SERINC5 protein inhibits HIV-1 fusion pore formation by promoting functional inactivation of envelope glycoproteins

The host proteins, SERINC3 and SERINC5, have been recently shown to incorporate into HIV-1 particles and compromise their ability to fuse with target cells, an effect that is antagonized by the viral Nef protein. Envelope (Env) glycoproteins from different HIV-1 isolates exhibit a broad range of sensitivity to SERINC-mediated restriction, and the mechanism by which SERINCs interfere with HIV-1 fusion remains unclear. Here, we show that incorporation of SERINC5 into virions in the absence of Nef inhibits the formation of small fusion pores between viruses and cells. Strikingly, we found that SERINC5 promotes spontaneous functional inactivation of sensitive but not resistant Env glycoproteins. Although SERINC5-Env interaction was not detected by co-immunoprecipitation, incorporation of this protein enhanced the exposure of the conserved gp41 domains and sensitized the virus to neutralizing antibodies and gp41-derived inhibitory peptides. These results imply that SERINC5 restricts HIV-1 fusion at a step prior to small pore formation by selectively inactivating sensitive Env glycoproteins, likely through altering their conformation. The increased HIV-1 sensitivity to anti-gp41 antibodies and peptides suggests that SER5 also delays refolding of the remaining fusion-competent Env trimers.

It has long been recognized that Nef enhances the HIV-1 infectivity measured by a single cycle infection assay (1)(2)(3)(4)(5)(6)(7)(8), but the underlying mechanism remained poorly understood. Nef has been reported to enhance an early step of HIV-1 entry upstream of reverse transcription (1,2); however, researchers disagreed as to whether Nef incorporation promotes the HIV-1 fusion step (3)(4)(5)(6)(7)9). Two recent papers have elucidated the infectivity-enhancing effect of Nef by showing that SERINC5 (SER5) and, to a lesser extent, SERINC3 expressed in infected cells incorporate into HIV-1 particles in the absence of Nef and potently inhibit infectivity of released virions (10,11). The antiviral activity of SER5 is antagonized by HIV/simian immuno-deficiency virus Nef and by the unrelated glycoGag protein of murine leukemia virus (MLV), 3 which are thought to diminish SER5 incorporation into virions by removing it from the plasma membrane (10,11). The importance of SER5 as a restriction factor is strongly supported by the observation that the potency with which Nef antagonizes this protein correlates with the prevalence of primate lentiviruses in the wild (12).
Little is known about SER5, except that this multipass transmembrane protein appears to be involved in serine incorporation into lipids and promotion of phosphatidylserine (PS) and sphingomyelin biosynthesis (13). Virus-incorporated SER5 has been shown to inhibit HIV-1 fusion (10,11), but the HIV-1 Env glycoproteins differ in their sensitivity to this factor, with the primary JRFL isolate being much less sensitive than the laboratory-adapted strains, such as NL4-3. Whereas MLV infection is also antagonized by SER5, fusion of HIV-1 particles pseudotyped with vesicular stomatitis virus (VSV)-G or Ebola virus glycoproteins is relatively resistant to this factor (10,11,14). In fact, Ebola pseudovirus infectivity has been reported to be enhanced by incorporation of SER5 (15).
Accumulating evidence implies that Nef is incapable of blocking SER5 incorporation into HIV-1 particles upon overexpression of this restriction factor (16,17) but that Env itself is a major determinant of SER5 sensitivity (16). Specifically, the Nef/glycoGag dependence of HIV-1 infectivity (and thus its sensitivity to SER3/SER5) has been mapped to the gp120 V1/V2 loops (18), while a recent study revealed a critical role of the V3 loop in modulating the Env sensitivity to SER5 (16). New evidence also suggests that, in addition to its role in SER5 internalization from the plasma membrane, Nef antagonizes the activity of virus-incorporated SER5 by a cryptic mechanism (17).
Although the mechanism by which SER5 inhibits HIV-1 fusion is not understood, the observation that SER5 more potently blocks HIV-1 infection than fusion led to a model that this protein impairs the enlargement of the fusion pore and thereby the release of viral nucleocapsid (10,11). Here, we investigated the effect of SER5 incorporated into HIV-1 particles on viral fusion in the absence of Nef. Functional virus-cell and cell-cell fusion assays and single virus imaging show that SER5 inhibits the formation of a small fusion pore between HIV-1 and a target cell and that this effect is Env-dependent. We demonstrate that HIV-1 fusion is inhibited through enhanced spontaneous inactivation of SER5-sensitive HIV-1 Env glycoproteins but not resistant Env. In addition, SER5 sensitizes HIV-1 to gp41-derived inhibitory peptides and neutralizing antibodies against cryptic functionally important gp41 domains. These effects are likely caused by SER5-mediated structural changes in HIV-1 gp41 and by slowing down its refolding that culminates in the formation of the final 6-helix bundle structure.

