The Late Stage of Human Immunodeficiency Virus Type 1 Assembly Is an Energy-Dependent Process

Abstract

Several recent studies have indicated the involvement of host cell factors in human immunodeficiency virus type 1 (HIV-1) assembly. To ascertain whether ATP-dependent factors play a role in this process, we quantified virus-like particle (VLP) production by ATP-depleted cells. Pharmacological ATP depletion abrogated VLP production without affecting cell viability or inducing degradation of HIV-1 Gag protein. This effect occurred even when the ATP-depleting agents were added 1 h into the assembly process, and it was reversed by removal of these agents. ATP depletion did not affect Gag membrane binding or multimerization. Density gradient analysis indicated that HIV-1 assembly intermediates were stalled late in the assembly process. This conclusion was further supported by electron microscopy analysis, which revealed a preponderance of plasma membrane-associated stalk-like structures in the ATP-depleted cells. Since no HIV-1 proteins bind or hydrolyze ATP, these findings indicate that an ATP-requiring cellular factor is an obligatory participant late in the HIV-1 assembly process.

Assembly of human immunodeficiency type 1 (HIV-1), as well as all other lentiviruses and retroviruses, is directed by the Gag protein (11–13). Cells expressing the Gag precursor in the absence of other viral proteins can produce virus-like particles (VLPs) (15), implying that Gag contains all of the determinants necessary for assembly. In addition, Gag recruits other HIV-1 proteins (11, 12, 45) and viral RNA (16) into nascent virions. Gag is synthesized as a polyprotein precursor, Pr55^gag, and is cleaved around the time of budding into its component subunits by HIV-1 protease (18, 21, 22). Cleavage of Gag generates four major subunits, which play structural roles in the virion. p17 MA lines the inner leaflet of the viral lipid envelope, p24 CA forms a cone-shaped capsid around the viral genome, and p7 NC coats the viral RNA. The structural role of the fourth domain, p6, is poorly understood.

Several regions of the Gag precursor are required for assembly and budding. The M domain, contained in the 31 N-terminal residues of p17 MA, mediates posttranslational binding to the inner leaflet of the plasma membrane (47). Both myristate, attached to the N-terminal glycine residue, and a cluster of basic amino acids between residues 15 and 31 are required for efficient plasma membrane targeting (19, 20, 47). Gag multimerization is mediated by the I domain, which extends from the C-terminal portion of CA to the N-terminal portion of NC (27, 35, 36). NC also contains two zinc finger motifs that bind to the viral RNA during assembly (2). Finally, the L domain, located in p6, is necessary for the pinching off of HIV-1 virions from virus-producing cells (17, 30).

Some of the first clues about the mechanism of HIV-1 assembly came from electron microscopy (EM) studies. Large Gag multimers can be visualized as electron-dense patches under the plasma membrane that deform the membrane outward as they grow (14). More-advanced intermediates appear as spheres connected to the cell by a thin stalk. Virions that have just pinched off from the cell retain the electron-dense layer under their lipid envelope and have a doughnut-shaped appearance. Cleavage of Gag triggers structural changes that produce mature virions containing the characteristic cone-shaped capsid.

We have recently described the use of pulse-chase analysis combined with density gradient centrifugation to dissect the HIV-1 assembly process into several stages (42). We showed that a subpopulation of newly synthesized Gag binds rapidly to membranes, forming membrane-bound assembly intermediates. The assembly process is accompanied by a progressive increase in size and/or density of the Gag complexes over a period of 4 to 6 h and can therefore be monitored by tracking Gag migration during density gradient centrifugation. The increase in mobility on the gradients likely reflects increasing Gag multimerization, although other processes may be involved as well. The final stages of assembly can be monitored by tracking the release of the labeled Gag into extracellular VLPs.

The host cellular requirements for the various stages of the HIV-1 assembly process have not been clearly defined. In vitro studies have suggested that RNA, ATP, Mg²⁺, and/or undefined cellular factors may be required (4, 23, 26), but the relevance of these findings must be demonstrated in vivo. As a first step toward this goal, the present study examines the effect of cellular ATP depletion on HIV-1 assembly, using biochemical techniques described previously (42). Here we show that ATP depletion rapidly and reversibly halts VLP production. In addition, biochemical and EM studies indicate that many of the assembly complexes are arrested at a late stage in the assembly process. These findings suggest that an ATP-requiring cellular factor is likely to participate in a late stage of the HIV-1 assembly process.

