HIV-1 Env associates with HLA-C free-chains at the cell membrane modulating viral infectivity

HLA-C has been demonstrated to associate with HIV-1 envelope glycoprotein (Env). Virions lacking HLA-C have reduced infectivity and increased susceptibility to neutralizing antibodies. Like all others MHC-I molecules, HLA-C requires β2-microglobulin (β2m) for appropriate folding and expression on the cell membrane but this association is weaker, thus generating HLA-C free-chains on the cell surface. In this study, we deepen the understanding of HLA-C and Env association by showing that HIV-1 specifically increases the amount of HLA-C free chains, not bound to β2m, on the membrane of infected cells. The association between Env and HLA-C takes place at the cell membrane requiring β2m to occur. We report that the enhanced infectivity conferred to HIV-1 by HLA-C specifically involves HLA-C free chain molecules that have been correctly assembled with β2m. HIV-1 Env-pseudotyped viruses produced in the absence of β2m are less infectious than those produced in the presence of β2m. We hypothesize that the conformation and surface expression of HLA-C molecules could be a discriminant for the association with Env. Binding stability to β2m may confer to HLA-C the ability to preferentially act either as a conventional immune-competent molecule or as an accessory molecule involved in HIV-1 infectivity.


Results
Surface expression of L31 reactive HLA-C molecules is increased in HIV-1 chronically infected cell lines. ACH-2 is a T-lymphocytic cell line chronically infected by HIV-1, derived from the A3.01 uninfected parental cell line 15 . HIV-1 expression can be activated in ACH-2 cells by TNF-α stimulation, through the induction of transcription factors binding the NF-kB site in the HIV-1 LTR region 16 . Both A3.01 and ACH-2 cells were treated with TNF-α and assessed for HIV-1 expression by flow cytometry using the anti-gp120 2G12 monoclonal antibody. A clear fluorescence increase was evident in ACH-2 cells, revealing a marked HIV-1 reactivation (Fig. 1a). Cells were stained in parallel with the DT9 antibody that recognizes HLA-C heterotrimers (HLA-C heavy chain, β 2 m and peptide) 17 and with the L31 antibody, recognizing β 2 m-free HLA-C molecules 18 . Following TNF-α stimulation, cytofluorimetric analysis showed that β 2 m-associated HLA-C molecules (DT9 reactive) were upregulated in both A3.01 and ACH-2 cells. Conversely, upregulation of HLA-C free chains (L31 reactive) was observed only in HIV-1-infected ACH-2 cells. As shown in Fig. 1b, L31 fluorescence in ACH-2 cells was significantly different between TNF-α stimulated and unstimulated cells (p = 0.002), as well as towards the stimulated parental A3.01 cells (p = 0.01). Upregulation of DT9-reactive HLA-C molecules after TNF-α stimulation in both cell types is expected, since TNF-α as well as interferons are known modulators of expression of MHC-I molecules in many cell types 19 . Conversely the selective cell membrane increase of HLA-C free chains by HIV-1 reactivation is an intriguing original finding. A time course analysis of HIV-1 Env, MHC-I and β 2 m expression was conduced in A3.01 and ACH-2 cells, comparing the RFI between TNF-α stimulated and unstimulated cells (Fig. 2). The results showed that the expression of HLA-C free chains (L31 antibody) at the cell surface correlated with the expression of HIV-1 Env (2G12 antibody), both decreasing at later time points. No changes were instead detected in the expression of total HLA-C molecules (L31 antibody after acid wash), HLA-C heterotrimers (DT9 antibody), MHC-I (W6/32 antibody) or β 2 m (NAMB-1 antibody) molecules. The correlation between HIV-1 infection and HLA-C free chains expression was also observed in an infection model using a different human T-lymphocytic cell line, PM1, chronically infected with HIV-1 III-B (Fig. 3a). PM1-IIIB cells have a constitutive baseline production of virus higher than ACH-2, as visualized by 2G12 staining of unstimulated cells and their downmodulation of MHC-I molecules (W6/32 antibody). When PM1-IIIB cells were stimulated with TNF-α , an increase of HLA-C free chains at the cell surface (L31 antibody) was observed (as in ACH-2 cells), which was barely detected in uninfected PM1 cells, further confirming a role of HIV-1 in promoting the formation of HLA-C free chains at the cell membrane. The mean RFI of L31 staining of three independent experiments was 3.2 and 6.5 in TNF-α -PM1 and TNF-α -PM1-IIIB cells, respectively.
