We show here that widely divergent human and yeast ubiquitin ligases of the NEDD4 family, and even a subset of isolated HECT domains, possess the intrinsic ability to function in HIV-1 release. The truncated C2 domain of NEDD4-2s provides a natural Gag-targeting module, which accounts for the unique ability of authentic NEDD4-2s to rescue HIV-1 budding defects. However, other NEDD4 family members, including yeast Rsp5, and in some cases even their isolated catalytic HECT domains, acquire the same ability if targeted to HIV-1 Gag. A common property that is shared by widely divergent NEDD4 family members is the preferential catalysis of K63-linked ubiquitin chains, and at least in the case of yeast Rsp5, the isolated HECT domain is sufficient to synthesize such chains 
. Our data support a model in which the ability to conjugate K63 chains to a viral or cellular substrate in the immediate vicinity of the emerging bud is central to the ability to stimulate virus release.
In our previous study, the unique potency of NEDD4-2s in the ΔPTAPP rescue assay did depend on its C2 domain being truncated, and was not shared by several other NEDD4 family members with intact C2 domains 
. One possible explanation for these observations was that the natural truncation of the C2 domain in NEDD4-2s relieves an autoinhibition, which would be consistent with a study showing that the catalytic activity of a subset of C2-WW-HECT E3s is regulated through an inhibitory interaction between their C2 and HECT domains 
. On the other hand, the residual C2 domain of NEDD4-2s was essential for activity in the ΔPTAPP rescue assay 
. This finding raised the possibility that the C2 domain remnant of NEDD4-2s, which corresponds to β-strands 7 and 8 of the intact domain 
, constitutes a functional domain on its own that plays an active role in the rescue of HIV-1 budding. The present study supports this notion by demonstrating that the residual C2 domain of NEDD4-2s is sufficient to transfer the ability to rescue HIV-1 budding defects to other NEDD4 family ubiquitin ligases, and even to a subset of isolated HECT domains.
Our previous results suggested that the C2 domain remnant of NEDD4-2s is required for activity in the ΔPTAPP rescue assay, because it mediates the association of the ubiquitin ligase with HIV-1 Gag 
. In support of this concept, we now show that other NEDD4 family members, and some isolated HECT domains, associate with HIV-1 Gag if tagged with the residual C2 domain of NEDD4-2s. Additional strong support is provided by the fact that we were able to functionally replace the C2 domain remnant of NEDD4-2s with CypA, an entirely unrelated protein that has long been known to specifically interact with HIV-1 Gag 
HECT E3s contain two broad functional regions: a large N-terminal region required for substrate recognition, and a C-terminal region (the HECT domain) which catalyzes the ubiquitination of bound substrates 
. Apart from the C2 domain, the N-terminal regions of NEDD4 family members harbor multiple WW domains, which we previously found dispensable for the rescue of HIV-1 budding by NEDD4-2s 
. In the present study, essentially the entire N-terminal substrate recognition portion of NEDD4-2s became dispensable in the presence of CypA, which served as a Gag-targeting module. The simplest interpretation of this result is that no substrate other than Gag needs to be recognized to stimulate virus release. However, if a transacting factor rather than Gag is the relevant substrate for ubiquitination as proposed 
, then the isolated catalytic HECT domain of NEDD4-2s must be sufficient to recognize that factor.
One potential transacting factor is ESCRT-I, because Sundquist and colleagues have demonstrated that the stimulation of HIV-1ΔPTAPP
release by NEDD4-2s depends on Tsg101/ESCRT-I 
. Furthermore, these authors showed that NEDD4-2s overexpression induces the ubiquitination of ESCRT-I complexes, particularly of those that contain MVB12B. They also reported that a PPQY sequence in MVB12B, which constitutes a potential binding site for WW domains, contributes to the ubiquitination of MVB12B/ESCRT-I complexes by NEDD4-2s. Based on these results, it was suggested that NEDD4-2s-mediated ubiquitination may activate ESCRT-I to function in HIV-1 release 
. If this hypothesis is correct, then our observations imply that NEDD4-2s must remain capable of recognizing ESCRT-I as a substrate even in the absence of its N-terminal substrate recognition domain.
