As discussed earlier, the lipid bilayers of many retroviruses are enriched in cholesterol, which supports their genesis in lipid raft microdomains [36,37,38]. This cholesterol enrichment appears to be important for viral infectivity. Depleting cholesterol from virions with ß-cyclodextrin (BCD) significantly inhibits the infectivity of HIV-1 virions [40,41,42,43]. A limitation of cholesterol depletion approaches is that removal of the majority of cholesterol from HIV-1 virions results in loss of viral components due to the formation of holes in the viral lipid bilayer [41]. However, modest depletion of cholesterol using milder cholesterol sequestering agents such as 2-hydroxypropyl-ß-cyclodextrin (2OHpßCD) does not cause substantial loss of viral proteins, but infectivity is still markedly reduced [40,41,42,43]. In this context, replenishment of cholesterol restores the infectivity of HIV-1 indicating that the virions were not completely disrupted after modest removal of cholesterol [40,41,42,43]. The effects of cholesterol depletion on virion density were also reversed by cholesterol replenishment [40]. Interestingly, replenishment with the cholesterol analog cholestenone reverses the shift in density induced by cholesterol depletion but does not restore virus infectivity [40]. The infectivity of VSV-G-pseudotyped HIV-1 virions was less susceptible to cholesterol depletion than the infectivity of virions bearing the native HIV-1 Env glycoproteins, suggesting that this sterol is critical for receptor-mediated HIV-1 entry [42]. Furthermore, a number of studies found that HIV-1 particles produced from cholesterol-depleted cells were poorly infectious compared to those produced from normal cells [44,45,46]. Because virions bud from cholesterol-rich microdomains, cholesterol depletion results in a reduced amount of cholesterol on the viral membrane, which in turn inhibits viral infectivity. Cholesterol-chelating compounds like nystatin have also been reported to inhibit HIV-1 infectivity [42,47]. Recently, we have shown that the cholesterol-binding compound amphotericin B methyl ester (AME), a water-soluble derivative of amphotericin B, inhibits HIV-1 entry [48,49,50,51].

Depletion of cholesterol from the target cell inhibits viral fusion and entry, resulting in impaired infectivity [41,52,53,54,55]. This inhibition could be due to perturbation of lipid rafts, resulting in dissociation of CD4 or co-receptors from these microdomains. It was reported that both CD4 and CCR5 are associated with lipid rafts [55,56], and that a CD4 mutant that is defective in raft association is unable to support HIV-1 infection in CD4⁺ T-cells [57]. However, studies from other groups suggested that HIV-1 entry is not dependent on the raft localization of CD4 or CCR5 [58,59]. It has been suggested that CXCR4, although normally not raft-associated, is recruited into raft aggregates after gp120 binding [55]. Interestingly, the inhibitory effects of cholesterol depletion can be overcome by coreceptor overexpression [60]. It has also been reported that a highly conserved motif, LWYIK, located near the N-terminus of the membrane-spanning domain of gp41, binds to cholesterol and that mutations in this domain inhibit HIV-1 infectivity [61-63]. In addition to acute cholesterol depletion with cyclodextrins, an alternative approach to limiting the formation of lipid rafts is to inhibit cholesterol biosynthesis using statin drugs, which target the enzyme 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA). For example, lovastatin treatment of macrophages significantly inhibits productive HIV-1 entry [64]. Further, drugs like nystatin and filipin that modify the cholesterol organization within cells without affecting overall cholesterol concentration also inhibit HIV-1 entry into macrophages [64]. The impairment of entry is correlated with the reduction in cell surface expression of CD4 and CCR5 [64]. This suggests that, at least in macrophages, cholesterol could potentially enhance the ability of the gp120-gp41 complex to cluster CD4 and coreceptors during the initiation of fusion.

Cholesterol and caveolae – flask-shaped invaginations in the plasma membrane that are categorized as a special type of lipid raft microdomain [65] – have also been implicated in MLV infection. The entry of amphotropic MLV was initially thought to occur directly at the plasma membrane, whereas in most cell types ectotropic MLV was reported to enter by pH-dependent fusion after endosomal uptake [66,67]. However, more recent studies show that amphotrophic MLV also enters target cells through caveolae-mediated endocytosis via its receptor Pit2 [68,69]. Beer et al. also found a direct association of Pit2 with caveolin, the protein component of caveolae, as demonstrated by coimmunoprecipitation and inhibition of amphotropic MLV infection by dominant-negative caveolin-1 [68]. The receptor for ectotropic MLV is the cationic amino acid transporter 1 (CAT1), and, based on its colocalization with caveolin, is thought to be associated with caveolae [66,70]. CAT1 is also detected in DRM fractions [71]. It has been reported that depleting cholesterol from target cells that express endogenous or over-expressed CAT-1 resulted in inhibition of MLV entry [71,72]. These studies indicate that cholesterol plays an important role in caveolae-mediated endocytosis of MLV.