MATERIALS AND METHODS

Plasmids and reagents.

Plasmid pHXB2gtpΔBal-D25S (pHXB2ΔBal) (47), a noninfectious HIV-1 proviral construct with a 2.2-kb deletion in pol and a protease-inactivating point mutation, was a kind gift from L. Parent and J. Wills (Pennsylvania State University Medical School, Hershey). ECL Western blotting reagents and secondary antibodies conjugated to horseradish peroxidase were purchased from Amersham-Pharmacia (Piscataway, N.J.). Tran³⁵S-label was obtained from ICN or NEN. Optiprep was obtained from Gibco Life Technologies (Rockville, Md.). 2-Deoxyglucose and NaN₃ were obtained from Sigma (St. Louis, Mo.).

Antibodies.

Rabbit anti-p24 CA antiserum or human anti-HIV immune globulin from the National Institutes of Health (NIH) AIDS Research and Reference Reagent Program was used to detect Pr55^gag.

Transfection, metabolic labeling, and VLP budding assays.

COS-1 cells were maintained as previously described (43). Confluent cells were transfected with pHXB2ΔBal by using Lipofectamine 2000 (Gibco Life Technologies). The cells were passaged approximately 24 h later and were harvested 36 to 48 h after transfection. Metabolic labeling was performed as previously described (43), using 50 μCi/ml Tran³⁵S-label. Cells were pulse-labeled for 20 min at 37°C and then chased for various lengths of time in chase medium (Dulbecco's modified Eagle medium containing 5% fetal bovine serum and 100 μM each cysteine and methionine). ATP depletion was performed by washing cells in glucose-free chase medium and incubating them in glucose-free chase medium containing 10 mM 2-deoxy-d-glucose and 10 mM sodium azide (Sigma). Cell viability was determined by the trypan blue exclusion assay.

VLPs were purified from the chase medium as previously described (10). The VLP pellet was solubilized in radioimmunoprecipitation assay (RIPA) buffer (150 mM NaCl, 1 mM EDTA, 0.1% sodium dodecyl sulfate [SDS], 0.5% deoxycholate, 1% Triton X-100, 10 mM Tris, pH 7.4) containing protease inhibitors and immunoprecipitated as previously described (42). Cells were harvested by washing them twice in saline-Tris-EDTA (STE) buffer (150 mM NaCl, 10 mM Tris, 1 mM EDTA, pH 7.4) and lysing them in RIPA buffer. The lysate was clarified at 100,000 × g and 4°C for 15 min and immunoprecipitated. SDS-polyacrylamide gel electrophoresis (PAGE) was performed as previously described (43). Analysis of radiolabeled Gag was performed by exposure to phosphorimager screens, which were scanned using a Storm apparatus (Molecular Dynamics, Sunnyvale, Calif.). Quantitation and preparation of visual images were performed with ImageQuant software (Molecular Dynamics).

Sucrose flotation assays and Optiprep gradient fractionations.

Flotation assays were performed by a previously described protocol with slight modifications (39, 42). Briefly, transfected cells were metabolically labeled for 5 min and chased for various lengths of time in the presence or absence of ATP-depleting agents. P100 fractions depleted of nuclei were adjusted to 72% sucrose and overlaid with 1.5 ml of 65% (wt/vol) sucrose and 1.5 ml of isotonic buffer. All solutions contained 1 mM EDTA, 10 mM Tris (pH 7.4), and protease inhibitors. Centrifugation was performed in an SW55 rotor for 2 h at 200,000 × g and 4°C. Fractions were collected from the top of the tube.

Optiprep gradient fractionations were performed as previously described (42). Briefly, nucleus-depleted P100 fractions were resuspended by Dounce homogenization in 1 ml of an isotonic buffer containing protease inhibitors and layered on top of a 0 to 18% Optiprep (Gibco) gradient containing 0.25 to 0.18 M sucrose, 1 mM EDTA, 10 mM Tris [pH 7.4], and protease inhibitors. For Gag multimerization assays, NP-40 was added to the isotonic buffer to a final concentration of 1% after resuspension of the P100 fraction, and the sample was layered over an Optiprep gradient containing 0.1% NP-40. The gradients were centrifuged for 3 h at 37,000 rpm (100,000 × g) and 4°C in an SW40 rotor (Beckman, Columbia, Md.), and 0.75-ml fractions were collected after the bottom of the tube was punctured. Aliquots from the gradient fractions and the gradient pellet were added to 1× RIPA buffer for immunoprecipitation or to SDS sample buffer for Western blotting.

