Differential dysregulation of β-TrCP1 and -2 by HIV-1 Vpu leads to inhibition of canonical and non-canonical NF-κB pathways in infected cells

ABSTRACT The HIV-1 Vpu protein is expressed late in the virus lifecycle to promote infectious virus production and avoid innate and adaptive immunity. This includes the inhibition of the NF-κB pathway which, when activated, leads to the induction of inflammatory responses and the promotion of antiviral immunity. Here we demonstrate that Vpu can inhibit both canonical and non-canonical NF-κB pathways, through the direct inhibition of the F-box protein β-TrCP, the substrate recognition portion of the Skp1-Cul1-F-box (SCF)β-TrCP ubiquitin ligase complex. There are two paralogues of β-TrCP (β-TrCP1/BTRC and β-TrCP2/FBXW11), encoded on different chromosomes, which appear to be functionally redundant. Vpu, however, is one of the few β-TrCP substrates to differentiate between the two paralogues. We have found that patient-derived alleles of Vpu, unlike those from lab-adapted viruses, trigger the degradation of β-TrCP1 while co-opting its paralogue β-TrCP2 for the degradation of cellular targets of Vpu, such as CD4. The potency of this dual inhibition correlates with stabilization of the classical IκBα and the phosphorylated precursors of the mature DNA-binding subunits of canonical and non-canonical NF-κB pathways, p105/NFκB1 and p100/NFκB2, in HIV-1 infected CD4+ T cells. Both precursors act as alternative IκBs in their own right, thus reinforcing NF-κB inhibition at steady state and upon activation with either selective canonical or non-canonical NF-κB stimuli. These data reveal the complex regulation of NF-κB late in the viral replication cycle, with consequences for both the pathogenesis of HIV/AIDS and the use of NF-κB-modulating drugs in HIV cure strategies. IMPORTANCE The NF-κB pathway regulates host responses to infection and is a common target of viral antagonism. The HIV-1 Vpu protein inhibits NF-κB signaling late in the virus lifecycle, by binding and inhibiting β-TrCP, the substrate recognition portion of the ubiquitin ligase responsible for inducing IκB degradation. Here we demonstrate that Vpu simultaneously inhibits and exploits the two different paralogues of β-TrCP by triggering the degradation of β-TrCP1 and co-opting β-TrCP2 for the destruction of its cellular targets. In so doing, it has a potent inhibitory effect on both the canonical and non-canonical NF-κB pathways. This effect has been underestimated in previous mechanistic studies due to the use of Vpu proteins from lab-adapted viruses. Our findings reveal previously unappreciated differences in the β-TrCP paralogues, revealing functional insights into the regulation of these proteins. This study also raises important implications for the role of NF-κB inhibition in the immunopathogenesis of HIV/AIDS and the way that this may impact on HIV latency reversal strategies based on the activation of the non-canonical NF-κB pathway.


