The Human Immunodeficiency Virus Type 1 (HIV-1) Vpu Protein Interferes with an Early Step in the Biosynthesis of Major Histocompatibility Complex (MHC) Class I Molecules

T_CD8+ cells were removed from buffy-coat cells of an HIV-negative HLA A-2,3–positive donor by magnetic beads coupled with an mAb specific for human CD8 (Dianova GmbH, Hamburg, Germany). The remaining CD8⁻ cells were stimulated with PHA (2 μg/ml) in the presence of 100 U/ml recombinant IL-2 for 2 d. 10⁶ stimulated cells were infected with 5 × 10⁴ tissue culture infectious doses, each of HIV-1_IIIB (41, 42), HIV-2_D205, simian immunodeficiency virus (SIV)_mac, or HIV-1_NL4-3 (43), and the isogeneic vpu mutants HIVNL-U₃₅ (24) and HIVNL4-3/U_2/6 (44). A3.01 cells (45) were cultivated in RPMI 1640/FCS. Routinely, 2 × 10⁶ reverse transcriptase units were used to infect 10⁷ A3.01 cells. After 15 h of absorption, cells were maintained at 10⁶ cells/ml as described previously (33, 46). Analyses of MHC-I expression was performed approximately at peak virus production, usually at day 12 after infection.

4 d after infection, aliquots from controls and infected PBL cultures were washed and incubated for 30 min on ice with mAb IOT 2 (Dianova GmbH). Cells were washed, incubated with FITC-conjugated goat anti–mouse Ig for 30 min on ice, washed, fixed in 3.5% formaldehyde, and analyzed on a cytofluorograph. To monitor infections, cells were fixed with 3.5% formaldehyde for 30 min at room temperature, washed, and treated for 20 min at room temperature with 0.25% Triton X-100 in PBS. Cells were then washed, incubated for 30 min at room temperature in 10% heat-inactivated sheep serum, washed, and incubated for 1 h at room temperature with purified polyclonal sheep antibody specific for HIV-1 p24^gag (Biochrom, Berlin, Germany). After extensive washing, cells were incubated with a 1:50 dilution of an FITC-conjugated polyclonal goat anti–sheep antibody and analyzed by flow cytometry. For detection of p27^gag, the FITC-conjugated mAb 2F12 (Biochrom) was used.

HeLa cells were infected for various times as indicated in the text with recombinant vaccinia viruses (rVVs) and incubated on ice for 30 min with FITC-conjugated antibodies specific for HLA class I molecules, W6/32 (Accurate Chem. & Sci. Corp., Westbury, NY), human β₂m (The Binding Site, Inc., San Diego, CA), mouse CD54 (PharMingen, San Diego, CA), mouse H-2 K^b, or human transferrin receptor (Becton Dickinson, San Jose, CA). Antibodies specific for mouse H-2 K^b were purified from the hybridoma cell line Y3 (American Type Culture Collection HB 176) and FITC-conjugated using standard procedures. After staining, HeLa cells were washed and resuspended in PBS supplemented with 10 μg/ml ethidium homodimer and analyzed by flow cytometry. Nonviable cells were excluded from analysis.

rVVs were generated and propagated as described (47). Vpu-expressing rVVs were produced by cloning either wild-type Vpu from HIV-1_NL4-3 (VV-Vpu; 43) or the mutant vpuDEL-1 (VV-U_DEL1; 31) behind the early/late VV p7.5 promoter. In brief, a 290-bp AflIII-KpnI fragment of the plasmid pSP-9 (24) or a 242-bp AflIII–KpnI fragment of the plasmid pSP-9/U_DEL1 (46) were cloned into the NcoI–KpnI sites of plasmid pSC11 (48) yielding plasmids pSCVpu and pSC-U_DEL1, respectively. VV-ICP47 (49), VV-K^d (50), VV-K^b (50), and vCB-3 (51) have been described. VV–intracellular adhesion molecule (ICAM) expressing mouse CD54 was created as described (47) by inserting cDNA (52) behind the early/late VV p7.5 promoter into plasmid pSC11 (48).

