In this study, we have used density gradient centrifugation to resolve distinct intermediates in the process of HIV-1 virion particle assembly. A combination of pulse-chase analysis and biochemical techniques allowed us to temporally order these intermediates and to characterize their sizes and intracellular localizations. Specifically, we have shown a number of differences between the behavior of the newly synthesized and steady-state (total) Gag populations. First, newly synthesized Gag migrates to lighter fractions of an Optiprep velocity gradient than does total Gag (Fig. A and B). In addition, a substantial portion of the newly synthesized Gag remains in the middle of the Optiprep gradient after treatment with nonionic detergent, displays protease sensitivity, and remains on the bottom of sucrose flotation gradients. In contrast, nearly all of the total Gag population shifts to the bottom of the Optiprep gradient after detergent treatment, is resistant to protease, and floats up through sucrose gradients (Fig. , B, and ). (For the sake of simplicity, the population of newly synthesized Gag that is protease sensitive, etc., will be referred to as “population A”; the population representative of steady-state Gag will be designated “population B”). We have also shown that population A disappears within 2 h of synthesis, with no corresponding increase in population B or in extracellular VLPs (Fig. ). Population B, by contrast, disappears with kinetics roughly paralleled by the appearance of Gag in VLPs. Finally, we have shown that, as a function of time, intracellular Gag migrates to increasingly dense fractions on Optiprep gradients and that after 6 h, the mobility of these fractions resembles that of VLPs (Fig. and ). Taken together, these data suggest that a substantial proportion of newly synthesized Gag exists in population A, which does not proceed to true assembly intermediates but instead is presumably targeted for degradation. Moreover, the remaining Gag protein, population B, consists of a number of membrane-bound assembly intermediates. These assembly intermediates can be resolved by velocity centrifugation. Nearly identical results were obtained in Jurkat cells, where the amount of Gag is 20-fold lower than in COS-1 cells. Thus, the level of Gag expression does not appear to influence the overall assembly pathway.
The majority of total Gag at steady state is membrane bound.
We have used three independent biochemical criteria to demonstrate that most of the total Gag population present in the P100 fraction is indeed membrane bound. First, addition of nonionic detergent resulted in a dramatic shift of nearly all the Gag to the bottom of an Optiprep velocity gradient, as expected for a large, membrane-bound protein complex (Fig. B). Second, the majority of the total Gag in the P100 fraction displayed protease resistance, as previously reported (16
), and became protease sensitive only when the membranes were permeabilized with detergent (Fig. ). These findings imply that Gag is enveloped in membrane vesicles. It is unlikely that cytosolic complexes of Gag are trapped inside membrane vesicles during homogenization, because β-Gal, a cytosolic protein which forms a tetramer of ~440 kDa, fractionates primarily in the S100 fraction (Fig. B). It is interesting to note that Gag was mostly protease resistant while another protein that is also bound to the inner surface of the plasma membrane, Raf–K-Ras (6
), was mostly degraded (not shown). The difference in the protease sensitivities of the two proteins may reflect Gag-induced curvature of the membrane, favoring right-side-out vesicularization of the Gag-containing vesicles. Regions of the plasma membrane containing Gag multimers would exclude other cellular membrane proteins and vesicularize separately, which also explains the ability of assembly domains at different stages to be resolved by centrifugation (see below). Protease resistance provides a conservative estimate of the proportion of Gag that is membrane bound, since some Gag is likely to be associated with inside-out plasma membrane vesicles. The third assay involved flotation through sucrose cushions. Nearly all the steady-state Gag floated to the interface, as expected for a membrane-bound protein (Fig. B). Taken together, these data clearly indicate that the majority of the total Gag population is membrane bound.
Several investigators have recently used detergent-resistant migration on sucrose gradients and failure to float on sucrose cushions to argue that a large proportion of Gag is present in intracellular cytosolic complexes (20
). Our data, as well as that of Spearman et al. (30
), strongly argue that the majority of the total Gag at steady state is membrane bound. The greater resolution afforded by gradients spanning a narrower range of densities increased the sensitivity of our assays. Other investigators used sonication during membrane preparation (23
), which may have resulted in the release of membrane-bound proteins. Moreover, some investigators analyzed only the pelleted material from sucrose gradient fractions (20
), whereas we utilized the entire sample for analysis. We cannot, however, exclude the possibility that differences in cell type, virus subtype, or membrane preparation methods account for differences in the experimental results. It is also possible that the presence of HIV-1 protease could influence the assembly pathway, although this would likely occur during the late stages of assembly (17
Newly synthesized Gag rapidly forms cytosolic complexes that copellet with the P100 fraction.
