The host-cell restriction factor SERINC5 restricts HIV-1 infectivity without altering the lipid composition and organization of viral particles

The host-cell restriction factor SERINC5 potently suppresses the infectivity of HIV, type 1 (HIV-1) particles, and is counteracted by the viral pathogenesis factor Nef. However, the molecular mechanism by which SERINC5 restricts HIV-1 particle infectivity is still unclear. Because SERINC proteins have been suggested to facilitate the incorporation of serine during the biosynthesis of membrane lipids and because lipid composition of HIV particles is a major determinant of the infectious potential of the particles, we tested whether SERINC5-mediated restriction of HIV particle infectivity involves alterations of membrane lipid composition. We produced and purified HIV-1 particles from SERINC5293T cells with very low endogenous SERINC5 levels under conditions in which ectopically expressed SERINC5 restricts HIV-1 infectivity and is antagonized by Nef and analyzed both virions and producer cells with quantitative lipid MS. SERINC5 restriction and Nef antagonism were not associated with significant alterations in steady-state lipid composition of producer cells and HIV particles. Sphingosine metabolism kinetics were also unaltered by SERINC5 expression. Moreover, the levels of phosphatidylserine on the surface of HIV-1 particles, which may trigger uptake into non-productive internalization pathways in target cells, did not change upon expression of SERINC5 or Nef. Finally, saturating the phosphatidylserine-binding sites on HIV target cells did not affect SERINC5 restriction or Nef antagonism. These results demonstrate that the restriction of HIV-1 particle infectivity by SERINC5 does not depend on alterations in lipid composition and organization of HIV-1 particles and suggest that channeling serine into lipid biosynthesis may not be a cardinal cellular function of SERINC5.