At day 4, p.i. aliquots of 10⁷ PBL were labeled with [³⁵S]methionine (2 mCi/ml) for 45 min, chased for the times indicated in the absence of radiolabeled methionine, and lysed in 90 μl of 3-(3-cholamidopropyl)diethylammonio-1 propanesulfonate (CHAPS) buffer (50 mM Tris, pH 8.0, 5 mM EDTA, 100 mM NaCl, 0.5% CHAPS) + 15 μl CHAPS/deoxycholate buffer (CHAPS buffer with 2% deoxycholate). For the immunocollection of MHC-I molecules, lysates were precleared with mouse IgG (1.5 h, 4°C) coupled to protein A–Sepharose followed by specific immunocollection with HLA-specific mAb IOT2 (2 h, 4°C) coupled to protein A–Sepharose. Immunocollected proteins were separated in a 10% SDS-PAGE and analyzed by quantitative fluorography. HeLa cells infected at rVVs with 3–5 PFU/cell were starved for 20 min in methionine-free DMEM and labeled for up to 15 min with [³⁵S]methionine (2 mCi/ml). Cells were washed, chased in DMEM containing 1 mg/ml l-methionine, and lysed in CHAPS/deoxycholate buffer. Cell lysates were precleared by incubation at 4°C for 1 h with GammaBind G–Sepharose preadsorbed with 20 μg of preimmune rabbit IgG, and incubated for 2 h with protein G–Sepharose preloaded with one of the following antibodies: mAb W6/32 (Harlan Sprague Dawley Inc., Indianapolis, IN), mAb SK81075 (Olympus Corp., Lake Success, NY), antihuman β₂m serum from rabbit (Dakopatts, Copenhagen, Denmark), mAb TW2.3 specific for vaccinia proteins E3L/E3L* (53), anti-Vpu sera from sheep (28) or rabbit (27), mAb 215 directed against conformed H chain K^d (54), polyclonal rabbit antibodies directed against exon 8 of H-2 K locus H chains (pAbs-ex8; reference 54), or anti-CD4 serum from rabbit (55). Immunoprecipitated proteins were separated in 10% or 12.5% acryl aide gels. Radioactive bands were visualized in fixed gels by fluorography and quantitated using a PhosphorImager.

We initially analyzed the potential contribution of Vpu to MHC class I cell surface downregulation by comparing the effects of infecting PHA/IL-2–activated, T_CD8+-depleted PBL with Vpu expressing HIV-1_IIIB or the non-Vpu expressing viruses HIV-2_D205, or SIV_mac (Fig. 1,A). Surface expression of class I molecules was determined by cytofluorographic analysis of cells indirectly stained using a mAb specific for native, peptide-bearing class I molecules. As previously reported (19, 20), HIV-1 infection resulted in an ∼10-fold decrease in the class I signal of a subpopulation of cells (Fig. 1,A d). This subpopulation correlates numerically with the number of cells expressing p24^gag, and double fluorescence confirmed that class I was selectively decreased in p24^gag-expressing cells (not shown). By contrast, infection with SIV or HIV-2 (Fig. 1,A, b and c) did not decrease class I expression to levels observed in HIV-1_IIIB–infected cells despite the fact that a comparable percentage of cells expressed viral p24^gag antigens at similar levels (Fig. 1 A, f–h).

To specifically test the effect of Vpu on cell-surface class I molecule expression in the context of HIV infection, we used the molecular clone HIV-1_NL4-3 (43) encoding either functionally characterized wild-type or mutant Vpus (24, 25, 31, 38, 39). T_CD8+-depleted, IL-2/PHA-activated PBL were infected with either wild-type HIVNL4-3 (43) or one of two isogeneic mutant viruses. The first (HIVNL-U₃₅) encodes a 35 amino acid inactive Vpu protein with a COOH-terminal truncation (24). The second (HIVNL43/U_2/6) encodes a full-length mutant, Vpu_2/6, that cannot be phosphorylated due to exchanging Asn for phosphoacceptor Ser⁵² and Ser⁵⁶ residues (29). Vpu_2/6 has lost the capacity to induce CD4 degradation, but maintains the ability to enhance virus secretion (44). Vpu_2/6 exhibits similar structure, stability, oligomerization, membrane integration, and subcellular localization as wild-type Vpu (46, 56).

