Like other enveloped viruses, HIV-1 uses cellular machinery to bud from infected cells. We now show that Tsg101 protein, which functions in vacuolar protein sorting (Vps), is required for HIV-1 budding. The UEV domain of Tsg101 binds to an essential tetrapeptide (PTAP) motif within the p6 domain of the structural Gag protein and also to ubiquitin. Depletion of cellular Tsg101 by small interfering RNA arrests HIV-1 budding at a late stage, and budding is rescued by reintroduction of Tsg101. Dominant negative mutant Vps4 proteins that inhibit vacuolar protein sorting also arrest HIV-1 and MLV budding. These observations suggest that retroviruses bud by appropriating cellular machinery normally used in the Vps pathway to form multivesicular bodies.

HIV release requires TSG101, a cellular factor that sorts proteins into vesicles that bud into multivesicular bodies (MVB). To test whether other proteins involved in MVB biogenesis (the class E proteins) also participate in HIV release, we identified 22 candidate human class E proteins. These proteins were connected into a coherent network by 43 different protein-protein interactions, with AIP1 playing a key role in linking complexes that act early (TSG101/ESCRT-I) and late (CHMP4/ESCRT-III) in the pathway. AIP1 also binds the HIV-1 p6(Gag) and EIAV p9(Gag) proteins, indicating that it can function directly in virus budding. Human class E proteins were found in HIV-1 particles, and dominant-negative mutants of late-acting human class E proteins arrested HIV-1 budding through plasmal and endosomal membranes. These studies define a protein network required for human MVB biogenesis and indicate that the entire network participates in the release of HIV and probably many other viruses.

The carboxyl-terminal domain, residues 146 to 231, of the human immunodeficiency virus-1 (HIV-1) capsid protein [CA(146-231)] is required for capsid dimerization and viral assembly. This domain contains a stretch of 20 residues, called the major homology region (MHR), which is conserved across retroviruses and is essential for viral assembly, maturation, and infectivity. The crystal structures of CA(146-231) and CA(151-231) reveal that the globular domain is composed of four helices and an extended amino-terminal strand. CA(146-231) dimerizes through parallel packing of helix 2 across a dyad. The MHR is distinct from the dimer interface and instead forms an intricate hydrogen-bonding network that interconnects strand 1 and helices 1 and 2. Alignment of the CA(146-231) dimer with the crystal structure of the capsid amino-terminal domain provides a model for the intact protein and extends models for assembly of the central conical core of HIV-1.

The host restriction factor TRIM5␣ mediates species-specific, early blocks to retrovirus infection; susceptibility to these blocks is determined by viral capsid sequences. Here we demonstrate that TRIM5␣ variants from Old World monkeys specifically associate with the HIV type 1 (HIV-1) capsid and that this interaction depends on the TRIM5␣ B30.2 domain. Human and New World monkey TRIM5␣ proteins associated less efficiently with the HIV-1 capsid, accounting for the lack of restriction in cells of these species. After infection, the expression of a restricting TRIM5␣ in the target cells correlated with a decrease in the amount of particulate capsid in the cytosol. In some cases, this loss of particulate capsid was accompanied by a detectable increase in soluble capsid protein.Inhibiting the proteasome did not abrogate restriction. Thus, TRIM5␣ restricts retroviral infection by specifically recognizing the capsid and promoting its rapid, premature disassembly.

SUMMARY The mature capsids of HIV and other retroviruses organize and package the viral genome and its associated enzymes for delivery into host cells. The HIV capsid is a fullerene cone: a variably curved, closed shell composed of approximately 250 hexamers and exactly 12 pentamers of the viral CA protein. We devised methods for isolating soluble, assembly-competent CA hexamers and derived four crystallographically independent models that define the structure of this capsid assembly unit at atomic resolution. A ring of six CA N-terminal domains form an apparently rigid core, surrounded by an outer ring of C-terminal domains. Mobility of the outer ring appears to be an underlying mechanism for generating the variably curved lattice in authentic capsids. Hexamer-stabilizing interfaces are highly hydrated, and this property may be key to forming quasi-equivalent interactions within hexamers and pentamers. The structures also clarify the molecular basis for capsid assembly inhibition, and should facilitate structure-based drug design strategies.