SER5 inhibits HIV-cell fusion and Env-mediated cell-cell fusion
To assess the effect of SER5 on HIV-cell fusion, we compared the fusion activity of pseudoviruses that contained or lacked SER5 in their membrane or contained the inactive SER2 variant as a control. SER5, but not SER2, inhibited HIV-1 fusion, as measured by the BlaM assay (Fig. 1A). In agreement with the previous reports (10,11), SER5 incorporation had a more pronounced effect on fusion of particles pseudotyped with HXB2 Env (HXB2pp) than those pseudotyped with JRFL Env A-C, Nef-negative HIV-1 pseudoviruses bearing full-length or cytoplasmic tail-deleted (⌬CT) HXB2 Env, JRFL Env, or VSV-G, either lacking or containing SER5-HA, were allowed to enter TZM-bl cells, and the resulting fusion was measured by the BlaM assay. N.S., not significant. D, comparison of virus fusion and infection in TZM-bl cells for HXB2pp and JRFLpp produced in cells transfected with increasing amounts of the SER5-HA plasmid and constant amount of the Env plasmid. E, effect of SER5 on FFWO mediated by HXB2pp. Increasing p24 quantities of HXB2pp produced in the presence or absence of SER5-HA were added to a mixed confluent monolayer of N4X4-DSP-1 and N4X4-DSP-2 cells by spinoculation. Fusion was allowed to proceed for 2 h at 37°C. Data are mean and S.E. of two independent experiments in triplicate. F, inhibition of HXB2 Env-mediated cell-cell fusion by SER5. The N4X4-DSP-2 cells were overlaid with 293-DSP-1 cells transiently transfected with equal amounts of full-length HXB2 Env and SER2-GFP, SER5-GFP, or empty vector or 4:1 ratio of Env/SER5 (penultimate bar). The fusion efficiency was measured after 2 h by dual-split luciferase assay, as in E. Viruses lacking SER5 (Vector) were additionally treated with 100 M AMD3100 as negative control. Data are mean and S.D. of two independent experiment in triplicates.
(JRFLpp), albeit JRFLpp fusion was significantly inhibited (Fig.  1, A and B). This inhibitory effect was not a result of gross SER5 overexpression in producer cells, because HXB2pp fusion could be partially or fully rescued by ectopic expression of HIV-1 Nef or MLV glycoGag in these cells (supplemental Fig.  S1). The inhibitory effect of SER5 was not caused by reduced incorporation of HIV-1 Env or by interference with its proteolytic processing (supplemental Fig. S2A).
SER5 also inhibited fusion mediated by other HIV-1 clade B and clade A Env glycoproteins to a varied degree (supplemental Fig. S3, A-C), and the inhibitory effect appeared to be largely independent of the target cells (supplemental Fig. S3, D and E). By contrast, and as reported previously (10,11), VSVpp fusion was relatively resistant to SER5 (Fig. 1C). However, in agreement with Ref. 11, VSVpp incorporated less SERINC proteins as compared with HIV-1 Env pseudotypes (supplemental Fig.  S2B). We also found that both SERINCs were somewhat less abundant in JRFLpp as compared with HXB2pp in two independent preparations (supplemental Fig. S2A and data not shown).
Parallel viral fusion experiments revealed that the cytoplasmic tail (CT) of Env did not consistently modulate the inhibitory effect of SER5. Although SER5 was somewhat less active against fusion of particles pseudotyped with the tail-deleted HXB2 Env (⌬CT) than against full-length Env (Fig. 1A), this effect was reversed for JRFL Env (Fig. 1B). Thus, the modest differences in SER5 sensitivity of the full-length and tail-deleted Env could be due to the efficiency of their incorporation into virions and/or proteolytic processing. SER5 suppressed HIV-1 fusion in a dose-dependent manner; progressive reduction in the fusion signal was observed upon increasing the ratio of SER5-to Env-expressing plasmid used to transfect the producer cells (Fig. 1D). Western blotting confirmed that viruses harvested from cells transfected with increased amounts of SER5 or SER2 plasmids incorporated proportionally more SERINC proteins (supplemental Fig. S2C). Interestingly, for a given SER5/Env ratio, SER5 more potently suppressed single cycle HIV-1 infection compared with fusion ( Fig. 1D). A stronger inhibition of infection versus fusion by SER5 has been reported in the recent studies (10,11) and interpreted as interference with the fusion pore enlargement and, thereby, with the release of HIV-1 capsid.
We next tested whether pseudoviruses carrying SER5 were impaired in their ability to mediate syncytia upon fusing with the plasma membranes of two adjacent cells, a phenomenon referred to as "fusion-from-without" (FFWO) (19). We have previously shown that FFWO is very inefficient and highly dependent on actin dynamics, in contrast to virus-cell fusion (20). Virus-mediated cell-cell fusion was assessed using a dual-split luciferase assay that produces a robust luciferase signal upon fusion between two target cell lines expressing the complementary split GFP-luciferase protein fragments (21). SER5-containing viruses less efficiently induced FFWO compared with control viruses (Fig. 1E). For the same ratio of SER5/Env plasmids used to obtain HXB2 pseudoviruses, virus-cell fusion and FFWO were inhibited to a comparable extent (Fig. 1, D and E).
Considering that HIV-cell fusion appears to occur in endocytic compartments (22)(23)(24), whereas FFWO results from virus fusion with the plasma membrane, the above findings imply that SER5 attenuates the virus' ability to fuse irrespective of the entry site.
To determine whether SER5 antagonizes the HIV-1 Env-mediated membrane fusion in the absence of other viral proteins, we measured fusion between CD4/coreceptor-expressing cells and cells expressing Env and SER5 (or SER2 as control). SER5 but not SER2 inhibited HXB2 Env-mediated cell-cell fusion (Fig. 1F) without significantly altering the expression of Env glycoprotein on the cell surface (supplemental Fig. S3, F and G). Comparable effects of SER5 on HXB2 Env-mediated virus-cell and cell-cell fusion (Fig. 1, A, D and F) argue against the involvement of other viral proteins in the SER5 restriction phenotype. To conclude, SER5 interferes with membrane fusion mediated by sensitive HIV-1 Env, while having a modest or no effect on resistant viral glycoprotein.

