The HIV-1 assembly machine
2001; Lippincott Williams & Wilkins; Volume: 15; Linguagem: Inglês
10.1097/00002030-200100005-00003
ISSN1473-5571
Autores Tópico(s)HIV/AIDS drug development and treatment
ResumoIntroduction HIV-1, like all retroviruses, is released by budding from the host cell's surface, and thereby acquires a plasma membrane-derived lipid envelope that provides a tight seal and shields the virus from the environment. Budding also offers a non-lytic pathway for virus egress, and thus has the advantage that the infected cell remains viable and can continue to produce progeny virions. HIV-1 particle assembly and budding are directed by the Gag polyprotein, the precursor for the internal structural components of the mature virion, which are matrix (MA), capsid (CA), nucleocapsid (NC), and p6. The Gag polyprotein coordinates the late stages of the replication cycle by bringing together all the building blocks of the virion, and is sufficient to organize the efficient assembly and release of virus-like particles even when expressed in the absence of other viral proteins. Initially, the polymerization of Gag polyproteins leads to the formation of an immature particle with a relatively stable spherical shell underneath the virion envelope. However, for the virion to become infectious, a maturation step is required that results in a suitably metastable capsid that can be disassembled within a target cell to release the viral genome. For maturation to occur, the Gag polyprotein must be cleaved by the viral protease, which triggers a large-scale rearrangement of the cleavage products within the assembled particle and a dramatic structural transition from a spherical to a conical capsid [1-4]. The structure of the Gag polyprotein and the precise nature of the interactions that enable it to assemble into a spherical protein shell capable of extruding through the plasma membrane remain unknown. However, substantial progress has been made in the recent past toward understanding the structure and function of each of the four major domains of the Gag polyprotein in virus morphogenesis. Matrix The N-terminal MA domain enables the Gag precursor to interact with the plasma membrane. The membrane targeting function of MA depends on the covalent attachment of myristic acid to an N-terminal glycine residue, which is absolutely essential for virus particle formation and HIV-1 replication [5,6]. However, since the membrane-binding energy contributed by the insertion of myristic acid into the lipid bilayer is small, additional determinants in MA are thought to contribute to the stable membrane association of Gag [7]. Indeed, the phenotypes of certain MA mutants indicate that the globular core of MA also has an important role in membrane binding [7-9]. The three-dimensional structure of HIV-1 MA reveals a globular 'head' formed by four α-helices and a C-terminal α-helix that projects away from the core domain [10-12]. The globular head forms trimers in all known crystal forms, and conserved basic residues cluster on the upper surface of the trimer, where they could potentially interact with acidic phospholipid head groups [12]. In support of this notion, the presence of a highly basic surface region is a common feature of retroviral MA proteins [13-15], and was recently shown to play a critical role in the plasma membrane localization of Rous sarcoma virus (RSV) Gag [16]. Interestingly, in the latter study, basic residues involved in membrane binding could be repositioned without affecting particle release [16], indicating that there is a certain degree of flexibility in the interaction of MA with anionic phospholipids. Accumulating evidence suggests a model in which the ability of HIV-1 MA to interact with membranes is regulated by a myristyl switch mechanism [17-20]. In this model, the myristic acid moiety would be sequestered by MA until a conformational change leads to its exposure and makes it available for membrane insertion. The model predicts that certain mutations should compromise the ability of MA to bury the myristyl group, which may explain why alterations within the α-helical core of MA often lead to substantial increases in membrane binding and, in some cases, clearly enhance viral particle yields [17,19,20]. Moreover, even deletions that remove the globular core of MA entirely can significantly increase HIV-1 particle production in cell culture [21]. Interestingly, deleting the globular core of MA also appears to facilitate assembly in vitro, because full-length HIV-1 Gag or a version lacking p6 formed only small spherical particles [22], whereas a Gag protein lacking both p6 and the globular core of MA formed particles that resembled authentic virus in size [23]. Taken together, these results support