Parallel cultures of PBL were infected with equal infectious doses of the three viruses, and cell surface expression of MHC-I molecules was analyzed 4 d after infection using a mAb that recognizes conformed class I molecules. Downregulation of surface class I molecules observed in the presence of wild-type Vpu (Fig. 1,B, b) was greatly compromised by either of the alterations in Vpu (Fig. 1,B, c and d). As above, the biphasic pattern of the histograms observed in the presence of Vpu reflects heterogeneity in HIV infection of the PBL cultures, as clearly revealed by cytofluorographic analysis of p24^gag expression (Fig. 1,B, f–h). Importantly, this demonstrated that there was no significant difference in either the number of p24^gag-expressing cells or the amount of p24^gag expressed by cells infected with the different clones. Thus, the inability of the Vpu mutants to decrease class I expression cannot be trivially attributed to impairment of viral replication and a concomitant decrease in the expression of other viral gene products. The residual slight decrease in class I expression exhibited by the Vpu mutants (Fig. 1 B, c and d) is consistent with recent reports that the HIV-1 accessory protein Nef induces endocytosis of cell surface MHC-I (57).

To further characterize the effect of Vpu on the expression of MHC-I molecules, PHA/ IL-2–activated, T_CD8+-depleted PBL from two donors were infected for 4 d with HIVNL4-3 or isogenic vpu mutants, and class I biogenesis was biochemically characterized (Fig. 2). After 45 min of radiolabeling with [³⁵S]methionine and chasing for various times at 37°C, class I molecules reactive with a mAb specific for conformed class I molecules were collected, separated by SDS-PAGE, and quantitated. Cells from either donor infected with wild-type virus demonstrated a 30–40% reduction in the amount of class I molecules recovered at the initiation of the chase, compared to uninfected or vpu-deficient cultures (Fig. 2, B and C). Additionally, an augmented decay of class I H chains was detectable in the presence of wild-type Vpu after a 4-h chase period (Fig. 2,A, 2) relative to either uninfected cells (Fig. 2,A, 1), cells infected with the Vpu-deficient mutant HIVNL-U₃₅ (Fig. 2,A, 3), or the Vpu phosphorylation mutant HIVNL4-3/U_2/6 (Fig. 2,C). Note that the effects of Vpu on class I biosynthesis are underestimated in these experiments because not all cells are infected in each culture (Fig. 1,B, f–h). Consistent with the Vpu-dependent effects in PBL (Fig. 2), class I molecules were also affected in a Vpu-dependent manner in HIV-1–infected CD4⁺ T cell line A3.01 (not shown). Neither the amount of β₂m recovered from cell lysates nor its secretion into the culture supernatant was altered by Vpu in HIV-1–infected PBL or T cells (not shown), indicating that Vpu does not prevent class I assembly by reducing β₂m to limiting levels. Rather, we conclude that the Vpu-dependent downregulation of class I cell surface expression in cells infected with wild-type HIV-1 largely reflects Vpu-mediated instability of MHC-I molecules due to their improper assembly or enhanced degradation.

Several characteristics of HIV make it less than optimal as a means of characterizing the effect of Vpu on class I molecules. First, Vpu enhances the release of HIV from the plasma membrane of infected cells (24, 25, 31–33). Since HIV particles can contain large amounts of class I molecules (58), this was a possible confounding factor in the observed effect of Vpu on recovery of class I molecules (Fig. 2). Second, since HIV only infects a subset of cells, the effects of Vpu are obscured by processes unaffected in uninfected cells. To circumvent these problems, we produced VV-Vpu. As a control for the possible effects of Vpu mRNA, we also produced a rVV, VV-U_DEL1, expressing mRNA from the gene from the natural vpu mutant vpuDEL-1 (31) that carries a 48-bp deletion within the Vpu transmembrane domain, resulting in the production of a 15-residue protein unrelated in sequence to Vpu (31).