Western blotting was performed as previously described (43). Western blots were exposed to BioMax MR film (Kodak, New Haven, Conn.) and scanned with an Epson scanner, and the bands were quantified using the MacBAS program.

Measurement of cellular ATP.

COS-1 cells were washed three times in STE at 4°C and then scraped into STE at a density of 5 × 10⁵/ml for analysis with the FL-ASC bioluminescent cellular ATP assay (Sigma) according to the manufacturer's instructions. Luminescence was measured in a Lumat LB 9507 luminometer. ATP levels were confirmed to be within the linear range of the assay.

EM analysis.

pHXB2ΔBal-transfected COS-1 cells were depleted of ATP for 3.5 h, washed three times in STE, scraped into STE, and pelleted for 5 min at 500 × g. The pellet was fixed with 2.5% glutaraldehyde in piperazine-N,N′-bis(2-ethanesulfonic acid) (PIPES) buffer, postfixed with 2% osmium tetroxide in PIPES, embedded in paraffin, and stained with 5% uranyl acetate and 0.4% lead citrate. Ultrathin sections (60 nm) were generated, and the samples were examined using a Jeol 1200 EX transmission electron microscope.

To calculate the frequency of HIV-1 assembly intermediates, single sections from approximately 20 untreated and 20 ATP-depleted cells were examined. The surface area of plasma membrane per cell profile was calculated by multiplying the circumference of the cells by the thickness of the sections. Untransfected cells, both untreated and ATP depleted, were also examined; no virus-like structures were observed.

RESULTS

Cellular ATP depletion reversibly halts the HIV-1 budding process.

As a first step toward identifying cellular factors required for the HIV-1 assembly and budding process, we examined whether cellular ATP is required for budding. COS-1 cells were transfected with the noninfectious HIV-1 proviral construct pHXB2ΔBal (47), which expresses Gag, Env, and most HIV-1 accessory proteins and contains a deletion in pol and a point mutation that inactivates the viral protease. The cells were incubated in glucose-free medium containing 2-deoxyglucose and NaN₃, a cocktail that has been shown to rapidly and reversibly deplete cellular ATP (44). As shown in Fig. 1A, cellular ATP levels dropped to 25% of their original level within 5 min of treatment and to 12% of the original level by 10 min. As previously reported (44), there was no change in cell viability after 4 h of treatment (data not shown).

FIG. 1.

Cellular ATP depletion inhibits HIV-1 VLP production. (A) Kinetics of cellular ATP depletion. COS-1 cells were incubated in glucose-free medium containing 2-deoxyglucose and NaN₃. Cellular ATP levels were quantified, as described in Materials and Methods, before treatment began and at various times after treatment. ATP levels are plotted as a percentage of the original level. (B) VLP production in untreated and ATP-depleted cells. Transfected COS-1 cells were pulse-labeled for 20 min with Tran³⁵S-label and chased for various time periods (5 min, 1 h, 2 h, or 4 h) either in glucose-containing medium (untreated; squares) or in glucose-free medium containing 2-deoxyglucose and NaN₃ (ATP depleted; diamonds). VLPs were isolated from the medium, and radioactivity from the labeled Gag protein in cells and VLPs was quantified by RIPA with anti-CA. VLP production was expressed as the ratio of the percentage of Gag radioactivity found in VLPs to the total Gag radioactivity (cells plus VLPs).

