Despite the importance of macrophages for the physiopathology of AIDS, and the initial interest after their identification as the second main target of the virus in vivo, very little is known about the HIV-1 cycle in macrophages. Most studies have been performed in non-macrophage cell lines. and it is unclear whether such results hold true in macrophages. Here, we will review the HIV assembly process within infected primary macrophages, i.e. most commonly, monocyte-derived macrophages.

In this section, we focus on the late events of viral replication in macrophages. Currently, it remains unclear how the various components of the viral particle are targeted to the assembly compartment of which the exact nature and localization remain elusive (see Figure 1 for a summary). Early studies showed that infected macrophages tend to accumulate intracellular vacuoles that contain numerous viral particles 128. Since budding events have been observed at the limiting membranes of these vacuoles, 910, they are generally considered as the site of HIV-1 assembly in macrophages. We will refer to these vacuoles as the viral assembly compartment in the present review.

The trafficking of viral components to the assembly site as well as their subsequent assembly and release in the form of an infectious particle are coordinated and regulated through interactions between viral structural proteins and cellular factors. The product of the gag gene has long been recognized as the main conductor of HIV-1 assembly since its expression alone gives rise to virus-like particles having the same spherical shell structure as immature viral particles 1112. The current view of HIV-1 assembly in T cells has been recently reviewed 1314, and we will only give here a brief overview of the process.

Gag is composed of three polypeptides-- the matrix, the capsid, the nucleocapsid; and three smaller peptides that function together to coordinate membrane binding and Gag-Gag lattice interactions in immature virions 15. One of the three peptides is called p6 or the "late domain" because it is required for virus budding and release 16. The Gag precursor is synthesized in the cytosol and co-translationally myristoylated at its N-terminus, which is required for stable membrane association. It is then targeted to the cytoplasmic leaflet of membranes through mechanisms that are not fully understood. There, Gag multimerizes into microdomains, which in turn stabilize its membrane association 17.

Gag can be found in the cytosol as small oligomers detected by immuno-EM 18, but it is not known whether Gag oligomerization is a prerequisite for the spherical Gag lattice formation. Similarly, it remains unclear whether the transport of the precursor relies on free cytoplasmic diffusion or if it requires trafficking along the cytoskeleton. It has also been suggested that RNA binding to Gag could play a role in the assembly process by providing a scaffold to stabilize intermolecular Gag interactions 1920. Where and when the interaction between Gag and viral RNA occurs is still debated, but the trafficking of genomic RNA may influence Gag cytosolic fate 21222324. Of note, the majority of data concerning intracellular Gag trafficking was obtained from immortalized cell lines and does not necessarily reflect the situation in infected primary macrophages.

Among the numerous cellular factors reported to be involved in HIV-1 assembly and budding, the ESCRT cellular machinery (Endosomal Sorting Complex Required for Transport) is recruited by the p6 domain and plays a key role in the formation and release of new particles. This complex has drawn a lot of attention, and much progress has been made in the last few years in understanding its way of functioning in three important processes: formation of intraluminal vesicles in multi-vesicular bodies (MVBs), HIV-1 budding and fission from membranes, and more recently in fission of the midbody during cytokinesis. The three processes have in common the need for severing a thin membrane to allow vesicles, nascent viral particles, or cells to be released. Since this large body of work has not been reproduced in macrophages and because the mechanisms involved have been thoroughly reviewed 131525, they will not be discussed here.

Additionally, Vpu, one of the accessory proteins of HIV-1, also plays a crucial role in the terminal step of particle release (see 13). Indeed, Vpu has been recently shown to counteract the activity of a restriction factor named tetherin/BST-2/CD317 2627282930. In the absence of Vpu, viral particles bud from the plasma membrane of T cells but cannot detach due to the presence of tetherin. The action of Vpu in T cells may rely on the down-regulation of BST-2 at the cell surface through both relocalization and degradation of this factor 313233. The molecular mechanism involved in this tetherin-mediated retention remains unknown as well as the exact role of Vpu in different cell types, especially in macrophages 32.

Many studies have been based on immuno-fluorescent staining of viral proteins such as Gag in infected macrophages. In fact, Gag has multiple localizations in infected cells (see an example in Figure 2, typical of day 7 post-infection). Gag goes from a diffuse cytosolic pattern to small dots in the periphery to large intracellular compartments. Moreover, the Gag staining pattern evolves with time post-infection. Two additional reasons render the interpretation of these staining even more complex:

i) The poor resolution of the epifluorescent microscopy technique does not allow one to distinguish mature or nascent viral particles from Gag aggregates. Note that the diameter of an immature viral particle is in the range of 100 to 200 nm (mean 129 nm) 59, which is below the resolution of epifluorescent microscopes.

ii) It is impossible to distinguish incoming virions, which may fuse or be internalized, from nascent viral particles eventually being secreted. Similarly, there is no way to know whether dots observed by immunofluorescence represent infectious or non-infectious particles. Finally, we do not know if all the synthesized Gag precursor has a homogeneous behavior or if several populations of Gag precursor exist with distinct fate and function. This idea is supported by Gousset et al. showing that only part of Gag was redistributed in infected macrophages towards the synapse formed with non-infected T cells 60.