Using a Vpu-specific antiserum, a protein with the predicted mobility of Vpu in SDS-PAGE was recovered from HeLa cells infected with VV-Vpu (Fig. 3,A), but not with VV-U_DEL1 or from the uninfected culture. The stability of Vpu expressed by VV-Vpu was comparable to the half life of Vpu previously reported for HIV-1–infected or –transfected human cell lines (25, 44, 46). To characterize the biological activity of Vpu in VV-infected cells, we examined its effects on CD4 biogenesis. Mouse L929 cells were coinfected with VV-Vpu and vCB-3 (expressing wild-type human CD4; reference 51), radiolabeled for 7 min with [³⁵S]methionine, and then chased at 37°C for up to 240 min (Fig. 3,B). Detergent extracts were immunoprecipitated with an anti-CD4 antiserum, and analyzed by SDSPAGE (Fig. 3,B, inset). The kinetics of CD4 decay were determined by calculating the levels of CD4 present at different times relative to the levels of CD4 present at the end of the pulse (0 min), which was defined as 100% (Fig. 3 B). In the presence of rVV expressed Vpu, the half-life of CD4 was ∼22 min. Similar Vpu activity was observed in CD4⁺ T cell lines and HeLa cells co-infected with VV-Vpu and vCB-3 (not shown). In the absence of Vpu, CD4 exhibited a t_1/2 ∼4 h which is consistent with previously reported half-lives of CD4 in human cell lines (39, 46). Therefore, rVV-expressed Vpu has biological activity comparable to Vpu expressed in human cell lines transfected with vpu⁺ HIV-1 subgenomic expression vectors (38, 44, 46).

The effect of Vpu on MHC-I molecules endogenous to HeLa cells was examined 2.5 h after infection with VVVpu or VV-U_DEL1 by radiolabeling cells for 10 min and chasing for up to 8 h. Class I molecules were collected from detergent lysates using either the mAb W6/32 specific for conformed class I molecules (Fig. 4,A), an antiserum specific for human β₂m (Fig. 4,B), or the mAb SK81075 specific for nonconformed human H chains (Fig. 4,C). Lysates were equalized to contain a comparable amount of incorporated [³⁵S]methionine. To control for possible differences between the infectivity of virus stocks, cell lysates were also subjected to immunocollection using a mixture consisting of polyclonal antibodies specific for Vpu and the mAb TW2.3 specific for vaccinia proteins E3L/E3L* (53; Fig. 4,D). Relative amounts of H chain and E3L proteins collected were quantitated by PhosphorImager analysis (Fig. 4, histograms). Expression of Vpu dramatically reduced the amount of class I H chains recovered with each of the antibody preparations used at each of the time points tested (Fig. 4, A–C). This cannot be trivially attributed to differences between the infectivity of VV-Vpu and VV-U_DEL1, as shown by the nearly identical amounts of VV proteins E3L/E3L* recovered by mAb TW2.3 from each of the time points (Fig. 4 D). Consistent with our previous finding that Vpu does not affect the stability of the transferrin receptor (CD72) in HIV-1–infected macrophages (33), similar amounts of CD72 were recovered from VV-Vpu– and VV-U_DEL1–infected cells pulse radiolabeled for 10 min with [³⁵S]methionine and chased for up to 8 h (not shown).

