Binding to DCAF1 distinguishes TASOR and SAMHD1 degradation by HIV-2 Vpx

Human Immunodeficiency viruses type 1 and 2 (HIV-1 and HIV-2) succeed to evade host immune defenses by using their viral auxiliary proteins to antagonize host restriction factors. HIV-2/SIVsmm Vpx is known for degrading SAMHD1, a factor impeding the reverse transcription. More recently, Vpx was also shown to counteract HUSH, a complex constituted of TASOR, MPP8 and periphilin, which blocks viral expression from the integrated viral DNA. In a classical ubiquitin ligase hijacking model, Vpx bridges the DCAF1 ubiquitin ligase substrate adaptor to SAMHD1, for subsequent ubiquitination and degradation. Here, we investigated whether the same mechanism is at stake for Vpx-mediated HUSH degradation. While we confirm that Vpx bridges SAMHD1 to DCAF1, we show that TASOR can interact with DCAF1 in the absence of Vpx. Nonetheless, this association was stabilized in the presence of Vpx, suggesting the existence of a ternary complex. The N-terminal PARP-like domain of TASOR is involved in DCAF1 binding, but not in Vpx binding. We also characterized a series of HIV-2 Vpx point mutants impaired in TASOR degradation, while still degrading SAMHD1. Vpx mutants ability to degrade TASOR correlated with their capacity to enhance HIV-1 minigenome expression as expected. Strikingly, several Vpx mutants impaired for TASOR degradation, but not for SAMHD1 degradation, had a reduced binding affinity for DCAF1, but not for TASOR. In macrophages, Vpx R34A-R42A and Vpx R42A-Q47A-V48A, strongly impaired in DCAF1, but not in TASOR binding, could not degrade TASOR, while being efficient in degrading SAMHD1. Altogether, our results highlight the central role of a robust Vpx-DCAF1 association to trigger TASOR degradation. We then propose a model in which Vpx interacts with both TASOR and DCAF1 to stabilize a TASOR-DCAF1 complex. Furthermore, our work identifies Vpx mutants enabling the study of HUSH restriction independently from SAMHD1 restriction in primary myeloid cells.