The host-cell restriction factor SERINC5 potently suppresses the infectivity of HIV, type 1 (HIV-1) particles, and is counteracted by the viral pathogenesis factor Nef. However, the molecular mechanism by which SERINC5 restricts HIV-1 particle infectivity is still unclear. Because SERINC proteins have been suggested to facilitate the incorporation of serine during the biosynthesis of membrane lipids and because lipid composition of HIV particles is a major determinant of the infectious potential of the particles, we tested whether SERINC5-mediated restriction of HIV particle infectivity involves alterations of membrane lipid composition. We produced and purified HIV-1 particles from SERINC5293T cells with very low endogenous SERINC5 levels under conditions in which ectopically expressed SERINC5 restricts HIV-1 infectivity and is antagonized by Nef and analyzed both virions and producer cells with quantitative lipid MS. SERINC5 restriction and Nef antagonism were not associated with significant alterations in steady-state lipid composition of producer cells and HIV particles. Sphingosine metabolism kinetics were also unaltered by SERINC5 expression. Moreover, the levels of phosphatidylserine on the surface of HIV-1 particles, which may trigger uptake into non-productive internalization pathways in target cells, did not change upon expression of SERINC5 or Nef. Finally, saturating the phosphatidylserine-binding sites on HIV target cells did not affect SERINC5 restriction or Nef antagonism. These results demonstrate that the restriction of HIV-1 particle infectivity by SERINC5 does not depend on alterations in lipid composition and organization of HIV-1 particles and suggest that channeling serine into lipid biosynthesis may not be a cardinal cellular function of SERINC5.
For efficient replication in natural target cells, HIV has to surmount numerous physical barriers posed by cell-intrinsic restriction factors such as Trim5␣, APOBEC3G, CD317/tetherin, Mx2, or SAMHD1 (1,2). Because their expression is often triggered or enhanced by interferons, restriction factors are considered an essential arm of cell-intrinsic innate immunity. HIV can overcome these barriers by genetic adaptation, preventing the recognition of subviral structures by restriction factors (e.g. modification of HIV-1 4 capsid to prevent interaction with human Trim5␣) or by expression of viral antagonists that counteract the antiviral activity of restriction factors. With Vif, Vpr, Vpu, and Nef, HIV-1 encodes four accessory proteins that are not essential for virus replication in cell culture but whose role in antagonizing host-cell restrictions to optimize virus spread in vivo is increasingly emerging. Although prominent restrictions antagonized by e.g. Vif (apolipoprotein B mRNA editing enzyme), Vpr (SLX4), and Vpu (CD317/tetherin) are well established (3)(4)(5)(6), it was only recently that serine incorporator 3 (SERINC3) and SERINC5 were identified as host cell restriction factors that potently impair the infectivity of HIV-1 virions and that are antagonized by HIV-1 Nef (7,8).
Nef is a myristoylated, 25-to 34-kDa protein that, in addition to HIV-1, is encoded by HIV-2 and simian immunodeficiency virus. In the infected host, Nef potently increases virus replication and thus serves as a pathogenicity factor that accelerates disease progression (9 -11). Nef affects many central processes in HIV target cells that may contribute to its role in AIDS pathogenesis, including down-regulation of an array of receptors from the surface of infected cells (12,13), alteration of signal transduction pathways such as T cell receptor signaling, (14 -20), as well as interference with host-cell actin dynamics and motility (21)(22)(23)(24)(25)(26)(27)(28). Nef also increases the infectivity of HIV-1 particles when expressed from proviral DNA in virus-producing cells (29 -31). Although this effect is mild in most cell types, production of HIV-1 variants lacking Nef expression (HIV-1⌬Nef) results in particles that are up to 100-fold less infectious than WT HIV-1 (32). These strong differences in virion infectivity correlate with high levels of SERINC5 expression, and Nef antagonizes the antiviral activity of SERINC5 under these conditions to restore full HIV-1 infectivity (7,8). The mechanisms by which Nef antagonizes SERINC5 restriction to HIV infectivity are not fully understood (33). Nef expression reduces cell surface exposure and virion incorporation of SERINC5; however, these effects are not sufficient to fully account for the antagonism, suggesting that Nef can also inactivate virion-incorporated SERINC5 molecules (7,8,34).
Although some insight has emerged regarding the mechanism of SERINC5 antagonism by Nef, molecular details about how SERINC5 interferes with HIV infectivity remain elusive. SERINC5 is a member of the SERINC protein family that is conserved from yeast to mammals and encompasses five members that are predicted to contain 10 -12 transmembrane domains (35,36). SERINC5 is efficiently incorporated into HIV particles, and its presence in the virion may be critical for its antiviral activity (7,8). In support of this model, SERINC5 reduces the efficacy of virus entry at the fusion step, and the Env glycoprotein contains a major determinant of HIV particles for sensitivity to SERINC5 restriction (7,8,37). A recent study suggested that the reduced fusogenicity of SERINC5-containing virions reflects a direct inactivation mechanism of the Env glycoprotein by the restriction factor (38), but the underlying molecular mechanism is still unclear. The overall biochemical properties and physiological roles of SERINC proteins are also not well characterized. The only report available to date describes that SERINC proteins, when ectopically expressed in Escherichia coli, yeast, and COS-7 cells, associate with key enzymes required for lipid and sphingolipid biosynthesis and incorporate serine into membranes, which coined the name serine incorporator (35). Reflecting that HIV buds from specialized plasma membrane microdomains, its lipid envelope displays a highly specialized lipid composition that is enriched in cholesterol and sphingolipids and that significantly differs from that of bulk plasma membrane (39,40). This specialized lipid composition is critical for virion infectivity because, e.g., alterations thereof caused by inhibition of sphingolipid biosynthesis significantly reduce the infectivity of HIV-1 particles (39). Considering that serine is an essential building block of sphingolipid biosynthesis and that this process is critical for full infectivity of HIV-1 particles (39), we hypothesized that the antiviral activity of SERINC5 may be linked to modulation of host-cell lipid biosynthesis and/or lipid composition of virus particles. We noted previously that HIV-1 Nef has the ability to slightly modify the lipid and sphingolipid content of HIV-1 particles when produced from the human CD4 T cell line MT-4, which expresses little or no SERINC5 (7,41), raising the possibility that this activity is emphasized in the presence of SERINC5 and allows Nef to antagonize the restriction. To test these hypotheses, we set out to define the impact of SERINC5 on HIV-1 particle lipid composition, lipid composition and sphingosine metabolism of HIV-producing cells, and lipid organization of HIV-1 virions.