SER5 incorporation does not trap HIV-1 at a dead-end hemifusion state or target virus to lysosomal degradation
A preponderance of evidence implies that viral and cellular protein-mediated fusion progresses through a hemifusion intermediate defined as lipid mixing without content transfer, which requires opening of a fusion pore (25). To determine whether SER5 blocks the transition from hemifusion to fusion, we treated cells at 30 min post-infection with chlorpromazine (CPZ) to mediate conversion of hemifusion to full fusion (26,27). This treatment did not affect the extent of fusion of control (or SER2ϩ) HXB2pp or rescue fusion of SER5ϩ pseudoviruses ( Fig. 2A). In contrast, VSVpp fusion was significantly enhanced by CPZ, irrespective of the presence of SER5 (Fig. 2B), suggesting the formation of a considerable fraction of dead-end hemifusion structures upon entry of this pseudovirus. Under these experimental conditions, exposure to CPZ did not adversely affect cell viability (Fig. 2C). We concluded that SER5 does not stall the transition from hemifusion to fusion and acts through an alternative mechanism.
Because HIV-1 appears to infect TZM-bl cells by an endocytic route (23,24), we asked whether diminished viral fusion observed in our experiments resulted from an accelerated degradation of SER5ϩ virions in late endosomes/lysosomes as compared with control viruses. The extent of lysosomal degradation of fusion-competent SER5ϩ viruses was evaluated by inoculating cells with pseudoviruses in the presence of BafA1 to block endosomal acidification and interfere with virus degradation. BafA1 only marginally (albeit significantly) increased the fusion efficiency of SER5ϩ HXB2pp compared with untreated control or to viruses lacking SER5 (Fig. 2D). This result argues against excessive virus degradation as the basis for the markedly reduced fusion efficiency of SER5ϩ particles.

SER5 but not SER2 inhibits the formation of small fusion pores
To determine which step of HIV-1 fusion downstream of hemifusion is antagonized by SER5, we imaged single virus fusion with living cells. HXB2pp were co-labeled with YFP-Vpr and the viral content marker Gag-iCherry containing an "internal" mCherry tag (28). Fusion of HXB2pp containing or lacking SER5-HA or SER2-HA culminates in the release of mCherry through fusion pores larger than 4 nm in diameter (Fig. 3, A and B) (28). As expected, SER2-HA-containing particles and control virions fused equally well, whereas fusion of SER5-HA particles was markedly inhibited (Fig. 3C). The extent of SER5 virus fusion (Ͻ0.2% of double-labeled particles) was indistinguishable from fusion of SER2-HA particles in the presence of the CXCR4 antagonist AMD3100. For the HXB2pp preparations used in the fusion experiments, incorporation of SER2 into virions was even more efficient than SER5 incorporation (Fig. 3D, inset), but only SER5 strongly inhibited infectivity (Fig. 3D).
The supplemental Fig. S2 shows that SER5 incorporation does not interfere with Env incorporation. We also verified the lack of SER5 effect on Env incorporation into fluorescently labeled pseudoviruses used for imaging. Immunostaining for Env (2G12 antibody) and for SER5 (anti-HA) revealed that the distribution of Env signal was not affected by the amount of virus-incorporated SER5 (Fig. 3, E and F). We thus concluded that SER5 specifically inhibits the formation of small fusion pores between HXB2pp and the target cell.

HIV-1 env does not stably interact with SER5 or form large aggregates in the plasma membrane
We next asked whether SER5 antagonizes the Env function by directly binding to the fusion protein. Lysates of mock-trans-fected 293T cells or cells transfected co-expressing SER5-HA or SER2-HA and HXB2 Env were immunoprecipitated using anti-HA tag antibody or HIV immune serum. This pulldown assay did not show any evidence for Env-SER5 interactions (supplemental Fig. S4). It is possible, however, that weak and/or transient interactions between SER5 and Env escape the detection by this assay.
A possible mechanism by which SER5 can antagonize HIV-1 fusion without directly interacting with Env is through modification of the viral membrane. For instance, the formation of large SER5 oligomers can inhibit fusion by restricting lipid diffusion and/or stiffening the viral membrane. Because protein oligomerization should be manifested in slower mobility in membranes, the lateral diffusion of SER5-GFP and GFPtagged CCR5 (control multi-transmembrane protein) was measured by fluorescence recovery after photobleaching (FRAP). We found that neither the calculated diffusion coefficient nor the immobile fraction of SER5-GFP was significantly different from those of CCR5-GFP (Fig. 4). Because SER5 diffuses freely in the plasma membrane, the reduction in membrane fluidity or the formation of large SER5-enriched domains are unlikely to underlie its ability to inhibit viral fusion. However, we cannot rule out the possibility that SER5 may be enriched in the viral membrane and thus be more prone to aggregate.