FIG 1
Vpu inhibits both the canonical and the non-canonical NF-κB pathways. (A) Graphical representation of the canonical NF-κB pathway, detailing events downstream of the activation of the IKK complex. Stimuli such as TNFα, tetherin activation (through the retention of budding virus particles) or MAVS activation (following upstream sensing of viral RNA) trigger signaling cascades that converge at the activation of the IKK complex. IKKβ phosphorylates inhibitors of NF-κB, most commonly IκBα (but also p105 and p100), on dual serine residues 32 and 36 in the degron sequence SGLDS, leading to recognition by the β-TrCP substrate adaptor portion of an E3 cullin-RING ligase (SCF β-TrCP ). Ubiquitination of IκBα on lysine residues (represented by red squares in the schematic) by SCF β-TrCP triggers proteasomal degradation, releasing the NF-κB transcription factor (in this example the p65/p50 heterodimer), which translocates to the nucleus and activates the expression of NF-κB-dependent genes. P105 also acts as a precursor for the p50 portion of the NF-κB transcription factor, and is converted to active p50 by partial proteasomal processing. (B) Graphical representation of the non-canonical NF-κB pathway. Stimuli such as lymphotoxin β (LTβ), CD40 ligand, or the synthetic compound AZD5582, lead to the activation of NIK, which in turn phosphorylates IKKα. Activated IKKα phosphorylates p100 on dual C-terminal serine residues, prompting its recognition by SCF β-TrCP , ubiquitination and partial proteasomal processing to form mature RelB/p52 dimers, able to translocate to the nucleus and activate transcription. (C) Transient NF-κB activation assays were performed in HEK293T cells by co-transfecting an NF-κB-dependent luciferase reporter construct (3xNF-κB pConA), a renilla luciferase control plasmid, a fixed dose of plasmid expressing an NF-κB stimulus (MAVS, tetherin, IKKβ or NIK), and an increasing dose of Vpu or A49 plasmid. Twenty-four hours after transfection, cells were lysed and luciferase activity was determined. Results are expressed as a percentage of normalized signal recorded in the absence of Vpu or A49 (% max). Means are presented from at least four independent experiments, with error bars showing ± SD. The 2_87 line is shown on all graphs for comparison (gray line, gray circles), with 2_87 S3/7A in green, NL4.3 in purple, NL4.3 S2/6A (Continued on next page) Research Article mBio (Emi1) being the only cellular target of β-TrCP to demonstrate requirement for both paralogues rather than redundancy (22). Through viral molecular mimicry, the HIV-1 accessory protein Vpu contains an SGxxS motif (DSGNES) akin to other targets of the SCF β-TrCP . Indeed, β-TrCP was first discovered through its interaction with Vpu (23), and was later ascribed its major function in the NF-κB pathway (24,25). The serines in the Vpu SGNES motif are highly conserved and are essential for the optimal execution of all known Vpu functions (6,7,12). Phosphorylation of the serines by casein kinase II (CKII/CK2) (26,27) creates a binding site for β-TrCP, which is then co-opted by Vpu both to inhibit the NF-κB pathway (11,(28)(29)(30)(31) and to induce the ubiquitination and subsequent degradation of the HIV-1 receptor CD4 and the antiviral protein BST2/tetherin (23,(32)(33)(34)(35)(36). Thus, unlike cellular proteins possessing SGxxS degrons that are themselves targeted for ubiquitination and degradation, Vpu acts as an adaptor protein to link the E3 ubiquitin ligase machinery to its target proteins.
Mechanistic insights into NF-κB inhibition by Vpu have been established from studies of the T cell line-adapted HIV-1 molecular clone, NL4.3 (29,30). These demonstrated a sequestration of β-TrCP, resulting in a block to the ubiquitination and degradation of IκB and downstream inhibition of NF-κB translocation. It has more recently been recognized that the potency of this activity has been underestimated, as the NL4.3 Vpu used for these studies has severely diminished activity compared with primary Vpus (11,31). Thus, a reassessment of the mechanism with primary Vpus is appropriate in order to fully understand the nature of the inhibition.
Here, we investigate the interaction between HIV-1 Vpu and β-TrCP and its down stream consequences, including previously uncharacterized effects of Vpu on p105. We further document effects on the non-canonical NF-κB pathway, which may have implications for HIV latency reversal strategies. We demonstrate that inhibition of both pathways by Vpu involves the simultaneous degradation and sequestration of β-TrCP1 and -2, respectively, revealing the potential for distinct activities by these two paralo gues, and illustrating the fine balance between exploiting and inhibiting the pathways.
(D) CD4+ Jurkat T cells were infected with recombinant NL4.3 proviruses engineered to express either highly active 2_87 Vpu, 2_87 S3/7A Vpu, NL4.3 Vpu or no Vpu (Δ Vpu) at an MOI of 3. Seventy-two hours after infection, the cells were treated with 5 ng/mL TNFα. Total RNA was isolated at the indicated time points after treatment and subjected to RT-qPCR to detect CXCL10 mRNA. Data are plotted as mean fold increase relative to uninfected and untreated cells in two independent experiments, with error bars showing SEM.
Research Article mBio leading to SCF β-TrCP recognition, ubiquitination, and subsequent proteasomal processing to p52 (Fig. 1B). All stimuli induced NF-κB activation when transfected into HEK293T cells, measured by luciferase reporter assay, to an average level of 122-(MAVS), 42-(tetherin), 110-(IKKβ), and 333-fold (NIK) above background. 2_87 Vpu and A49 potently inhibited NF-κB induced by all four stimuli, demonstrating that these viral antagonists can inhibit both the canonical and non-canonical NF-κB pathways (Fig. 1C). In contrast, NL4.3 was significantly impaired across all concentrations (Fig. 1C). Compared to their wild-type counterparts, both double serine mutants (2_87 S3/7A and NL4.3 S2/6A) were significantly defective against all stimuli, with 2_87 S3/7A showing some activity at higher concentrations. Inhibition of NF-κB induced by tetherin revealed differences between the antagonists, with the defective Vpus 2_87 S3/7A, NL4.3 and NL4.3 S2/6A all showing increased inhibitory activity, corresponding to the fact that Vpu has an independent direct antagonistic effect on tetherin, ultimately inducing its degradation. Conversely, A49, which potently inhibited NF-κB activity induced by MAVS and IKKβ, was less effective at inhibiting tetherin-mediated NF-κB stimulation. The inhibition of IKKβ-induced signaling confirms previous findings that Vpu and A49 inhibit the NF-κB pathway downstream of the activation of the IKK complex (29,30,37,38); while the inhibition of NIK implies a similar block to the non-canonical pathway, both consistent with inhibition occurring at the β-TrCP level of the pathway. In order to determine if enhanced NF-κB suppression by the 2_87 Vpu was also observed in the context of replicating virus, NL4.3 viruses were engineered to express heterologous 2_87 Vpu, and mutants thereof, at endogenous levels. CD4+ Jurkat T cells were infected with viruses expressing either 2_87, 2_87 S3/7A, NL4.3 or no Vpu for 72 h, treated with TNFα and examined for CXCL10 mRNA expression over 24 h (Fig.  1D). Induction of CXCL10 was similar in uninfected cells and those infected with the Vpu-defective mutant. Whilst there was a blunting of CXCL10 induction at late time points in cells infected with virus expressing NL4.3 Vpu, this was much more pronounced and sustained in cells infected with a virus expressing the 2_87 Vpu. This was reversed with mutation of the phosphorylated serines. Thus, the 2_87 Vpu exhibits the enhanced suppression of NFκB-dependent responses in HIV-1 infected cells.

Primary Vpu induces the degradation of β-TrCP in infected cells
Previous work on the mechanism of NF-κB inhibition by HIV Vpu has shown that β-TrCP is sequestered and stabilized (29,30). To investigate whether a direct effect on β-TrCP could be visualized, endogenous levels of β-TrCP1 were examined under conditions of natural infection by HIV-1. Forty-eight hours following infection of cells with viruses expressing either 2_87, 2_87 S3/7A, NL4.3 or no Vpu, β-TrCP1 levels were examined by western blot (Fig. 2). Incongruous with the notion that Vpu sequesters and utilizes β-TrCP for the degradation of its target proteins, we observed a significant and consis tent depletion of β-TrCP1 in HEK293T, primary CD4 + T cells, and CD4 + Jurkat T cells infected with 2_87-expressing virus ( Fig. 2A through C). Viruses expressing NL4.3 Vpu exerted a similar but lesser effect. Degradation was rescued by treatment with protea somal inhibitor MG132 and the NEDD8-activating enzyme (NAE) inhibitor MLN4924, which specifically blocks the activation of cullin-RING ligases (CRLs) by inhibiting their activation through neddylation ( Fig. 2D; Fig. S1A), indicating degradation through a proteasomal and CRL-dependent pathway. Treatment with an inhibitor of lysosomal degradation, concanamycin A, did not rescue degradation (Fig. S1A).