Introduction
Human Immunodeficiency viruses type 1 and 2 (HIV-1 and HIV-2), responsible for Acquired Immunodeficiency Syndrome (AIDS), appeared in humans after cross-species transmission of non-human primate viruses (Simian Immunodeficiency viruses or SIV). They encode for viral auxiliary proteins, which play a major role in helping the virus to evade hurdles represented by "Restriction factors" [1]. Commonly, they act as viral antagonists engaging specific ubiquitin ligases to induce the ubiquitination and subsequent degradation of the restriction factor and, in turn, enabling the virus to bypass a specific block along the viral life cycle [2]. The Cullin-RING-type class of E3 ubiquitin ligases, constituted of a central Cullin scaffold protein and a catalytic RING subunit [3], represent the major class of ubiquitin ligases hijacked by lentiviral proteins including Vpr and Vpx [2]. All extant lentiviruses encode for Vpr, while Vpx is only encoded by the HIV-2/SIVmac/SIVsmm (infecting human, macaque and sooty mangabey, respectively) and SIVrcm/mnd-2 (infecting red-capped mangabey and mandrill) lineages [4][5][6]. Both Vpr and Vpx are incorporated into virions [7,8] and present structural (3α-helices and unstructured N-and C-termini tails) and functional similarities, likely due to an ancestral Vpr gene duplication or a recombination event, which has given rise to Vpx [4][5][6]9]. To ensure their functions, both viral proteins bind to the DDB1-and Cul4A-associated factor 1 (DCAF1) substrate adaptor of the host Cullin4A-RING ubiquitin ligase [10][11][12][13][14][15][16][17][18][19]. While Vpr exhibits a wide range of cellular substrates [20][21][22][23][24][25][26][27][28], reviewed in [29], Vpx seems to target only a few pathways via DCAF1 [30][31][32][33]. At the forefront, HIV-2/SIVsmm Vpx induces the degradation of SAMHD1, relieving a block at the reverse transcription step [31,32]. SAMHD1 is a deoxynucleotide triphosphate (dNTP) triphosphohydrolase that restricts HIV replication by lowering the pool of dNTP and thereby inhibits the synthesis of viral DNA in non-dividing cells, macrophages and quiescent CD4+ T cells [34][35][36][37][38]. Vpx bridges SAMHD1 to DCAF1, leading to SAMHD1 poly-ubiquitination and subsequent degradation [31,32,39]. Intriguingly, Vpx/ Vpr proteins from divergent viruses, despite their structural homology, target entirely different domains of SAMHD1 in a species-specific manner [40][41][42]. This difference in SAMHD1 recognition is evolutionarily dynamic and is further witnessed by sites of positive selection in both N-and C-terminal domains of the host protein [9,40,43]. Crystal structures of SIVsmm Vpx, the C-terminal WD40 domain of DCAF1 and the C-terminal region of human SAMHD1 complex, and between SIVmnd-2 Vpx, the C-terminal WD40 domain of DCAF1 and the Nterminal region of mandrill SAMHD1 highlight a conserved mode of interaction of Vpx with DCAF1, as well as providing clues as to how a conserved DCAF1-Vpx module can bind different SAMHD1 from different host-species [41,42]. In addition to SAMHD1, we and others uncovered the ability of HIV-2/SIVsmm Vpx to induce the degradation of HUSH, an epigenetic complex repressing the expression of transgenes, retroelements and hundreds of cellular genes [30,33,[44][45][46][47]. The HUSH complex is constituted of three subunits, TASOR, MPP8 and Periphilin, with TASOR acting as a core member by interacting with both MPP8 and periphilin [47,48]. TASOR contains at its N-terminus part an inactive poly ADP-ribose polymerase (PARP)-like domain essential for HUSH-mediated repression [48].
By degrading HUSH, Vpx favours viral expression in a model of HIV-1 latency, which is referred to "viral reactivation" thereafter [30,33]. Altogether, Vpx induces the degradation of two antiviral proteins acting at two different steps of the viral life cycle: SAMHD1, at the reverse transcription step, and HUSH, at a post-integration step.
As of today, the molecular determinants of SIVsmm Vpx involved in SAMHD1 antagonism are well characterized. Indeed, Vpx interacts directly with SAMHD1 and DCAF1, with residues located in the α-helices 1 and 3 for SAMHD1 and several residues along Vpx for DCAF1 as described in [5,39,42]. However, the molecular details of how Vpx promotes HUSH degradation have been poorly investigated. Our previous results suggest that HUSH destabilization by Vpx occurs within a TASOR-MPP8 complex, with TASOR being the most efficiently impacted by Vpx. Several results initially led us to hypothesize that the mechanism of HUSH degradation by Vpx is likely to follow the mechanism of SAMHD1 degradation. Indeed, we showed that DCAF1 depletion or mutation of the Q76 residue in SIVsmm Vpx, critical for binding to DCAF1 [14], prevented TASOR degradation. In addition, SIVsmm Vpx Q76R interacted with TASOR. On the other hand, SIVsmm Vpx Q47A-V48A was identified as a TASOR-binding-deficient mutant that failed to induce TASOR degradation, but still degraded SAMHD1. Altogether these results suggested a "ubiquitin ligase hijacking model", in which Vpx would bridge DCAF1 to the HUSH complex. Here we have challenged this model by performing additional interaction and functional experiments and by characterizing new HIV-2 Vpx mutants. Our results demonstrate that TASOR binds DCAF1 independently from Vpx, while Vpx reinforces the strength of interaction between the two cellular proteins. Furthermore, the ability of Vpx to interact with DCAF1 correlates with TASOR degradation by Vpx, while degradation of SAMHD1 is efficient even when binding of Vpx to DCAF1 is strongly impaired. These results led us to speculate that a novel rearrangement of the HUSH-DCAF1 interaction by Vpx is necessary for HUSH degradation.