SER5 enhances HIV-1 sensitivity to neutralizing antibodies and inhibitory peptides that recognize conserved gp41 domains transiently exposed during viral fusion
To elucidate the mechanism by which SER5 inhibits HIV-1 fusion, we probed the effect of this protein on exposure of cryptic Env domains during viral fusion. The extent of exposure of the functionally relevant Env domains in control and SER5ϩ viruses was examined by comparing the sensitivity to neutralizing antibodies against the gp41 membrane-proximal extracellular region (MPER) and the HR1 domain that are transiently exposed during HIV-1 fusion (29 -32). Incorporation of SER5 sensitized the HIV-1 fusion to 4E10, a broadly neutralizing antibody against the gp41 MPER (Fig. 5A). The Fab fragment of 8k8 antibody against the gp41 HR1 domain (33) also more potently inhibited fusion of SER5ϩ than control viruses (Fig. 5B), although the effect was not as strong as for 4E10 antibody. SER2, which does not significantly affect HXB2 fusion (Fig. 1), did not impact the virus' sensitivity to 4E10 and 8k8 (Fig. 5, A-F). SER5 also enhanced the neutralizing activity of 4E10 against JRFLpp and BaLpp (Fig. 5, E and F), in excellent agreement with the previously reported potent neutralization of Nefdeficient HIV-1 and of SER5-containing virions by this antibody (16,34).
By contrast, SER5 did not affect the inhibitory activity of the broadly reactive 2G12 antibody that recognizes a conformation-independent glycan cluster on the gp120 (Fig. 5, C and D) (35,36). Furthermore, it did not noticeably alter the potency of several other anti-gp120 antibodies, including the broadly neutralizing PG16 antibody against a quaternary glycan-containing . SER5 restricts fusion pore formation between single HXB2pp and target cell. HXB2pp co-labeled with YFP-Vpr and Gag-iCherry were produced in cells transfected with SER5-HA, SER2-HA, or an empty vector. Virions were pre-bound to CD4/CXCR4-expressing CV1 cells in the cold, and their fusion was initiated by shifting to 37°C. A and B, images and single particle tracking results for fusion of SER2-HA containing HXB2pp showing the release of the mCherry marker. C, analysis of the effect of SER5 and SER2 on single HXB2pp fusion in the absence or in the presence of 10 M AMD3100. Data are means and S.D. from 4 to 5 independent experiments. The numbers of double-labeled particles analyzed for each condition are shown above the graph. D, infectivity of the viral preparations analyzed in C in TZM-bl cells. Inset, percent fluorescent virions after immunostaining for HA tag to control for SER2/SER5 incorporation. E, SER5-HA fluorescence intensity distribution from imaging single pseudoviruses immunostained with anti-HA and 2G12 antibodies. Arbitrary classification of particles based on their low and high SER5 content is shown (black and red lines). F, Env content (2G12 staining intensity) distribution for low versus high SER5 particles (black and red lines, respectively).

SERINC5 blocks HIV-1 fusion pore formation
epitope on the native Env trimer, as well as the antibodies m36 and 17b to CD4-induced gp120 epitopes (supplemental Fig.  S5). These results indicate that SER5 does not modulate antigenic properties of gp120 or enhance the exposure of cryptic gp120 epitopes during fusion.
Enhanced HIV-1 neutralization by antibodies recognizing cryptic gp41 epitopes can result from exposure of these epitopes on unliganded Env or from slower gp41 refolding that prolongs the lifetime of the gp41 pre-hairpin intermediate. To determine whether SER5 alters the gp41 structure prior to Env-CD4 engagement, 4E10 binding to single coverslip-adhered pseudoviruses was measured by indirect immunofluorescence. The background-corrected fluorescence intensities were measured for each mCherry-Vpr-labeled particle lacking or containing SER5-GFP or SER2-GFP. Specific infectivity for the analyzed preparations was determined to ensure proper restriction of the Env function by SER5 (Fig. 5G). To account for variations in Env incorporation between different preparations and for possible gp120 shedding, median 4E10 intensity values were normalized to values obtained in parallel experiments using a control 2G12 antibody. Incorporation of SER5 into HXB2pp did not significantly increase the 4E10/2G12 binding ratio ( Fig. 5H and supplemental Fig. S6). By contrast, the relative 4E10 binding was significantly enhanced for SER5ϩ JRFLpp compared with vector or SER2 viruses ( Fig. 5H and  supplemental Fig. S6). In control experiments, treatment of either HXB2pp or JRFLpp with soluble CD4 (sCD4) enhanced the relative 2G12 binding, in agreement with a previous report (37). Our results suggest that the MPER is more accessible on SER5ϩ JRFLpp, in line with a more efficient 4E10-mediated neutralization of this virus as compared with control JRFLpp (Fig. 5E). The lack of SER5 effect on the 4E10 binding to HXB2pp could reflect the natural tendency of this Env to expose MPER or result from possible functional inactivation of Env that sequesters this region (see under "Discussion") (30).
To assess whether the increased potency of anti-gp41 antibodies could be due to a prolonged exposure of the HR1 domain during fusion, we assessed the virus' sensitivity to the gp41 HR2-derived peptides C34 and T-20 (referred to as C-peptides). These peptides bind to the HR1 domain transiently exposed on pre-hairpin intermediates (similar to the 8k8 antibody (33)) and block the 6-helix bundle formation (38). Fusion of SER5-containing HXB2pp and JRFLpp was inhibited by 2-4fold lower doses of C-peptides than fusion of control or SER2ϩ viruses (Fig. 6, A-D). The higher sensitivity to C-peptides is consistent with the prolonged exposure of gp41 pre-hairpins in the course of viral fusion (39 -44). Collectively, the above results suggest that SER5 may cause structural changes in the native HIV-1 gp41 trimer and potentially slow down the gp41 refolding into the final 6-helix bundle structure.