Vpu has differential effects on β-TrCP1 and -2
The observation that Vpu leads to β-TrCP degradation is at odds with the essential role of β-TrCP in the Vpu-mediated degradation of CD4 and other cellular targets. There are two paralogues of β-TrCP -β-TrCP1 (BTRC) and β-TrCP2 (FBXW11)-encoded on separate chromosomes, each with several isoforms (21). Previous reports have implicated β-TrCP2 in the degradation of BST2/tetherin, while β-TrCP1 was dispensable for Vpu function (33,35). We sought to reconcile previous reports of selective β-TrCP usage by Vpu with our Research Article mBio observations of β-TrCP1 degradation. The lack of an antibody suitable for the detection of endogenous levels of β-TrCP2 led us to take a molecular approach. Transient expres sion assays were performed by co-transfecting β-TrCP1 or -2 with Vpu, in the presence or absence of active NF-κB signaling (+/− IKKβ), harvesting at 24 h and western blotting cell lysates ( Fig. 3A and B). Levels of β-TrCP1 were depleted by an average of 52% in the presence of 2_87 Vpu (74% in the presence of IKKβ), with S3/7A showing a modest reduction. In contrast, NL4.3 Vpu caused more than a 4-fold increase in β-TrCP1 levels (2.7-fold in the presence of IKKβ). In agreement with previous studies espousing sequestration of β-TrCP through molecular mimicry by Vpu (29,30), we demonstrate that both 2_87 and NL4.3 robustly stabilize β-TrCP2, both in the presence and absence of stimulus ( Fig. 3A and B; Fig. S1B). Control experiments demonstrate that GFP levels remained unchanged under the same conditions (Fig. S1B). These data corroborate our findings in infected cells, and highlight the differential effects on β-TrCP, both in terms of β-TrCP paralogues and when comparing 2_87 and NL4.3 Vpu. Given the simultaneous reduction of β-TrCP1 and stabilization of β-TrCP2, and the involvement of a CRL-and proteasomal-dependent degradation pathway for the former (Fig. 2D), we hypothesized that Vpu might exploit an SCF β-TrCP2 for the downregulation of β-TrCP1 (40). However, siRNA knockdown of β-TrCP2 had no apparent effect on the reduced levels of β-TrCP1 in infected cells (Fig. S1C). Co-immunoprecipitation experiments were performed in order to establish whether the dichotomous effects on β-TrCP were due to obvious differences in binding ability of 2_87 and NL4.3 Vpu (Fig. 3C). In the case of β-TrCP1, MG132 was added prior to co-immunoprecipitation in order to mitigate degradation by 2_87 Vpu. Both 2_87 and NL4.3 were able to bind β-TrCP1 and -2, with no discernible difference in binding ability. As expected, the serine mutants of both Vpus were unable to bind both β-TrCPs (Fig. 3C). The binding, degradation, and sequestration patterns seen in Fig. 2A through D and 3A through C were next corroborated by confocal microscopy. As shown previously (21), β-TrCP1 and -2 are found both in the nucleus and cytosol, but with predominant nuclear localization (Fig. 3D). In the presence of 2_87 Vpu, the previously observed contrasting effects on β-TrCP1 and -2 can be seen, with levels of β-TrCP1 severely depleted, while β-TrCP2 was dramatically re-localized to the cytosol and sequestered predominantly in perinuclear regions, consistent with typical trans-Golgi network (TGN) localization of Vpu (41). β-TrCP2, on the other hand, was re-localized and sequestered by both 2_87 and NL4.3 Vpu (Fig. 3D).

Infection with HIV-1 leads to tonic activation of NF-κB, with stabilization of p105 (NFkB1) in the presence of primary Vpu
Such a direct and dramatic effect on β-TrCP led us to investigate the effect of primary Vpu on the β-TrCP substrates IκBα and p105, alongside all other major components downstream of the IKK complex. Cells were infected with viruses expressing the indica ted Vpu and NF-κB pathway components were examined by western blot (Fig. 4A). Total IKKβ, p65, and IκBα remained unaffected under all infection conditions, while phos phorylated p65 and IκBα were not detected. Unexpectedly, phosphorylated p105 was detected under all infection conditions but not in uninfected control cells, indicative of a vestige of ongoing NF-κB activation due to virus infection. Phosphorylated p105 was significantly increased in cells infected with 2_87 Vpu virus (Fig. 4A). This was also evident in both primary and Jurkat CD4 + T cells ( Fig. 4B and C). In the primary CD4 + T cells, phospho-p105 was detected in all conditions, including uninfected cells, due to NF-κB activation induced by the CD3/CD28 co-stimulation conducted prior to HIV-1 infection (Fig. 4B).
We next investigated the effect of infection under conditions of active NF-κB signal ing. Infected cells were treated with TNFα, subjected to 4 h time courses, then examined for components downstream of the IKK complex (Fig. 4D). Typical cyclical profiles were seen for phosphorylated IκBα in the uninfected cells, with phosphorylation detected at 15 min, degradation at 30-60 minutes, and renewed detection of p-IκBα at 2 and 4 h due Results thus far show that levels of phosphorylated p105 are increased in cells infected with viruses expressing 2_87 Vpu, both at steady state and following TNFα treatment. Considering that p105 has a complex role in the NF-κB pathway, both as the precursor to p50 and as a non-classical IκB, with phosphorylation important for both processes, this could have important implications. We therefore sought to clarify the effect of Vpu on p105 in transient p105 processing assays. N-terminally HA-tagged p105 constructs were transfected alongside an NF-κB stimulus (IKKβ) in the presence or absence of Vpu (Fig. 4E). An increase in p50 levels consistent with signal-induced processing was seen in response to co-expression of IKKβ, along with the appearance of an upper p105 band indicative of phosphorylated or monoubiquitinated p105. In contrast to Fig. 4A through D, however, the presence of 2_87 Vpu did not result in the stabilization of p105; rather, the upper p105 band was no longer present, and p50 levels were depleted. p105 and p50 levels in the presence of S3/7A and NL4.3 were similar to stimulated p105 processing in the absence of Vpu. The same pattern was observed when using TNFα as a stimulus (Fig. S2B). Again, increased processing of p105 to p50 in the presence of active NF-κB signaling was significantly diminished in cells co-expressing 2_87 Vpu. To demonstrate that Vpu specifically affects processing of p105 to p50, rather than acting directly on the p50 protein, identical assays were performed using HA-tagged p50 constructed specifically to test this, rather than p105. Levels of p50 were maintained under all conditions (Fig. S2B). As Vpu has a marked effect on p105 and a downstream effect on p50, we performed p50 nuclear translocation assays as an alternative to more traditional p65 translocation assays. P105 constructs were N-terminally tagged with mCherry and upon co-transfection with IKKb, p105 processing led to p50 translocation to the nucleus (Fig. 4F). As expected, the presence of 2_87 Vpu inhibited p50 translocation, while S3/7A and NL4.3 Vpus were unable to inhibit nuclear translocation, or in some cases demonstrated an intermediate phenotype (Fig. 4F). (Continued) or no Vpu (Δ Vpu), were used to infect HEK293T cells (MOI 5) for 48 h. Cells were harvested and western blotted for Hsp90 (loading control), IKKβ, phospho-p105 (Ser932), total p105, p50, p65, phospho-p65 (Ser536), total IκBα and phospho-IκBα (Ser32/Ser36). HIV-1 Gag (p55 and p24) and Vpu were blotted as controls for infection levels. Western blots for phospho-p105 were quantified, normalized to Hsp90 levels for each lane and to the uninfected sample for each experiment, and plotted as averages of at least three separate experiments (bars). Individual data points are shown as dots. Error bars represent ± SEM. Unpaired one-tailed T tests were performed for each condition, with P-values indicated by asterisks: ns, not significant (P > 0.05); * < 0.5, **< 0.05). (B), as for A, but in primary CD4+ T cells. (C) as for A, but in CD4+ Jurkat cells. Note that T cell experiments are the same as those in Fig. 2B