It has also been suggested that NEDD4 family E3s interact through their HECT domains with as yet unknown components of the ESCRT pathway, because several NEDD4 family members, and the isolated HECT domain of WWP1, localized to aberrant endosomal class E compartments induced by dominant-negative VPS4 
. We have now observed that Rsp5, the single C2-WW-HECT E3 of Saccharomyces cerevisiae
, can strongly stimulate HIV-1ΔPTAPP
release and Gag processing when its C2 domain is replaced. Thus, if an interaction with an ESCRT pathway component is required for activity in the ΔPTAPP rescue assay, such an interaction and the interfaces involved must be conserved between yeast and man. One reported interaction that potentially meets these criteria is that between NEDD4 and ALIX or their yeast homologues Rsp5 and Bro1 
. Notably, the protein regions involved in the interaction appear conserved, because yeast Rsp5 co-immunoprecipitated with mammalian ALIX 
. However, the NEDD4-ALIX interaction may depend on WW domains 
, which are dispensable for the ability of NEDD4-2s to rescue HIV-1ΔPTAPP 
In principle, ESCRT pathway components could also be recruited via ubiquitinated Gag, because the upstream ESCRT complexes each possess at least one component that binds ubiquitin 
. For instance, the human ESCRT-I components Tsg101 and VPS37A contain ubiquitin-binding domains 
. It was also recently reported that the ESCRT-associated protein ALIX specifically binds to ubiquitin 
.However, at least the ubiquitin-binding activity of Tsg101 is not required for the rescue of HIV-1 budding by NEDD4-2s 
. Also, there is evidence that the ubiquitination of Gag is dispensable, because the PPxY-dependent budding of a foamy virus Gag protein completely devoid of ubiquitin acceptors could be stimulated by catalytically active WWP1 
. However, in the latter case, Gag-associated WWP1 could have served as an alternative ubiquitin acceptor, since the enzyme is capable of auto-ubiquitination.
K63-linked ubiquitin chains are required for the transport of at least some cargo into MVB 
, and have also been implicated in the function of PPxY-type L domains 
, which act by recruiting NEDD4 family members 
. Several NEDD4 family members have indeed been shown to preferentially synthesize K63-linked ubiquitin chains, including mammalian NEDD4-1 and ITCH 
, as well as yeast Rsp5 
. In contrast, E6AP, another HECT domain E3, preferentially synthesizes K48-linked chains, which provide a signal for proteasomal degradation 
. At least in the case of ITCH, chain type specificity is determined by the C lobe of the HECT domain 
. For instance, the replacement of the C lobe of ITCH with that of HUWE1 considerably reduces the preference for the synthesis of K63 chains 
. In the present study, we observed that an ITCH/HUWE1 C lobe chimera did not efficiently rescue HIV-1 budding. Interestingly, the C lobe chimera retained the ability to efficiently ubiquitinate Gag, but lacked the ability of the parental ITCH construct to induce the attachment of K63-linked ubiquitin chains to Gag. Taken together, these data indicate that the ability to synthesize K63 chains is crucial for the stimulation of HIV-1 budding.
Structural studies indicate that the conformations of K63- and K48-linked chains are markedly distinct. Specifically, K63-linked di- or tetraubiquitin chains exhibit an extended conformation in which functionally important surface hydrophobic residues are constitutively exposed, whereas K48-linked chains can adopt a closed conformation in which these hydrophobic surface residues are sequestered 
. It is thus likely that linkage-specific conformations provide a basis for the recruitment of distinct cellular recognition factors. Interestingly, it has recently emerged that K63-linked ubiquitin chains serve as specific signals for the ESCRT-dependent sorting of cargo into MVBs 
. For instance, in the case of the yeast membrane protein Gap1, monoubiquitination is sufficient for its efficient endocytosis 
. However, the presence of short K63-linked chains is required for the entry of Gap1 into the MVB pathway 
, suggesting preferential recognition of K63-linked chains by some component of the ESCRT machinery. The results presented here imply that HECT ubiquitin ligase-stimulated virus budding, which is also ESCRT-dependent 
, is governed by the same type of ubiquitin modification.