SER5 specifically promotes inactivation of sensitive HIV-1 Env
We next examined the effect of SER5 on the functional stability of Env. HXB2pp preincubation at 37°C for 4 h promoted spontaneous loss of the fusion activity and infectivity, and this loss was markedly accelerated in the presence of SER5 (Fig. 7, A  and B). In contrast, a 37°C preincubation only modestly diminished the ability of JRFLpp and VSVpp to fuse or infect target cells, irrespective of the presence of SER5. SER5 incorporation also accelerated inactivation of other HIV-1 strains, albeit to a lesser extent than HXB2 Env (Fig. 7, C and D). As expected, SER2 did not significantly modulate the loss of HXB2pp fusion activity. Unlike the fusion activity, the infectivity of HXB2pp and JRFLpp containing SER2 was modestly reduced compared with control viruses following a preincubation at 37°C (Fig. 7, A  and B).
The accelerated functional inactivation of HXB2 Env, but not JRFL Env or VSV-G, in SER5-containing virions phenocopies the differential sensitivity of these fusion proteins to SER5 (Figs. 1, A-C, and 7) (10, 11). To conclude, the above results establish an important link between the inherent stability of Env trimers and sensitivity to SER5, because the resistant fusion proteins appear to be less prone to spontaneous inactivation.

Discussion
Our results demonstrate that SER5 inhibits HIV-cell fusion and Env-mediated cell-cell fusion at the stage of small pore formation and link this effect to the following: 1) promotion of spontaneous inactivation of sensitive (HXB2) but not of resistant (JRFL or VSV) viral fusion proteins; 2) alteration of the gp41 structure; and 3) delayed gp41 refolding into the final trimerof-hairpins structure. Importantly, these SER5 effects are specific, as the inactive SER2 variant does not accelerate the loss of the HIV-1 fusion competence, alter the virus sensitivity to fusion inhibitors, or increase the accessibility of the gp41 MPER to neutralizing antibodies.
Selective inactivation of sensitive HIV-1 Env by SER5 phenocopies the known restriction specificity (10,11,18), suggesting a connection between the SER5 restriction efficiency and inherent stability of native Env trimers (45)(46)(47). The differences in the predicted stoichiometry of oligomeric Env complexes required for productive fusion/infection (48) can exacerbate the SER5 effect on the sensitive HIV-1 strains. A greater number of SER5-sensitive NL4-3 and SF162 Env per virion is required to mediate productive entry compared with the less sensitive JRFL Env (4 -7 versus 2 Env, respectively (48)). Thus, partial inactivation of JRFL Env should be less consequential for fusion than inactivation of Envs on NL4-3 or HXB2 virions that must stochastically assemble larger Env complexes to promote fusion.
Although we found that SER5 inhibits small pore formation, excessive functional inactivation of Env is also in line with interference with pore enlargement required for the viral capsid release, as proposed based upon a more potent inhibition of Figure 5. SER5 potentiates the neutralizing activity of antibodies and peptides against cryptic gp41 epitopes exposed during fusion. A-F, HIV-1 particles pseudotyped with HXB2, JRFL, or BaL26 Env, lacking or containing SER5-HA or SER2-HA, were used to inoculate TZM-bl cells, and the resulting fusion was measured by the BlaM assay. Indicated concentrations of neutralizing antibodies were added after virus pre-binding in the cold, immediately before raising the temperature. Data are means of two independent experiments performed in triplicate. The IC 50 values obtained by curve-fitting and statistical significance of the effects of SER5 on HIV-1 sensitivity to neutralization (sum of squares reduction test) are shown on the plots. G and H, specific infectivity (G) and 4E10/2G12 median single-particle staining intensity ratio (H) for HXB2pp and JRFLpp. Data are means and S.D. from two independent experiments. The sCD4-induced increase in 4E10 binding was statistically significant for JRFLpp but not HXB2 particles due to a large variance between the immunofluorescence experiments. See also supplemental Fig. S6.  . SER5 selectively promotes functional inactivation of HXB2 Env. A, Nef-negative HIV-1 particles pseudotyped with HXB2 Env, JRFL Env, or VSV-G, containing or lacking SER5-HA or SER2-HA, were preincubated in growth medium containing 10 mM HEPES (pH 7.3) for 4 h at 37°C or used immediately after thawing the samples (fresh). Viruses were bound to TZM-bl cells in the cold by spinoculation, and cells were washed and incubated at 37°C for 90 min. The fusion efficiency measured by the BlaM assay was normalized to the respective freshly initiated virus. B, pseudovirus infectivity was measured using conditions described in A. C and D, SER5 promotes spontaneous inactivation of pseudoviruses bearing BaL and R3A Env, as measured by a virus-cell fusion assay after a 4-h preincubation at 37°C, as above. Data are mean and S.E. of two independent experiments in triplicates.
HIV-1 infection than fusion by this protein (10,11). Fewer functional Env glycoproteins on SER5-containing viruses are less likely to drive pore dilation, which is a highly energetically unfavorable step of membrane fusion (49,50). The development of direct viral pore enlargement/core release assays is needed to determine whether SER5 incorporation inhibits HIV-1 core release.
We found that SER5 but not SER2 incorporation facilitates the binding of anti-MPER 4E10 antibody to the native JRFL Env, in agreement with the recent studies (16,34) and consistent with the enhanced exposure of the gp41 MPER. Strangely, Beitari et al. (16) did not detect SER5 effects on HIV-1 neutralization by other anti-MPER antibodies, such as 2F5 and 10E8. In this context, it is worth noting that the 4E10 binding to virions is lipid-dependent, with considerable affinity to negatively charged lipids, including PS, whereas the binding of 2F5 and 10E8 is less affected by this lipid (51). Thus, the enhanced 4E10 binding to SER5ϩ JRFLpp may be caused by the increased PS content and not just by the MPER exposure. Whereas this model could reconcile discrepant findings regarding the effect of SER5 on binding of anti-MPER antibodies (16), the lack of effect on binding to HXB2 pseudoviruses (Fig. 5H) and the apparently similar levels of PS in control and SER5ϩ particles (15) are inconsistent with this possibility. We surmised that the similar extents of 4E10 binding to control and SER5ϩ HXB2pp may be caused by the excessive HXB2 gp41 MPER exposure in the absence of the restriction factor and/or from the functional inactivation of this Env on SER5ϩ virions, which may sequester the MPER domains (30).
Finally, SER5 appears to slow down gp41 refolding triggered upon gp120-CD4/coreceptor engagement, as evidenced by the increased sensitivity to C34 and T-20 and to neutralizing antibodies against the gp41 HR1 domain. Although a recent study did not observe an SER5 effect on the anti-HIV-1 potency of the T-20 peptide (16), the discrepancy could be due to the use of a V3 loop Env chimera in the latter study and/or to the differences in the virus-cell fusion protocols. Collectively, our results are consistent with both SER5-mediated structural changes, at least within the gp41 MPER domain, and with slowing down the gp41 refolding.
Although direct SER5-Env interaction could not be detected by a co-immunoprecipitation assay, we surmise that SER5 may partition into the same lipid domains in the viral membrane as Env and act as a physical barrier for functional Env clustering. The unusually high cysteine content of SER5 supports its affinity for lipid rafts into which Env glycoproteins tend to partition. The fact that SER2 lacks 11 of the Cys residues present in SER5 supports the possibility of selective SER5 interference with functional Env oligomerization and/or modulation of lipid composition of the viral membrane. Further studies are needed to determine how SER5 destabilizes the native Env glycoprotein and inhibits the viral fusion pore formation and possibly pore dilation. Toward this goal, it is essential to delineate the SER5 effect on the viral lipid composition and the propensity of Env trimers to form functional clusters in the viral membrane.
EnduRen TM and Bright-Glo luciferase were from Promega (Madison, WI). Heat-inactivated fetal bovine serum (FBS), poly-L-lysine, calf skin collagen, and AMD3100 were from Sigma. Live Cell Imaging Buffer and FluoroBrite TM DMEM were from Life Technologies, Inc. The BMS-626529 was purchased was from AURUM Pharmatech LLC (Franklin Park, NJ). The gp41-derived C34 peptide was a gift from Dr. L. Wang