To measure the dependence of HIV-1 budding on cellular ATP, we utilized a quantitative budding assay that we have previously described (42). Transfected cells were pulse-labeled with [³⁵S]Met-Cys for 20 min to label newly synthesized Gag and then chased for various lengths of time (5 min, 1 h, 2 h, and 4 h) in the presence of excess unlabeled Met-Cys and either glucose-containing medium (untreated) or glucose-free medium containing 2-deoxyglucose and NaN₃ (ATP depleted). This protocol ensures that identical amounts of radiolabeled Gag protein are present in both the treated and untreated samples at the start of the chase period. VLPs were isolated, and labeled Gag protein in cells and VLPs was quantified. As depicted in Fig. 1B, VLP production was decreased by 72% at 2 h and by 83% at 4 h in the ATP-depleted cells (n = 9). To ensure that the treatment with 2-deoxyglucose and NaN₃ was not having effects in addition to ATP depletion, cells were treated with one or the other agent alone. A 4-h treatment with either agent had a partial effect on VLP production, with 2-deoxyglucose causing a 42% decrease and NaN₃ causing a 23% decrease, respectively (data not shown). Likewise, neither agent alone decreased ATP levels to the same extent as they did in combination. Thus, both agents are required to block budding.

To ascertain whether the block in budding was reversible, cells were incubated with ATP-depleting agents for the first 2 h of the chase period and then washed and incubated in glucose-containing medium for an additional 4 h (reversed samples). Cellular ATP was restored to approximately 32% of the original level within 45 min and to 52% of the original level after 4 h in reversal medium (Fig. 2A). VLP production was restored in the “reversed” samples, with Gag levels in VLPs reaching 57% of the levels in untreated control cells (Fig. 2B). The restoration of VLP production reflects the partial recovery of ATP levels. Taken together, the data from Fig. 1 and 2 indicate that ATP depletion induces a reversible block in HIV-1 budding. This finding implies that cellular ATP is required for one or more stages in the HIV-1 assembly and budding process.

FIG. 2.

VLP production is restored by reversal of cellular ATP depletion. (A) Rebound in cellular ATP levels. COS-1 cells were depleted of ATP for 2 h as described in the legend to Fig. 1A and then washed and incubated in glucose-containing medium. Cellular ATP was quantified at various time points after addition of the glucose-containing medium and is plotted as a percentage of the original ATP level (time zero). (B) VLP production after removal of ATP-depleting agents. Transfected COS-1 cells were pulse-labeled for 20 min with Tran³⁵S-label, chased for 2 h in the presence of ATP-depleting agents, washed, and chased for an additional 2 or 4 h in glucose-containing medium (depleted/reversed; circles). For untreated (squares) or ATP-depleted (diamonds) cells, the entire chase was performed in glucose-containing medium or with ATP-depleting agents, respectively. VLP production was expressed as described in the legend to Fig.1. For the depleted/reversed samples, VLPs were isolated from both the chase medium from the first 2 h of the chase and the glucose-containing chase medium from the subsequent chase time. Counts of labeled Gag from the two medium samples were combined and then divided by the total Gag counts (cells plus VLPs) to obtain the value for VLP production.

To investigate the possibility that ATP depletion alters the level of intracellular Gag protein, the amounts of Gag in untreated and ATP-depleted cells were compared. Cells were pulse-labeled and then chased either in complete medium or in glucose-free medium containing ATP-depleting agents, as described for Fig. 1B. As we and others have previously shown (37, 42), approximately 80% of the total cellular Gag protein was degraded in the first 2 h in untreated cells. Interestingly, there was less degradation of intracellular Gag in the ATP-depleted cells (Fig. 3). At the 2-h time point, cells that had been depleted of ATP contained 33% more Gag than untreated cells. Between 2 and 4 h, intracellular Gag levels declined at approximately the same rate in untreated and ATP-depleted cells (Fig. 3). This finding suggests that ATP depletion partially prevents intracellular degradation of Gag. Thus, the inhibition of budding by ATP depletion is not due to a loss of intracellular Gag protein.

FIG. 3.

Cellular Gag counts in untreated and ATP-depleted cells. Transfected COS-1 cells were pulse-labeled and chased in control medium (untreated; squares) or in medium containing ATP-depleting agents (depleted; diamonds) as described in the legend to Fig. 1B. Cell lysates were immunoprecipitated with anti-CA and analyzed by SDS-PAGE and exposure to phosphorimager screens. Counts from labeled cellular Gag were expressed as a percentage of the counts in the pulse-labeled cells.

Gag membrane binding is unaffected by cellular ATP depletion.

