HIV-1 Vpu utilizes both cullin-RING ligase (CRL) dependent and independent mechanisms to downmodulate host proteins

Hijacking of the cullin-RING E3 ubiquitin ligase (CRL) machinery is a common mechanism employed by diverse groups of viruses for the efficient counteraction and degradation of host proteins. In particular, HIV-1 Vpu usurps the SCFβ-TrCP E3 ubiquitin ligase complex to mark CD4 for degradation by the 26S proteasome. Vpu also interacts with and downmodulates a number of other host proteins, including the restriction factor BST-2. However, whether Vpu primarily relies on a cullin-dependent or -independent mechanism to antagonize its cellular targets has not been fully elucidated. We utilized a sulphamate AMP analog, MLN4924, to effectively block the activation of CRLs within infected primary CD4+ T cells. MLN4924 treatment, in a dose dependent manner, efficiently relieved surface downmodulation and degradation of CD4 by NL4-3 Vpu. MLN4924 inhibition was highly specific, as this inhibitor had no effect on Nef’s ability to downregulate CD4, which is accomplished by a CRL-independent mechanism. In contrast, NL4-3 Vpu’s capacity to downregulate BST-2, NTB-A and CCR7 was not inhibited by the drug. Vpu’s from both a transmitted founder (T/F) and chronic carrier (CC) virus preserved the ability to downregulate BST-2 in the presence of MLN4924. Finally, depletion of cellular pools of cullin 1 attenuated Vpu’s ability to decrease CD4 but not BST-2 surface levels. We conclude that Vpu employs both CRL-dependent and CRL-independent modes of action against host proteins. Notably, we also establish that Vpu-mediated reduction of BST-2 from the cell surface is independent of β-TrCP and the CRL- machinery and this function is conserved by Vpu’s from primary isolates. Therefore, potential therapies aimed at antagonizing the activities of Vpu may need to address these distinct mechanisms of action in order to achieve a maximal effect.


Background
Cullin-RING ligases (CRLs) constitute an important group of ubiquitin ligases and play a prominent role in the efficient regulation of protein turnover and homeostasis [1]. In particular, a recurring theme among viruses from distant families is their common ability to usurp CRL complexes with the aim of evading host control mechanisms. Notably, the HIV-1 accessory protein Vif hijacks a cullin-5 containing ubiquitin ligase complex (CRL5) to target cytidine deaminases of the APOBEC3 family for proteasomal degradation [2][3][4][5]. Similarly, the HIV-2 accessory protein Vpx relies on a CRL4 complex to degrade the restriction factor SAMHD1 [6,7].
Activation of CRLs is dependent on a process known as neddylation. This post-translational modification involves the covalent addition of the NEDD8 protein, a relative of ubiquitin, onto a lysine residue on the cullin backbone. Neddylation induces a conformational change in the CRL complex that turns the enzyme catalytically active, allowing the transfer of ubiquitin to a substrate [8]. MLN4924, a potent inhibitor of the E1 neddylation enzyme NAE (Nedd8-activating enzyme), blocks the activity of all CRLs but does not affect non-cullin ubiquitin ligases [9]. Previous studies have shown that MLN4924 can potently block Vif-mediated proteasomal degradation of APOBEC3G [10]. Furthermore, in the context of HIV-2, MLN4924 inhibited the degradation of SAMHDI induced by Vpx, phenocopying the absence of Vpx in HIV-2 and restoring efficient restriction of the virus in myeloid cells [11][12][13].
In this study, we asked whether pharmacological inhibition of the SCF β-TrCP complex by MLN4924 would reveal whether cullin activity is important for Vpu to downmodulate its cellular targets. We hypothesized that downregulation of BST-2, CCR7 and NTB-A by Vpu would not be impacted by MLN4924 treatment. Moreover, we predicted that Vpu downregulation (and degradation) of CD4 would be relieved by MLN4924. Finally, we sought to determine whether Vpu-mediated cell surface downregulation of BST-2 is a function that can be dissociated from BST-2 degradation and that is cullin-independent.