SERINC5 blocks HIV-1 fusion pore formation
To obtain PBJ5-SER2-HA plasmid, the PBJ6-SER2-HA was digested with NotI and EcoRI restriction enzymes, and the fragment SER2-HA was purified on 1% agarose gel and ligated with PBJ5 digested and purified in similar manner. Construction of PBJ5-SER2-GFP and PBJ5-SER5-GFP was done as follows. The SER2-GFP gene fragment was amplified using TaqDNA polymerase high fidelity (Invitrogen), CMV-SER2-GFP plasmid as a template, and the forward and reverse primers 5Ј-GCG-CTCGAGCGGCCGCCATGGACGGGAGGATGATGAG-3Ј and 5Ј-GCTGCGGCCGCTTACTTGTACAGCTCGTCCAT-GCCGA-3Ј, respectively. The amplified fragment was inserted into pCR2-Topo vector using a TOPO cloning kit (Invitrogen). The SER2-GFP gene was digested with XhoI and NotI restriction enzyme sites (corresponding to the italicized regions in primers), purified, and ligated into PBJ5 digested and purified in an identical manner. The SER5-GFP was amplified using CMV-SER5-GFP plasmid as a template, and the forward 5Ј-GCGCTCGAGCGGCCGCCATGTCAGCTCAGTGCTG-TGC-3Ј and the same reverse primer as for SER2-GFP. The amplified fragments were purified and cloned into pCR2-Topo vector followed by restriction digestion with XhoI and NotI restriction enzymes and cloned into PBJ5 digested with the same restriction enzymes.