We next examined the stage at which the assembly and budding process required ATP. Previous work in our laboratory had shown that the first stage in the HIV-1 assembly process is membrane binding of newly synthesized Gag protein (42). To determine whether Gag membrane binding requires ATP, transfected cells were metabolically pulse-labeled for 5 min and then chased for 2, 20, or 60 min in the presence or absence of ATP-depleting agents. P100 fractions were further fractionated by sucrose flotation to separate membrane-bound proteins, which float to the interface, from cytosolic proteins, which remain at the bottom of the tube (20, 28, 29, 39). Most of the labeled Gag from the cells chased for 2 min remained at the bottom of the tube (Fig. 4A, top panel), indicating that newly synthesized Gag is cytosolic. Longer chases of the control cells resulted in a progressive increase in the amount of Gag that floated to the 10 to 65% sucrose interface (Fig. 4A, untreated), as previously reported (42), reflecting an increase in the percentage of Gag that was membrane bound. Gag protein from the ATP-depleted cells exhibited essentially the same pattern as the untreated-cell Gag (Fig. 4). These results indicate that ATP depletion has no effect on Gag membrane binding.

FIG. 4.

ATP depletion of cells does not affect Gag membrane binding assay. Transfected COS-1 cells were pulse-labeled for 5 min and chased for 2, 20, or 60 min in control medium (untreated) or in medium containing ATP-depleting agents (ATP depleted). Denucleated P100 fractions were prepared and subjected to sucrose flotation as described in Materials and Methods. Labeled Gag in the gradient fractions was quantified after immunoprecipitation, SDS-PAGE, and phosphorimaging. (A) A representative experiment showing labeled Gag in each gradient fraction. (B) Graphic depiction of data from three independent experiments. The counts from the labeled Gag in the interface fractions (fractions 4 and 5) were divided by the total counts in the gradient fractions to determine the percentage of Gag that was membrane bound.

Gag multimerization is unaffected by cellular ATP depletion.

The next stage of HIV-1 assembly is multimerization of Gag protein at the plasma membrane (42). A density gradient centrifugation assay that resolves Gag multimers was used to determine whether Gag multimerization was impaired upon ATP depletion. Cells were pulse-labeled for 5 min and either harvested immediately or chased for 2 h in the presence or absence of ATP-depleting agents. P100 fractions were generated and then treated with 1% NP-40 to release membrane-bound material, as described previously (42). Gag multimers were separated by density on linear Optiprep gradients. Most of the Gag from the pulse-labeled cells migrated about one-third of the way through the gradient. The remainder, 14% of the total counts, migrated to the bottom fraction (Fig. 5, top panel). In the untreated cells chased for 2 h, most of the labeled Gag was found in the bottom fraction, reflecting increased Gag multimerization (Fig. 5, middle panel). The pattern of Gag distribution in the gradients from the ATP-depleted, chased cells was similar to that of the untreated, chased cells (Fig. 5; compare middle and bottom panels). Quantitation revealed no significant change in the proportion of Gag in the bottom fraction (mean ± standard deviation, 60% ± 13% for the depleted cells versus 62% ± 0.4% for the untreated cells). The steady-state distribution of Gag protein in the gradients was determined by Western blotting with anti-CA antiserum. This analysis revealed no discernible difference between treated and untreated cells. Most of the total Gag protein was at the bottom of the gradient (data not shown). These findings suggest that most or all of the Gag multimerization process is ATP independent. However, since this assay is rather insensitive, it is possible that subtle changes in Gag multimerization may not have been detected.

FIG. 5.

ATP depletion does not affect Gag multimerization. Transfected COS-1 cells were pulse-labeled and either harvested immediately (pulse) or chased for 2 h in control medium (untreated) or in medium containing ATP-depleting agents (ATP depleted). P100 fractions were prepared and fractionated over Optiprep gradients containing a nonionic detergent, as described in Materials and Methods. Labeled Gag in the gradient fractions was quantified by analyses of immunoprecipitates by SDS-PAGE followed by phosphorimaging.

ATP depletion halts the HIV-1 assembly process at a late stage.