Pharmacological inhibition of CRL-activity disables NL4-3 Vpu's ability to downregulate CD4, but not BST-2, CCR7 or NTB-A
To determine whether Vpu can act as multifunctional protein capable of downregulating host proteins in the absence of neddylation and a functional SCF β-TrCP complex, primary CD4 + T cells were infected with either an HIV-1 NL4-3 -derived, replication-defective virus carrying GFP in place of Nef (DHIVGFP(Vpu+/Nef−); Fig. 1a), or with an isogenic virus lacking both Nef and Vpu (DHIVGFP(Vpu−/Nef−); Fig. 1b) [21]. All viruses were pseudotyped with the vesicular stomatitis virus glycoprotein G (VSV-G). We utilized the above nef-and env-deficient viruses such that the known activities of Nef and Env on CD4 would not interfere with that of Vpu (reviewed in [22]). Two days post infection, cells were incubated in either DMSO (solvent) or MLN4924 and protein surface expression analyzed by flow cytometry 24 h later. As expected, the virus devoid of Nef and Vpu (DHIVGFP(Vpu−/Nef−)) showed similar surface levels of CD4, BST-2, CCR7 and NTB-A when comparing GFP-negative (uninfected) and -positive (infected) cells (Fig. 2a, panels ii, viii, xiv, xx). Downregulation of CD4, BST-2, CCR7 and NTB-A was apparent in cells that were infected with DHIVGFP(Vpu+/Nef−) and treated with DMSO (Fig. 2a, panels iii, vix, xv, xxi). However, MLN4924 relieved downmodulation of CD4 in a dosedependent manner (Fig. 2a, panels iii-vi, b). In contrast, downregulation of BST-2, CCR7 and NTB-A was unaffected by MLN4924 treatment (Fig. 2a, panels vii-xxiv, b). These results indicate that Vpu utilizes both cullindependent and -independent mechanisms for down modulating host proteins.

MLN4924 relieves NL4-3 Vpu, but not Nef mediated, degradation of CD4
To determine whether MLN4924 also prevented the degradation of CD4, primary CD4 + T cells were infected as described in Fig. 2 but were instead permeabilized, fixed and stained for total levels of CD4. Figure 3 shows that inhibition of neddylation rescued CD4 from Vpuinduced degradation (Fig. 3a, panels ix-xii, b). As a further control to show specificity of cullin inactivation by MLN4924, primary CD4 + T cells were infected with either env-defective HIV-1 (DHIV; Fig. 1e), lacking Vpu but expressing Nef (DHIV Vpu−/Nef+; Fig. 1f ) or an isogenic virus lacking Nef and expressing Vpu (DHIV Vpu+/Nef−; Fig. 1g). Nef accelerates the endocytosis of target CD4 molecules present on the plasma membrane via clathrin and Adaptor-Protein 2 (AP-2), ultimately shuttling CD4 for lysosomal degradation in a multivesicular body (MVB) dependent manner [23,24]. We therefore hypothesized that a virus encoding only Nef (DHIV Vpu−/Nef+) would be able to downmodulate CD4 in a manner that would be insensitive to MLN4924 treatment. This expectation was corroborated as shown in Figs. 3c (panels vii and viii), d.