the view that MA can assume a conformation that interferes with assembly, and that a conformational switch is needed to trigger efficient Gag polymerization. If this switch regulates the availability of the myristyl group for membrane insertion, then disrupting the globular core of MA may cause increased Gag membrane binding because the myristyl moiety is now constitutively exposed. This model could also accommodate the observation that even conservative single amino acid substitutions near the N-terminus of MA can severely reduce Gag membrane binding and viral particle formation without affecting Gag myristylation [19,20,24]. The effects of these N-terminal mutations on Gag membrane binding resemble those seen in the absence of the myristyl acceptor site, indicating that they interfere with the exposure of the myristyl group. Remarkably, the defects in membrane binding, assembly, and virus replication can all be reversed by second-site mutations in the globular core of MA that increase Gag membrane binding [19,20], perhaps because they disrupt the ability of MA to sequester the myristyl group. A conformational switch that activates Gag membrane binding could, in principle, account for the selective targeting of Gag to the plasma membrane. However, recent results suggest that there is no direct relationship between Gag membrane binding and Gag targeting [25]. One et al. observed substitutions in MA where retargeted assembly to intracellular membranes did not always affect the extent of Gag membrane binding. Conversely, a mutation in the basic region of MA did not induce Gag re-targeting even though it increased the fraction of Gag that was membrane associated [25]. Interestingly, whereas large deletions in MA apparently lead to promiscuous assembly at the most abundant cellular membrane, the endoplasmic reticulum [26], single amino acid substitutions in MA can re-target HIV-1 assembly to the Golgi and/or Golgi-derived vesicles [25]. These observations raise the possibility that an association of Gag with Golgi-derived compartments prior to assembly at the plasma membrane may play a previously unsuspected role in the viral life cycle. Besides its key role in Gag membrane targeting, a second essential function of MA during assembly is to allow the incorporation of the HIV-1 envelope (Env) glycoprotein spikes. These spikes consist of an oligomeric complex formed by the surface glycoproteins and the transmembrane glycoproteins (TM). Although HIV-1 Env incorporation into assembling particles is often blocked by mutations in MA or in the cytoplasmic domain of TM [27-32], it is clear that a specific interaction between these two components is not always necessary. Indeed, several lines of evidence support the concept that MA is primarily required to passively accommodate the long cytoplasmic tail of HIV-1 TM into the particle. First, MA mutants that exhibit an absolute block in the incorporation of the HIV-1 Env complex readily accept heterologous viral Env glycoproteins with short cytoplasmic domains [30,33]. Second, the incorporation of the autologous HIV-1 Env complex into MA mutants can be fully rescued by second-site mutations that shorten the cytoplasmic tail of TM [30,33]. Third, at least in transient expression systems, the entire MA domain becomes dispensable for efficient Env incorporation if the cytoplasmic tail of HIV-1 TM is deleted [21]. Remarkably, if TM is truncated, efficient HIV-1 replication can be observed in the complete absence of MA in at least one human T-cell line [21]. On the other hand, new studies provide direct support for a stable interaction between TM and Gag in immature HIV-1 particles [34] and for a cell-type dependent requirement for the MA-TM interaction in HIV-1 Env incorporation [35,36]. Removing the TM cytoplasmic tail reduced Env incorporation more than 10-fold in the majority of human T-cell lines and in primary cells, where the truncation prevents HIV-1 replication [35]. Similarly, cell-type-specific defects in Env incorporation were seen in the presence of small deletions within a putative α-helical region of the TM cytoplasmic tail [36]. Moreover, the Env incorporation defect of one of these TM mutants could be reversed by a single amino acid change (V34I) in MA, supporting the view that an interaction between MA and TM drives Env incorporation in infected T cells [36]. Intriguingly, the V34I substitution was independently observed in a revertant derived from a virus that combines wild-type gag and env genes from different strains of HIV-1 (A. Borsetti, T. Dorfman, H. Göttlinger, unpublished observation). Although the chimeric virus replicates efficiently in primary cells, it fails to spread in most transformed T-cell lines unless the V34I substitution in MA is present. Taken