TASOR interacts with DCAF1 and this association is stabilized by HIV-2 Vpx
The study here is dedicated to HIV-2 Vpx from the Ghana-1 strain [49]. In agreement with our previous results obtained using SIVsmm Vpx proteins [30], immunoprecipitation of HAtagged Vpx from HIV-2 with anti-HA antibodies pulled-down TASOR-Flag (Fig 1A, lane 5). The DCAF1 binding-deficient Vpx Q76R mutant has kept its ability to bind TASOR (Fig 1A,  lane 6). Of note, HIV-2 HA-Vpx was detected under two forms of different molecular weight, at about 15-kDa and 30-kDa (Fig 1A or 1B, input). The 15-kDa form corresponds to the expected size of monomeric Vpx, while the 30-kDa form is likely a dimer of the protein. This 30-kDa form is non-denaturable, neither by DTT or β-mercaptoethanol reducing agents nor by 8M Urea-containing lysis buffer as previously characterized in [50] (S1A Fig). Both the 15-kDa and 30-kDa forms were recognized by anti-Vpx antibodies (S1B Fig). The 30-kDa Vpx form was better immunoprecipitated with TASOR than the monomeric form following anti-Flag-mediated immunoprecipitation of TASOR-Flag (Fig 1B, lane 6). Nonetheless, the interaction was confirmed between Flag-tagged Vpx, which is unable to form an apparent dimer, and HA-TASOR, suggesting that monomeric Vpx binds TASOR efficiently (S1C Fig). To our surprise, endogenous DCAF1 was immunoprecipitated with TASOR-Flag in the absence of Vpx (Fig 1B, lane 5), though this TASOR-DCAF1 interaction was reinforced in the presence of Vpx (Fig 1B, compare lanes 5 and 6, DCAF1 panel). In contrast, DCAF1 was immunoprecipitated with SAMHD1-Flag only in the presence of Vpx as expected [39] (Fig 1C, lane 6). Interaction between TASOR and DCAF1 isoform 1 or DCAF1 isoform 3, which lacks the chromodomain present in isoform 1 [51], was confirmed following overexpression of the proteins (S2A Fig). Thus, while Vpx bridges SAMHD1 to DCAF1, TASOR interacts with DCAF1 independently from Vpx. This result raised the possibility that DCAF1, as an adaptor of a ubiquitin ligase complex, could be involved in TASOR turnover. Nonetheless, in our conditions, we could not observe any endogenous TASOR levels modulation upon DCAF1 depletion (S2B Fig). We

PLOS PATHOGENS
Mechanism of TASOR degradation by HIV-2 Vpx further wondered whether binding of Vpx to TASOR could depend on a pre-existing TASOR-DCAF1 complex. The interaction between TASOR-Flag and HA-Vpx or between HA-TASOR and Flag-Vpx was unaffected by inhibition of DCAF1 expression (Figs 1D, compare lanes 10 to 8 and S2C). These results, together with the finding of an association between Vpx Q76R and TASOR, suggest that Vpx can interact with TASOR independently from DCAF1.
Because Vpx stabilizes the TASOR-DCAF1 interaction, we expected Vpx and DCAF1 to bind distinct domains of TASOR. Fig 2A depicts the TASOR constructs we tested for their interaction with DCAF1: 1-1670 and 1-1512 represent two distinct TASOR isoforms that differ only in their C-terminal part, with 1-1512 being the one we have used in this manuscript.