Purification of HIV particles produced in the presence of SERINC5
Ectopic expression of SERINC5 in 293T cells, which express low levels of endogenous SERINC5 that are not sufficient to exert significant antiviral activity (7), potently suppresses the infectivity of nef-negative HIV-1 (⌬Nef), whereas Nef expression by WT HIV-1 antagonizes particle infectivity restriction by SERINC5 (7,8,34,42). We therefore used ectopic expression of SERINC5 in 293T cells to test whether the presence of SERINC5 during the production of HIV particles affects the lipid composition of viral progeny. 293T cells were transfected with proviral plasmids together with an empty control or a SERINC5 expression plasmid (Fig. 1) in a large-scale format. Quantification of reverse transcriptase activity in the cell culture supernatants revealed that comparable amounts of virus particles were produced under all conditions (Fig. 1A). SERINC5 markedly reduced the relative infectivity of HIV-1 ⌬Nef particles, whereas expression of Nef from HIV-1 WT counteracted the antiviral activity of SERINC5 (Fig. 1B). Virus particles were purified from these cell culture supernatants via velocity gradient centrifugation using a protocol that yields HIV particles without substantial contamination with extracellular vesicles (39,41). These concentrated virus preparations contained significant amounts of SERINC5 in the case of HIV-1 ⌬Nef, whereas only very low levels of the restriction factor were detectable in preparations of HIV-1 WT (Fig. 1C). SDS-PAGE and silver staining of these virus preparations confirmed that HIV-1 p24 was the main protein component of these preparations, which displayed comparable purity (Fig. 1D).

SERINC5 does not alter the lipid composition of HIV-producing cells or HIV particles
We next analyzed these HIV particle preparations and the corresponding producer cells by quantitative lipid mass spectrometry. To this end, we subjected purified viral particles and whole producer cells to lipid extractions in the presence of internal lipid standards. Lipid profiles of HIV-1 particles ( Fig.  2A, primary data in supplemental Table S1) and producer cells (Fig. 2B) were determined by a bottom-up lipidomics approach based on precursor ion and neutral loss scanning for lipid classspecific fragments (43). In line with previous reports (39, 40, 44 -47), HIV-1 WT particles were enriched in cholesterol, sphingomyelin (SM), hexosylceramide (HexCer), and phosphatidylserine (PS) compared with the producing cells (Fig. 2, A  and B). No significant differences in lipid composition were observed between HIV-1WT and ⌬Nef virions ( Fig. 2A, major and rare lipid classes are shown in the left and right panel, respectively; compare light blue and dark blue columns). Presumably reflecting differences in the overall lipid composition of the cell lines used, these particles, produced from 293T cells, did not display the Nef-mediated enrichment in SM observed previously in virions produced from MT-4 cells (40, 41). Considering the suggested potential of SERINC proteins to feed

Lack of lipid modulation in antiviral activity of SERINC5
serine into sphingolipid (SP) biosynthesis, we expected that expression of SERINC5 during virus production may affect levels of host-cell and/or virion lipids that require serine during biosynthesis, such as SM, HexCer, and PS. However, and irrespective of the presence of Nef, expression of SERINC5 during virus production had no significant impact on the abundance of individual lipid classes in virions ( Fig. 2A) or producer cells (Fig. 2B). Likewise, the overall category profile, including glycerophospholipids (GP), SP, and sterols (ST), was indistinguishable in virus producer cells and virions for all tested conditions (Fig. 2C).
Although overall lipid class profiles remained unchanged by the presence or absence of SERINC5 during virus production, differences might occur for individual lipid species within lipid classes. To test for changes at the level of molecular lipid species compositions, we first compared their distributions on a global level. In agreement with previous reports (39, 40, 44 -47), we observed, for HIV-1 particles, an increase relative to producer cells in GP, predominantly with a total of 30 and 32 fatty acyl carbon atoms ( In turn, the global length distribution of GP and SP species was overall similar in producer cells and virions under all conditions tested. As the only exception, GP-34 was slightly more abundant in cells than in virions (Fig. 3B).
Another profound difference between HIV-1 particles and membrane lipids of producer cells is the enrichment of saturated lipid species, contributing to the higher lateral order of particles (39,40,48). Accordingly, HIV-1 particles (WT as well as ⌬Nef) were enriched in GP with saturated fatty acyl moieties (Fig. 3C, left panel, e.g. GP-0), whereas only slight differences between producer cells and HIV particles were observed for the saturation of SP (Fig. 3C, right panel). For both lipid categories, absence of Nef and presence of SERINC5 did not impact the global GP and SP compositions. Thus, SERINC5 did not influence the partitioning of saturated lipid species or individual lipid classes into the viral membrane envelope.
A striking feature of HIV-1 and some other viruses is the enrichment of dihydrosphingomyelin, such as N-palmitoyl-dihdrosphingomeylin (annotated as SM 34:0;2 in Fig. 3A, right panel) in viral particles compared with producer cell lines (39,40,48,49). We showed previously that dihydrosphingomyelin is not enriched in HIV-1 particles produced from HEK293T cells (47). Consistently, we did not observe an increase in dihy-