together, these results indicate that the MA-Env interaction can substantially influence the cell tropism of HIV-1, and that cell-type specific factors play a crucial role in Env incorporation. It is conceivable that the role of MA in Env incorporation is linked to its specific membrane targeting function, because recent studies suggest that both Gag and Env associate with detergent-insoluble membrane domains known as lipid rafts [37,38]. Capsid CA, which directly follows MA in the context of the HIV-1 Gag polyprotein, forms the core of the mature virion. Structural studies indicate that CA has two distinct, largely α-helical domains that are connected through a flexible linker region [39-41]. Consistent with this model, cryo-electron microscopy of immature HIV-1 or Moloney murine leukemia virus reveals a radial arrangement of protein layers, two of which can be assigned to the N-terminal and C-terminal domains of CA [42-44]. The two CA domains have different roles in virus morphogenesis. The larger N-terminal CA domain is required for the formation of the cone-shaped core of the mature virion but is dispensable for the assembly of immature viral particles [45-48]. In contrast, the C-terminal domain is crucial for Gag polyprotein multimerization and HIV-1 particle assembly [41,46-49]. The N-terminal CA domain interacts specifically with the peptidyl-prolyl cis-trans isomerase cyclophilin A (CyPA) and mediates its uptake into HIV-1 virions at a CA : CyPA ratio of approximately 10 : 1 [50-52]. Pharmacological agents or CA mutations that disrupt the interaction with CyPA have no evident effects on HIV-1 particle assembly, Gag processing, or virion maturation [51-53]. However, viral infectivity is reduced, indicating that the incorporation of CyPA is functionally relevant [51,52]. Recently, the regulatory role of CyPA in HIV-1 infectivity was formally demonstrated in a variant of the Jurkat T-cell line that lacks CyPA [54]. In contrast to HIV-1, closely related primate immunodeficiency viruses such as HIV-2 and simian immunodeficiency virus SIVmac do not require CyPA. However, the transfer of HIV-1 CA, or of the CyPA-binding site within CA, to SIVmac can confer an HIV-1-like sensitivity to CyPA-binding compounds, which strongly suggests that CA is not only the interaction partner, but also the functional target of CyPA [55,56]. It has been proposed that CyPA promotes HIV-1 core disassembly following entry into the host cell by destabilizing CA interactions [57], but a new study shows that CyPA does not efficiently dissociate in vitro assembled CA cylinders [58]. Furthermore, recent image reconstructions of CA tubes suggest that the CyPA-binding loop is on the exterior surface of the viral core, and that isolated CyPA molecules can be modeled onto the surface without steric clash and are thus unlikely to promote disassembly [59]. Based on the observation that CyPA at the same molar ratio to CA as in the virion increases the efficiency of CA assembly in vitro, it was recently proposed that CyPA acts as a core assembly chaperone during virus maturation [58]. The C-terminal CA domain begins with the major homology region (MHR), a stretch of about 20 amino acids that is uniquely conserved among retroviral CA proteins [41]. The C-terminal CA domain dimerizes in solution with nearly the same efficiency as intact CA; however, the MHR does not contribute to the dimer interface [41]. In the context of the Gag precursor, the CA dimerization domain may extend into the adjacent p2 'spacer' peptide, which is predicted to form part of an α-helix that begins in CA [60]. Mutations that disrupt the putative α-helical region at the CA-p2 boundary lead to the assembly of large electron-dense patches or long tubular structures at the cell membrane, indicating that p2 is required for the induction of curvature [5,60-62]. Cleavage at the CA-p2 junction during virus maturation, which is expected to break the helix and to weaken CA-CA interactions, is required for the rearrangement of the spherical CA shell into a cone-shaped core [60,63]. In a recent study, the C-terminal CA-p2 domain was sufficient to direct efficient VLP formation when combined with a minimal membrane anchor, a heterologous protein-protein interaction domain, and a short peptide that promotes virus release [64]. This result points to a high degree of plasticity in the assembly process and confirms that the C-terminal CA domain is at the heart of the HIV-1 assembly machinery. In vitro, CA spontaneously assembles into long helical tubes and into cones that resemble mature viral cores [59,65-67]. However, the addition of as few as four MA residues to the N-terminus of CA prevented tube formation and instead resulted in the assembly of spherical structures more