There is a considerable amount of evidence to support the hypothesis that HTLV-1 entry is dependent on cholesterol-rich domains. Niyogi and Hildreth reported that monoclonal antibodies that bind to raft-associated proteins, including several GPI-anchored proteins, inhibit HTLV-1-induced syncytium formation [73]. Depletion of cellular cholesterol with BCD inhibited syncytium formation [73] and binding of HTLV-1 virions to T cells [74]. Two of the reported HTLV receptors, GLUT-1 and neuropilin-1, have been shown to colocalize with raft-associated markers and to cluster in specific membrane microdomains [75,76]. Cholesterol-depletion experiments suggest that cholesterol promotes efficient HTLV entry into target cells without affecting receptor binding [76]. These studies support the importance of cholesterol and raft-associated receptors in HTLV-1 entry.

The importance of cholesterol has also been demonstrated for other retroviruses. For EIAV, BCD treatment of the virion but not the target cell interferes with viral infectivity [77]. SIV infectivity is also susceptible to cholesterol depletion [41]. Finally, we have shown that the cholesterol-binding compound AME inhibits the infectivity of SIV by more than 20-fold, similar to its effect on HIV-1 [49,50].

Sphingolipids, like cholesterol, are an important constituent of lipid rafts, and have also been reported to play an important role in viral entry. Several lines of evidence suggest that GSLs in the cell membrane serve to promote HIV-1 binding. It was reported that the V3 loop of gp120 interacts with galactoceramide (GalCer) to facilitate infection of cells lacking CD4 [78,79,80,81,82]. Yahi et al. reported that a synthetic peptide derived from the V3 loop consensus motif (GPGRAF) inhibited HIV-1 infection both in CD4-negative and CD4-positive cells [83]. Synthetic soluble analogs of GalCer, and anti-GalCer antibodies, reportedly inhibit HIV-1 entry when present during infection [80,82,84]. Hammache et al. demonstrated specific interactions of HIV-1 and HIV-2 Env glycoproteins with GSLs GalCer, GM3, and globotriaosylceramide (Gb3) [81]. A similar study showed preferential interactions of gp120 from CXCR4- and CCR5-tropic viruses with Gb3 and GM3, respectively [85]. Several studies suggest that besides GM3, GM1 also functions as a cofactor for HIV-1 Env-mediated fusion [81,86,87,88,89]. Interestingly, reconstitution of GSLs such as GM3 and Gb3 in GSL-depleted GM95 cells restored HIV-1 Env-mediated fusion [88,90]. The galactosyl ceramide/sulfatide was shown to bind directly to gp120 [78,79]. Pretreating CD4⁺ cells with compounds that inhibit GSL synthesis has a significant effect on HIV-1 fusion [88]. The requirement for GSLs in target cells for promoting entry could not be rescued by overexpression of CD4 or coreceptors [87]. Nguyen and Taub showed that co-treatment of target cells with sphingomyelinase (SMase) and cholesterol oxidase reduced HIV-1 infectivity by several fold [91]. SMase treatment of target cells restricted the lateral diffusion of CD4 and inhibited HIV-1 fusion [92]. The upregulation of ceramide, a by-product of SMase treatment, significantly reduced HIV-1 infectivity in CD4⁺ lymphocytes and in monocyte-derived macrophages without overt toxicity [93]. Thus, GSLs may potentially enhance HIV-1 Env-mediated fusion by serving as direct binding sites for Env on the cell surface [89,94]. Furthermore, GSLs may promote the clustering of CD4 and coreceptors in specific domains and regulate their mobility in the target cell membrane, perhaps promoting HIV-1 entry by stabilizing fusion intermediates [89]. In contrast, CD4⁺/coreceptor⁺ derivatives of a mouse melanoma cell line (B16) that expresses exceptionally high levels of GM3 [95] were resistant to CD4-dependent fusion [96]. Interestingly, this block was rescued by pretreating the target cells with GSL inhibitors such as 1-phenyl-2-palmitoylamino-3-morpholino-1-propanol (PPMP); recovery of fusion correlated with downregulation of GM3 to normal levels [96]. The authors proposed that the elevated expression of GM3 in B16 cells could restrict the mobility of CD4 and/or chemokine receptors, thereby blocking the interactions between the different cell surface receptors required for fusion [96]. These observations suggest that GSLs can enhance or inhibit HIV-1 fusion and entry depending on their levels of expression. GSLs could act at different steps in the virus entry process; i.e., they could stabilize virus binding to the cell surface, or could promote the association of CD4/gp120 complexes with co-receptor [97]. Although a role for GSLs in HIV-1 entry is compelling, the mechanism underlying the GSL requirement for viral entry remains to be elucidated.