Notably, the effects of Vpu in class I biogenesis were evident immediately upon pulse labeling, indicating that Vpu affects early events in the assembly of class I molecules. To facilitate biochemical evaluation of Vpu effects on H chain synthesis, we examined the biosynthesis of mouse K^d class I H chain expressed in HeLa cells co-infected with VV-K^d, VV-U_DEL1, or VV-Vpu (Fig. 5). The relatively high levels of K^d expression allowed for shorter labeling times with [³⁵S]methionine. Cells were radiolabeled for 4 min, and immediately placed on ice to stop cellular metabolism. Cells were then either maintained on ice or incubated for up to 2 h at 37°C. Cell lysates were equalized for [³⁵S]methionine incorporation, and class I molecules were collected with either mAb 215 specific for conformed K^d (Fig. 5,A, mAB 215; reference 54) or with rabbit serum specific for exon 8 (Fig. 5,B, pAb-ex8) that reacts with both folded and unfolded K^d molecules (54). To ensure complete recovery of H chains, supernatants from the first round of immunocollection were subjected to a second round of precipitation with the same Ab (Fig. 5, A and B, 1^st and 2^nd). The relative amounts of H chains recovered after sequential collection were determined by PhosphorImager analysis (Fig. 5, A and B, left histograms). In the absence of Vpu, the amount of K^d recovered by mAb 215 doubled over the first 20 min of chase, and then decayed slowly. This ∼10 min half-life of K^d folding is consistent with previous findings regarding class I folding (3). We also detected a similar, but less marked, increase in K^d class I molecules collected by pAb-ex8. The increase in this case may be due to initial masking of exon 8 by interaction with molecular chaperones (60). We observed a number of remarkable effects of Vpu on K^d biogenesis. First, approximately two- to threefold less K^d reactive with either mAb 215 or pAb-ex8 was recovered from the unchased samples. Second, instead of an initial increase, the amount of K^d recovered precipitously declined over the first 20 min of chase period. Vpu had no significant effect on the stability of K^d that survived the initial effect, which are transported through the Golgi complex with kinetics similar to K^d synthesized in the control infected cells, as indicated by the recovery of a K^d form with lower mobility on SDS-PAGE (indicated by arrows in Fig. 5, A and B) that probably represents K^d with terminally sialyated N-linked oligosaccharides.

The fate of K^d that escaped the early action of Vpu was further examined in an experiment in which K^d H chains, collected from detergent lysates with a mixture of mAb 215 and pAb-ex8, was digested with endo H (Fig. 6). K^d possesses three N-linked oligosaccharides. Endo H removes high mannose N-linked oligosaccharides, but not N-linked oligosaccharides that have been acted on by Golgi mannosidase II. The rate of H chain transport to the mannosidase II–containing compartment was determined by calculating the ratio of H chain partially or completely resistant to endo H digestion (Fig. 6,A, +CHO) to endo H–sensitive H chains (Fig. 6,A, −CHO). Vpu expression had no significant effect on the kinetics of acquisition of endo H resistance (Fig. 6 B). This confirms that K^d molecules that escape the early destructive effect by Vpu in the ER are exported and processed in a manner indistinguishable from H chains produced in cells not expressing Vpu.

The effects of Vpu on endogenous class I molecules present on the cell surface were investigated by testing rVV-infected HeLa cells for expression of conformed HLA molecules using either the mAb W6/32 or anti-β2m antibodies directly conjugated to FITC (Fig. 7, A and B). Infection with rVV resulted in a 20–30% decrease in binding of either antibody, probably due to ongoing turnover of class I molecules combined with reduced replacement due to VV suppression on host cell protein synthesis. Staining was reduced a further 5–10% by Vpu, which is likely due to inhibition of expression of the reduced amounts of class I molecules produced in the face of the ongoing VV infection. This latter effect is selective for class I molecules since VV-Vpu and VV-U_DEL1 infections reduced surface levels of endogenous CD72 nearly identically (Fig. 7,C), the reduction again probably related to VV induced suppression of host protein synthesis. Furthermore, Vpu had no effect on cell surface expression of mouse ICAM-1 (CD54) in HeLa cells co-infected with VV-Vpu and VV-ICAM (Fig. 7,D). In contrast, Vpu reduced cell-surface expression of K^b class I molecules co-synthesized in HeLa cells infected with VVVpu and VV-K^b (Fig. 7 E). These findings further demonstrate that Vpu selectively inhibits the cell surface expression of newly synthesized class I molecules, but has no or very minor effects on preexisting cell surface class I molecules.