Some of these problems can be, in theory, circumvented by ultrastructural approaches. So far, only Immuno-electron microscopy (Immuno-EM) allows one to distinguish viral particles, from viral buds, and from non-assembled Gag. However, this technique remains tedious, difficult to master, and only works with very few antibodies on fixed samples.

EM studies have greatly influenced our view of the viral cycle in macrophages. Early work revealed the existence of large intracellular vacuoles in which viral particles tend to accumulate. Raposo et al. showed by immuno-EM that these vacuoles contained not only virions, but also endosomal markers such as MHC II and CD63. Based on EM profiles they also proposed that viral budding takes place at the limiting membrane of the compartments, and that fusion of these compartments can occur at the plasma membrane leading to the release of their contents; HIV-1 particles and exosomes 9. Pelchen-Matthews et al. confirmed these results and provided additional biochemical evidence that viral particles originate from late endocytic compartments and carry markers from these compartments 1061.

To our knowledge, only one team observed by Immuno-EM some ESCRT-related specific staining at the limiting membrane of these compartments 62. However, these ESCRT-components were also present elsewhere in the cell and did not appear to be relocated to the site of viral assembly upon HIV infection 62. In our preparations of macrophages, Alix and CHMP4 were present mainly in virions, but also at the limiting membrane of the viral assembly compartment (Figure 3). However, we did not succeed in finding other ESCRT-specific antibodies effective for immuno-EM despite testing a large collection. This difficulty may reflect the tightness of the ESCRT multi-protein complex. This also points to the limitations of the immuno-EM studies for which few antibodies can be used on ultrathin sections. Nevertheless, it is now well-accepted that the ESCRT machinery is recruited by HIV-1 in macrophages as well as in T cells at their respective locations for HIV-1 assembly, either inside the cell or at the plasma membrane.

Initial studies suggested the existence of an intracellular compartment specialized in the assembly and storage of viral particles. Ultrastructural studies revealed budding profiles at the limiting membrane of internal compartments 63 in a process and topology similar to the biogenesis of internal vesicles or exosomes in MVBs, which are late endosomes 9. Similar profiles were reported later 106465. Proteomic analysis of the host cell proteins incorporated into highly purified virions produced by macrophages revealed the presence of many late endosomal proteins such as MHC II, CD63, and tetraspanins 52 which is in agreement with immuno-EM studies 91061. Moreover, such virions and macrophage-derived exosomes had similar protein compositions 66. Using recombinant viruses in which a tetracysteine tag was introduced at the C-terminus of the matrix domain of Gag, it has been possible to visualize Gag trafficking in living macrophages. Accumulations of Gag were observed both at the plasma membrane and in internal compartments carrying late endosome/MVB markers 60.

Other arguments supporting the idea that productive intracellular assembly takes place in MVB-like compartments are weak as they come from studies performed in cell lines such as HeLa, HEK 293 T, or COS cells 676869. Viral budding was observed in MVBs from such cells 70, while Gag was found to be transported to CD63+ MVBs in an AP3-dependent manner 44. It has been also suggested that Gag transiently traffics through MVB-like compartment to recruit the ESCRT machinery before reaching the plasma membrane in these cell lines 71. Recently, Joshi et al. used a HIV-1 carrying a Gag-matrix mutant (29/31KE) which localizes to MVBs in all cell types, thus showing that efficient intracellular assembly and release of viral particles occurred not only in macrophages but also in T cells 72. This study therefore establishes that endosomal compartments can serve as productive sites for HIV-1 assembly in both T cells and macrophages.

A characteristic of the endocytic pathway is its progressive acidification which allows the activation of degradative enzymes. Endosomes would therefore constitute a hostile environment for HIV-1 which is a fragile virus sensitive to low pH and proteases 73. However, HIV-1 remains infectious in macrophages, even after residing in macrophages for long periods of time 3. Simultaneous identification by immuno-EM of viral assembly compartments and estimation of their pH were carried out on infected macrophages 65. While the extended network of lysosomes present in infected macrophages was correctly acidified, viral compartments were not. Endosomal acidification is required for maturation along the endocytic pathway and fusion with lysosomes. Therefore, HIV-1 may have evolved a strategy for survival in macrophages.

It has been proposed that intracellular virions observed in HIV-1-infected macrophages represent endocytosed particles produced by neighboring cells 74. Several arguments can be put forward to rule out this hypothesis: 1) Immuno-EM profiles obtained by several teams show viral particles at various stages of budding at the limiting membrane of the compartment 291065. Moreover, the viral particles seen in these compartments were often immature virions, as judged by their electron lucent material at the core and electron dense material at the periphery (see Figure 1 a schematic representation). 2) Shortly after exposure of macrophages to HIV-1, most virions are found in macropinosomes or in acidic endosomes and are subsequently degraded 6575. 3) In all the studies mentioned, the HIV-1 strains used expressed Vpu, which promotes virus release but also inhibits virus uptake by endocytosis 2876. Taken together, this strongly suggests that the majority of viral particles detected in intracellular compartments of HIV-1 infected macrophages have been de novo produced rather than recently endocytosed.