Of interest, TNF-α stimulation counteracted the downregulation of MHC-I molecules induced by HIV-1 infection in PM1-IIIB cells. Given the correlation between increased HLA-C free chains and HIV-1 virion production in both ACH-2 and PM1-IIIB cellular models, we sought to investigate which HIV-1 component was involved in this phenomenon, starting from Env. We thus transfected 293T cells with an HIV-1 Env plasmid and a plasmid encoding an Env defective full length HIV-1 genome (pSG3 Δenv ) (Fig. 3b). An increase of L31 staining Scientific RepoRts | 7:40037 | DOI: 10.1038/srep40037 was detected after transfection of HIV-Env and not after transfection of the pSG3 Δenv plasmid, whereas no differences were detected in L31 staining after acid wash. These data suggest that the observed increase of HLA-C free chains at the cell membrane is not due to an increased HLA-C expression level or membrane translocation, but rather it may reveal a switch in HLA-C conformation from heterotrimer to free chain on the cell membrane in the presence of the HIV-1 Env protein.
None of the HIV-1 proteins is responsible for the appearance of HLA-C free chains at the cell surface. To investigate if any of the HIV-1 proteins is instead responsible for the translocation of HLA-C free chains at the cell membrane, plasmids encoding Gag, Env, Vpu, Vif, Nef, Tat, as well as pSG3 Δenv (containing the Env defective full length HIV-1 genome) were transfected in CRISPR/Cas9 β 2 m knockout 293T cells. Cells were treated with an acid wash, before labelling with L31 antibody, in order to exclude that the potential intracellular association between HLA-C and any of the tested HIV-1 protein could mask the L31 epitope 18 . Control transfection with β 2 m expressing vector increased the surface expression of HLA-C free chains, while no HLA-C upregulation was detected on the membrane of cells transfected with all the HIV-1 proteins encoding constructs (Fig. 4). Moreover, we prepared β 2 m knockout variants of both HeLa and HeLa-Lai cell lines using the CRISPR/Cas9 technique. Both parental and β 2 m negative cells were analysed by flow cytometry for the presence of HLA-C free chains at the cell surface using the L31 antibody (Fig. 5). We observed that both HeLa and HeLa-Lai β 2 m negative cells lacked the expression of HLA-C at the cell membrane, indicating that the presence of HIV-1 Env in HeLa-Lai was not sufficient for HLA-C translocation, which strictly requires β 2 m for MHC-I assembly and transportation to the surface, where the presence of Env might favour the formation of HLA-C free chains.

HIV-1 Env and HLA-C interaction: BiFC assay. Previous studies demonstrated the association between
Env and HLA-C 3,4 . To analyse Env and HLA-C interaction, the Bimolecular Fluorescence Complementation (BiFC) assay was exploited. BiFC assay is based on the association between two non-fluorescent fragments of a fluorescent protein which have low affinity for each other, but when they are fused to interacting proteins, the fluorophore is reconstituted and a fluorescent signal is produced by bimolecular complementation 20 . Several constructs were developed, cloning the N-and C-terminal fragments (Yn and Yc) of the Yellow Fluorescent Protein (YFP) to the C-terminus of Env and HLA-C proteins. The system was validated in HeLa cells, assessing the complementation between Jun-Yn and Fos-Yc, which resulted in a green signal in the nucleus (Fig. 6a), as previously described 21 . Since it is known that Fos is efficiently translocated from cytoplasm to the nucleus only after interacting with Jun through their leucine zippers and by the use of the nuclear localization signal of Jun 22 , the nuclear BiFC signal confirms the Fos/Jun association. For the same reason when Fos-Yc, which in the absence of Jun localizes in the cytoplasm, was co-transfected with Env-Yn and the result was a BiFC negative signal (Fig. 6b). Finally, HeLa cells were co-transfected with Env-Yn and Env-Yc and stained with the anti-β 2 m antibody (red signal). The BiFC green signal was detected at the plasma membrane, due to Env-Env interaction (Fig. 6c,e) 23 . Similarly, co-transfection of Env-Yn and HLA-C-Yc reconstituted the YFP fluorophore at the cell membrane (Fig. 6d,f). All the z-stacks from a single cell were further investigated (see Supplementary Movie S1). The BiFC green signal was not widespread, but rather polarized in specific membrane clusters. The Env/ HLA-C complex did not appear to co-localize with β 2 m, indicating that HLA-C molecules may associate at the cell membrane either with Env or with β 2 m, resulting in green or red fluorescence, respectively. The absence of a clear co-localization signal suggests that the association between Env and HLA-C involves HLA-C free chains.

