multimerization. In contrast, membrane flotation analysis as well as confocal imaging indicated that the IL-333,334AA mutant was clearly plasma membrane bound, in the contexts both of Gag itself and a CD4T(−)Gag chimera. Several lines of evidence support the notion that IL-333,334AA Gag is also capable of forming Gag-Gag multimers. This construct could be rescued into VLPs by coexpression with wild-type GagGFP. The extent of rescue was as great as or greater than that observed for the rescue of nonmyristoylated G2A Gag into VLPs by wild-type GagGFP (Fig. ). Moreover, the localization patterns of IL-333,334AA Gag as well as the CD4T(−)IL-333,334AA Gag chimera were clearly punctate, consistent with the presence of multimerized Gag (19
). It remains possible that the IL-333,334AA mutant has a subtle effect on multimerization that cannot be detected by these individual assays. Alternatively, the defect in VLP production exhibited by IL-333,334AA Gag could be due to defective association with components of the endocytic pathway.
There are at least three well-described mechanisms for internalizing proteins and lipids from the plasma membrane, including endocytosis via clathrin-coated pits, caveolae, and rafts. Our finding that coexpression of Gag with dominant-negative dynamin had no effect on Gag localization or VLP production implies that Gag internalization does not proceed via clathrin-coated pits or caveolae. Disruption of lipid rafts by treatment of cells with cyclodextrins is known to inhibit viral particle production (27
). However, rafts are required for the HIV-1 assembly process (19
), and it is not known whether the cyclodextrin effect is due to the inhibition of raft- and/or non-raft-mediated endocytosis (39
), inhibition of Gag-mediated assembly, or both. It is interesting that multiple new pathways for internalization, including dynamin-independent pathways, have been described recently (for example, see reference 25
One of the unanswered questions in the field of virus assembly is how the location for particle production is chosen. It is well established that, in many cell types, HIV-1 Gag is primarily localized at the plasma membrane. However, recent evidence clearly implicates the MVB as a site for Gag localization and particle production in macrophages. Our data (Fig. ) strongly suggest that Gag distribution is regulated by the intracellular distribution of cholesterol. Treatment of cells with U18666A had a striking effect on Gag localization, causing a dramatic accumulation of Gag in CD63-positive MVBs. Particle production occurred at levels equivalent to those for untreated cells. These results imply that a COS-1 cell can be induced to behave like a macrophage with respect to virus assembly by redistributing cholesterol from the plasma membrane to the MVB.
Three HIV-1 proteins have been shown to interface with the cellular endocytic machinery: Env, Nef, and Vpu. Env has a tyrosine-based internalization motif (10
). In cells that express Env alone, Env internalization is rapid, whereas in HIV-1-infected cells, Env internalization is very slow. The presence of Gag is both necessary and sufficient for the inhibition of Env internalization (10
). It is possible that interaction of Gag with endocytic sorting factors reduces the availability of these components for binding to other proteins, resulting in a reduced rate of Env internalization.
In conclusion, the data presented in this paper strongly suggest that HIV-1 Gag contains a dileucine-like signal that mediates the trafficking of Gag from the plasma membrane to the MVB. This signal is independent of the L domain, which has been shown to interact with endosomal proteins. The steady-state distribution of Gag in COS-1 cells is primarily at the plasma membrane, implying that either the rate of Gag endocytosis is slow or the rates of trafficking out of MVBs (recycling to the plasma membrane or entry into lysosomes) are rapid. The situation is different for macrophages, in which Gag is found in MVBs. We propose that the steady-state subcellular localization of Gag, and therefore the site of VLP formation, can be regulated by the relative distribution of cellular cholesterol in a particular cell type. Moreover, these results imply that the internalization of Gag is functionally important, as mutation of the endocytosis signal blocks the ability of Gag to target MVBs and also inhibits VLP production.