The results presented above suggested that ATP depletion stops budding at a late stage in the assembly process. To test this possibility, we fractionated cellular membrane fractions by centrifugation through an Optiprep gradient; to preserve cellular membrane-derived vesicles, this was performed in the absence of detergent. We have previously shown that these gradients resolve HIV-1 assembly intermediates (42). Transfected cells were treated for 0, 2, or 4 h with ATP-depleting agents and were then fractionated over Optiprep gradients as described in Materials and Methods. The total Gag protein was detected by Western blotting. Gag protein from the untreated cells was broadly distributed throughout the gradient, indicating the presence of Gag in assembly domains at various stages in the assembly process (Fig. 6A, top panel). In contrast, in the samples from ATP-depleted cells, there was an accumulation of Gag in the bottom fraction of the gradient (Fig. 6A, middle and bottom panels). Quantitation of several experiments (Fig. 6B) revealed that 36% ± 7% and 45% ± 4% of the Gag protein was in the bottom fraction after 2 h and after 4 h of ATP depletion, respectively, compared to 17% ± 4% for the untreated cells. This finding suggests that ATP depletion results in the accumulation of Gag in dense structures that correspond to late assembly intermediates.

FIG. 6.

ATP depletion results in accumulation of dense Gag complexes at the bottom of Optiprep gradients performed in the absence of detergent. (A and B) Fractionation of Gag protein from untreated and from ATP-depleted cells on detergent-free Optiprep gradients. Transfected COS-1 cells were subjected to pulse-chase and ATP depletion as described in the legend to Fig. 5. P100 fractions were prepared and fractionated over detergent-free Optiprep gradients, as described in Materials and Methods. Gag in the gradient fractions was detected by Western blotting with anti-p24 CA antiserum. (A) A representative experiment. (B) Percentages of Gag in the bottom two fractions (density, ≥1.11 g/ml) in several experiments. Similar results were obtained by labeling cells at steady state with Tran³⁵S-label, chasing in the presence or absence of ATP-depleting agents, and visualizing labeled Gag protein (data not shown). (C) Results of a VLP budding experiment like that shown in Fig. 1B, except that the ATP-depleting agents were not added until 1 h into the chase period (arrow).

If ATP were required at a late stage in the assembly and budding process, one would expect ATP-depleting agents to halt budding even when added at later time points during the assembly process. To test this hypothesis, we performed a variation of the experiment depicted in Fig. 1B. Instead of adding 2-deoxyglucose and NaN₃ at the beginning of the chase period, the ATP-depleting agents were added 1 h into the chase. Since there is little budding by 1 h (Fig. 1B) and ATP depletion is rapid (Fig. 1A), we reasoned that cellular ATP would be depleted before the conclusion of the assembly and budding process. The results of this experiment are depicted in Fig. 6C. Even when the ATP-depleting agents were added 1 h into the chase period, VLP production was sharply reduced, to a level similar to the reduction seen when these agents were added at the beginning of the chase (Fig. 1B). This finding implies that ATP depletion 1 h into the assembly process is sufficient to halt HIV-1 budding.

Late HIV-1 assembly intermediates can be visualized by EM (14). We therefore used this technique to examine untreated and ATP-depleted cells to determine whether ATP depletion altered the visible pattern of assembly intermediates. The frequency of assembly intermediates was determined as described in Materials and Methods. In the untreated, transfected cells, assembly intermediates were occasionally detectable at the cell surface, at a frequency of 9 per 100 μm² of plasma membrane (Fig. 7A). In addition, immature VLPs were seen in the extracellular space (Fig. 7A). Since the cells were washed extensively before they were harvested, these structures probably represent nascent VLPs that had almost pinched off from the cells and were sheared off during sample preparation. Many more assembly intermediates were observed in the ATP-depleted cells (approximately an eightfold-higher frequency). Slightly over half (56%) of the intermediates seen in the depleted cells were stalk-like structures, indicative of a very late stage in HIV-1 assembly (Fig. 7B and C). In addition, VLPs were observed in the extracellular space (Fig. 7B and C). These images indicate that in ATP-depleted cells, assembly is arrested at a late stage. Taken together, our data indicate that an ATP-requiring factor(s) participates in the HIV-1 assembly and budding process. Furthermore, the data presented in Fig. 4 to 7 suggest that the ATP requirement occurs at a late stage in assembly, most likely immediately before the final step of pinching off from the cell.

FIG. 7.