Given the effect of Vpu on early events in class I biogenesis, it was of interest to compare the effects of Vpu with the ICP47 protein of HSV. ICP47 is a competitive inhibitor of TAP that greatly reduces the supply of class I–binding peptides (16–18, 61, 62). In several of the experiments described above, we included an additional sample in which VV-ICP47 was substituted for VV-Vpu. In the pulse chase experiments (Fig. 5, A and B, +ICP47), the effects of ICP47 and Vpu were quite disparate; In contrast to Vpu, ICP47 did not cause an immediate reduction in the amount of H chains recovered at the end of the pulse-labeling time. In fact, the amount of H chains precipitated with the conformation-sensitive mAb 215 increased slightly over the first 10 min of chase period (Fig. 5,A, +ICP47). Consistent with previous findings (16–18, 61, 62), ICP47 expression results in the recovery of less conformed and total K^d over the chase period. Notably, the initial loss of immunoreactive class I induced by ICP47 occurs with a half-life approximately twice as long as the Vpu-induced loss (Fig. 5, A and B). Another important difference is that ICP47 delays the transport of class I molecules from the ER, as demonstrated by the failure of class I molecules to demonstrate the decrease in mobility associated with siaylation (Fig. 5, A and B, +ICP47). ICP47 and Vpu have a similar effect on class I surface expression (Fig. 7,E), blocking expression of VV-encoded K^b to the same extent as Vpu, while, like Vpu, having little effect on preexisting surface HLA moleclues (Fig. 7, A and B).

As with many viruses, HIV-1 induces a rapid T_CD8+ response in the host (63, 64). However, unlike most viruses, HIV-1 is not eradicated completely and continues to replicate for many years during the asymptomatic period of infection before the onset of AIDS. During this latent period, HIV-specific T_CD8+ are active and present in PBL at very high levels relative to other viral infections (constituting up to 1% of circulating T_CD8+; reference 65), only gradually declining in parallel with T_CD4+ during the progression to AIDS. The presence of a vigorous T_CD8+ response raises the question of how the virus can persist in the face of an immune effector mechanism that is usually highly effective at eradicating virus. One possibility is that a subset of virusproducing cells are resistant to T_CD8+-mediated anti-viral effects.

Several mechanisms, including antigenic drift (63, 66, 67, 68), exhaustion of T_CD8+ response by high level of antigen exposure (69), or deficiencies in HIV-1 antigen processing and presentation (11, 19–21, 57) have been proposed for resistance to T_CD8+-mediated lysis. In the present paper, we show that Vpu interferes with the expression of MHC-I molecules on the surface of HIV-1–infected cells. Indeed, viruses lacking Vpu demonstrate only a minor decrease in class I expression. It seems, therefore, that the Tat-mediated decrease in class I transcripts (11) plays little role (23) in the decrease in class I expression. Rather, we conclude that Vpu is required for most of the inhibition of cell surface MHC-I expression seen in HIV-infected cells, and propose that this may contribute to the inability of HIV-specific T_CD8+ to eradicate the virus in infected individuals.

When expressed in the context of HIV-1 infection, we observed two effects of Vpu on class I molecules: (a) decreased recovery of class I H chains collected with immature N-linked oligosaccharides immediately after [³⁵S]methionine labeling and (b) decreased stability of class I molecules with mature N-linked oligosaccharides. Only the former effect was observed using rVV to express Vpu. There are several explanations for this discrepancy. First, Vpu may act in concert with other viral gene products to affect the stability of class I molecules expressed on the cell surface. In this regard, it was recently reported that the HIV Nef gene product induces the degradation of cell surface class I molecules (57). Possibly, the activity of Nef and/or Vpu is increased by coexpression of both proteins. Second, alterations in cellular metabolism induced by rVV infection may interfere with the ability of Vpu to affect cell surface class I molecules. Third, Vpu is known to augment by two to threefold virus release from PBL (33). Since in HIV-1 particles, class I molecules constitute up to ∼20% of the concentration of Gag protein on a molar basis (58), it is likely that at least some of Vpu-dependent loss in class I molecules with mature N-linked oligosaccharides is due to increased virion production with its attendant increase in class I shedding.