PLOS PATHOGENS
Mechanism of TASOR degradation by HIV-2 Vpx DCAF1 interacted preferentially with the 930 first amino acids of TASOR (Fig 2B, compare lanes 9 and 10). In contrast to DCAF1, Vpx perfectly binds the 630-1512 C-terminal fragment of TASOR (Fig 2C). The N-terminus part of TASOR contains the PARP-like domain with no catalytic activity, but required to maintain transgene repression by HUSH [48]. Depletion of this domain reduced binding of DCAF1 to TASOR both in the absence or in the presence of Vpx (Fig 2D, compare lanes 9 to 8 and 12 to 11). In contrast, TASOR-ΔPARP was still able to bind equivalently to MPP8 in the absence of Vpx in agreement with results from Douse et al. [48] (Fig 2D, lanes 8,9), supporting the fact that the binding affinity of DCAF1 to TASOR-ΔPARP is weaker. Both TASOR WT and TASOR-ΔPARP over-expressions tend to increase MPP8 protein levels ( Fig 2D, lanes 1-3). However, upon Vpx expression, endogenous MPP8 accumulation could only be observed with the PARP-like truncated version of TASOR and not with TASOR WT, suggesting that TASOR N-terminus part is necessary for Vpx-mediated HUSH degradation (Fig 2D, lanes 4-6 compared to lanes 1-3). Thereby, these results suggest that DCAF1 binds the N-terminus part of TASOR. Of importance, Vpx binds TASOR-ΔPARP as well as TASOR-WT ( Fig 2D, lanes 11,12) suggesting that DCAF1 and Vpx interact with different regions in TASOR. However, the TASOR-DCAF1 interaction was no more stabilized by Vpx when TASOR PARP-like domain was removed ( Fig 2D, compare lanes 11 to 8, 12 to 9, quantification under the Figure).
Altogether, our results suggest the existence of a ternary complex between TASOR, DCAF1 and Vpx with independent possible binary interactions between the three partners.

Vpx Q47AV48A and Vpx V48A have an unexpected defect in DCAF1 binding
To further study the link between TASOR binding, TASOR degradation and increase of HIV-1 minigenome expression, we investigated the impact of the individual Q47A and V48A mutations in HIV-2 Vpx, keeping in mind that we previously found SIVsmm Vpx Q47A-V48A to be impaired in TASOR binding and degradation [30]. Vpx proteins (WT or mutants) were incorporated into Viral-Like-Particles (VLPs). Their incorporations were checked and adjusted in order to deliver about the same quantity of viral proteins into J-Lat A1 T cells [52] (S3A Fig). These cells harbor a latent HIV-1-LTR-Tat-IRES-GFP-LTR cassette, which GFP expression is being cooperatively increased upon Vpx-mediated TASOR degradation along with Tumor Necrosis Factor alpha treatment (TNFα) [30]. As in SIVsmm Vpx, mutation of both Q47 and V48 residues to alanines (QV mutant) in HIV-2 Vpx, strongly, but not fully, impaired TASOR degradation and viral reactivation (Fig 3A and 3B). Because J-Lat T cells do not express SAMHD1, we analyzed SAMHD1 degradation in the THP-1 myeloid cell line. The QV mutant was able to induce SAMHD1 degradation ( Fig 3C). Vpx V48A, alike Vpx QV, was impaired in both TASOR degradation and viral reactivation, but not in SAMHD1 degradation. In contrast, Vpx Q47A could degrade both TASOR and SAMHD1, and was able to increase the GFP expression derived from the integrated HIV-1 minigenome (Fig 3A, 3B and 3C). Therefore, the V48A mutation is at stake in the loss of HUSH antagonism by the Vpx Q47A-V48A mutant.
Interaction experiments were further undertaken with proteins expressed from transfected vectors. Of note, the three HA-tagged Vpx mutants, Q47A, V48A and QV, were detected as monomers, but did not produce the 30-kDa form (Fig 3D, input, left). Following an anti-HA (Vpx) immunoprecipitation, none of them were found in association with TASOR, in contrast to WT Vpx (Fig 3D, TASOR-Flag panel). As this was inconsistent with Vpx Q47A being functional as well as WT Vpx (Fig 3A and 3B), we performed the reverse co-immunoprecipitation experiment pulling down first TASOR-Flag ( Fig 3E). In these conditions, we could detect an interaction between TASOR and monomeric Vpx proteins, including Vpx Q47A ( Fig 3E). However, Vpx V48A and the QV mutant did not promote TASOR-DCAF1 interaction as well as Vpx Q47A, in agreement with the functional data ( Fig 3E, DCAF1 panel and quantification under the Figure). Thus, we also tested the ability of these Vpx mutants to bind DCAF1. In our previous study [30], we assumed Vpx Q47A-V48A would bind DCAF1 as it was able to degrade SAMHD1. However, to our surprise, interaction affinity of Vpx Q47A-V48A or Vpx V48A to DCAF1 was strongly reduced, while Vpx Q47A could bind DCAF1 as WT Vpx ). Hence, the loss of activity of Vpx Q47A-V48A or Vpx V48A could result from lower binding affinity to DCAF1 and not necessarily from a loss of binding to TASOR.