Primary Vpu isolates maintain the ability to downregulate BST-2 in the absence of CRL-activity
A recent study showed that Vpu alleles from field strains of HIV-1 have the capacity to modulate host proteins, in particular CD4 and BST-2, to a greater extent than the prototypical HIV-1 NL4-3 Vpu [25]. We therefore wished to determine whether the CRL-dependent and independent mechanisms observed with HIV-1 NL4-3 Vpu ( Fig. 2a) would be maintained with Vpu alleles from primary isolates. To address this, we replaced Vpu in DHIV-GFP(Vpu+/Nef−) with Vpu's from either a transmitted founder (T/F; WITO) [26][27][28][29][30] or from a chronic carrier (CC; WARO) [31] virus. These new viruses were termed DHIV-GFP WITO (Vpu+/Nef−; Fig. 1c) and DHIV-GFP WARO (Vpu+/Nef−; Fig. 1d). We observed that Vpu's from both a T/F (DHIV-GFP WITO) and CC (DHIV-GFP WARO) exhibited an enhanced ability to downregulate CD4 when compared to HIV-1 NL4-3 Vpu (Fig. 4a, panels iii-v, b upper left). This effect was previously reported by Jafari et al., who proposed that primary Vpu isolates may adopt an optimized tertiary structure better suited to counteract CD4 as a result of differences between their The lentiviral vector DHIV, or "defective" HIV, was derived from the HIV-1 NL4-3 sequence and cut between two BglII sites to efficiently delete envelope/gp120 (gray box with dashed lines, red X marks) but maintain in-frame Tat, Rev and RRE ORFs. The constructs used in this study were derived from the DHIV backbone and are as follows: i the GFP gene in place of Nef. ii Replacement of NL4-3 Vpu with a primary Vpu isolate or iii introduction of a frame shift mutation within Vpu and Nef. a DHIV-GFP ( h Amino acid sequence alignment of Vpu proteins from a transmitted founder (T/F;WITO) and chronic carrier (CC; WARO) compared to NL4-3 Vpu. An asterisk indicates fully conserved residues; colon represents amino acid conservation with strongly similar properties; period designates amino acid conservation with weakly similar properties. Highlighted residues mark amino acid differences between the three strains.
amino acid sequence and that of HIV-1 NL4-3 Vpu [16,32,33] (Fig. 1h). MLN4924 treatment had a dramatic effect on the ability of both a T/F and CC Vpu to downregulate CD4 (Fig. 4a, panels viii-x, b lower left). Both a T/F and CC Vpu were able to decrease the cell surface density of BST-2 to about the same extent as HIV-1 NL4-3 Vpu (Fig. 4a, panels xiii-xv, b upper right). However, the addition of MLN4924 in cells expressing Vpu did not restore surface levels of BST-2 to any significant degree (Fig. 4a, panels xviii-xx, b lower right). These results reinforce the fact that primary Vpu isolates also possess the ability to counteract host proteins by CRL-independent mechanisms.

