For some time, it has been known that in human MDMs, HIV-1 buds into and accumulates in surface-connected intracellular compartments, or IPMCs (also termed virus-containing compartments or VCC [15, 27]). Although their origin, organization and function is poorly understood, much of our current knowledge of these compartments derives from EM studies, in which various techniques, including serial sectioning, electron tomography or ion abrasion scanning electron microscopy [10, 11, 13, 15] have indicated that IPMCs consist of complex intracellular networks of membranes, with interconnected vacuole-like and tubular components, and channel-like connections to the cell surface. However, these morphological techniques are limited to the analysis of small portions of the total volume of fixed cells and do not provide information on the dynamics of the compartments in real time. Here we have used fluorescent membrane labels - FM 4–64, CellMask and the PI(4,5)P2 probe PH-GFP, in combination with confocal z-series imaging, three-dimensional volume reconstructions and live cell imaging - to study the properties of IPMCs. In agreement with and extending previous studies [10, 11, 13, 15], we show that both uninfected and HIV-1-infected MDMs contain morphologically similar IPMCs that appear as dynamic networks of vacuoles of various sizes, connected to each other and to the cell surface by thinner tubules or closely apposed membrane sheets. Furthermore, we show that the normal morphology of IPMCs is dependent on the integrity of the actin cytoskeleton and that disrupting this integrity can stimulate the release of mature, IPMC-sequestered HIV-1.
EM analysis has previously shown individual HIV-containing vacuoles and/or CD81-, CD9- or CD44-labeled structures within MDMs, suggesting that a single cell may contain several IPMC structures [10, 11, 13, 15]. Our live cell imaging studies, which avoid fixation-induced fragmentation of membrane compartments, indicate that these vacuoles are in most cases sub-domains of single, larger IPMCs. Moreover, although most MDMs contained a single IPMC, three-dimensional reconstruction of complete cells and IPMCs highlighted the complexity and extensive cell-to-cell variability in the size and morphology of IPMCs. The rapid labeling of IPMCs (within minutes), even at 4°C when endocytosis is inhibited, and the visualization of at least one, and frequently more, connections from IPMCs to the cell surface provides additional evidence supporting the notion that the compartment is continuous with the cell surface and accessible to small molecule tracers, as suggested by previous work [11, 15, 16]. Thus, the term IPMC accurately reflects the fact that these compartments are connected to the cell surface and that the IPMC membrane can be regarded as a sub-domain of the plasma membrane. Although accessible to small tracer molecules, it has been suggested that IPMCs are not accessible to antibodies and that this might protect IPMC-sequestered virus from recognition by neutralizing antibodies [13, 28]. However, in our hands at least, IPMCs are accessible to antibodies fed from the cell surface at 37°C  and we find that many IPMCs can be accessed by high concentrations of antibodies or the fluid tracer horseradish peroxidase when incubated for 1 hour at 4°C (AP-M, unpublished data).
Significantly, many of the studies described here were performed on uninfected MDMs, demonstrating that IPMCs are not induced by HIV infection, although, as previously shown, the compartment expands in size upon HIV-1 infection . IPMCs are therefore likely to have some as yet unidentified function(s) in uninfected macrophages [11, 15]. HIV appears to use pre-existing IPMCs for assembly, suggesting that there is an advantage to the virus to exploit these compartments or that key components required for virus assembly are located within these plasma membrane sub-domains. Currently, it is not clear how HIV targets IPMCs, though the lipid PI(4,5)P2, which binds directly to the HIV matrix basic domain and plays a key role in Gag recruitment to the plasma membrane, is likely to be involved [20, 29]. We analyzed the distribution of PI(4,5)P2 in MDMs, either using the PH-GFP probe or immunostaining with a PI(4,5)P2-specific antibody [23, 24]. PH-GFP labeled the cell surface as well as IPMCs, indicating that PI(4,5)P2 is abundant in these locations. Labeling of IPMCs with anti-PI(4,5)P2 antibody required permeabilization with Triton X-100, and was poor after saponin treatment, perhaps indicating the presence of detergent-resistant membranes in IPMCs.
Given that the PH-GFP probe strongly stained IPMC membranes, it could be used to follow the compartment in live cell imaging studies. This allowed, for the first time, studies of the behavior and dynamics of IPMCs in uninfected macrophages. IPMCs labeled with PH-GFP were essentially stable throughout the time of recording, that is, for at least one hour. Similar observations were made with MDMs transfected with Gag-GFP, where IPMCs also appeared stable, though we could occasionally observe changes in the subcellular distribution of Gag-GFP in IPMCs and in IPMC-associated channels. Our studies therefore complement previous analyses with Gag-GFP or biarsenical-tetracysteine-tagged fluorescent Gag [13, 30], where accumulations in MDMs were also seen to be comparatively stable. FRAP and FLIP analyses of the PH-GFP probe demonstrated that IPMC membranes are able to rapidly exchange PH-GFP with surrounding membranes and therefore they are dynamic structures. Interestingly, we did not observe any kinetic differences in the behavior of plasma membrane or IPMC-bound PH-GFP in our experiments.
Because IPMCs are coated with actin filaments and, in mature MDMs, the structure of the IPMCs is at least in part stabilized via β2 integrin-containing focal adhesion-like complexes linking to the actin cytoskeleton , we investigated the role of actin in the organization of the IPMCs. Inhibitors of actin polymerization (latrunculin, cytochalasin D and cytochalasin E) caused the intracellular accumulations of HIV particles to disperse into smaller pockets of viruses, an effect similar to that seen after depletion of β2 integrins in MDMs . A similar effect was previously described in dendritic cells  and may explain the reduction in intracellular Gag accumulation after cytochalasin D treatment of 7-day-old MDMs . Staining with CellMask demonstrated that, although dispersed, the IPMCs remained connected to the plasma membrane in the drug-treated cells. In addition, when MDMs were transfected with PH-GFP and treated with latrunculin, the IPMCs appeared as a meshwork of membranes with clear connections to the cell surface (Additional file 12). We also showed that latrunculin treatment did not inhibit virus assembly in MDMs, but instead enhanced the release of preformed HIV-1 particles, presumably through the dispersed membrane channels. Together these experiments suggest that an intact actin cytoskeleton is required both to maintain the structure of IPMCs and regulate the release of HIV from MDMs. A recent study showed that microtubules also affect the distribution of VCCs in MDMs, and suggested that kinesin family-3A complexes may drive IPMCs toward the plasma membrane and stimulate virus release .