Integrity of a set of Vpx exposed residues is required for HUSH antagonism
In an attempt to map Vpx residues important for TASOR binding, we put forward the hypothesis that these residues would not be in contact with DCAF1 or SAMHD1 (see Fig 4A for such contacts). SIVsmm Vpx lacking the C-terminal poly-proline tail is still able to assemble into a ternary complex with SAMHD1 and DCAF1 and to induce SAMHD1 ubiquitination [39,53]. Thus, we first focused on this flexible C-terminal tail but found that the apparent dimeric Vpx, deleted of this tail, could still interact with TASOR (S5 Fig). We then paid attention to exposed residues alike Vpx R42, which is present in most Vpx from the HIV-2/SIVsmm lineage and Vpr proteins from the SIVagm lineage able to counteract HUSH, but not in some lineages defective for HUSH degradation [30]. When looking at the published crystal structure of the complex between SIVsmm Vpx, the C-terminus domains of DCAF1 and SAMHD1, the residue R42 appeared accessible, i.e. in contact neither with SAMHD1 nor with DCAF1, and located in a charged area with several Arginine and Acid glutamic residues (Fig 4A, based on [42]). Therefore, we decided to substitute R42 and these other charged residues, namely E30, R34 from α-helix 1; E43, R51, R54 and D58 from α -helix 2. In this second helix, we also mutated other exposed residues alike V37, N38, F46, W49, Q50, and did modify the C89 and L90 residues in the C-terminal tail ( Fig 4A). All mutants were tested for TASOR and SAMHD1 degradation in J-Lat A1 T and THP-1 cells (Fig 4B and 4C). The quantity of VLP was adjusted to transduce the same quantity of Vpx proteins (S3A Fig). Some Vpx mutants were impaired in the degradation of both TASOR and SAMHD1 (W49A, C89A, D58A) and were not further considered. Among mutations that do not impact SAMHD1 degradation, some (R42A, V37A, N38A or L90A and R42A-QV (RQV) triple mutant) strongly impaired TASOR degradation and reactivation in J-Lat A1 T cells, while others (R34A, E43A, F46A, R51A) only reduced TASOR degradation and viral reactivation. Other Vpx mutants had no impact (E30A, Q50A, R54A). Altogether, our results suggest that the integrity of a set of Vpx exposed residues (residues in red on Fig 4A) is important for HUSH antagonism. Importantly, a good correlation was obtained between Vpx ability to degrade TASOR and Vpx-mediated reactivation in J-Lat A1 T cells (Fig 4B).