siRNA mediated knockdown of cullin 1 reduces CD4, but not BST-2, surface levels in the presence of Vpu
Although MLN4924 is primarily used and known as an NAE (and thus pan-CRL) inhibitor, at IC 50 values of 1.5 and 8.2 μM the drug can also block the functions of the NAE-related enzymes, ubiquitin-activating enzyme (UAE) and SUMO-activating enzyme (SAE) [9]. Therefore, as an alternative approach to chemical inhibition, we assessed Vpu function in cells depleted of cullin 1. HeLa-CD4 cells were transfected twice (24 h apart) with either a non-targeting siRNA or siRNA targeting cullin 1, followed by infection with either DHIV-GFP WARO (Vpu+/Nef−) or DHIV-GFP (Vpu−/Nef−). As shown in Fig. 5, knockdown of cullin 1 attenuated Vpu's capacity to downregulate CD4 from the cell surface (Fig. 5a, compare panels ii and v), but had no effect on downregulation of BST-2 (Fig. 5a, panels viii and xi). We conclude that Vpu's ability to decrease surface levels of BST-2 is independent of CRL-activity.
Numerous studies have shown that the di-serine motif of Vpu, which mediates interaction with β-TrCP when phosphorylated, is required for downmodulation and/or degradation of both CD4 and BST-2 [17,[48][49][50][51][52]. Thus, mutation of the di-serine motif abrogates degradation of CD4 [16,53], confirming a role for β-TrCP. On the other hand, Tervo et al. found that depletion of β-TrCP2 or the simultaneous depletion of β-TrCP1 and 2 did not block the ability of Vpu to promote HIV-1 release and failed to restore surface levels of BST-2 [43]. These results, taken together, suggest that the di-serine motif of Vpu is directly involved in interaction with β-TrCP, but that mutations in this motif affect the downregulation of BST-2 through possibly a more general effect on Vpu protein conformation or perhaps the binding of another cellular factor implicated in the mislocalization of host proteins [43]. Therefore, our observation that downregulation of BST-2, CCR7 and NTB-A is CRL-independent does not contradict the notion that the di-serine motif of Vpu is required for this function [15,21,54].
(see figure on previous page.) Fig. 2 HIV-1 Vpu utilizes both cullin dependent and independent mechanisms to downregulate host proteins. a Primary CD4 + T cells were either mock infected or infected at an MOI of 1 with DHIVGFP(Vpu+Nef−) or DHIVGFP(Vpu−Nef−). 2 days post infection, either DMSO or increasing concentrations of MLN4924 were added to cell cultures. 24 h later, surface expression of CD4, BST-2, CCR7 or NTB-A was analyzed by flow cytometry. Histograms depict a comparison of GFP negative (blue line) and GFP positive (red line) cells along with an IgG isotype control (gray shaded histogram). Unless otherwise noted, all experiments involving primary CD4 + T cells are representative of three separate experiments performed in three different healthy donors. b Relative mean fluorescence intensity (MFI) values of surface expression of CD4, BST-2, CCR7 or NTB-A from DHIVGFP(Vpu+Nef−) infected cells (a). Data was normalized by setting the MFI values from uninfected (mock) cells to 100% and is depicted graphically as ±SEM. Unless otherwise noted, all experiments including statistics were performed through a pairwise Student's t test comparing GFP positive and GFP negative cells to assess significance: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Fig. 3 MLN4924 alleviates Vpu-but not Nef-mediated degradation of CD4. a Cultured T CM were infected as described in Fig. 2a. To assess total levels of CD4, cells were permeabilized and stained 24 h after addition of MLN4924 and analyzed by flow cytometry. Histograms are color-coded as described in Fig. 2a. b MFI values of total (intracellular) CD4 expression levels from DHIVGFP(Vpu+Nef−). Data was normalized and statistical significance obtained as described in Fig. 2b. n.s. not significant. c Primary CD4 + T cells were either mock infected or infected at an MOI of 1 with DHIV WT, DHIV Vpu−/Nef+ or DHIV Vpu+/Nef−. 2 days post infection, cell cultures were treated with either DMSO or 500 nM MLN4924. 24 h post MLN4924 treatment, cells were assessed for surface levels of CD4 between p24Gag neg (blue line) and p24Gag pos (red line). Gray shaded histograms represent an IgG matched isotype control. Shown is one representative experiment out of three. d Relative CD4 surface levels were quantified from data obtained in Fig. 3c and are depicted graphically as ±SEM of either cells treated with DMSO (left) or 500 nM MLN4924 (right). Statistical significance between p24Gag neg and p24Gag pos populations was determined as in Fig. 2b.  Fig. 4 Primary Vpu isolates maintain the ability to decrease surface expression of CD4 and BST-2 in a CRL-dependent and independent manner. a Primary CD4 + T cells were either mock infected or infected at an MOI = 1 with DHIV-GFP (Nef−/Vpu−), DHIV-GFP (Nef−/Vpu+) or viruses encoding Vpu taken from either a transmitted/founder (T/F) (DHIV-GFP WITO (Nef−/Vpu+)) or a chronic carrier (CC) primary isolate (DHIV-GFP WARO (Nef−/ Vpu+)). The cells were treated with either DMSO or 500 nM MLN4924 48 h post infection. 24 h post MLN treatment, the cellular GFP negative and positive populations (red and blue histograms) were assessed for surface levels of CD4 and BST-2 via flow cytometry. An IgG control was included to set the baseline for positive staining (gray shaded histogram). b Relative CD4 and BST-2 surface levels were quantified, normalized, and scored statistically as described in Fig. 2b. Structurally, phosphorylation of Vpu at serine 52 and 56 induces a conformational change in Vpu's C-terminus: the formation of a β-strand within residues 50-59 and displacement of the third alpha helix (h3) away from the phosphorylation site [55]. The structural rearrangements induced by phosphorylation result in the emergence of acidic side chains surrounding serine 52 and 56, creating an electronegative binding region upon which a protein exhibiting electropositive potential can bind [55]. In particular, phosphorylated serines 52 and 56, glycine 53, asparagine 54 and the hydrophobic residues isoleucine 46 and leucine 45 were shown to be involved in the binding to the F-box protein β-TrCP [56]. Our observations with MLN4924 and cullin knockdown argue against a role for cullin mediated ubiquitination in the downregulation of BST-2 from an independent line of experimentation other than mutagenesis.
Given that Vpu does not induce the degradation of CCR7 or NTB-A but rather retention within the TGN, our data suggest that these molecules are also downregulated in a CRL-independent manner (Fig. 2a) [20,21,54]. A recent report by Bachle et al. identified a conserved C-terminal AWF motif present within HIV-1 subtype B Vpu isolates that influences the ability of Vpu to downregulate the lipid antigen receptor CD1d [57]. Therefore, whether Vpu's AWF or ELV motifs, which are CRL-independent, are required for downmodulation of CCR7 and NTB-A remains to be determined. It will also be compelling to investigate the requirement for β-TrCP and the SCF β-TrCP complex in downmodulation of other reported targets of Vpu, such as CD1d, CD155, CD62L and members of the tetraspanin family [58][59][60][61].