HIV-1 Env interacts with HLA-C at the cell membrane.
To further investigate the association between Env and HLA-C at the cell membrane, HeLa-Lai cells were treated with the cell membrane insoluble, thiol-cleavable DTSSP reagent 24 , which was used to crosslink protein complexes at the cell surface. Proteins were purified on a GN lectin column 25 which specifically binds D-mannose groups present on the HIV-1 Env glycoprotein and several fractions were obtained by elution with increasing concentrations of methyl α -D-mannopyranoside. Purified complexes heavier than 100 kDa were analysed by western blot after reduction (Fig. 7a). As a negative control, identical amounts of proteins from HeLa cells were analysed following the same procedure. To exclude any excessive cross-linking of membrane proteins due to DTSSP treatment, flotillin-1, a membrane protein, was tested: no specific copurification in the presence of HIV-1 Env was observed, and no differences in its purification were detected between the two cell lines. HLA-C molecules produced in HeLa cells were eluted in the first fractions as they contain few mannoses 26 , whereas HLA-C molecules produced in HeLa-Lai cells were co-purified, in parallel with Env, in higher amounts and in all the fractions. In addition, co-immunoprecipitation of protein complexes from the cell surface of both Hela and HeLa-Lai cells with the 2G12 antibody (anti-Env), further confirmed the association between Env and HLA-C (Fig. 7b), in agreement with previously reported data 3 . Incubation with the DT9 antibody did not result in Env immunoprecipitation, excluding the involvement of HLA-C heterotrimers in the association with Env. Overall, these results indicated that HLA-C free chains are co-purified with Env at the cell membrane. Functional effect of β 2 m on HIV-1 infectivity. Previous results showed that HIV-1 envelope glycoprotein specifically associates with HLA-C, improving the ability of virions to infect target cells 4 . To discriminate the role of β 2 m in virion infectivity, Env-pseudotyped viruses (QHO and pRHPA) were produced in parental 293T and in CRISPR/Cas9 β 2 m negative 293T cells. An infectivity assay was performed with TZM-bl cells, in which the luciferase expression signal reflects virus infectivity 27 . Equivalent amounts of pseudoviruses, determined by p24 titration, were used. Luciferase expression resulted significantly higher following infection with pseudoviruses originating from 293T-β 2 m positive cells than with the ones derived from 293T-β 2 m negative cells (p < 0.0001) (Fig. 8). Analysis of the interpolating curve, best fitting the Relative Luminescence Units (RLU) values, indicated that in the absence of β 2 m, the virus was about 3 times less infectious. No difference in infectivity was observed using Vesicular Stomatitis Virus (VSV) G protein pseudotyped viruses produced in both 293T cells types. These results suggested that β 2 m increases HIV-1 infectivity ensuring HLA-C translocation at the cell surface, where the association with Env occurs.