Pseudovirus production and characterization
HIV-1 pseudoviruses used in the BlaM assay and single virus experiments were produced by transfection of HEK293T/17 cells, as described previously (23,28). Cells were transfected with the following amounts of plasmids per 100-mm dish: 3 g of Env, 4 g of pR9⌬Env or pR9⌬Env⌬Nef, 1.5 g of pBJ5 or pBJ5-SER2-HA or pBJ5-SER5-HA (unless indicated otherwise), 2 g of BlaM-Vpr, and 0.5 g pcRev. To examine the effects of Nef or glycoGag, cells were transfected as above, except 2 g of pBJ5-Nef LAI or pBJ5-glycoGag was added to the mix. To produce virus for imaging assays, HEK293T/17 cells on a 60-mm dish were transfected with 0.6 g of Env, 0.2 g of pcRev, 0.3 g of pR9⌬Env⌬Nef, 0.3 g of psPAX2-Gag-imCherry, 0.3 g of YFP-Vpr, and 0.3 g of either pBJ5-SER2-HA, pBJ5-SER5-HA, or empty pBJ5 vector. Transfection was carried out using JetPrime transfection reagent. Forty eight hours post-transfection, the supernatant was collected and stored at Ϫ80°C. For concentrated viral stock, the viral supernatants were either pelleted through a 20% sucrose cushion by centrifugation at 100,000 ϫ g for 2 h at 4°C or using Lenti-X concentrator (Clontech).

Virus-cell fusion (BlaM), FFWO, and infectivity assays
Virus-cell fusion was measured using the BlaM assay, as described previously (22,23). Briefly, the target cells were grown in 96-well black clear-bottom plates (Corning) to 95-100% confluency. Viruses (equal amounts of p24) were bound to target cells by centrifugation at 4°C for 30 min at 1550 ϫ g followed by incubation in DMEM without phenol red growth medium at 37°C, 5% CO 2 for 90 min. The fusion reaction was stopped by placing the plates on ice, and the medium was replaced with the BlaM substrate, CCF4-AM (Invitrogen). Cells were incubated at 12°C overnight, and the BlaM activity was determined from the ratio of coumarin (blue) and fluores-cein (green) fluorescence signals, using a SpectraMaxi3 fluorescence plate reader (Molecular Devices, Sunnyvale, CA). Cell viability was measured using a colorimetric CellTiter Blue assay (Promega) according to the manufacturer's recommendations.
The FFWO assay was carried out using the double-split GFPluciferase reporter system, as described previously (20). Twenty four hours prior to the experiment, N4X4-DSP-1 and N4X4-DSP-2 cells were seeded at a 1:1 ratio in collagen-coated 96-well black clear-bottom plates (Corning). The confluent co-cultures were preincubated with 40 M of the cell-permeable Renilla luciferase substrate EnduRen TM in HBSS, 10% FBS for 2 h, centrifuged with viruses for 30 min at 4°C (1550 ϫ g) to allow virus binding, and washed once with ice-cold HBSS, 10% FBS. FFWO was initiated by incubation at 37°C, and the resulting luciferase signal was measured using TopCount NXT reader (Perkin-Elmer Life Sciences). For virus infectivity, the target cells were inoculated with viruses, as described above, and cultured at 37°C, 5% CO 2 for 24 -36 h. The luciferase signal was detected using the Bright-Glo luciferase substrate.

Virus inactivation assay
To assess virus inactivation over time, pseudoviruses (normalized for p24) were suspended in phenol-free growth medium supplemented with 10 mM HEPES and preincubated at 37°C for 4 h. Virus-cell fusion was then measured by the BlaM assay.

Cell-cell fusion and flow cytometry
For the dual-split protein-based cell-cell fusion assay, 293T-DSP-1 (DSP-1) cells in 100-mm dishes were transfected with 5 g of pHXB2-Rev (a gift from Dr. Anna Cereseto, University of Trento, Italy) and 5 g of either pBJ5 (empty vector), pBJ5-SER2-GFP, or pBJ5-SER5-GFP, using JetPrime transfection reagent that was replaced with DMEM after 8 h. Concurrent with 293T-DSP-1 transfection, NP2/CD4/CXCR4/DSP-2 (DSP-2) cells were seeded onto a black wall clear-bottom 96-well plate coated with collagen. Thirty four hours post-transfection, confluent DSP-2 cells were incubated with 60 M EnduRen TM for 2 h. Next, DSP-1 cells were non-enzymatically dissociated from plates, using Cellstripper TM (Mediatech, Manassas, VA), and 2 ϫ 10 6 DSP-1 cells were overlaid onto treated DSP-2 in each well. To initiate cell-cell fusion, plates were transferred to 37°C and incubated for 2 h, after which time the luciferase activity was measured using a TopCount NXT plate reader. In parallel, 4 ϫ 10 6 cells from each 293T-DSP-1 transfection were cooled on ice and resuspended in ice-cold PBS, 15% FBS for 2 h. Half of each sample was resuspended in 5 g/ml 2G12 and the other half in PBS (control) and incubated overnight at 4°C. The following day, all samples were washed three times in ice-cold PBS, resuspended in 2 g/ml goat anti-human Alexa-647 (Life Technologies, Inc., and Thermo Fisher Scientific, Waltham, MA), incubated on ice for 2 h, and washed three times in icecold PBS. Samples were treated for 5 min with propidium iodide solution (Sigma), and single cell propidium iodide, GFP, and Alexa-647 fluorescence were detected using a BD LSR II flow cytometer (BD Biosciences).