The time course experiments shown in Fig. demonstrate that newly synthesized Gag rapidly localizes to the P100 fraction in a manner that is independent of other viral proteins and downstream Gag oligomerization domains. Several criteria were used to distinguish membrane-bound Gag from cytosolic Gag complexes in the P100 fraction. Approximately 20% of the pulse-labeled Gag floated to the interface of a discontinuous sucrose gradient (Fig. ) and shifted to the pellet of a continuous Optiprep gradient when detergent was added (Fig. ). These are characteristics of membrane-bound proteins and likely reflect the presence of a population of plasma membrane-bound protein among the newly synthesized Gag. However, the majority of pulse-labeled Gag exhibited sensitivity to protease digestion and remained at the bottom of the tube in sucrose flotations (Fig. and ), suggesting that most of the newly synthesized Gag exists in cytosolic complexes that are detergent resistant (Fig. ).
Most of the newly synthesized Gag is degraded intracellularly.
Quantitation of the levels of Gag at different times after synthesis revealed that ~80% of the Gag disappeared within 2 h, followed by slower disappearance of the remaining Gag (Fig. B and C). The degradation of most of newly synthesized Gag explains the low efficiency of HIV-1 budding that has been observed (34
). The rapid disappearance in the first 1 to 2 h occurred selectively in population A. In the sucrose flotation assay, the labeled Gag at the bottom of the tube was depleted within 2 h (Fig. A). Similar trends were observed in the Optiprep gradients. In the presence of detergent, the Gag population in the middle fractions disappeared almost entirely within 4 h, while in the absence of detergent, Gag counts per minute in the lighter fractions of the gradient (~1.06 g/ml) were selectively depleted (Fig. A). These data indicate that population A has a relatively short half-life (less than 1 h) while population B has a half-life of several hours. Most of the total Gag population is present in population B.
Optiprep gradients resolve membrane-bound assembly intermediates.
Pulse-chase analysis combined with Optiprep gradients (Fig. A) revealed that Gag progressively migrated to denser gradient fractions as a function of time. It should be noted that a subtle density difference between total and newly synthesized Gag was previously reported (32
) but was not further characterized. Several lines of evidence indicate that increasing assembly drives the progression to denser gradient fractions. (i) An assembly-deficient Gag mutant fails to display the same behavior (Fig. B). (ii) Gag takes 4 to 6 h to migrate through the gradient, similar to the kinetics of its appearance in VLPs (Fig. A and D). (iii) At later chase points, Gag migrates almost identically to VLPs, suggesting that the Gag-containing cellular domains resemble VLPs (Fig. B). (iv) The Gag in the denser fractions exists entirely in large complexes (Fig. B). We therefore believe that these gradients separate complexes based on their degrees of assembly. Chemical cross-linking with a homobifunctional cross-linker confirmed that Gag multimers were present in the P100 fraction (Tritel and Resh, unpublished).
Taken together, these data are consistent with the following model for lentivirus assembly. Most of the newly synthesized Gag proceeds to population A, which consists of cytosolic complexes that copellet with the P100 fraction. These complexes are characterized by protease sensitivity, resistance to detergent, and failure to float through a sucrose cushion. Population A is rapidly degraded, and the rate of disappearance is not paralleled by the appearance of Gag in VLPs. The rest of the newly synthesized Gag is present in population B, consisting of large, membrane-bound complexes that also pellet with the P100 fraction. Gag appears in VLPs at approximately the same rate as the loss of population B (Fig. B and D). Thus, it is likely that population B represents true assembly intermediates, although the possibility that a minority of the Gag in population A proceeds to population B or VLPs cannot be excluded. At the plasma membrane, Gag complexes undergo increasing multimerization, producing large, dense arrays of Gag protein under the membrane. As Gag multimerization continues, the membrane is deformed outward, host cell proteins are excluded, and other viral components are recruited to the site of assembly. As a result, the protein-to-lipid ratio, and therefore the density of the Gag-containing membrane assembly domains, increases. The sequential assembly complexes isolated on the Optiprep gradients (Fig. ) therefore likely reflect increasing formation of Gag multimers within assembly domains at the membrane.
In conclusion, we report the novel separation and characterization of sequential in vivo HIV-1 assembly intermediates. These intermediates are large, membrane-bound, Gag-containing complexes. The length of time required for assembly may reflect the transport of Gag to specialized domains in the plasma membrane, Gag multimerization, or recruitment of other viral components into the budding particle. Clearly, the processes involved represent potential targets for therapeutic intervention. Future studies will exploit this system for monitoring HIV-1 assembly to test the role of energy metabolism and identify the cellular and viral components required for assembly.