In this study we have demonstrated that LPS-activated DCs concentrate HIV into a single intracellular compartment along with a subset of cell surface proteins, most strikingly CD81, as has been reported elsewhere 
. Surprisingly, the compartment remained accessible to surface-applied ligands, indicating that the concentrated HIV did not reside in an endocytic compartment but instead was sequestered in an invaginated, plasma membrane derived pocket-like structure. Accessibility of HIV to surface-applied probes was maintained throughout extended culture, suggesting that the pocket-like structure was not an intermediate vesicle destined for the endocytic pathway.
Live cell analysis showed that after contact with CD4 T cells, the HIV compartment was rapidly polarized to the cell–cell junctions, and transmission of viral particles occurred in discreet, relatively infrequent events. During prolonged T cell contact, the compartment appeared to be quite dynamic, releasing and re-incorporating individual virions and at times even splitting into two distinct structures and reforming as one, suggesting that the compartment was not an vesicle with a single limiting membrane but instead might consist of multiple compacted membrane domains that can stretch and re-form in the cell.
summarizes two competing models of trans
-infection that have been presented in the literature. The exosome model () proposes that HIV is sequestered into a subcellular compartment in mature DCs that is endocytic in origin 
and likely to be a late endosomal/multivesicular body structure 
. In support of this model, HIV is released from DCs in association with exosomal structures formed in the MVB mediated pathway of exocytosis 
. This model proposes that trans
-infection occurs by re-exposure of the contents of the HIV containing MVB/late endosome at the cell surface, an event requiring fusion of the MVB and plasma membranes.
The exosome model was recently challenged by Cavrois et al., who used viral infectivity assays to demonstrate that trans-infection occurred primarily by surface-accessible HIV. In their central experiment, HIV bound at 4°C prior to co-culture was entirely inhibited by a soluble CD4 (sCD4) protein, and when the DCs were shifted to 37°C before sCD4 addition, the inhibitor still blocked infection. Cavrois et al. reasoned that sCD4 should not have access to the internalized HIV and therefore that the internalized virus does not substantially contribute to productive transfer of infection. They therefore proposed that virus particles bound to the external plasma membrane are the primary source of trans-infection, and that the internalized virus observed by others likely was bound for lysosomal degradation ().
The results presented here reconcile these two models by demonstrating that the intracellular, apparently endocytosed HIV remains fully accessible to a surface-applied inhibitor (). In this model, HIV is taken up by DCs and sequestered in the cytoplasm by invagination of the plasma membrane to form a pocket-like, intracellular compartment that remains contiguous with the cell surface. Individual virus particles can escape from the pocket-like structure and infect target cells at the infectious synapse without the need for exocytic delivery. Membrane invagination is likely to engage the endocytic machinery; however, fusion and endocytic maturation is arrested, resulting in a membrane enclosed intracellular structure that is not subject to endosomal degradation.
Although we have been able to confirm the virus neutralization results of Cavrois et al., the imaging-based approaches presented here are inconsistent with their conclusion that only the minority of virus particles that remain on the cell surface are responsible for trans
-infection. Virus particles that remain on the cell surface are likely to remain infectious, and undoubtedly can contribute to trans
-infection if the DC encounters a T cell before sequestration of that virus. However, we have demonstrated that the sequestered virus is also able to infect T cells after re-emerging from the compartment at the infectious synapse. We hypothesize that HIV is sequestered within the DCs in order to prevent surface-mediated transfer, however a small amount of egress from the compartment results in efficient infection of the target cells and results in higher levels of infection than that generated by an equivalent amount of cell free HIV. Consistent with this idea, Cavrois et al. demonstrated a progressive loss of infectious transfer over time after shifting HIV-loaded DCs from 4°C to 37°C 
. They concluded that internalization of the HIV at 37°C resulted in endocytosis and degradation of the virus. We have observed similar loss of infectivity concomitant with sequestration into the invaginated compartment over time. Under those conditions, however, we did not observe substantial loss of the fluorescent HIV signal and the virus remained accessible to surface applied probes (data not shown). We therefore believe that what Cavrois et al. were observing was the loss of infection arising from sequestration and not endocytic degradation of the virus.
Similar intracellular structures have been recently described in HIV infected macrophages, which concentrate virions in CD81-positive, plasma membrane-derived intracellular invaginations during productive infection 
. EM analysis revealed that the intracellular HIV-containing compartment, previously identified as a multivesicular body, consisted of a network of invaginated structures contiguous with the plasma membrane that were induced following HIV infection. The structures were accessible by membrane impermeant probes applied at 4°C, similar to the experiments presented in this study 
. Those reports resolved the conflicting models of virus assembly sites in macrophages and identified a previously unknown mechanism for physical sequestration of viral particles in a non-endocytic, surface exposed cellular compartment. Interestingly, a recent report demonstrated that sequestered HIV was rapidly translocated to the virological synapse formed between infected macrophages and uninfected T cells 
We propose that dendritic cells use similar mechanisms to sequester HIV when it is taken up from the extracellular milleau and concentrated into the invaginated plasma-membrane derived structure described in this study. After sequestration, the virus can remain intact for an extended period of time and either traffic into the cell for endocytic degradation or transfer to other cells to share antigenic processing and presentation functions. Because dendritic cells constantly interact and stimulate CD4 T cells, even infrequent transmission of intact HIV particles during cellular communication events can result in efficient dissemination of HIV from the very cells that are designed to control its infections.