Primary Vpu inhibits the non-canonical NF-κB pathway
The non-canonical equivalent of p105 is p100, which becomes phosphorylated by IKKα following stimulation and is partially proteasomally processed to p52 (Fig. 1B). Analogous to p105, it also functions as an IκB and can assemble into high molecular weight complexes containing multiple NF-κB dimers (kappaBsomes) (42)(43)(44)(45). It has also been implicated in downstream signaling following activation of cytosolic DNA sensing pathways (46,47). P100 processing assays, conducted by co-transfecting N-terminallytagged p100 with NIK, demonstrated that 2_87 Vpu was able to inhibit the processing of p100 to p52 (Fig. 5A), while both S3/7A and NL4.3 were defective. Effects on both the non-canonical and canonical pathways, induced by AZD5582 (a SMAC mimetic currently under investigation as a potential latency reversal agent [48]) and TNFα, respectively, were next compared under conditions of natural infection, using NL4.3 viruses express ing either 2_87, 2_87 S3/7A or NL4.3 Vpu in HEK293T cells (Fig. 5B), CD4 + T cells (Jurkat, Fig. 5C), and HeLa cells (TZMbl, Fig. S2). Cells were infected for 42 h then treated for 6 h before harvest. In uninfected cells, the induction of the non-canonical pathway by AZD5582 was indicated by the detection of phospho-p100 and the increased processing of p100 to p52, while TNFα stimulation resulted in increased levels of p100 and the detection of phospho-p105. As previously shown in Fig. 4, infection with the 2_87 Vpu virus resulted in increased levels of p-p105 in untreated cells, and this was much more pronounced upon treatment with TNFα; for the S3/7A and NL4.3 viruses, p-p105 levels were similar to uninfected cells upon treatment. Indicative of the interdependence of the two pathways (49), AZD5582 stimulation also lead to the stabilization of p-p105, and this was higher in cells infected with the 2_87 Vpu virus. Under the same conditions, a strong and striking stabilization of p-p100 is seen, accompanied by a reduction in p100 processing, represented by the reduction of p52 levels, and increase in p100 levels, back to those seen in uninfected, unstimulated lanes. P-p100 and p100/p52 levels in cells infected with S3/7A Vpu virus were similar to uninfected cells, while NL4.3 Vpu virus gave an intermediate phenotype. As also shown in Fig. 2, β-TrCP was diminished in cells infected with 2_87-expressing virus. Overall, these results demonstrate that infection with viruses possessing optimal Vpu function causes significant dysregulation of both canonical and non-canonical NF-κB pathways.

Both β-TrCP1 and -2 must be knocked down to phenocopy the effects of 2_87 Vpu
The reduction of β-TrCP1 and the stabilization of β-TrCP2 by 2_87 Vpu prompted us to question whether there was a hierarchy in these actions for the inhibition of NF-κB by Vpu. As previously reported, Vpu specifically co-opts the SCF β-TrCP to target tetherin for degradation in the host cell (33,35). In agreement with these studies, knocking down β-TrCP1 (BTRC) expression by siRNA had no effect on the ability of 2_87 and NL4.3 Vpus to down-regulate cell surface CD4 expression, whereas β-TrCP2 (FBXW11) knockdown alone, or in combination with β-TrCP1, led to a restoration, albeit partial, of cell surface CD4 levels (Fig. 6A). In contrast, individual knockdowns had a marginal effect on the canonical and non-canonical pathways as measured by p105 and p100 phosphorylation (Fig. 6B), whereas the double knockdown of β-TrCP1 and -2 phenocopied the hallmarks of NF-κB inhibition by 2_87 Vpu, with the significant stabilization of p-p105 and p-p100, indicating that the inhibition of both β-TrCP paralogues is required for the inhibition of NF-κB by Vpu.