Discussion
HLA-C molecules are selectively incorporated into the HIV-1 envelope during virus budding from the host cell membrane 2 . Several evidences of the interaction between Env and HLA-C indicate that the presence of HLA-C is necessary to facilitate HIV-1 infection. The two molecules have been found associated and in the absence of HLA-C both fusion efficiency and virus infectivity are reduced [3][4][5] . In this work, a significant up-regulation of surface-HLA-C free chains, reactive to the L31 antibody 18 was observed in HIV-1 chronically infected cells (ACH-2) following virus reactivation by TNF-α stimulation. This was not observed in the corresponding uninfected parental cell line (A3.01). On the contrary, cellular staining with the DT9 antibody, which detects HLA-C heterotrimers 17 , did not reveal any difference between the two cell lines, suggesting that HIV-1 reactivation selectively enhances HLA-C free chains expression at the cell surface. Moreover, the amount of HLA-C free chains at the cell surface correlated with the expression of HIV-1 Env, decreasing after 48 and 72 h from TNF-α stimulation. On the contrary, no differences in the cell membrane expression of total HLA-C molecules, HLA-C heterotrimers, MHC-I or β 2 m were detected, indicating that HIV-1 infection may preferentially increase the amounts of HLA-C free-chain conformation. This observation was further confirmed in another cell line (PM1), where the TNF-α -induced increase of HIV-1 replication correlated with the upregulation of HLA-C free chains at the cell membrane. In addition we observed that Env-transfected 293T cells expressed more HLA-C free chains on their membrane than 293T cells transfected with the Env defective full length HIV-1 genome (pSG3 Δenv plasmid), suggesting that Env may promote the formation of HLA-C free chains at the cell surface. HIV-1 infection does not directly cause upregulation of HLA-C, neither at the expression level nor by modulating its stability or membrane translocation. We hypothesize that the presence of HIV-1 Env favours the dissociation of β 2 m from HLA-C, thus resulting in a higher amount of HLA-C free chain molecules at the plasma membrane, but this point requires further investigation. β 2 m is essential to translocate HLA-C at the cell surface, where β 2 m light chain might dissociate from HLA-C heavy chain and the association of HLA-C free chain with Env may take place. β 2 m is necessary for ensuring HLA-C translocation at the cell surface, indicating that HLA-C molecules strictly require β 2 m for reaching the plasma membrane, following the classical MHC-I pathway and that neither the presence of Env, nor the presence of other HIV-1 proteins, is responsible for HLA-C free-chains translocation. Our results with the BiFC technique provided experimental evidence of a very close association between Env and HLA-C free chains at the cell membrane. The association was further confirmed both by proteins purification on GN lectin columns after membrane proteins crosslinking and by co-immunoprecipitation of membrane protein complexes from Hela and HeLa-Lai cells. HLA-C molecules are characterized by a post-translational biosynthetic bottleneck due to an inefficient assembly with the peptide, which leads to accumulation in the endoplasmic reticulum of both β 2 m-free and β 2 m-associated folding intermediates 28 . This may be the reason why HLA-C could more easily dissociate from β 2 m compared to HLA-A and -B. Due to their long half-life, MHC-I free chains have been shown to cis-associate both with themselves and with a variety of membrane receptors, such as CD3, CD8α β , CD25, Ly49A and IL-15Rα 2,29-31 . It is therefore plausible that HLA-C free chains may also have the propensity to associate with Env. Since β 2 m is essential for the assembly of MHC-I molecules and their translocation to the plasma membrane, we investigated the role of β 2 m in HIV-1 infectivity. In particular, we perfomed an infectivity assay in TZM-bl cells, demonstrating that the infectivity of HIV-1 pseudoviruses produced in β 2 m negative 293T cells was roughly 3 times lower than the infectivity of pseudoviruses produced in the corresponding β 2 m positive parental 293T cells. A similar 3-fold reduction in HIV-1 infectivity was also reported by Cosma et al. 3 . Our data are also supported by previous data showing that silencing of HLA-C expression significantly reduces HIV-1 infectivity 4 . This result further confirms the involvement of HLA-C in increasing HIV-1 infectivity 4 , indicating that β 2 m is essential for the translocation of HLA-C to the cell surface where, after β 2 m dissociation, HLA-C free chains might associate with Env. However, since β 2 m knockout in 293T cells also highly affects the expression of other classical and non-classical HLA-Class I molecules, the observed reduction in infectivity of the virions may not be exclusively due to the absence of HLA-C on the cell surface. The experiments described, together with previously reported data [3][4][5] , demonstrate that HLA-C interacts with Env; in particular in this study we showed that the enhanced infectivity conferred to the virus by HLA-C, previously reported by our group 4 , involves HLA-C free chains. Overall, the data presented in this work support the hypothesis that HLA-C involvement in HIV-1 infection may confer some advantages to the virus. HLA-C may be involved in several HIV-1 processes, such as improving the virion assembly process, reducing the shedding of viral glycoproteins from the cell membrane, stabilizing, through a cis-interaction, the Env proteins on the virion in a favourable conformation for cell attachment and fusion. The presence of HLA-C also reduces HIV-1 virions susceptibility to neutralizing antibodies 3 . On the other hand, it has been reported that HLA-C is involved in the immune control of HIV-1 infection, contributing to Western blot analysis of immunoprecipitated samples shows that HLA-C is associated either with β 2 m or with Env. HLA-C and α /β -tubulin expression were detected as loading control.