Lack of lipid modulation in antiviral activity of SERINC5
drosphingomyelin, neither in HIV-1 WT nor in ⌬Nef particles produced in the absence or presence of SERINC5 (Fig. 3A, right panel, compare light blue and dark blue columns of the 34:0;2 species).
Taken together, expression of SERINC5 did not significantly alter the lipid composition of HIV-producing cells or the resulting virus particles. We conclude that, in the experimental system used, restriction of HIV particle infectivity by SERINC5

Lack of lipid modulation in antiviral activity of SERINC5
and antagonism by Nef occur in the absence of overall differences in lipid composition.

SERINC5 does not alter sphingosine metabolism
Our previous observation that expression of SERINC5 did not affect the global steady-state lipid composition of HIV particles did not exclude that the restriction factor alters the kinetics of lipid biosynthesis in functionally relevant ways. We therefore assessed next whether expression of SERINC5 alters the metabolism of sphingosine, the main building block of serinedependent SP synthesis, which begins with the condensation of serine and CoA-activated palmitate, followed by the subsequent generation of sphinganine, ceramide, and key end products such as SM and glucosylceramide (Fig. 4A). To this end, we made use of a functionalized sphingosine (pacSph) containing a photoactivatable diazirine group and a clickable alkyne terminus to enable copper-catalyzed alkyne-azide cycloaddition. In cells, pacSph undergoes metabolism like its natural counterpart (50,51), leading to the synthesis of complex functionalized sph-ingolipids. In the presence of sphingosine-1-phosphate lyase, an enzyme involved in degradation of sphingosine, the hydrocarbon backbone of pacSph, which contains the photoactivatable and clickable functionalities, is also fluxed into the glycerol-and phosphoglycerolipid biosynthetic pathway (50, 51) (Fig. 4A). Thus, the kinetics of SP, GP, and glycerolipids (GL) can be monitored at the same time. Cells cultured in the presence of pacSph were lysed after different times of incubation. Cell lysates were subjected to lipid extraction and coppercatalyzed alkyne-azide cycloaddition of a fluorescent dye (Fig.  4B). Lipids were separated by TLC to monitor the kinetics of SP, GP, and GL (Fig. 4C). Quantitative analysis of lipid classes showed that sphingosine metabolism in 293T cells were not affected by the presence of SERINC5.

SERINC5 restriction is not linked to altered phosphatidylserine exposure on HIV particles
As an alternative explanation for the effects of SERINC5 on HIV particle infectivity, we considered that the restriction fac-