Determinants of Vpu required for binding and degradation of β-TrCP1
We next focused on features of Vpu that contribute to the inhibition of NF-κB activity, in particular the contribution of individual serine residues. Following initial reports demonstrating that both serines in the SGNES motif are phosphorylated (26,27,50), S52/56 or S53/57 have traditionally been mutated together, and their contribution to Vpu function is rarely investigated individually, particularly in the context of primary Vpu.
Therefore, all three serines in the cytoplasmic tail of 2_87 Vpu (Fig. 7A) were individually mutated to alanines. We found that mutating serine 57 had a greater effect on NF-κB Research Article mBio inhibitory activity than mutating serine 53, and that the S57A mutant closely resembled wild-type NL4.3 in its inhibitory profile (Fig. 7B). As reported previously for other functions of Vpu, mutating serine 65 enhanced the NF-κB inhibitory function of Vpu (51). In contrast to the 2_87 profiles, mutating either serine 52 or 56 in NL4.3 had a similar negative impact on function with no dominant effect of either serine (Fig. 7B). A previous study in which we compared the ability of 304 primary Vpus to counteract physical virus restriction by tetherin with the inhibition of tetherin-mediated NF-κB signaling revealed regions of Vpu that were specifically required for the counteraction of NF-κB activation. All residues were located in regions flanking the SGNES site. These naturally-occurring mutations were introduced into a 2_87 Vpu background and tested for their ability to inhibit MAVS-stimulated NF-κB. All mutants were compared at a dose of 10 ng, at which input 2_87 Vpu reduces NF-κB activation induced by MAVS by 90% (Fig. 1C and Fig. 7C). As shown in Fig. 7C, individually the mutants had a partial impact on NF-κB inhibitory function but none more so than the serine mutations. Mutants that affected the charge of an acidic patch C-terminal to the SGNES, G59R and E62G, had the greatest effect on function, potentially through disrupting the putative function of this region as a CK2 priming site (52). Mutating the double aspartic acid residues to alanine in this acidic patch (EE62/63AA) demonstrated a similar phenotype.
In order to determine whether differences in β-TrCP binding accounted for the results shown in Fig. 7B and C, co-IP experiments were performed with the single serine mutant Vpus. Mutating serine 57 completely abolished binding of 2_87 Vpu to both β-TrCP1 and -2, in keeping with the more potent effect of this mutation on Vpu function, whereas mutating serine 53 had no effect on binding (Fig. 7D). This result suggests hierarchical serine phosphorylation, which is consistent with CK2 requiring an acidic amino acid (aspartic or glutamic acid) or a phosphorylated serine/threonine at position +3 of the phosphorylation site (minimal consensus sequence [S/T]xx[E/D/S p ]; [52]), and agrees with original reports of preferential CK2 phosphorylation of serine 56 of NL4.3 Vpu in vitro (26). Interestingly, the co-IP experiments revealed another prominent difference between the NL4.3 and 2_87 Vpus; while the presence of serine 57 was sufficient for the binding of 2_87 Vpu to β-TrCP1 and -2, mutation of either of the serine residues in NL4.3 Vpu completely abolished binding to both β-TrCP paralogues (Fig. 7D).
We reasoned that disrupted phosphorylation of NL4.3 Vpu compared to 2_87, on either of the serines but in particular S56, might account for the requirement for both serines for binding to β-TrCP and for the resemblance of the 2_87 S57A NF-κB inhibitory profile to that of NL4.3 (Fig. 7B). To investigate this, we performed phosphate-affinity PAGE, which specifically resolves phosphorylated proteins (Fig. 7E). In agreement with previous studies (53), the profile for 2_87 Vpu showed four clearly defined phosphoryla tion states: double phosphorylation (lane 1), single phosphorylation on S57 (lane 2), single phosphorylation on S53 plus unphosphorylated (lane 3), and unphosphorylated (lane 4). Consistent with the results from co-IP experiments and with sequential phos phorylation, the phosphorylation state of the S57A mutant was partial, whereas the phosphorylation of the S53A mutant, although a smaller gel shift, was total. NL4.3 showed a similar profile, with the caveat that bands were less resolved for NL4.3 than for 2_87. Thus, serine phosphorylation differences are unlikely to account for the reduced potency of NF-κB inhibition by NL4.3 Vpu or the total inhibition of binding to β-TrCP upon mutation of either serine; it is more likely that differences in the regions flanking the SGNES contribute to the defect.