Despite the numerous evidence showing that HIV-1 assembly occurs in macrophages in MVB-related compartments, recent studies have challenged this view. They were based on the usage of the ruthenium red (RR), which is a membrane-impermeant dye added during the fixation of infected macrophages and before their analysis by electron microscopy. Deneka et al. suggested that at least some of the virus-positive, "intracellular" structures in macrophages were actually connected to the plasma membrane via very thin microchannels allowing access of the RR dye 77. Another team achieved similar results 64, and both concluded that the viral assembly compartment originates from the plasma membrane in infected macrophages. We also observed in our macrophage preparations that some viral compartments were RR+; however, 80% of them remained negative (Figure 4 and 65). Interestingly, we frequently noticed in the vicinity of the viral compartments numerous electron-dense lipid droplets that were heavily stained by the RR dye (Figure 4A, see white asterisks) in agreement with the known capacity of RR to bind lipids and suggesting their connection to the extracellular space. As previously reported for other cell types 7879, our pictures on Figure 4 reveal however the presence of electron-dense RR+ areas in the cytoplasm and mitochondria near lipid droplets, and thus indicates that the RR dye is not totally membrane-impermeant in macrophages.

A very recent study based on ion-abrasion scanning electron microscopy indicates that HIV-1-infected macrophages possess an extensive network of tubules occasionally connecting virus containing compartments with the cell surface 80. These virion-containing tubules have a diameter of 150-200 nm and thus may differ from the narrow (< 20 nm) virion-free microchannels mentioned above. Future work will aim at confirming and quantifying the presence of these microchannels or tubules using alternative techniques.

It is currently not known whether these connections to the plasma membrane are transient or permanent. However, they may account for the lack of acidification of the viral compartment mentioned above. They could also occur as an early event during the establishment of the intracellular vacuole; or on the contrary, they may precede an exocytosis process of the viral particles, although the diameter of the microchannels appears too small to accommodate virus trafficking (around 20 nm, 77).

Despite hundreds of EM profiles of HIV-1-infected macrophages analyzed, we never saw any budding event taking place at the plasma membrane like we observed in T cells (M. J. and P. B., unpublished observations). Importantly, three studies on macrophages showed that the viral compartments were accessible to Transferrin, but not to BSA-gold or immunoglobulin-coated gold beads added to the extracellular medium 26465, supporting the concept of a compartment separated from the endocytic pathway but capable of exchanges with the recycling compartment. Alternatively, Transferrin access may be due to the microchannel connections to the plasma membrane.

Altogether it remains unclear whether the viral compartments observed in HIV-infected macrophages correspond to invaginations of the plasma membrane. We favor the notion of an intracellular compartment separated from the endocytic pathway, possessing a neutral pH and transiently connected via microchannels to the plasma membrane. However, more work is needed to resolve the nature of the viral compartment in macrophages.

The limiting membrane of the compartment where viral budding takes place will eventually wrap the nascent viral particle. Therefore the lipid and protein composition of the viral membrane may reflect the origin of the assembly compartment (see 81). The HIV-1 membrane is enriched in cholesterol, GM1 and tetraspanins, supporting the idea that HIV-1 budding could take place on lipid raft-like membranes. However, several proteins known to be normally associated with rafts like CD14 and CD45 are not found in viral envelope, whereas some proteins present in HIV-1 envelope appear excluded from lipid-rafts 66.

Tetraspanins such as CD9, CD53, CD81 and CD82 were enriched both in the compartment and in the viral membrane 106182. Although CD63 was specifically associated with HIV-1 assembly compartments in macrophages, it was dispensable for the production of infectious virus 82. However, opposite results were obtained also in macrophages 83. Learning more about the function of CD63, which remains elusive, will probably help to solve this discrepancy.

The limiting membrane of the viral compartment often appears to contain molecular coats (see 77 and Figure 3B') of which, the composition remains elusive. These coats are reminiscent of flat clathrin lattices found in MVBs 84 but they appear less flat, do not contain clathrin and are also observable on the microchannels connecting the compartments to the plasma membrane 77.

Deneka et al. reported the frequent presence of "sponge-like" structures in the immediate vicinity of viral assembly compartments in infected macrophages 77. These structures are very rich in highly interconnected membranes and accessible to the RR dye. We also observed in our macrophage preparations such RR+ structures (Figure 4C), of which the nature and function remain so far unknown. As previously noticed 77, their morphology appears similar to structures observed in primary macrophages that have been exposed to aggregated low-density lipoproteins and that are also efficiently stained by RR (see 85).

HIV-1 is known to wrap into cholesterol-rich membranes that are required for viral production and infectivity. Since cholesterol efflux is inhibited in HIV-1-infected macrophages through a Nef-dependent mechanism 35, this accumulation of lipids may contribute to the appearance of the sponge-like structures. However, Nef does not promote the intracellular accumulation of viral particles in macrophages 3 and is dispensable for effective HIV-1 replication in macrophages 8687. Future work will elucidate the connection between lipid homeostasis, Nef and the assembly process in macrophages.