Lack of lipid modulation in antiviral activity of SERINC5
tor may affect membrane organization rather than overall composition. In this context, the phospholipid PS was of particular interest. PS is generally present at the inner, cytoplasmic side of the plasma membrane of mammalian cells; however, specific triggers, such as the onset of apoptosis, lead to the translocation of PS to the outer leaflet of the plasma membrane and, thus, exposure at the cell surface (52). Physiologically, PS externalization serves as an uptake signal, e.g. for apoptotic cells by phagocytic cells. Such cells display specialized PS receptors, such as, e.g., T cell immunoglobulin and mucin domain 1 (TIM) and Tyro-3, Axl, and Mer (TAM) proteins, at their surface that recognize membranes enriched in PS and whose engagement triggers the internalization pathways (53). Numerous viruses, including vesicular stomatitis virus, vaccinia virus, and hepatitis C virus (HCV), expose considerable amounts of PS on infectious particles, allowing them to hijack PS pathways for uptake and productive infection in a process referred to as apoptotic mimicry (54,55). Although PS receptor-mediated uptake may lead to productive infection for viruses whose glycoproteins, e.g., require activation by low pH, we reasoned that such entry pathways may be abortive for viruses such as HIV-1 that infect target cells in a mostly pH-independent manner. In this scenario, elevation of PS levels exposed at the surface of HIV-1 particles by SERINC5 could enhance the use of such abortive uptake and explain the observed restriction to particle infectivity.
To test this hypothesis, we first sought to quantify surface PS levels on HIV-1 particles and employed flow cytometry, following protocols for the analysis of liposomes of the size of HIV-1 particles (56) or of purified HIV-1 particles (57). As a positive control, liposomes with defined lipid composition and the approximate size of HIV particles (100 -150 nm) were used. Liposomes only containing PC or a mixture of PC and PS (50% PS) were undistinguishable by size and granularity and did not display fluorescence when unstained (data not shown). Staining with a PS-specific antibody (anti-PS) (Fig. 5A) or with Mfge8.EGFP, a protein that binds to PS (58, 59) (Fig. 5B), readily detected PS-positive populations in PC/PS liposomes, and, in line with previous results, significantly reduced background binding was observed with PC liposomes (60). Using this setup, purified HIV particles were detectable by FACS and could be identified, e.g., by the incorporation of Vpr.GFP (data not shown). We therefore stained purified HIV-1 WT and ⌬Nef particles produced in the absence or presence of SERINC5 with Mfge8.eGFP (Fig. 5, C and D). Surprisingly, a robust population of particles produced in the absence of SERINC5 was found to expose PS independently of Nef. The frequency and intensity of this PS positivity, however, were unchanged for virions produced from SERINC5-positive cells (Fig. 5, C and D). Although the percentage of HIV-1 particles detected as PS-positive varied between virions produced and stained on different days (60% for the particle preparations shown in Fig. 5C, ranging from 50 -90%), virion preparations generated and analyzed in parallel displayed comparable PS positivity irrespective of expression of Nef or SERINC5. Comparable frequencies of PS-positive HIV-1 particles produced in the absence or presence of SERINC5 and/or Nef were also detected with the anti-PS antibody (Fig. 5E, 10 -98% of HIV-1 particles detected as PS-positive). We conclude . pacSph fed to cells can either enter the biosynthetic (orange) or degradative (purple) pathway. In SP biosynthesis, ceramide serves as a branching point for the synthesis of phosphosphingolipids such as SM and glycosphingolipids such as glucosylceramide (GlcCer). Breakdown of sphingosine yield shuttles the hydrocarbon backbone via palmitoyl-CoA into biosynthesis of GL and GP. B, pacSph contains a photoactivatable diazirine group and a terminal alkyne moiety. pac-Sph is metabolized like its endogenous counterpart, entering both the biosynthetic and degradation pathways, giving rise to functionalized cellular lipids. Following lysis at different times of labeling, cells are subjected to a click reaction to couple a fluorescent reporter to alkyne-containing lipids. Lipid extracts were analyzed by TLC and fluorescence imaging to monitor Sph metabolism. C, cells were labeled with pacSph for 15 min, 60 min, or 18 h, lysed, and subjected to a click reaction using an azide-activated fluorophore. Following TLC separation, lipids were visualized by fluorescence imaging. Lipid classes were distinguished based on co-migrating standards, except for neutral lipids. Fluorescence signals were quantified, and relative contributions of lipid classes were calculated. Cer, ceramide; DAG, diacylglycerol; GlcCer, glucosylceramide; TAG/CE, triacylglycerol/cholesteryl ester.

Lack of lipid modulation in antiviral activity of SERINC5
that HIV-1 particles carry a substantial amount of externalized PS independent of SERINC5 expression.

Competing with PS binding sites on target cells does not affect SERINC5 restriction
To test whether SERINC5 may introduce qualitative rather than quantitative changes to PS levels exposed on HIV particles, we next tested whether saturation of PS receptors on target cells affects SERINC5 restriction to virion infectivity. To this end, PC-, PE-, or PS-containing liposomes were incubated with target cells to block PS receptors and interfere with PS-dependent virus uptake. As a control, we used infection of Huh7 cells with HCV and scored the expression of the viral NS5A protein as a marker for productive infection by Western blotting. In line with a recent report (61), preincubation of Huh7 cells with PS but not PC or PE liposomes efficiently reduced HCV infection (Fig. 5F). In contrast, parallel treatment of TZM-bl cells had no effect on HIV-1 infection (Fig. 5G). In particular, the strong inhibition of HIV-1⌬Nef infectivity by SERINC5 was maintained even in the presence of 100 M PS receptor-competing PS liposomes. Together, these results illustrate that restriction of HIV-1 infectivity by SERINC5 does not involve targeting of HIV-1 particles to non-productive apoptotic mimicry entry routes.