Phosphatidylserine (PS) is one of the membrane phospholipids that is enriched in HIV-1 virions relative to the host cell plasma membrane [36,37,38]. PS is normally present in the inner leaflet of the lipid bilayer; however, during apoptosis, PS is exposed in the outer leaflet [98]. Following infection with HIV-1, T cells undergo programmed cell death, resulting in translocation of PS to the outer leaflet. Differentiated macrophages, but not monocytes, constitutively express PS in the outer leaflet of their plasma membranes [99]. HIV-1 particles produced from macrophages thus incorporate relatively high levels of PS in the outer leaflet of their envelopes [100]. Recognition of PS in the viral envelope by macrophages is not required for virus binding; however, PS reportedly stabilizes virus-cell interactions, favoring more efficient fusion [100]. Miller and co-workers found that treating cells with PS-containing liposomes increased infection by multiple enveloped viruses by up to 20-fold [101,102]. Treating with PS-enriched liposomes does not alter receptor levels or Env binding to cells, suggesting that PS enhances virus fusion [100]. In some settings, glycosylation of retroviral receptors blocks virus entry, apparently serving as a host-cell defense mechanism against retroviral infection [103]. In this context, treatment of cells with PS liposomes was observed to overcome the infectivity block [101]. Although the mechanism behind this phenomenon awaits clarification, PS treatment does not appear to act by altering the glycosylation state of the receptor. Unlike the rapid enhancement in infectivity conferred by PS treatment in cells containing functional virus receptors, the effect in cells expressing glycosylated receptors occurs with markedly delayed kinetics [101].

The finding that the envelopes of many retroviruses are markedly enriched in cholesterol relative to the host cell plasma membrane suggests that retroviral budding takes place in cholesterol-enriched membrane microdomains [33,36,37,38,107,108]. The hypothesis that retroviral assembly takes place in lipid rafts is also supported by biochemical analyses showing that the Gag proteins of several retroviruses, including HIV-1 [45,46,109,110,111,112,113,114], MLV [115], and HTLV-1 [116] are associated with DRMs. Microdomains containing highly multimerized Gag are likely to be denser than classical lipid rafts because of the tight packing of large numbers of Gag molecules [109,111,112]. The exact nature of membrane microdomain(s) with which Gag associates, and how Gag binding alters the properties of such microdomains, remains to be fully understood.

There are several motifs in retroviral Gag proteins required for membrane binding and DRM association. In the case of HIV-1 and many other retroviruses, the association of the Gag precursor with membranes is mediated by a myristic acid moiety covalently linked to the N-terminal Gly of the MA domain, and a positively charged cluster of amino acids located near the N-terminus of the MA domain. Early reports showed that myristylation of Gag is critical for Gag targeting to membrane, as mutating the N-terminal Gly to which the myristate is covalently attached blocks Gag-membrane association [117-119]. It was proposed that the membrane binding of Gag is regulated by a myristyl switch mechanism, whereby the myristate is in equilibrium between a sequestered and an exposed conformation [120,121]. This model is supported by both structural and biochemical evidence [122,123,124,125,126]. Summers and co-workers have demonstrated that the myristyl switch is triggered by multimerization of Gag [126]. Mutating residues near the N-terminus of MA blocks the myristyl switch [122,127]. Pulse-chase analysis suggests that Gag binding to membrane is rapid but DRM association is slower, and Gag association with lipid rafts is enhanced or stabilized by Gag-Gag interactions [45,112]. Although Gag multimerization appears to stabilize Gag-DRM association, studies with an assembly-deficient NC mutant demonstrated that lower-order Gag multimerization is sufficient for Gag binding to membrane and DRM [128].