EM images of Gag-producing untreated and ATP-depleted cells. Transfected untreated (A) or ATP-depleted (B and C) cells were washed in STE and then fixed, stained, and examined by transmission EM as described in Materials and Methods. HIV-1 budding structures are indicated by arrowheads. Cross-sections of ∼20 untreated and ∼20 ATP-depleted cells were examined to obtain the estimated frequencies of the budding structures in the untreated and depleted cells that are stated in the text. Bars, 500 nm.

DISCUSSION

Little is known about host cell requirements for HIV-1 assembly and budding (12). We report here that cellular ATP depletion reversibly halts the HIV-1 assembly and budding process (Fig. 1B and 2B). The observed inhibition in budding due to ATP-depleting agents was not a consequence of cell death. The viability of the cells during the time course of these experiments was unaffected by ATP depletion, as evidenced by both trypan blue viability assays and a rapid rebound in cellular ATP levels when the ATP-depleting agents were removed. This is the first report to show that cellular ATP is required for lentivirus assembly in vivo. Our data are consistent with those of a previous report demonstrating an ATP requirement in a cell-free HIV-1 assembly system (23). Other viruses that assemble via mechanisms different from those used by HIV-1 have recently been shown to require cellular ATP for assembly. For example, in type D retroviruses such as Mason-Pfizer monkey virus, capsids first assemble intracellularly and are then transported to the plasma membrane (34). Both of these processes have been shown to require cellular ATP (44). An ATP requirement for assembly has also been shown for influenza virus, vesicular stomatitis virus, African swine fever virus, and herpes simplex virus type 1 (5–7, 9).

It was recently demonstrated that a substantial fraction of newly synthesized Gag protein is degraded intracellularly (37, 42). We therefore examined, by measuring the amount of cellular Gag in untreated and depleted cells, the possibility that ATP depletion affects the level of intracellular Gag protein. Interestingly, the disappearance of cellular Gag was slowed in the ATP-depleted cells (Fig. 3). This is consistent with the fact that much of the Gag degradation occurs in the proteasome (37), which requires ATP for its function (1).

Despite the fact that there was more intracellular Gag in the ATP-depleted cells than in the untreated cells, the efficiency of VLP production was decreased by 70 to 80% in the former (Fig. 1B). One might argue that the presence of increased intracellular Gag artificially reduced the VLP production measurement, which is expressed as the percentage of Gag in VLPs divided by the total amount of Gag in cells plus VLPs. However, even when the raw Gag counts in VLPs were compared, without normalization to total Gag, VLP production in ATP-depleted cells was decreased by 52% ± 13% at 2 h and by 71% ± 9% at 4 h relative to that of the control cells.

Our laboratory recently showed that newly synthesized Gag is found in two pools, membrane-bound protein complexes that proceed through the assembly process and cytosolic complexes that are rapidly degraded (42). The experiment depicted in Fig. 4 shows that ATP depletion does not affect the kinetics of membrane association or the proportion of Gag protein that is membrane bound. The rapid reversibility of the block in budding (Fig. 2B) also supports the contention that the accumulated Gag in ATP-depleted cells is present in membrane-bound protein complexes. Taken together, the data represented in Fig. 1 to 4 suggest that cellular ATP depletion results in the accumulation of membrane-bound Gag complexes stalled in the assembly process.

Gag multimerization also proceeds without cellular ATP (Fig. 5). The ATP independence of Gag membrane binding and multimerization was not unexpected, given the intrinsic affinity of Gag for binding to membranes in vitro (47, 48) and the propensity of Gag molecules to self-assemble in vitro (4, 23, 26). However, these data do not definitively exclude the possibility that Gag membrane binding and/or multimerization requires ATP in vivo. A subset of newly synthesized Gag protein is already membrane bound within 10 min of synthesis (42), at which time cellular ATP depletion is not complete (Fig. 1A). This might preclude the detection of an ATP requirement for membrane binding. Although no block in multimerization was detected (Fig. 5), this assay may lack the discrimination to resolve large Gag multimers of different sizes and therefore might not detect a multimerization block occurring late in assembly. Nevertheless, the data shown in Fig. 4 and 5 indicate that the stalled assembly complexes observed in our system are membrane bound and have proceeded through most or all of the multimerization process.