DISCUSSION
Viruses have evolved to maintain an optimal balance between exploiting cellular processes for virus replication while inhibiting those that cause obstructions. This balance is exemplified by HIV requiring NF-κB for viral transcription, while also employ ing multiple strategies to temporally inhibit NF-κB signaling to avoid the induction of inflammatory responses at specific stages in its lifecycle. It is further illustrated by the coopting of SCF β-TrCP2 by HIV-1 Vpu for the ubiquitination and subsequent degradation of its cellular targets, while inducing the degradation of β-TrCP1 to inhibit NF-κB. By inserting a potent primary Vpu (2_87) and mutants thereof into the NL4.3 provirus/GFP system we have demonstrated that Vpu both degrades and sequesters β-TrCP, and through these interactions has multi-layered consequences for the infected cell. Beyond HIV-1 Vpu, other proteins from diverse viral families contain decoy degrons that bind and disable β-TrCP, including rotavirus NSP-1 (DSGIS; [54][55][56][57][58][59]), EBV LMP1 (DSGHES; [60]) and vaccinia virus A49 (YSGNLES [37]), all of which are mechanistically distinct. Following phosphorylation of its C-terminal degron by CK2, NSP-1 binds to β-TrCP and recruits a Cul3 CRL complex via its N-terminal RING domain, triggering the poly-ubiquitination and degradation of β-TrCP (54)(55)(56)(57)(58)(59). Thus, NSP1 acts as the CRL substrate adaptor while β-TrCP becomes the substrate. A49, on the other hand, is not associated with β-TrCP degradation but acts as a transdominant decoy -binding, and sequestering β-TrCP via its phospho-serine motif (37). A49 employs a further regulatory step through the phosphorylation of its decoy degron by the IKK complex itself, thus only being activated upon triggering of the signaling cascade it subsequently inhibits (38). For EBV Lmp1, certain variants of this oncogenic protein demonstrate a biphasic activation of NF-κB, activating at moderate levels then inhibiting at high expression levels, due to dominant negative inhibition of β-TrCP through the LMP1 decoy degron (60). As demonstrated here and previously, for HIV-1 Vpu, the binding of the SGNES degron to β-TrCP has dual functionality: to act as a dominant negative decoy molecule, binding β-TrCP and preventing its participation in the NF-κB pathway (28)(29)(30), and to recruit the SCF β-TrCP ligase for the ubiquitination and subsequent degradation of its target cellular proteins, including CD4 and tetherin (23,(32)(33)(34)(35)(36). We further demonstrate that this involves inducing the degradation of the β-TrCP1 paralogue in a CRL-dependent manner, while exploiting β-TrCP2 for the recruitment of the Cul1 CRL complex. Thus, while evolving an SxxxS motif might seem a simple act of molecular mimicry adopted by multiple virus families to inhibit the NF-κB pathway, the regulation and specific mechanisms behind such inhibitory strategies are varied and complex.
To date, Vpu remains one of the few substrates of β-TrCP that distinguishes between the two paralogues. In agreement with previous studies demonstrating that degradation of tetherin requires only β-TrCP2 (33,35), but in conflict with the requirement for both β-TrCP1 and β-TrCP2 for the degradation of CD4 (32), we show that only β-TrCP2 is required for the downregulation of CD4. Despite undetectable levels of β-TrCP1 in siRNA experiments, both 2_87 and NL4.3 Vpus were able to achieve full downregulation of CD4 from the cell surface (Fig. 6). This is logical given the significant depletion of β-TrCP1 seen in cells infected with virus expressing the 2_87 Vpu ( Fig. 2; Fig. 5B and C; Fig. S2C) and the disparity between β-TrCP1 and -2 levels in the presence of 2_87 Vpu in transient assays ( Fig. 3; Fig. 7F). Thus, for the ubiquitination and degradation of Vpu's cellular targets, as for rotavirus NSP1, Vpu becomes the substrate adaptor, connecting the SCF β-TrCP2 ligase Research Article mBio machinery to CD4 and tetherin, without being degraded itself. For the β-TrCP1 depletion, it remains to be determined whether this involves an active degradation similar to that seen with NSP1, with β-TrCP1 becoming the substrate, or whether Vpu exploits a feature of β-TrCP1 regulation and turnover that differs from β-TrCP2. Of note, we have excluded the possibility that Vpu co-opts a β-TrCP2-specific SCF CRL to degrade β-TrCP, as there was no discernible difference in β-TrCP levels in cells treated with β-TrCP2 siRNA (Fig.  S1C).
The simultaneous inhibition and co-option of β-TrCP by Vpu has multi-layered consequences for the infected cell, including the inhibition of both the canonical and non-canonical NF-κB pathways, in addition to effects on other myriad targets of β-TrCP such as CDC25A and β-catenin (16,(61)(62)(63). In cells infected with viruses expressing primary Vpu proteins we demonstrate potent stabilization of both phosphorylated p100 and p105. These proteins have a complex role in the NF-κB pathway, acting both as IκBs (IκBδ and γ respectively) and precursors to mature NF-κB subunits p52 and p50. As IκBs, they are able to form high-molecular weight complexes-in the case of p100 these are known as kappaBsomes-that sequester multiple NF-κB subunits, estimated to inhibit up to 50% of the cellular NF-κB (44,45). Thus, their inhibition likely accounts for the increased potency of NF-κB inhibition by primary 2_87 Vpu, and indeed, perhaps that of other viral proteins that target β-TrCP.
Whether the inhibition of the non-canonical pathway is directly beneficial for HIV-1, or whether it is simply a side effect of inhibiting the canonical pathway, requires further investigation. Non-canonical signaling has been demonstrated to occur in a cGAS/STING-dependent pathway following the sensing of cytoplasmic DNA (46,47); therefore, the targeting of this pathway by Vpu may serve to further evade undesirable signaling events in the infected cell. Furthermore, promising latency reversal strategies that use SMAC mimetics to activate the non-canonical NF-κB pathway and re-activate integrated viral genome transcription, while avoiding the more pleiotropic effects of canonical NF-κB agonists such as PKC activators (48,64), would need to take into account the potent inhibitory effects of primary HIV-1 Vpu proteins on this pathway, as demon strated here. Humanized mouse experiments using the cell line-adapted JR-CSF Vpu, and macaque experiments using SIVmac, however, may underestimate the counteractive effect of primary Vpus.
As previously noted by us and Sauter et al. (11,31), the binding of Vpu to β-TrCP does not strictly track with the ability of Vpu to inhibit NF-κB. Primary Vpus with double serine mutations have residual NF-κB inhibitory function, despite being unable to bind β-TrCP (11,31). We extend those findings by demonstrating that only S57 is essential for binding to β-TrCP1 and -2, yet this mutant had NF-κB inhibitory activity equivalent to that of NL4.3 Vpu. Interestingly, despite no apparent differences in phosphorylation status (Fig.  7E), NL4.3 is significantly diminished for NF-κB inhibitory activity. Single serine mutations of either serine 52 or 56 completely abolishes binding of NL4.3 Vpu to β-TrCP, yet the overall ability of wild-type NL4.3 to bind β-TrCP1 and -2 does not appear impaired in co-immunoprecipitation or immunofluorescent experiments (Fig. 3C and D and Fig. 7D). The obvious region of Vpu to account for such differences is the second alpha helix of the cytoplasmic tail, where there are multiple differences between 2_87 and NL4.3 Vpu, including the lack of additional acidic residues in NL4.3, predicted to act as a CK2 prime site. Indeed, some of the natural mutations found to specifically impact NF-κB inhibition mapped to this region (31).
Early studies demonstrated constitutive phosphorylation of Vpu (26,27,50). Here, we demonstrate clear differences between NL4.3 and 2_87 Vpus in the requirement for phosphorylation on one or both serines for the binding of β-TrCP, with 2_87 Vpu only requiring the phosphorylation of S57. We further demonstrate that, while only S57 is required for 2_87 Vpu binding to β-TrCP1 and -2 and the stabilization of β-TrCP2, both serines are required for degradation of β-TrCP1 (Fig. 7F). Proteomics studies have indicated that Vpu can potentially interact with (65) and be dephosphorylated by (66) the phosphatase PP2A. Furthermore, an additional serine at position 61 (65 in 2_87 Vpu) has been shown to regulate Vpu function and lead to its proteasomal degradation via an unidentified CRL (51). The phosphorylation state of Vpu has also been predicted to determine its oligomerization status (53). Thus, further studies are required to under stand the precise phosphorylation status of all three serines in the Vpu cytoplasmic tail in infected cells, and how this may relate to the regulation of its myriad functions.
We have demonstrated that infection with HIV-1 leaves behind a tell-tale trace of NF-κB perturbation, in the form of phosphorylated p105, even in the absence of exogenous stimuli (Fig. 4 and Fig. 5B and C). In all infection conditions, including those in the absence of Vpu or presence of suboptimal Vpus, p-p105 could be detected. This rose to significantly higher levels of p-p105 in the presence of the highly active 2_87 Vpu. No such indication of signal activation was found in phospho-IκBα or p65 after 48 h of viral infection, suggestive of long-term rather than acute activation of the pathway. It is unclear what provides the initial stimulus for this activation. Potentially some level of viral sensing may be at play, or this may reflect the activity of viral proteins shown to boost NF-κB signaling such as gp41 (67) or Nef (11). Thus, while the detection of phospho-p105 reveals the NF-κB activation that occurs due to HIV infection and is exploited for viral transcription, the inhibition of the pathway mediated by Vpu is in turn apparent in the significant enrichment of p-p105. As such, p-p105 has potential as a convenient marker for NF-κB status in infected cells.
In summary, we provide a detailed view of the consequences of β-TrCP inhibition in the HIV-1-infected cell, including previously undocumented interference with the non-canonical NF-κB pathway. We underscore the importance of using Vpu proteins representative of natural infection and studied in the context of actively replicating virus.