The functional relevance of the association of Gag with lipid rafts has been demonstrated by studies performed with cholesterol-depleting agents (e.g., BCD), cholesterol biosynthesis inhibitors (e.g., statins), cholesterol sequestering agents (e.g., nystatin), and a cholesterol-binding fungal antibiotic (AME) [45,46,49,129]. Treating virus-producing cells with these cholesterol-disrupting agents impairs HIV-1 particle production [45,46,49,129]. The inhibition of HIV-1 production by cholesterol depletion is due to reduced Gag binding to the plasma membrane and defective higher-order Gag multimerization [114]. Treating Gag-expressing cells with an unsaturated fatty acid that replaces the N-terminal myristate reduces Gag association with DRMs and virus production [130]. Feng et al. reported that treating virus-producing cells with α-interferon inhibited HTLV-1 assembly by preventing Gag association with rafts [116]. These studies indicate that lipid rafts play an important role in retroviruses assembly and release.

The HIV-1 [46,113,131] and MLV [46,68,132] Env glycoproteins are reportedly DRM/lipid raft associated. Acylation of proteins, especially involving multiple palmitates, often leads to raft association. Although the Env glycoproteins of HIV-1 [131,133,134,135], MLV [136,137], and RSV [138] are palmitoylated, there are contradicting reports in the literature regarding the relationship between retroviral Env palmitylation and DRM association. Some studies reported that palmitylation of two Cys residues in the gp41 cytoplasmic tail (CT) is required for Env-DRM association [131,133]; other reports, however, found that palmitylation is not required for Env-DRM association, Env incorporation into virions, or viral infectivity [133,134]. Studies from Clapham’s group showed that Gag-Env interactions are required for Env association with lipid rafts, as mutations in MA or the gp41 CT that block the putative Gag-gp41 interaction prevent Env association with DRM [139]. Furthermore, these authors showed that Gag coexpression is required for DRM association of Env, suggesting that Gag recruits Env into rafts [139]. By using protein colocalization and DRM fractionation assays, Nabel and co-workers observed that HIV-1 Env and the Ebola glycoprotein localize to distinct domains with HIV-1 Gag [140]. Johnson and co-workers using scanning electron microcopy showed that the glycoproteins of HIV-1, VSV, and MLV are recruited to budding sites formed by either HIV-1 or RSV Gag whereas RSV Env is more selective, exhibiting recruitment only to RSV budding sites [141]. In this regard, it has been demonstrated that the CT of gp41 is required for the Env incorporation into virions in a number of T-cell lines and primary cell types [142]. Mutating the palmitylation site in MLV Env diminishes the association of Env with DRM and significantly reduces Env incorporation [132,136]. However, although the palmitylation of Env reduces its expression on the cell surface, palmitylation has no significant impact on syncytium formation or virus replication [132,136].

Nef, an accessory protein encoded by HIV-1, HIV-2 and SIV, plays an important role in the pathogenesis of these primate lentiviruses. The functions of Nef include enhancement of viral infectivity and downregulation of CD4 and MHC class I molecules from the surface of infected cells [143]. Nef undergoes N-terminal myristylation required for membrane binding and DRM association [144,145]. In addition to N-terminal myristylation, basic residues in Nef also contribute to membrane binding and raft association [146]. Nef expression has been reported to increase the levels of cholesterol and GM1 in virions, perhaps contributing to the enhancement of infectivity induced by Nef [145,147]. In macrophages, Nef interacts with ABCA1 (ATP-binding cassette transporter A1) and impairs ABCA1-mediated cholesterol efflux, thereby increasing virion cholesterol content and infectivity [148]. In contrast to this study, Brugger and colleagues observed that Nef increased SM and reduced PC without affecting the levels of cholesterol in virions produced from the MT-4 T-cell line [149]. Nef associates with lipid rafts in which kinases like Lck, Fyn and LAT are enriched, and stimulates p21-activated kinase (PAK) [144,150]. Fackler and co-workers demonstrated that PAK binding and infectivity enhancement require raft association, whereas CD4 downregulation does not; nevertheless, membrane binding is required for Nef function [146,150]. In contrast, Schwartz and co-workers found that not only CD4 downregulation but also viral infectivity is independent of Nef association with lipid rafts [151]. The differences in these studies could be attributed to the use of different cell types or Nef alleles. Additional studies will be required for a more complete understanding of the role of raft association in Nef function.

Vpu is another accessory protein that associates with membranes, though unlike Nef it is not incorporated into viral particles. Vpu significantly enhances HIV-1 release in certain cell types; in these cell types, in the absence of Vpu, virions remain tethered to the plasma membrane after completion of the budding process [152]. Recent studies have identified CD317/BST-2 (also named tetherin) as the host restriction factor that tethers budded virions to the cell surface and is counteracted by Vpu [153,154]. It has been observed that tetherin and Vpu are both raft-associated [155,222], further supporting the concept that HIV-1 assembly takes place in lipid rafts.