Three additional lines of evidence indicate that assembly complexes in ATP-depleted cells are stalled late in the HIV-1 assembly process. First, after ATP depletion, Gag accumulates at the bottom of Optiprep gradients at a density characteristic of late assembly intermediates (42) (Fig. 6A and B). Second, the ATP-depleting agents still stop budding when added 1 h into the assembly process, indicating that the ATP-dependent step occurs after this time (Fig. 6C). Since most HIV-1 budding occurs within 2 to 4 h after synthesis (42) (Fig. 2B), much of the labeled Gag population is likely to form productive assembly complexes within the first hour. Finally, stalk-like structures, which correspond to a very late stage in assembly, are enriched in cells treated with ATP-depleting agents (Fig. 7). It is likely, however, that a subset of the HIV-1 assembly structures in the depleted cells is arrested at an earlier stage of assembly. This possibility is consistent with the observed enrichment of crescent-shaped structures underlying the plasma membrane in ATP-depleted cells (Fig. 7C) and the presence of Gag protein in the middle fractions of detergent-free Optiprep gradients prepared from ATP-depleted cells (Fig. 6A).

Although our data suggest that cytosolic Gag complexes are degraded in ATP-depleted cells, it remains possible that misfolded Gag protein accumulates in ATP-depleted cells as a result of proteasome inhibition. It has been suggested that misfolded Gag protein incorporated into assembly complexes might dominantly arrest assembly by perturbing the structure of the complexes (38). However, this mechanism would not predict the rapid reversal of the block in budding that was observed when the ATP-depleting agents were removed (Fig. 2B). We therefore favor the hypothesis that ATP depletion blocks budding by inhibiting an ATP-requiring process in the cell.

It is also possible that ATP depletion halts budding indirectly by arresting protein translation during the chase period, thereby reducing the level of a viral or cellular protein with a rapid turnover rate. Consistent with this possibility, cycloheximide treatment caused a drop in VLP production that was almost as large as the decrease occurring in the ATP-depleted cells (data not shown). This result is difficult to interpret because cycloheximide frequently exhibits pleiotropic effects on cellular metabolism (41, 46).

The other mechanism that explains the block in assembly is inhibition of the activity of a protein that requires ATP binding and/or hydrolysis. Since none of the HIV-1 genes encodes an ATPase or GTPase, host cellular ATP-dependent processes are likely to be involved in HIV-1 assembly. This possibility is supported by recent reports stating that HIV-1 fails to bud in an insect cell line (31) and in murine cells (3, 24), implying that these cells may lack a host cell factor(s) required for HIV-1 assembly.

Taken together, the data presented in this study suggest that HIV-1 assembly complexes in ATP-depleted cells are arrested immediately before the final step of pinching off from the cell. This step involves the breaking and rejoining of the cellular and viral lipid envelopes and is mediated by the L domain of Gag proteins (17, 30). EM images of cells transfected with mutant Gag proteins lacking the L domain (17) resemble the EM images of ATP-depleted cells (Fig. 6B and C), implying that assembly is arrested at similar steps in the two systems. The L domains of several retroviruses have recently been shown to interact with the cellular ubiquitination machinery (32, 38, 40). Interestingly, retroviral budding is halted when cells are treated with proteasome inhibitors, which decrease the intracellular concentration of free ubiquitin. Moreover, stalk-like structures are enriched in the proteasome inhibitor-treated cells (38). Since proteins in the ubiquitination apparatus utilize ATP, they are good candidates for an ATP-requiring cellular factor needed in HIV-1 assembly.

It is also tempting to speculate that the pinching-off process may mechanistically resemble the budding and/or fusion of intracellular vesicles. These processes involve cellular proteins that require ATP or GTP for their function (8, 25, 33). Since intracellular GTP levels are closely linked to ATP levels, depleting cells of ATP could exert an indirect effect by lowering cellular GTP levels. The proteins involved in vesicle budding and fusion thus represent another class of candidates for proteins mediating the ATP requirement in HIV-1 assembly. In conclusion, this study has shown that cellular ATP is required for efficient HIV-1 particle production and that the ATP requirement most likely occurs during the late stage of HIV-1 assembly. Ultimately, the identification of host factors required for budding will likely provide new targets for antiviral therapy.