Scientific RepoRts | 7:40037 | DOI: 10.1038/srep40037 recognition by CTLs and lysis of HIV-infected cells. Protective genotypes have been correlated with high levels of HLA-C surface expression 6,32 : higher HLA-C surface expression has been associated with a stronger control of viral load 33 . A recent study 34 reported that the different amounts of surface-expressed HLA-C is dependent on the allelic variant and confirmed the correlation between high HLA-C levels and control of HIV-1 viremia. However, HLA-C does not efficiently present antigen to CTLs, while it is an extremely good ligand for KIR receptors, protecting target cells from lysis mediated by NK cells 35 . This may imply that other properties and functions of HLA-C alleles could be involved 14 . Recently, the relation between HLA-C alleles and expression levels has been questioned by a study in which the reported association between HLA-C cell surface expression and the − 35 Kb polymorphism was not confirmed. Rather a high-allelic variability in HLA-C mRNA expression was reported 11 . The protective role of highly expressed alleles towards infection is in apparent contrast with our results, which show a role for the Env/HLA-C association in enhancing viral infectivity. Interestingly, in this work we showed that the interaction between HIV-1 and HLA-C selectively involves HLA-C free chains (L31 reactive). HLA-C expression observed by Apps et al. 34 was assessed using a different antibody (DT9), which recognizes β 2 m-associated HLA-C and cross-reacts with HLA-E 36 . In our study, the L31 antibody, specific for the HLA-C free chains, allowed the definition of the HLA-C conformation as an important discriminatory factor for the association with HIV-1 Env. This may confer to HLA-C allelic variants a selective capacity to function as accessory molecules in the process of virus assembly and budding 14 . Differences reported among HLA-C alleles might rely on intrinsic differences in their lower or higher ability to bind β 2 m. It is known that murine MHC-I molecules show allelic differences in their interaction with β 2 m 37 . In addition, it should be taken into consideration the possibility that HIV-1 could have the capacity to directly displace β 2 m from HLA-C and/or bind β 2 m, as previously reported for CMV 38 . β 2 m has been detected on HIV-1 virions 39 , as well as on HTLV-I 40 , echoviruses 41 and coxsackievirus 42 viral particles. β 2 m bound to these viruses is involved in modulating viral infectivity as well as in conferring protection from neutralizing antibodies. Furthermore, β 2 m bound to HIV-1 virions is recognized by a specific antibody, R7V, which neutralizes different HIV-1 isolates but does not bind to the cell surface; this epitope appears to be exposed only when β 2 m is bound to HIV-1 particles 43 . In conclusion, the alternative conformations of HLA-C may directly affect the balance between protection and susceptibility outcome. In particular, its prevalence at the cell membrane either as heterotrimer or as free chain could determine whether HLA-C is more advantageous to the host or to the virus, respectively. Despite differences in the expression of HLA-C alleles do exist, due to the action of redundant regulatory mechanisms, a common feature is the poor HLA-C assembly 28 . This conformational instability suggests that HLA-C allotypes weakly binding β 2 m might be mainly present as free chains, resulting in a selective propensity for association with other proteins, including HIV-1 Env. The stability of different HLA-C alleles was assessed in a previous work by measuring the relative abundance of both free and β 2 m-associated HLA-C molecules at the plasma membrane 28 . A correspondence may be observed between highly expressed alleles, according to Apps' definition 34 , and a more stable association with β 2 m, and between low expressed alleles and prevalence as free chains. In particular, low expressing alleles, such as C * 04 and C * 07, showed a strong prevalence as L31-reactive free chains, while high expressing alleles, such as C * 06 and C * 08, showed an increased β 2 m-association, detected with the W6/32 antibody 28 . A complete characterization of HLA-C conformations on infected cells is required, in relation to their expression level. We are currently attempting to investigate if the different amounts of HLA-C detected at the cell membrane may also be attributed to different predominant conformations. Different HLA-C alleles should also be evaluated for possible differences in the interaction with HIV-1 Env. This study suggests that HLA-C expression may go beyond its immunological role; rather it could have a deeper implication during HIV-1 infection. Our results help to elucidate the interaction between HLA-C and Env and provide new insights into the complex role played by HLA-C in HIV-infection. The understanding of this interaction may assist in the design of new therapeutic strategies aimed at controlling HIV-1 infection.