Discussion
Based on the reported ability of SERINC proteins to feed serine into SP biosynthesis pathways, an attractive and often suggested model predicts that its pronounced negative effect on virus infectivity reflects alterations in virion lipid composition (7,8,34,38,62). The goal of this study was to test this hypothesis and assess whether expression of SERINC5 alters the lipid composition and organization of HIV-1 particles and whether such putative effects contribute to the antiviral activity of SERINC5. Our results do not support these hypotheses because, under experimental conditions where infectivity of HIV-1⌬Nef was potently suppressed, expression of SERINC5 during HIV production did neither affect sphingosine metabolism of producer cells nor the overall lipid composition or PS externalization of HIV particles.
Together, these results seem to refute the idea that the restriction of SERINC5 to HIV infectivity is mediated by alterations in HIV lipid composition. Although our quantitative lipidomics analysis covered a comprehensive range of all major lipid classes, these analyses cannot exclude the involvement of rare lipid species we were not able to detect by our mass spec-trometry approach. However, it seems unlikely that a serine incorporation function of SERINC5 would be required for such an effect because we did not observe any difference in the abundance of any of the key products of serine-dependent lipid biosynthesis that would also be required for the generation of such still to be determined lipids. Our bulk analysis of whole cell membrane lipids also does not exclude that SERINC5 may have subtle effects on the lipid composition of the plasma membrane, from which HIV particles bud. Such changes, however, should become apparent in virions whose lipid composition was unaffected by SERINC5 in our analysis. We therefore favor the idea that SERINC5 suppresses HIV virion infectivity by alternative mechanisms. In line with a previous report on the effect of Nef on the accessibility of Env in virions to antibody neutralization (63), such effects could include alterations in organization and abundance of Env-containing microdomains at the plasma membrane and in virions. In this scenario, SERINC5 may create a local microenvironment that impairs the fusogenic potential of Env by physical association or indirect mechanisms (38). Such effects of SERINC5 may well be independent of its putative serine incorporator activity. Alternatively, serine incorporation may allow SERINC5 to affect metabolic processes other than lipid biosynthesis that impact HIV particle infectivity.
In addition to the mechanism of antiviral activity of SERINC5, the findings presented here expand our knowledge of the general properties of this restriction factor. The namegiving study by Inuzuka et al. (35) characterized SERINC proteins as transmembrane proteins that have the ability to associate with key enzymes of SP biosynthesis and to incorporate serine into membranes, which suggests that they may facilitate the synthesis of PS and SP. Our comparison of SERINC5-negative 293T cells with 293T cells that overexpress functional levels of SERINC5 did not reveal significant differences in abundance of SP and PS. Serine incorporation by SERINC5 therefore does not seem to contribute to the biosynthesis of these lipids in this experimental system. Similar findings were made recently in a mouse strain in which expression of SER-INC1 is abolished by retroviral insertion but did not affect the composition of serine-derived lipids in lymphocytes and macrophages (64). The fact that no significant changes in cellular SP composition were observed upon manipulation of individual SERINC proteins may reflect compensation of the serine incorporator function by other members of the SERINC pro-

Lack of lipid modulation in antiviral activity of SERINC5
tein family. Cell system-specific differences in the relative contribution of individual SERINC proteins to SP synthesis may also exist. Nevertheless, it needs to be considered that the serine incorporator activity demonstrated by Inuzuka et al. (35) does not primarily feed into SP synthesis or may not even represent the key physiological activity of SERINC proteins.
Together, the results presented here demonstrate that the potent antiviral activity of SERINC5 does not rely on alterations of host cell and virion lipid composition and warrant further investigation into the biological activities of SERINC proteins.