Vpx mutants defective for TASOR but not SAMHD1 degradation show a defect in DCAF1 binding
Strikingly, all the Vpx mutants proficient for SAMHD1 degradation, but partially or totally deficient for TASOR degradation, were able to bind TASOR (S6A and S6B Fig), except monomeric V48A, as shown in Fig 3D, and RQV mutants, when pulling down first HA-Vpx. In contrast, we noticed that some of them, alike R34A, R42A, R51A, bound less efficiently to DCAF1 (S6A Fig). The reduced binding affinity of Vpx R34A and Vpx R42A for DCAF1 was reproducible ( Fig 5A and 5B). The double mutant Vpx R34A-R42A (RR) showed a dramatic decrease of its affinity for DCAF1, while it was still interacting with TASOR ( Fig 5A, lane 12 and 5B). Again, similar results were obtained with SIVsmm Vpx mutated on R42 residue (S4A Fig, lane 11) and with HIV-2 Flag-Vpx mutants (S4B Fig, TASOR degradation, S4C, interaction with DCAF1, lanes 10-13, and S4D, interaction with TASOR, lanes 10-13). In addition, the interaction between TASOR-Flag and DCAF1 was not stabilized by Vpx RR or Vpx R42A (Fig 5C, lanes 11 and 12, DCAF1 panel). In turn, Vpx R42A and Vpx RR could neither degrade TASOR, nor reactivate HIV-1 in J-Lat-A1 cells, while they could degrade SAMHD1 in THP-1

PLOS PATHOGENS
Mechanism of TASOR degradation by HIV-2 Vpx cells, though a little less efficiently for the RR mutant (Figs 5D, 5E, 5F and S3B for VLPs incorporation). Of note, Vpx R34A showed a reduced binding to DCAF1, but still retained some ability to induce TASOR degradation and HIV-1 reactivation. Altogether, these observations suggest the necessity of a strong binding affinity between Vpx and DCAF1 to stabilize the interaction between TASOR and DCAF1 and to induce TASOR degradation. Such strong binding would not be as much requested for SAMHD1 degradation.

Vpx RR and Vpx RQV degrade SAMHD1 but not TASOR in macrophages
Up to now, TASOR degradation was analyzed in the J-Lat A1 T-cell line and SAMHD1 degradation in the THP-1 myeloid cell line. Because degradation of one protein could impact degradation of the other, we questioned the phenotype of Vpx mutants in primary macrophages, in which SAMHD1 restriction operates. In macrophages, Vpx-mediated SAMHD1 depletion was very efficient: SAMHD1 was still not detected 7 days after Vpx delivery (Fig 6A). In contrast, TASOR protein levels reappeared at day 1 or day 2 following Vpx addition, depending on the efficacy of Vpx delivery by VLPs (Figs 6A and S3C for VLPs incorporation). In consequence, SAMHD1 and TASOR degradation by Vpx and mutants were monitored between 0 and 24 hours following Vpx addition. While all mutants efficiently induced SAMHD1 degradation, differences were observed regarding TASOR degradation. Vpx QV, Vpx Q47A, Vpx V48A and Vpx R34A were all able to degrade TASOR in macrophages, even if degradation was slightly less efficient with Vpx V48A and Vpx Q47A-V48A for some donors (Figs 6B and S7A). Vpx R42A could also induce TASOR degradation, but less efficiently (Figs 6B and S7A). Only Vpx RR and Vpx RQV were strongly impaired in TASOR degradation (Figs 6B and S7A). As a control, the Vpr protein from Vervet African Green monkey SIV induced TASOR but not SAMHD1 degradation as expected (S7B Fig) [9,30,33]. Altogether, two Vpx mutants (Vpx RR and Vpx RQV), characterized by a decreased affinity for DCAF1 binding, are no more able to degrade TASOR but are proficient for SAMHD1 degradation in macrophages.