Cells
HEK293T cells and Jurkat T cells were obtained from American Type Culture Collection (ATCC). HeLa TZMbl were obtained through the NIH HIV Reagent Program, Division of AIDS, NIAID, NIH, kindly provided by John C. Kappes. Primary CD4+ T cells were purified from freshly isolated PBMCs from healthy donors. PBMCs were isolated by density gradient using Lymphoprep (Axis-Shield) and CD4 + T cells purified by negative selection using the Dynabeads Untouched Human CD4 + T Cell Isolation kit (Invitrogen) according to the manufacturer's instructions. CD4 + T cells were activated using Human T-Activator CD3/CD28 Dynabeads (Invitrogen) according to the manufacturer's instructions and maintained in RPMI GlutaMax supplemented with 10% FCS and 30 U/mL recombinant IL-2 (Roche).

Phosphate affinity PAGE
10% polyacrylamide gels were prepared containing 50 uM Phostag (Alpha Laboratories) in the separating gel. Cell lysates were first diluted 1:10 in llaemmli buffer, and SDS-PAGE was performed using standard protocols using methanol-based transfer.

Proviral constructs
An HIV-1 proviral construct (HIV-1 NL4-3 IRES-eGFP infectious molecular clone that encodes the full length HIV-1 NL4.3 genome with the nef open reading frame augmen ted by an IRES-eGFP (kindly provided Drs. Munch, Schindler and Kirchhoff via the NIH HIV Reagent Program, Division of AIDS, NIAID, NIH [pBR43leG-nef+, cat #11349 (71)]) was used as the basis of all viruses described. This proviral genome was rendered Vpu-defective as described previously (70). To make NL4.3 IRES-eGFP with different Vpu alleles, SnaB1 and Xba1 sites were inserted 5′ and 3′, respectively, of the vpu gene. Vpus were PCR amplified with flanking SnabI and XbaI sites and inserted. After sequence confirmation, the restriction sites were reverted by site directed mutagenesis to preserve cis-acting regulation of Vpu and Env translation (72). Site-directed mutations of the serine codons to alanines at positions 52/53 and 56/57 in NL4.3 and 2_87 Vpu were performed by quick-change.

Virus production
Subconfluent HEK293T cells in 10 cm plates were co-transfected with 10 µg of proviral plasmid and 2 µg of pCMV-VSV-G plasmid using 1 mg/mL polyethyleneimine (PEI). Media was changed 6-12 h post transfection. Cell supernatant was harvested 48 h after transfection, filtered, and ultracentrifuged over 20% sucrose in PBS at 28,000 rpm for 2 h.
Pellets were resuspended in serum-free RPMI medium, aliquoted and stored at −80°C. Titres (infectious units/mL) were determined on HeLa-TZMbl reporter cells.

p105 or p100 processing assays
For p105 processing assays, subconfluent HEK293T cells plated in 24-well plates were co-transfected with 100 ng pCR3.1-HA-p105 plus 20 ng pCR3.1-IKKβ-flag or empty vector plus 50 ng pCR3.1 Vpu HA or empty vector and harvested 24 h later for western blot analyses. In the case of TNFα stimulation, cells were transfected as above but omitting IKKβ flag, and treated with 10 ng/mL TNFα 18 h after transfection. Cells were harvested at 5, 15, 30, 60, 120, 240, and 360 min after TNFα addition and analyzed by western blot. For p100 processing assays, cells were co-transfected with 100 ng pCR3.1-HA-p100 plus 100 ng pCR3.1 NIK or empty vector plus 50 ng pCR3.1 Vpu HA or empty vector and harvested 24 h later for western blot analyses.

siRNA knockdown assays
Cells were pretreated with siRNA prior to CD4 downregulation assays, TNFα or AZD5582 treatment or infection. ON-TARGETplus SMARTpool human BTRC siRNA and human FBXW11 siRNA (Dharmacon) were used to target β-TrCP1 and -2, respectively. TZMbl or HEK293T cells were reverse transfected with 20 pmol siRNA per well of a 24-well plate using Lipofectamine RNAiMAX transfection reagent (ThermoFisher Scientific) according to the manufacturer's instruction. Twenty-four hours later, cells were trypsinized, split into three and the reverse transfection process was repeated. Twenty-four hours after the second reverse transfection, cells were transfected, treated with TNFα or AZD5582 or infected.

CD4 downregulation
Cells were pretreated with siRNA prior to CD4 downregulation assays, as detailed above. CD4 downregulation assays were performed as described previously (31). Briefly, 24 h after the second siRNA treatment, subconfluent TZMbl cells were co-transfected with 200 ng pCR3.1-GFP or empty vector control and 40 ng pCR3.1 Vpu or empty vector control in 24-well plates. Twenty-four hours after transfection, cells were harvested and stained for cell surface CD4 expression using antihuman CD4 APC (RPA-T4, eBioscience, ThermoFisher Scientific) and analyzed by flow cytometry on a BD FACSCanto II system (BD Biosciences) using FlowJo software. Cells were gated for high GFP expression and CD4 levels were determined as median fluorescence intensity in the absence of Vpu expression, with CD4 levels in the presence of Vpu expressed as a percentage of this.