Virus production and purification of HIV-1 particles
Virus was generated by transfection (via calcium phosphate) of proviral HIV-1 NL43 plasmids into HEK293T cells. For this, 5 ϫ 10 6 293T cells were seeded per 150-cm 2 flasks 1 day before co-transfection of proviral DNA (70 g) and plasmid DNA for expression of internally HA-tagged SERINC5 (7 g) or a vector control. HIV-1 particles were purified as described previously (39,41). Briefly, 2 days after transfection, culture supernatants were harvested, cleared, and concentrated through a 20% (w/v) cushion of sucrose in PBS. Concentrated virus was further purified via velocity gradient centrifugation on an OptiPrep gradient (Axis-Shield, Oslo, Norway). The visible virus band was collected and pelleted (44.000 rpm, 45 min). The amount of released, purified viral particles was determined using a onestep SYBR Green I-based, product-enhanced reverse transcriptase assay (67). For that, 5 l of virus-containing supernatant was lysed in 2ϫ virus lysis buffer for 10 min and diluted with 90 l of PCR buffer, and 10 l thereof was used for the RT-PCR reaction. Absolute values of the virus suspensions tested were retrieved from a standard curve created using serial dilutions of a previously characterized virus stock. The relative infectivity of purified virus particles was determined by infection of TZM-bl reporter cells cultured in 96-well format. Infectivity was determined 48 h after infection by analysis of firefly luciferase activity (measured in relative light units) as described previously (68). The purity of the virus particle preparation was determined via silver staining of SDS-polyacrylamide gels. For Western blot analysis of virions, virus particle preparations were lysed in 2-6ϫ SDS sample buffer containing 50 mM tris(2carboxyethyl)phosphine (0.5 M stock (pH 7.0), Sigma).
Single-cell suspensions of human hepatoma-derived Huh7.5 cells were prepared by trypsinization and washed once with PBS. Cells were resuspended at a concentration of 1 ϫ 10 7 cells/ml in Cytomix (69) containing 2 mM ATP and 5 mM glutathione. 10 g of in vitro transcribed RNA (70) was mixed with 400 l of the cell suspension and transfected by electroporation with a Gene Pulser system (Bio-Rad) in a cuvette with a gap width of 0.4 cm (Bio-Rad) at 975 microfarad and 270 V. Cells of two transfections were immediately transferred into medium and seeded into a 15-cm-diameter culture dish. Supernatants were collected 48 h after electroporation and filtered through a 0.45-m-pore membrane, and virus particles were concentrated by ultrafiltration using a centrifugal filter device (Centricon Plus-70, Millipore). Concentrated culture supernatant was stored at Ϫ70°C. Infectivity titers of virus stocks were determined by limiting dilution assay using Huh7.5 cells as described elsewhere (71).

Quantitative lipid analysis
Cells and viral particles were subjected to acidic Bligh and Dyer lipid extractions (72). Lipid standards were added prior to extractions using a master mix containing phosphatidylcholine  43). Evaporated lipids were redissolved in 60 l of 10 mM ammonium acetate in methanol and analyzed on a QTRAP6500ϩ mass spectrometer (Sciex) with chip-based (HD-D ESI Chip, Advion Biosciences) electrospray infusion and ionization via a Triversa Nanomate (Advion Biosciences) as described previously (43). Redissolved lipid extracts were diluted 1:10 in 96-well plates (Eppendorf Twintec 96, colorless, Sigma, Z651400-25A) prior to measurement. Precursor and neutral loss scanning in positive ion mode was employed to measure glycerophospholipids as described previously (43). The remaining samples were subjected to cholesterol determination as described previously (73). Data evaluation was done using LipidView (ABSciex) and a software developed in-house (ShinyLipids).