Methods
Cell lines. HeLa and 293T cells were obtained from the American Type Culture Collection (ATCC). HeLa-Lai cells were kindly provided by Dr. Uriel Hazan, Institut Cochin, Paris, France. Briefly, these cells were obtained by transfection of a HIV-1 provirus in which gag and pol genes were deleted, while nef was replaced by the dhfr drug resistance 44 . TZM-bl cells were provided by the EU Programme EVA Centre for AIDS Reagents, NIBSC (ARP5011) 45 . A3.01 46 is a CD4 + T-lymphoma cell line, from which the HIV-latently infected cell line (ACH-2) has been derived by in vitro infection with HIV-1 IIIB. The viral expression is induced in ACH-2 cells by stimulation with TNF-α 16,47,48 . PM1 (from Dr. P. Lusso, AIDS Research and Reference Program, Division of AIDS, NIAID, NIH, USA) is a clonal derivative of the human T lymphocytic HUT 78 cell line and PM1-IIIB is the chronically-infected counterpart. All the cell lines were grown at 37 °C in a humidified atmosphere with 5% CO 2. HeLa, HeLa-Lai, 293T, TZM-bl cell lines were cultured in Dulbecco's modified Eagle's medium, high glucose (Euroclone), while A3.01, ACH-2 and PM1 cell lines were cultured in RPMI-1640 medium (Euroclone). Culture medium was supplemented with 10% FBS, 2 mM L-Glutamine, 100 U Penicillin/ml and 100 U Streptomycin/ml (Lonza). Flow cytometry analyses. A3.01, ACH-2 and PM1 cells were cultured either in the absence or in the presence of TNF-α (10 ng/ml) for the indicated time points. Cells were stained for 30 minutes at 4 °C with 2.5 μ g/ml of DT9, 1 μ g/ml of L31 and W6/32, 5 μ g/ml of 2G12 and 1:200 dilution of ascitic fluid for NAMB-1. After PBS washes, cells were labelled with phycoerythrin-conjugated anti-mouse or Alexa Fluor 488 conjugated anti-human antibodies. Cells were fixed with 1% formaldehyde prior to acquisition with a Gallios flow cytometer (Beckman Coulter) and analysed using the FlowJo software (TreeStar, San Carlos, CA). Data for the time course analysis between TNF-α stimulated and unstimulated A3.01 and ACH-2 cells were expressed as fluorescence fold change between RFI (Relative Fluorescence Intensity) of TNF-α stimulated and unstimulated cells. β 2 m negative 293 T cells (transfected with the different HIV-1 plasmids), as well as parental and β 2 m negative HeLa and HeLa-Lai cells, were labelled with 1 μ g/ml of L31 antibody for 30 minutes at 4 °C, followed by staining with Alexa Fluor 488-conjugated goat anti-mouse antibody. Data were acquired with a FACSCanto flow cytometer (BD Bioscience) and analysed using the FlowJo software (TreeStar, San Carlos, CA). For HLA-C free chains determination (L31 Scientific RepoRts | 7:40037 | DOI: 10.1038/srep40037 acid wash), cells were treated with RPMI-1640, 20% FBS, pH 2.5 for 3 minutes in ice to displace β 2 m, prior to washes and staining with L31.