Conclusions
This work shows that cullin inactivation, through either pharmacological inhibition or depletion of cullin 1, does not render Vpu unable to downregulate BST-2, CCR7 or NTB-A. This highlights that facts that Vpu is multifunctional and that therapeutic targeting of neddylation, while potent and specific against CD4 downmodulation by Vpu, would still allow other targets to be downregulated.

Antibodies and reagents
Antibodies used in this study were as follows: APC-

Cells and plasmids
Human embryonic kidney 293T (HEK293T) cells were maintained in Dulbecco's Modified Eagle's Medium (DMEM) containing 10% fetal bovine serum (Atlas Biologicals) supplemented with 100 U/ml penicillin, 100 μg/ ml streptomycin and 2 mM l-glutamine (Life Technologies). HeLa-CD4 + clone 1022 (obtained through the NIH AIDS Reagent Division of AIDS, NIAID, NIH (Cat. #1109; Dr. Bruce Chesebro) were cultured in RPMI complete media in the presence of 1 μg/ml G418 (Life Technologies) while CCRF-CEM and primary CD4 + T cells were cultured in RPMI complete media only. All cells were maintained at 5% CO 2 at 37°C. For all experiments involving primary CD4 + T cells, coverage was maintained under protocol #IRB_00067637 approved by the University of Utah Institutional Review Board. The generation of cultured T CM has been described previously [21].

Flow cytometry
Surface levels of CD4, BST-2, CCR7 and NTB-A were assessed by staining cells with their appropriate antibodies at 4°C for 30 min in buffer (1 × PBS + 3% FBS). An additional step including a secondary antibody was necessary to detect BST-2 surface levels. A viability dye, eFlour 450 (eBioscience) was then used to distinguish live from dead cells. Fixation was achieved using 0.5% Paraformaldehyde (PFA).
In experiments involving surface analysis of CD4 and detection of intracellular p24, cells were first probed with anti-APC-CD4, stained with eFluor 450, permeabilized (Cytofix/Cytoperm: BD Biosciences) and then stained with mouse-anti-FITC-p24. Total levels of CD4 in primary CD4 + T cells were measured by staining cells with eFlour 450, permeabilization and then probing with anti-APC-CD4. All data was collected on a BD FACS CantoII and analyzed with FlowJo software.

Viruses and infections
Viral stocks were generated by co-transfection of 20 μg DHIV along with 5 μg of a construct expressing vesicular stomatitis virus G-protein (VSV-G) by calcium phosphate mediated transfection into HEK293T cells. Media (DMEM) was replaced after 16 h and the cellular supernatant collected, aliquotted and stored at −80°C 48 h post-transfection. Viral titers and MOI were determined via infection of CCRF-CEM cells. Primary CD4 + T cells generated as described above were infected 5 days postactivation at an MOI of 1 via spinoculation: 10 6 cells (1 ml final volume) for 2 h at 37°C in the presence of 8 μg/ml Polybrene (Sigma). After infection, cells were then resuspended in RPMI complete medium supplemented with IL-2 (30 U/ml).

siRNA mediated cullin 1 depletion
HeLa cells (5 × 10 5 ) were transfected twice (24 h apart) with either control siRNA or siRNAs against cullin 1 at a final concentration of 100 nM using Lipofectamine 2000 (Invitrogen) according to the manufacturers' instructions. The medium was changed 4 h after each transfection. 4 h post the second transfection, the cells were either mock infected or infected with DHIV-GFP (Vpu−/ Nef−) or DHIV-GFP WARO (Vpu+/Nef−) for 4 h at 37°C at an MOI = 0.8. After 48 h, the cells were either processed for flow cytometry or lysed and subjected to Western blot to detect cellular levels of cullin 1 between samples.