Analysis of sphingosine metabolism
3 ϫ 10 5 293T cells were seeded in 6-well plates and transfected with 2.5 g of GFP or SERINC5.GFP expression plasmids for 24 h. For pacSph labeling, medium was removed, and cells were washed with PBS and labeled with DMEM containing 10% delipidated FCS and 2 M pacSph for 18 h, 60 min, 15 min, or 0 min. Labeling medium was removed, and cells were detached using cell dissociation buffer (Thermo Fisher, 13151014), transferred into 1.5-ml tubes, and pelleted (300 ϫ g, 5 min, 4°C). Cell pellets were washed three times with PBS and resuspended in 300 l of PBS. For lipid extractions, 600 l of MeOH and 150 l of CHCl 3 were added, and samples were mixed by vortexing. Proteins were pelleted (14,000 ϫ g, 5 min), and supernatants were transferred into 2-ml tubes. Following addition of 300 l of CHCl 3 and 600 l of 0.1% acetic acid in H 2 O, samples were vortexed and centrifuged (14,000 ϫ g, 5 min). The lower organic phase was transferred into a 1.5-ml tube, and the solvent was evaporated in a SpeedVac at 30°C. Lipids were redissolved in 7 l of CHCl 3 and 30 l of freshly prepared click mixture containing 3 l of 44.5 mM 3-azido-7hydroxycoumarin (Jena Bioscience, CLK-FA047-1) in DMSO, 250 l of 10 mM Tetrakis(acetonitrile)copper(I) tetrafluoroborate (Sigma-Aldrich, 677892) in MeCN, and 1 ml of EtOH. Click reactions were performed in a SpeedVac at 45°C for 20 min. Samples were dissolved in 20 l of CHCl 3 :MeOH:H 2 O: AcOH (65:25:4:1, mobile phase 1) and spotted on a HPTLC silica gel 60 plate (Merck, 1.05641.0001) using a CAMAG Linomat 5 applicator. HPTLC plates were developed to 60% with mobile phase 1 and to 100% in mobile phase 2 (hexane:ethyl acetate, 1:1) in a CAMAG ADC2 system. For enhanced fluorescence of the coumarin derivative, the plates were sprayed with 4% (v/v) Hünig's base (Sigma-Aldrich, 387649) in hexane. HPTLCs were imaged with the Amersham Biosciences Imager 600 at 460 nm.

Generation of liposomes
Lipid stocks solutions of DOPC, DOPE and DOPS (Avanti Polar Lipids, 10 mg/ml in CHCl 3 ) were prepared. Lipids were mixed and dried under nitrogen and a vacuum for 1 h. Lipids were solubilized in PBS to obtain a final concentration of 1 mM. Liposomes were prepared by 10 freeze-thaw cycles of warming to 42°C and freezing in liquid nitrogen (Ϫ196°C) and sized by extrusion through a 50-nm polycarbonate filter using a mini extruder device (Avanti Polar Lipids). Size distributions of liposomes were measured by dynamic light scattering (Wyatt Technology).

FACS analysis of HIV particles or liposomes
Staining of PS on liposomes (100 l; total lipid concentration, 1 mM) or HIV particles (50 l of 5 ϫ 10 10 -1 ϫ 10 11 picounits RT/l) was performed by staining with a mouse anti-PS antibody (1:100 in PBS) or Mfge8.eGFP (1:100 in PBS) for 1 h at 4°C. Liposomes treated with anti-PS antibody were stained with an APC-coupled anti-mouse secondary antibody (1:200). The liposome antibody or Mfge8.eGFP mixture was filled up to 1 ml with PBS to wash out unbound antibody or Mfge8.eGFP. Liposomes and virus particles, respectively, were pelleted through a cushion of 20% sucrose in PBS (44.000 rpm for 45 min) and resuspended in 300 l of PBS. Stained HIV-1 particles were inactivated with paraformaldehyde (3% final concentration) for at least 90 min at room temperature. HIV-1 particles and liposomes were analyzed with a FACS Calibur instrument (BD Biosciences), and data were processed with FlowJo software.

Liposome competition experiments
For HCV infections in the presence of liposomes, Huh-7.5 cells were incubated with 10 or 30 M liposomes or medium as control for 30 min at room temperature and then infected with HCV Jc1 at a multiplicity of infection of 10 in the presence of liposomes (10 or 30 M) at 37°C for at least 8 h. 48 h after infection, cell lysates were harvested for detection of HCV NS5A by Western blot analysis using an NS5A-specific monoclonal antibody. Transferrin receptor (TfR) was used as a loading control. For HIV-1 infections in the presence of liposomes, TZM-bl cells were incubated with 10 or 30 M liposomes or medium as control for 30 min at room temperature and then infected with HIV-1 (2 ϫ 10 8 picounits RT/l) for 6 h. 48 h after infection, cells were lysed, and the infectivity of HIV-1 particles was determined by analysis of firefly luciferase activity.

Statistical analysis
Statistical analysis of datasets was carried out using Microsoft Excel and GraphPad Prism. Statistical significance of parametrically datasets was analyzed by unpaired two-tailed Student t test: *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001.