In the experiments described here, we have presented evidence that endophilins interact with the matrix or MA domain of the Mo-MuLV Gag protein and may contribute to the process of virion production. We detected an interaction between Mo-MuLV Gag and endophilin 2 in the yeast two-hybrid system, and in vitro, and in vivo. The analogous interaction was not detected for all retroviral Gags, but was seen for Mo-MuLV and RSV Gags. Further tests showed that exogenously expressed endophilin 2 is associated with virion particles, and is protected within the viral envelope. Several fragments of endophilin 2 were also incorporated into virions; these experiments did not identify any specific domain of endophilin as essential for the process, and it is possible that more than one domain can direct incorporation. The endophilin fragments may have been targeted to virions by dimerization with endogenous endophilin, or by indirect interactions with other proteins. About 0.1-0.2% of the intracellular levels of endophilin were recovered in virus, even for those fragments which bound poorly to Gag in yeast. Titrating the levels of endophilin expression showed that the binding sites for endophilins during virion formation are limited and the level of incorporated protein is saturable. These observations suggest that the presence of endophilins within virion particles is not simply attributable to mass action, but that the incorporation might be mediated through specific contacts, with only a limited number of sites for Gag-endophilin association. It is not clear whether the incorporation per se is involved in virion production.
If the interaction is crucial for virion production, we thought it possible that overexpression of full-length or fragments of endophilins might interfere with this process by perturbing the correct stoichiometry of the interaction between endophilin and Gag, or other proteins required for this process. Consistent with this notion, overexpression of full-length endophilin 2 did act in a dominant-negative fashion to significantly reduce Mo-MuLV virion production. The inhibition occurred in a dose-responsive manner. Overexpression of endophilin 2 or N156, a fragment that contains a coiled-coil region, had no effect on the level of expression of Gag precursor or a reporter gene within the cells. Moreover, overexpression of endophilin or its fragments had no or little effect on the production of HIV virus-like particles, correlating with our observation that there is no direct interaction between HIV-1 pr55Gag and endophilin. These data further support the notion that the inhibition of Mo-MuLV virion production we observed is not a nonspecific consequence of overexpression on cell viability or physiology. Rather, overexpression could titrate out Mo-MuLV Gag or other interacting proteins that are specifically required for Mo-MuLV and not HIV production.
We used the siRNA method to knock-down endogenous endophilin 2. This technique has been successfully used to document the requirement of TSG101 in virion assembly . In our case, however, the knock-down of endogenous endophilin 2 had no significant effect on virion yield. One possibility is that the levels of endophilins remaining inside cells after knock-down are sufficient to execute the required functions; indeed, the levels required for virion production may be very low, as only a very small proportion of the intracellular protein is incorporated. The other possibility is that other endophilin family members could compensate for the loss of expression. Indeed, this is very likely because endophilin 1 is expressed in 293T cells (data not shown) and we have shown that it can interact with Gag. It is not clear if the various endophilin family members are fully interchangeable for this or even host functions. We do know that the sequence of the siRNA oligonucleotides used in these experiments is specific to endophilin 2, and would not be able to affect the levels of endophilin 1. We were unable to identify a sequence of suitable length that was an identical match between the two mRNAs.
Retrovirus budding is the topological reverse of endocytosis , and membrane curvature changes dramatically from the relatively planar plasma membranes during formation of 100 nm virions. So far it has not been documented whether proteins that modify membrane curvature contribute to retroviral budding. In this study, we have shown that overexpression of N125, a fragment of endophilin that binds liposomes and induces tubules of diameter 20-100 nm , or of full-length endophilin, a protein that presumably promotes both positive and negative membrane curvature changes during the formation of endocytic vesicles [48, 51], causes a significant reduction of virion production. These observations raise the possibility that our overexpression disrupts normal endophilin functions, and that the binding of lipids and subsequent promotion of changes in plasma membrane curvature could be one of the functions normally exerted by endophilin to assist Mo-MuLV virion formation. It is noteworthy that clathrin, like Gag, was long thought to be sufficient to drive membrane budding by itself. More recently it has become evident that groups of proteins have to work together to bend a biological membrane . Endophilin is not the only protein that can induce membrane curvature. At least three other proteins (dynamin, amphiphysin and epsin) that are involved in clathrin-medicated endocytosis can independently trigger membrane tubulation in vitro [53–55]. These proteins might function at multiple stages during vesicle formation. Morphological analyses indicate that endophilin A is required for clathrin-mediated synaptic vesicle endocytosis at multiple stages, including the following: the early stage of endocytosis, the formation of shallow pits; late stages, with the formation of deeply invaginated, elongated pits; and fission [45, 46]. Proteins such as endophilin, which can help generate membrane curvature, might be involved in virion production at several stages, from formation of slightly curved membrane structures to stalk-like structures until viral fission.
Endophilin may act as part of a large complex, and may associate with many other proteins. The SH3 domain of endophilin has been shown to bind to the proline-rich domain of dynamin , amphiphysin  and synaptojanin, a phosphatidylinositol 5'-phosphatase implicated in synaptic-vesicle uncoating . We found that AP-2 and clathrin, but not dynamin 2, are significantly incorporated into virion particles. The incorporation of any one of the various proteins involved in vesicular trafficking may depend on the behavior of the particular complex in which it resides; one explanation for the lack of incorporation of dynamin 2 is that it may interact with a distinct pool of endophilins, one that does not interact with Gag.
Among the proteins associated with Gag are several other proteins that are implicated in retrovirus budding. Tsg101 is known to interact with the PTAP motif of the L domain of HIV-1 and many other retroviruses, and is required for efficient virion release. Members of the Nedd4 family, involved in endocytosis and recycling of membrane proteins, interact with the PPPY motif of the L domain of many other retroviruses and play a similar role. Dominant-negative fragments of Nedd4-like family members inhibit the release of viral particles much as we have seen for endophilin [28, 29]. These observations suggest that a completely functioning vesicular trafficking pathway is required for retroviruses budding [50, 56, 57]. Interactions with vesicles may also be involved in earlier stages of trafficking of genomic RNA, Env and Gag to the cell surface ; possibly the binding of endophilins to Gag can promote their association with endosomal vesicles.
Very recently, HIV-1 Gag has been shown to associate with the endocytic protein AIP-1/Alix through specific contacts with the Gag p6 domain [59, 60]; Alix is known to interact with endophilins . This observation, along with our results, suggests that endophilin could be another component that is hijacked by retroviruses to promote virion production.