Discussion
The mechanism of Vpx-mediated degradation of HUSH relies on the use of the DCAF1 ubiquitin ligase adaptor suggesting the existence of a classical ubiquitin ligase hijacking model, in which Vpx would bridge HUSH to DCAF1, as it is the case for SAMHD1. Nonetheless, our findings here suggest that HUSH and SAMHD1 mechanisms present notable differences. Indeed, firstly, TASOR can interact with DCAF1, thanks to its PARP-like domain located in its N-terminal part, in the absence of Vpx, while SAMHD1 interacts with DCAF1 only in the presence of Vpx; secondly, the Vpx-mediated degradation of TASOR is less efficient than that of SAMHD1 (in agreement with results from our SILAC screen published in [30]); thirdly, an The integrity of a set of Vpx exposed residues from α-helix 1 and 2 and the C-ter tail is required for HUSH antagonism. (A) The representation of the crystal structure of CtD-huDCAF1/Vpx SIVsmm/CtD-huSAMDH1 ternary complex, resolved by Schwefel et al in 2014 [42] (PDB: 4CC9), has been adapted here to highlight Vpx residues that we have tested in this study regarding SAMHD1 and TASOR degradations. Top: sequence alignment of HIV-2.Gh1 Vpx and SIVsmm Vpx. Our study is dedicated to HIV-2 Vpx, while the structure was obtained with SIVsmm Vpx. The two sequences share 79.81% identity and 92% similarity. Green and Yellow marks indicate residues of SIVsmm Vpx involved in DCAF1 and SAMDH1 binding respectively, according to Schwefel et al. [42] Stars: interaction mediated by the lateral chain. Dot: Interaction mediated by the principal chain. The integrity of residues shown in red is important for HUSH degradation, while this is not the case for residues in blue. Residues in bold dark are important for HUSH and SAMDH1 degradation.

PLOS PATHOGENS
Mechanism of TASOR degradation by HIV-2 Vpx apparently weaker interaction between Vpx and DCAF1 has no impact on the efficiency of SAMHD1 degradation, while it seems critical to stabilize the TASOR-DCAF1 interaction and thus the degradation of TASOR (model Fig 7).

TASOR interaction with DCAF1 in the absence of Vpx
DCAF1/VprBP has been mainly studied as a component of an E3 ubiquitin ligase machinery playing a role in various cellular processes [54]. We showed that DCAF1 does not regulate However, it could be that DCAF1 controls TASOR turnover in a specific cellular context, for instance in a specific window along cell cycle progression or upon DNA damage. Further investigations are required to determine whether DCAF1 could serve as an ubiquitin ligase adaptor for TASOR degradation in physiological conditions. Interestingly DCAF1 has been shown to negatively regulate transcription and to help the formation of repressive chromatin by binding histone H3 tails protruding from nucleosomes [55]. Moreover, DCAF1 was shown to possess an intrinsic protein kinase activity and is capable of phosphorylating histone H2A on threonine 120 (H2AT120p) in a nucleosomal context [56]. A role of DCAF1 in gene expression has also been uncovered with the discovery of DCAF1 working in conjunction with the Enhancer of Zeste homolog EZH2, a histone methyl transferase associated with transcriptional repression [51]. Therefore, it would be interesting to test the possibility that TASOR works with DCAF1 to repress gene expression. Of note, we found some of the genes regulated by DCAF1 in Kim et al. [56] to be upregulated following TASOR depletion in an RNA-seq analysis (unpublished results), supporting the idea of a possible repressive activity of TASOR and DCAF1 on common genes. One may also wonder

PLOS PATHOGENS
Mechanism of TASOR degradation by HIV-2 Vpx whether DCAF1, together with TASOR, could regulate HIV transcription. While DCAF1 is well-known as the ubiquitin ligase adaptor hijacked by Vpr and Vpx, a direct role of DCAF1 in viral transcription has not been investigated yet. In addition to a potential role in transcriptional repression, a cooperation between TASOR and DCAF1 might be at stake in the response to DNA damage. Indeed, DCAF1 interacts with Damage specific DNA Binding protein 1 (DDB1), which is found in complex with the PARP-domain containing PARP1 protein, a sensor of DNA damage (reviewed in [57]). The HUSH complex playing an important role in the epigenetic repression of integrated HIV and recently integrated retroelements, it would be of interest to investigate a role of TASOR in controlling expression of genes near DNA breaks.