reminiscent of immature capsids [66,67]. Based on the three-dimensional structure of the N-terminal CA domain, it was proposed that cleavage at the MA-CA junction creates a new CA-CA interface essential for core assembly by allowing Asp51 of CA to form a salt bridge with the N-terminal CA proline residue [66]. Nevertheless, a new study shows that this salt bridge is not absolutely necessary for tube formation in vitro if additional domains of the Gag precursor are present [23]. A Gag precursor that lacked only a portion of MA and the C-terminal p6 domain assembled into either spheres or tubes and cones, depending on the pH of the reaction [23]. Interestingly, sphere formation was blocked by deleting the p2 spacer peptide that separates CA and NC, whereas short tubes and cones could still be assembled [23]. In agreement with the data obtained in the viral context [60], these in vitro results suggest that p2 functions as a molecular switch region that controls the transition from sphere to cone formation required for virus maturation. Ganser et al. used synthetic cores assembled from HIV-1 CA-p2-NC to test their model that retroviral cores are composed of closed hexagonal lattices [68]. In this model, the conical shape of HIV-1 cores is determined by the presence of 12 pentameric defects. In addition to sheets and cylinders, all of the five discrete cone angles allowed by the model were in fact observed in synthetic core preparations [68]. On the other hand, the cone angles of authentic HIV-1 cores isolated from detergent-stripped virions reveal considerable variability around the narrowest of the angles allowed by the hexagonal lattice model [69-71]. Recent cryo-electron microscopic image reconstructions of synthetic CA tubes confirm that these are composed of hexameric rings [59]. Molecular modeling indicates that the rings on the tube exterior are formed by the N-terminal domains of CA, whereas the C-terminal CA domains correspond to densities at a lower tube radius that connect each ring to its six nearest neighbors [59]. The three-dimensional structures of different types of tubes indicate considerable flexibility in the orientation of the bridging C-terminal domains relative to the N-terminal hexameric rings. This flexibility allowed the construction of a model in which the HIV-1 core is made up from hexameric CA rings, as observed in the tubes, and includes 12 pentons to close the ends of the cone [59]. An interesting aspect of this model is that the core should be permeable for small macromolecules, such as for nucleotides required for reverse transcription of the viral genome. Nucleocapsid The NC domain of the HIV-1 Gag precursor contains two copies of a conserved zinc finger-like motif that are required for the selective encapsidation of the genomic viral RNA into nascent particles [72]. NC also contains clusters of basic residues, and these are crucial for its non-specific nucleic acid-binding activity. It has long been known that retroviral assembly does not depend on the presence of packageable viral RNA. However, as shown recently, retroviruses package cellular RNA in place of genomic viral RNA if the latter is not available [73]. It is now becoming increasingly clear that the ability of NC to bind nucleic acid plays a central role both in the initiation of Gag multimerization and in the maintenance of the retroviral core. An early mutagenic analysis of NC revealed that separate point mutations affecting both zinc fingers simultaneously can significantly reduce HIV-1 particle production in transfected cells [74]. Subsequently, in-frame NC deletion mutants were shown to be drastically defective in assembly [75,76]. Although small amounts of particulate Gag are released from transfected cells in the absence of NC, the released material has a lower density than wild-type HIV-1 virions [75,77], and the extreme N-terminus of NC is particularly critical for normal density [78]. While the zinc fingers appear dispensable for assembly in cells that express high levels of Gag, clusters of basic residues in NC remain crucial, indicating that the non-specific RNA-binding activity of NC is required [79]. Basic residue mutations that interfere with the production of extracellular particles also disrupt detergent-resistant intracellular Gag polyprotein complexes [79], which may represent assembly intermediates [80,81]. However, despite their critical role in assembly, a recent study shows conclusively that basic residues in NC do not determine the density of HIV-1 virions [82]. Interestingly, the assembly function of HIV-1 NC in a proviral context can be fully replaced by heterologous polypeptides that promote interprotein contacts, including dimer-forming coiled coil domains that are not known to bind RNA [64,83]. The central role of RNA in the assembly function of NC is illustrated by the fact that nucleic acid dramatically induces the in vitro assembly of purified HIV-1 or RSV Gag protein into cylindrical, spherical, or cone-shaped particles [22,23,65,68,84,85]. In remarkable accordance, the effect of nucleic acid on in vitro assembly depends on clusters of basic residues within NC that are also important for Gag assembly in vivo[85,86]. The products of in vitro assembly reactions are sensitive to nuclease treatment, implying that direct Gag-Gag interactions are too weak to maintain the oligomeric complexes [22,65]. Particle formation proceeds with similar efficiency whether viral or non-viral RNA is used, indicating that the non-specific RNA-binding activity of NC is sufficient to drive assembly [22]. Indeed, it was recently shown that single-stranded DNA oligonucleotides, linear double-stranded plasmid DNA, and even the unrelated polyanion heparin can effectively substitute for RNA, indicating that nucleic acid promotes assembly primarily through electrostatic interactions of the phosphate backbone with basic residues in NC [85]. Oligodeoxynucleotides must be at least 15-20 nucleotides long to work in in vitro assembly reactions [22,85]. This size limit suggests that two Gag molecules need to be able to bind, and the primary role of nucleic acid in the assembly may thus be to induce the dimerization of Gag molecules. Gag dimers may then be capable of completing assembly through protein-protein interactions, as indicated by the observation that leucine zipper dimerization domains can fully substitute for the assembly function of NC in vivo[64,83]. The concept that the basic assembly unit is a Gag dimer is also supported by recent chemical cross-linking experiments on the products of in vitro assembly reactions [22]. p6 The p6 domain, which is at the C-terminus of the HIV-1 Gag polyprotein, is only found in primate immunodeficiency viruses. In addition to its essential role in the incorporation of the regulatory protein Vpr [87], p6 facilitates a late step in the budding process [88-90]. This step requires a membrane fusion event at the neck of the bud to release the assembled particle from a tether that connects it to the cell surface. In an initial analysis of the function of HIV-1 p6, vigorous budding on the cell surface was observed in the absence of p6, but the budding virions remained attached to the plasma membrane via a thin stalk [88]. Within p6, the region required for efficient virus release maps to a P(T/S)APP motif that is conserved among lentiviruses, even in those that lack a p6 domain [88,89]. The only exception is equine infectious anemia virus (EIAV), which uses a YxxL motif within its unique C-terminal Gag domain for virus release [91]. Highly conserved Gag regions that act late in assembly, commonly referred to as L domains, have also been identified in oncoretroviruses [92-95]. In these viruses, the L domains are located upstream of the CA domain and contain a PPxY motif at their core. Remarkably, the unrelated L domains of HIV-1 and of oncoretroviruses are exchangeable and function largely independent of their position [96,97], suggesting that they represent docking sites for cellular factors. It has been noted that the PPxY motif matches the core consensus of WW domain ligands [98], and certain WW domains (including several from members of the Nedd4 ubiquitin ligase family) indeed interact with oncoretroviral L domains in vitro[98-100]. Interestingly, retroviruses are known to contain ubiquitin, both in free form and conjugated to Gag [101,102]. Ubiquitin is a highly conserved polypeptide that is conjugated to lysine residues in various substrates, which include ubiquitin itself. Whereas the formation of polyubiquitin chains is required to target proteins to the proteasome for degradation [103], mono-ubiquitination is sufficient to trigger the budding of endocytic vesicles from the plasma membrane [104]. Both in HIV-1 and in Moloney murine leukemia virus particles, a fraction of the Gag products that harbor the L domain core motifs are mono-ubiquitinated [105], suggesting that L domains come into close contact with ubiquitinating enzymes. Three recent papers now directly implicate the cellular ubiquitination machinery in retroviral L domain function. These studies show that the budding of three different retroviruses, including that of HIV-1, is suppressed by proteasome inhibitors [106-108]. These compounds reduce the mono-ubiquitination of HIV-1 Gag by depleting the levels of free cellular ubiquitin. Inhibition of the proteasome arrests the budding of HIV-1 or RSV at a late stage, similar to what is seen in the absence of an L domain [106,108]. In the case of RSV, proteasome