Bimolecular Fluorescence Complementation (BiFC) assay. HeLa cells were transfected with BiFC
constructs and then switched to 30 °C for 2 hours to allow the fluorophore maturation 21 . Cells were fixed with 4% paraformaldehyde and labelled with the anti-β 2 m antibody (Abcam). After PBS washing, cells were incubated with Texas red-labelled sheep anti-rabbit antibody and nuclei were stained with Hoechst. Coverslips were mounted with Dako fluorescence mounting medium. The images were captured using a confocal laser-scanning fluorescence microscope Leica TCS SP5 and analysed with the LAS AF software (Leica). For each experiment, the Jun-Yn and Fos-Yc complementation was assessed as a positive control 21 , while the complementation of Fos with either HLA-C or Env was tested as a negative control.
Env/HLA-C co-purification. HeLa and HeLa-Lai cells were fixed with 1 mM DTSSP (ThermoScientific) according to manufacturer's instructions. Cells were washed with PBS and resuspended in lysis buffer (2.5 mM Hepes, 145 mM NaCl, 1% Triton, 0.1 mM PMSF). The cell lysate was centrifuged to pellet nuclei and the supernatant was passed over a snowdrop lectin column (Galanthus nivalis lectin, Sigma) 25,59 , previously equilibrated with dialysis buffer (2.5 mM HEPES, 145 mM NaCl, pH 7.4, and 0.1% Triton). Proteins were collected in five elution fractions containing increasing concentrations (0.25, 0.40, 0.55, 0.70 and 1 M, one column volume) of methyl α -D-mannopyranoside (Sigma). Each eluted fraction was concentrated using a 100 KDa Amicon Ultra Centrifugal Filter (Millipore). Lysate (25 μ g), not bound (25 μ g) and 1/4 of the total volume of concentrated fractions were incubated for 10 minutes at 98 °C in the presence of 3% β -Mercaptoethanol, to cleave the DTSSP thiol links, and then analysed by western blot.
Co-immunoprecipitation of protein complexes at the cell surface. Co-immunoprecipitation was performed as previously described 60 . Briefly, HeLa and HeLa-Lai cells were labelled at the cell surface with 2G12, DT9, or anti-β 2 m (Abcam) antibodies, for 45 minutes at 4 °C. After PBS washing, cells were resuspended in 0.5% NP-40 non-denaturing lysis buffer. Cellular nuclei were removed by centrifugation at 3000 rpm for 3 minutes.
Cell surface labelled proteins complexes were incubated at 4 °C with Dynabeads ® protein G (Life Technologies) for 45 minutes. Beads were washed and resuspended in elution buffer containing 2% SDS and 40 mM DTT. After boiling at 98 °C, proteins were separated by SDS-PAGE and analysed by western blot.

Production of Env-pseudotyped viruses and TZM-bl assay. Pseudoviruses expressing the Rev/Env
sequences of different HIV-1 subtype-B strains (QHO and pRHPA) were produced by co-transfecting 293 T cells with the pSG3 Δenv backbone plasmid and collecting the pseudoviruses-containing supernatant after 48 hours, as previously described 27 . VSV-G was used as a control. Pseudoviruses were quantified using the p24 ELISA kit (Aalto Bio Reagents Ltd, Dublin, Ireland) and the same amount of virus was used to infect 1 × 10 4 TZM-bl cells in the presence of 20 μ g/ml DEAE-dextran 27 . Luminescence was measured 48 hours later, using the Victor TM 3 luminometer (Perkin Elmer). Each assay was performed in quadruplicate.

Statistical analysis.
Results were analysed using two-way ANOVA using StataMP 13.0 (Stata Corp. 2013, College Station, TX, USA). A significance level of p < 0.05 was accepted as statistically significant.