When expressed by rVV, Vpu reduced the amount of class I molecules recovered after brief labeling with [³⁵S]methionine, and, during a 20-min chase period, Vpu also decreased the metabolic stability of the H chains with immature N-linked oligosaccharides that escaped the immediate effects of Vpu. We interpret the latter phenomenon to represent the continuation of the same process resulting in decreased recovery of pulse-labeled heavy chains. Importantly, the effect of Vpu on class I molecules was clearly distinguished from that of ICP47 which acts by blocking the peptide transporting activity of TAP. This indicates that the effects of Vpu on class I biogenesis do not result from a decrease in the quantity or quality of peptides available for class I binding. We have not formally demonstrated that Vpu induces the degradation of MHC-I molecules. It is also possible that the decreased recovery of class I molecules is due to Vpu-induced decreases in solubility of class I molecules in the detergents used for extraction, or to the association with proteins that block antibody accessibility to class I molecules.

How might Vpu compromise biogenesis of class I molecules? It was recently demonstrated (46) that the two previously defined biological functions of Vpu, CD4 degradation and regulation of virus release, are controlled by two separable structural and functional domains. CD4 degradation requires phosphorylation of the Vpu cytoplasmic tail (44) and the binding of Vpu to CD4 (70) trapped in the ER by complex formation with the HIV-1 envelope precursor gp160 (40). In contrast, the effect of Vpu on virus particle release is regulated by an ion channel activity of its transmembrane domain (71, 72), occurs in a post-ER compartment (44), and does not require phosphorylation of Vpu (44). Vpu, therefore, possesses at least two distinct activities: first, alteration of the ionic milieu in a post-ER compartment, and second, targeting of proteins located in the ER for destruction. Our proposal that Vpu induces a rapid decay of nascent class I H chains is more consistent with the known ER activity of Vpu, and, based on the ability of Vpu to induce proteolysis of nascent CD4 in the ER (38, 39), it is likely that Vpu acts similarly on MHC class I molecules.

It was recently demonstrated that the US2 and US11 gene products of human cytomegalovirus are able to induce the rapid degradation of class I H chains (15, 73). Both act by routing newly synthesized class I molecules to the cytosol, where degradation requires the activity of proteasomes, in as much as it was blocked by proteasome inhibitors (15). The interaction of H chains with one of the components of the translocon suggest that US11 directs H chains in a retrograde fashion through the sec61-containing pore used to insert proteins into the ER (73). Indeed, based on this and other recent findings in Saccharomyces cerevisae (74), it seems likely that many cases of presumed degradation of membrane proteins by ER proteases are, in fact, due to routing of proteins to the cytosol where they are degraded in a process requiring active proteasomes. It is necessary to reexamine if the Vpu-induced CD4 degradation involves the ubiquitin-proteasome pathway and whether a similar process is used for class I molecules. It will also be of great interest to determine the shared features of class I H chains and CD4 that target them for Vpu-induced destruction while other glycoproteins such as Env, CD8 (75), CD54, and CD72 are spared.

In summary, our data indicate that Vpu is the major factor in the reduction of MHC class I molecules on HIV-1 infected cells. This adds another function to the two previously ascribed to Vpu, CD4 degradation and virus release. It is possible that one or more of these functions is an unintended byproduct of the other(s). However, it is equally plausible that each contributes to the evolutionary fitness of the virus. The relative contributions of these effects may vary depending on the nature of other viral genes or the host response to the virus. It will be a challenge to determine the biological significance of these three functions of Vpu.

This work was supported by grant BGA II.031-87 of the Bundesgesund-heitsministerium and the Fonds der Chemischen Industrie. U. Schubert was a recipient of a fellowship from the AIDS-Research Program of the Krebsforschungszentrum Heidelberg, Germany, and was supported by grant PO-396/3-1 from the Deutsche Forschungsgesellschaft.