Single virus immunofluorescence staining and analyses
Immobilized particle immunofluorescence experiments were performed on pseudoparticles diluted in ice-cold PBS ϩϩ , filtered through a 0.2-m filter, and allowed to attach onto poly-L-lysine-coated 8-well chamber slides (Lab-Tek, Nalge Nunc International, Penfield, NY) for 30 min at 4°C. Wells were washed with cold PBS ϩϩ to remove unbound virus. Samples were blocked for 2 h at room temperature in PBS ϩϩ , 15% FBS before addition of 5 g/ml 2G12 or 2 g/ml 4E10 with or without 10 g/ml sCD4 and incubated for 1 h at 37°C. Samples were washed with PBS ϩϩ , fixed in 2% fresh paraformaldehyde for 20 min at room temperature, and blocked for 30 min at room temperature in PBS ϩϩ , 15% FBS before incubation with 2 g/ml goat anti-human Alexa-647 for 1 h at room temperature. Samples were washed three times with PBS ϩϩ and imaged immediately or stored at 4°C for imaging the following day. Stained virus samples were imaged with a Zeiss LSM780 using a Plan-Apo ϫ63/1.4NA oil-immersion objective in a single Z-plane at 6.3-s pixel dwell time (for each sample 10 -25 fields were imaged containing Ͼ1000 particles). GFP, mCherry, and Alexa-647 were excited at 488, 561, and 633 nm, respectively. After acquisition, virus particles were identified as mCherryϩ spots using the spot detector algorithm in Volocity. Intensities in GFP, mCherry, and Alexa-647 channels corresponding to two-pixel dilation around mCherryϩ spots were background subtracted. Particles were filtered by mCherry intensity to exclude the bottom and top 10% of particles. Using thresholds derived from negative controls, the remaining particles were identified as GFPϩ and/or Alexa-647ϩ. The median value was used to parameterize the single particle staining intensity distributions.

Live cell single virus imaging
Single viral fusion experiments were performed with CV1. CD4.CXCR4 cells plated on collagen-coated glass-bottom dishes (MatTek, Ashland, MA) in FluoroBrite DMEM and grown to 70% confluency. Before imaging, the cells were chilled on ice, washed with ice-cold PBS ϩϩ , and spinoculated with freshly thawed pseudovirus at 4°C for 20 min at 1500 ϫ g. After spinoculation, cells were washed with ice-cold PBS ϩϩ to remove unbound virus. Virus entry was initiated by addition of pre-warmed live cell imaging buffer. Images were acquired in a single axial plane with a Personal DeltaVision imaging system (Applied Precision, GE Healthcare) using a UPlanFluo ϫ40/1.3 NA oil objective (Olympus, Tokyo, Japan) and a GFP/Cherry standard filter set (Chroma, Bellows Falls, VT). Two-channel fluorescence emission was recorded in series by an EM-CCD camera (Photometrics, Tucson, AZ) for a single field-of-view imaged every 3 s. During time-lapse imaging, an environmental chamber was used to maintain samples at 37°C and high humidity, and the UltimateFocus module (Applied Precision, GE Healthcare) was used to compensate for axial drift.

Event annotation, curve fitting, and single particle tracking
Single particle color change (content release) events were annotated using the region of interest manager in ImageJ (National Institutes of Health). Annotated particles were tracked in Volocity (PerkinElmer Life Sciences). The local background subtraction at each time point was performed, using a custom script that created a region of interest encompassing a particle by one-or two-pixel dilation of the tracked object, as described previously (63). A script in Microsoft Excel was designed to collate the tracking data, combining the tracked particle and the dilated object to correct for the local background at each time point.

FRAP
CV-1 cells were seeded onto collagen-coated glass-bottom dishes in FB, grown to 40 -50% confluency, and transfected with 100 l of mixture consisting of 1 g of either CCR5-GFP or SER5-GFP plasmid and 2 l of JetPrime transfection reagent, added directly to the microwell. Eight hours post-transfection, the transfection mix was removed, and 2 ml of pre-warmed FB was added to the dishes. Twenty four hours post-transfection, the samples were cooled to room temperature and washed with PBS ϩϩ . Bleaching and single plane time series imaging at room temperature were acquired with a Zeiss LSM780 using a C-Apo ϫ40/1.2NA water-immersion objective. Circular regions 2.5 or 4 m in diameter near the periphery of flat GFPϩ cells were photobleached. Fluorescence recovery in the photobleached region was monitored, along with intensity in nearby identically shaped regions, both on and off the cells, to control for inadvertent photobleaching and background signal, respectively. FRAP traces corrected for photobleaching and background were fit to a single-exponential rise-to-maximum curve to extract the half-time (t1 ⁄ 2 ) and immobile fraction. Diffusion coefficient was estimated as D ϭ 0.25⅐r 2 /t1 ⁄ 2 (64), where r is the radius of the bleached region.

Statistical analyses
Unless stated otherwise, statistical analysis was performed using the Student's t test or non-parametric Man-Whitney test, as appropriate. Triplicate data for each independent experiment (usually 2-3 experiments) performed under identical conditions were normalized to the internal control and pooled to calculate the mean and standard error. Statistical comparison of dose-response curves for the fusion inhibitors for the control and SER5 viruses was done using the sum of squares reduction test. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001, N.S., not significant.
Author contributions-C. S. and M. M. designed, performed, and analyzed the experiments and co-wrote the paper. A. C. and M. P. constructed and tested pertinent expression vectors and co-wrote the manuscript. G. B. M. conceived and coordinated the study, analyzed the results, and wrote the paper. All authors reviewed the results and approved the final version of the manuscript.