Virus infection assays
HEK293T cells were infected in 24-well plates at 300,000 cells per well at an MOI of 3 or 5. Jurkat and CD4 + T cells were infected in 48-well plates at 500,000 cells per well at an MOI of 3 or 5. Forty-eight hours after infection, cells were harvested for western blot analyses. A minimum of three independent experiments were performed, as specified in figure legends. In the case of primary CD4+ T cells, experiments were performed on cells from three different donors. Western blots for multiple target proteins were performed on individual experiments and internally controlled, then compared using statistical methods, as specified in figure legends and detailed in the Statistics section below. For drug treatment of infected cells, a final concentration of 10 μM Mg132, 100 μM MLN4924, 50 nM concanamycin A or mock treatment of DMSO or water as appropriate was added to the cell culture medium 6 h before harvest.

TNFα and AZD5582 NF-κB activation assays
HEK293T or Jurkat CD4 + T cells were infected as above and treated with TNFα or AZD5582 for 6 h before harvest at 48 h post infection. For TNFα timecourse assays, cells were infected in bulk at an MOI of 3, then divided into separate wells (300,000 cells per well) for TNFα or control treatment. TNFα was added 42 h after infection, to a final concentration of 10 ng/mL, and cells were harvested at 0, 15, 30, 60, 120, and 240 min post-TNFα addition for western blot analyses.

Quantitative RT-PCR
CD4 + Jurkat T cells were infected as above for 72 h, treated with TNFα (5 ng/mL), then harvested at 0.5, 2, 8, and 24 h post-treatment. RNA was extracted from cells using a QIAGEN RNeasy kit, reverse transcribed using random hexamers, and assayed for CXCL10 and GAPDH mRNA expression by RT-qPCR, as described previously (39).

Immunofluorescence
For p50 nuclear translocation assays, subconfluent HEK293T cells plated in 24-well plates were co-transfected with 100 ng pCR3.1 CHE p105, 25 ng pCR3.1-IKKβ-flag or empty vector, and 20 ng pCR3.1 Vpu HA (2_87, 2_87 S3/7A or NL4.3) or empty vector. For β-TrCP localization assays cells were co-transfected with 150 ng of pCR3.1 GFP BTRC or 100 ng of pCR3.1 GFP FBXW11 and 20 ng of pCR3.1 Vpu HA (2_87, 2_87 S3/7A or NL4.3) or empty vector. For microscopy, glass coverslips were placed in 24-well plates and treated with 400 uL of 10% gelatin in PBS (prewarmed at 37°C to liquify) for 30 min at room temperature. To obtain optimal cell density for microscopy of individual cells, 24 h after transfection each well was trypsinized, split 1:4 to 1:7 and re-seeded onto the pretreated glass coverslips. Remaining cells were replated into 24-well plates for parallel western blot analysis as required. Cells were allowed to adhere to the glass cover slips overnight at 37°C, then fixed with 4% formaldehyde in PBS for 15 min at room temperature, washed once with PBS then with 10 mM glycine in PBS. To permeabilize, cells were treated with 0.1% Triton X-100 and 1% BSA in PBS for 15 min at room temperature, before incubation with mouse anti-HA antibody (anti-HA.11 clone 16B12, BioLegend) in 0.01% Triton X-100 in PBS for 45 min at room temperature. Cells were washed three times with 0.01% Triton X-100 in PBS, incubated with Alexa Fluor 488, 594 or 647 anti-mouse secondary antibody (Molecular Probes, Invitrogen) and washed again three times. Cover slips were mounted on slides with ProLong diamond antifade mountant with DAPI (Invitrogen) and imaged on a Nikon Eclipse Ti inverted microscope with Yokogawa CSU-X1 spinning disk unit. Image analyses were performed with NIS Elements Viewer and Fiji software.

Immunoprecipitation
Subconfluent HEK293T cells were co-transfected with 600 ng pCR3.1-β-TrCP1/BTRC-, β-TrCP2/FBXW11-HA or empty vector control, plus 500 ng pCR3.1 Vpu flag or empty vector control per well of a 6-well plate. Twenty-six to twenty-eight hours after trans fection, cells were lysed in IP buffer (50 mM Tris pH 7.5, 100 mM NaCl, 1 mM EDTA, 2 mM DTT, 0.1% Nonidet P40 substitute, supplemented with cOmplete protease inhibitor cocktail [Roche] and PhosSTOP phosphatase inhibitor tablets [Roche]), incubated for 10 mins on ice, sonicated and centrifuged for 5 mins at 13,000 rpm at 4°C. Lysates were incubated with mouse anti-HA antibody (anti-HA.11 clone 16B12, BioLegend) or rabbit anti-Flag antibody (F7425, Sigma) for 1 h at 4°C with rotation. Sixty microliters of washed protein G agarose beads were added to each sample and incubated at 4°C with rotation overnight. Beads were then washed with IP buffer and resuspended in Laemmli buffer for western blot analysis. In the case of β-TrCP1 (BTRC)/Vpu CoIPs, cells were treated with 10 µM MG132 for 6 h before harvest, to avoid degradation of β-TrCP1 by Vpu.

Statistics
Statistical analyses were performed in Graphpad Prism v 9. Unless stated otherwise, all graphs show means from at least three independent experiments with errors bars indicating ± SD. Transient NF-κB reporter assays and CD4 downregulation assays with siRNA treatment were analyzed using two-way ANOVA with multiple comparisons and mixed-effects analyses. Western blot intensities were calculated by first normalizing to the Hsp90 loading control for each lane, and calculating the percentage band inten sity relative to the uninfected control. For phosphorylated targets, bands were further normalized within each gel to a positive control band containing stimulated, uninfected cell lysate (p-IκBα, p-p105 or p-p100). Normalized values from at least three independent experiments were then compared using unpaired one-tailed T tests (p-p105) or unpaired two-tailed T tests (β-TrCP). P-value > 0.1 (ns), < 0.1 (*), < 0.01 (**), < 0.001 (***), < 0.0001 (****).

ETHICS APPROVAL
Ethical approval to draw blood from healthy donors as a source for primary lymphocytes was granted by the KCL Infectious Disease BioBank Local Research Ethics Committeeapprovals SN1-100818 and SN1-160322.