PI(4,5)P₂ plays an important role in the association of HIV-1 Gag with the plasma membrane. PI(4,5)P₂ belongs to the phosphoinositide family of lipids, the members of which differ in the number and position of phosphates on their inositol head groups [156]. The different phosphoinositide family members are distributed at specific membranes within the cell. For example, PI(3)P is concentrated in early endosomes, PI(3,5)P₂ is most abundant in late endosomes and PI(4,5)P₂ is concentrated primarily on the inner leaflet of the plasma membrane [157]. The phosphoinositides help regulate the location of a number of cellular proteins; for example, those that contain pleckstrin homology (PH) and epsin N-terminal homology (ENTH) domains bind directly to the head-group of PI(4,5)P₂ through their positively charged cluster of basic amino acids [156]. PI(4,5)P₂ levels can be regulated by either overexpression of a constitutively active ADP-ribosylation factor 6 (Arf6) mutant, Arf6/Q67L, that translocates PI(4,5)P₂ from the plasma membrane to PI(4,5)P₂-enriched endosomal structures [158,159] or by phosphatases such as polyphosphoinositide 5-phosphatase IV (5ptaseIV) that hydrolyzes the phosphate at position 5, thereby reducing the levels of PI(4,5)P₂ in the plasma membrane [160].

It was proposed that the highly basic domain of the HIV-1 MA domain binds to negatively charged lipids in the inner leaflet of the plasma membrane [121,161,162]. Mutating basic residues in the MA domain mistargets Gag, such that it associates with internal compartments rather than the plasma membrane [122,163-169]. These compartments were defined as late endosomes or MVBs [165]. It was shown that overexpression of Arf6/Q67L retargets Gag to PI(4,5)P₂-enriched endosomal vesicles, whereas overexpression of 5ptaseIV relocalizes Gag from the plasma membrane to late endosmes/MVBs [170,171]. In both cases, due to retargeting of Gag, HIV-1 release is severely reduced. These findings demonstrated that PI(4,5)P₂ in the plasma membrane is required for efficient HIV-1 release.

Several groups, using different approaches, have demonstrated a direct interaction between MA and PI(4,5)P₂[38,124,170,172]. Summers and co-workers used nuclear magnetic resonance (NMR) spectroscopy to solve the structure of a complex between myristylated MA and a soluble derivative of PI(4,5)P₂[124,173]. This study suggested that besides electrostatic interactions between the negatively charged headgroup of PI(4,5)P₂ and basic residues in MA, the 2’-unsaturated acyl chain of PI(4,5)P₂ binds to a hydrophobic cleft within the globular core of MA (Figure 6) [124,173]. The extrusion of the unsaturated acyl chain from the lipid bilayer upon Gag binding could promote the partitioning of Gag to the more saturated lipid raft microenvironment [124]. Furthermore, these NMR data also suggested that PI(4,5)P₂ binding to MA triggers the myristyl switch, leading to myristate exposure. Thus, according to this model, PI(4,5)P₂ acts as both a membrane anchor and a trigger for myristate exposure. Shkriabai et al. used mass spectrometric protein footprinting analysis with a lysine-modifying agent to identify MA residues 29 and 31 as being important for the MA-PI(4,5)P₂ interaction [172]. Notably, mutation of these basic residues mistargets Gag and virus assembly to MVBs [165,167,174,175]. Ono’s group used liposome-binding assays to show that full-length myristylated Gag binds to PI(4,5)P₂-enriched vesicles, and the residue 29/31 mutant showed reduced PI(4,5)P₂-dependent liposome binding, again highlighting the importance of these residues in the Gag-PI(4,5)P₂ interaction [170]. Chan et al. reported that HIV-1 and MLV virions are enriched in PI(4,5)P₂ relative to the plasma membrane, and mutating the basic residues in MA significantly reduced the incorporation of PI(4,5)P₂ into particles [38]. These observations suggest that MA interacts with PI(4,5)P₂ not only in vitro but also in cells. Direct binding of PI(4,5)P₂ is also reported for other retroviruses, e.g., HIV-2 [176], EIAV [177], and MLV [178]. Barklis and co-workers recently reported that myristylated MA organizes into hexameric rings of trimers on artificial membranes containing 60% PC, 20% PI(4,5)P₂, and 20% cholesterol [179]. Interestingly, cholesterol was observed to enhance the selectivity of MA for PI(4,5)P₂ in this in vitro system. Collectively, these studies demonstrate that diverse retroviruses exploit MA-PI(4,5)P₂ interactions for Gag trafficking and virus release.