inhibition induced impressive crystalline-like clusters of viral particles on the cell surface [96]. The particles appeared linked via membranous connections, indicating that ubiquitin is required to separate emerging particles both from the cell and from each other. Remarkably, the effect of proteasome inhibition on RSV budding could be suppressed by overexpressing ubiquitin, providing strong evidence that ubiquitin plays a role in retrovirus budding [96]. Independently, evidence for such a role emerged from a study of the minimal requirements for HIV-1 particle formation. Strack et al. observed that the presence of L domains in minimal HIV-1 Gag constructs induced the appearance of modified Gag species, and these turned out to be Gag-ubiquitin conjugates [107]. Importantly, the conjugation of ubiquitin to Gag strictly depended on the presence of a functional L domain and was induced by the unrelated L domains of HIV-1 and RSV. Particularly robust Gag ubiquitination and enhancement of virus-like particle release were obtained with the candidate L domain of Ebola virus, which is in the putative matrix protein VP40 and contains overlapping P(T/S)AP and PPxY L domain core motifs [107]. In agreement with these findings, Harty et al. recently showed that the Ebola PPxY motif interacts with a ubiquitin ligase and is important for the self-exocytosis or budding of VP40 [109]. PPxY motifs have also been shown to regulate ubiquitin-dependent endocytosis [110], which suggests that L domains may engage components of the endocytic machinery to stimulate exocytosis. A link between endocytosis and virus budding is also suggested by the observation that the critical YxxL motif in the L domain of EIAV recruits clathrin-associated adaptor complexes during virion assembly [111]. Consistent with the evidence that EIAV uses a unique pathway to promote virus release, EIAV budding is not sensitive to proteasome inhibitors (J. Wills, personal communication, 2001). It is not yet known whether Gag-ubiquitin conjugates are functionally relevant or merely byproducts that reflect the presence of ubiquitin ligase activity at the site of virus budding. On the one hand, a functional role is suggested by the finding that RSV budding from cells depleted for ubiquitin could be rescued by fusing ubiquitin directly to Gag [108]. On the other hand, the conserved lysine residues in HIV-1 p6 that serve as acceptor sites for mono-ubiquitination are not required for virus replication [105]. Thus, it will be important to determine whether L domains engage the ubiquitination machinery to modify a cellular target at the budding site. Further work is also needed to determine the identity of the ubiquitinating enzymes that are recruited by L domains. Members of the Nedd4 family of ubiquitin ligases seem like good candidates because they harbor WW domains that are known to bind PPXY motifs [99]. However, Strack et al. tested several Nedd4 family members for their effects on RSV L-domain-mediated budding, with negative results [107]. Another candidate that has recently emerged from a two-hybrid screen is the putative ubiquitin regulator Tsg101, which is homologous to E2-ubiquitin conjugases but lacks the defining cysteine residue required for the function of these enzymes (C. Carter, personal communication, 2001). However, there is evidence that non-canonical E2 proteins can participate in ubiquitin conjugation as components of a protein complex [112]. While it seems clear that TSg101 interacts specifically with the HIV-1 L domain in a P(T/S)APP-dependent manner, the significance of the interaction remains to be established. Concluding remarks A general theme emerging is that HIV-1 assembly is controlled through a series of conformational changes that regulate the affinity of Gag for membranes, allow the formation of a spherical immature capsid shell, and determine the rearrangements required to produce a mature core. Some of these changes are likely to be modulated by host factors, as it is becoming clear that human cells contain essential co-factors for HIV-1 assembly [113]. The availability of cell lines that do not allow HIV-1 replication because of a block in assembly offers the prospect that such co-factors can be identified by complementation cloning [113-115], which may provide a basis for novel antiviral strategies designed to target the late phase of the replication cycle. With this in mind, it will be particularly important to determine what controls the membrane binding and targeting functions of MA, why the requirements for the MA-TM interaction are cell-type dependent, how CA can switch from a sphere to a cone assembly mode, how nucleic acid controls the assembly function of NC, and what role ubiquitin plays in virus release.
Referência(s)