Vpx-mediated TASOR degradation: Binding to HUSH
We were able to highlight several positions in Vpx that are important for HUSH, but not for SAMHD1 antagonism, mainly in α-helices 1 and 2 and in the C-terminal tail of the viral protein. The integrity of several of these residues appears to be important for DCAF1 binding, In the absence of Vpx, TASOR can be found in association with DCAF1. TASOR PARP-like domain is involved in this interaction. By co-binding to TASOR and to DCAF1, Vpx WT would create another indirect contacting point between TASOR and DCAF1, increasing the interaction affinity between TASOR and DCAF1. Structural conformation changes would trigger the degradation of TASOR by the whole ubiquitin ligase complex and TASOR poly-ubiquitination. Vpx RR or Vpx RQV mutants bind TASOR but present a lower affinity for DCAF1, which is represented by a black line; in turn, the interaction affinity between TASOR and DCAF1 is not sufficient to orientate the whole ubiquitin ligase complex in order to trigger TASOR degradation. Of note, whether TASOR-DCAF1 and TASOR-Vpx interactions are direct or not is unknown. (B) SAMHD1 antagonism. SAMDH1 does not interact with DCAF1 in the absence of Vpx. Vpx bridges the two cellular proteins, allowing the recruitment of the whole ubiquitin ligase complex, SAMHD1 ubiquitination and degradation. Vpx RR or RQV are both able to induce SAMHD1 degradation, despite their apparent low binding affinity for DCAF1. https://doi.org/10.1371/journal.ppat.1009609.g007

Mechanism of TASOR degradation by HIV-2 Vpx
whereas none of them seem required for TASOR binding. This result, together with the ability of TASOR to interact with DCAF1, led us to question whether the interaction of Vpx with DCAF1 could by itself trigger HUSH degradation. However, we do not favor this hypothesis because the interaction between Vpx and TASOR does not depend on DCAF1. Indeed, on the one hand, the Vpx Q76R mutant deficient in binding to DCAF1 is able to interact with TASOR and on the other hand, Vpx interacts with TASOR even when DCAF1 is depleted. Altogether, the Vpx viral determinants involved in HUSH binding remain to be discovered. Furthermore, whether this interaction is direct needs to be investigated.

Vpx-mediated TASOR degradation: Binding to DCAF1
Our results show that DCAF1 is better immunoprecipitated with TASOR in the presence of Vpx. Is it that Vpx induces a conformational rearrangement of the TASOR-DCAF1 complex allowing TASOR to better interact with DCAF1? Does Vpx create new contact points between a preexisting DCAF1-TASOR complex, modifying TASOR/DCAF1 complex conformation allowing TASOR targeting by the ubiquitination complex? Does Vpx bring more DCAF1 in the vicinity of TASOR? In other words, we do not yet know whether Vpx uses the DCAF1 molecule already in association with TASOR or whether Vpx reprograms a new DCAF1 molecule in order to promote TASOR degradation. Nonetheless, we rather favor the hypothesis that Vpx helps TASOR to interact better with DCAF1, since the interaction between TASOR-ΔPARP and DCAF1 is no more stabilized in the presence of Vpx, while TASOR-ΔPARP binds properly to Vpx (model Fig 7). In turn, we speculate that a conformational change could lead to efficient ubiquitination of TASOR, and thus to its efficient degradation by the proteasome. Strikingly, several Vpx mutants impaired for HUSH antagonism had a defect in DCAF1 binding, while none of them were impaired in TASOR binding. Namely, Vpx R42A, Vpx RR and Vpx V48A show reduced binding affinity to DCAF1, but not to TASOR, while being impaired in TASOR degradation in the T-cell line. Vpx R34A presents a moderate phenotype, with reduced affinity for DCAF1 like Vpx R42A, but not as defective in TASOR degradation. Consistently, Vpx R42A and Vpx RR do not stabilize the interaction between DCAF1 and TASOR, whereas Vpx R34A can still do so. Thus, determinants other than DCAF1 binding could be at stake to explain loss of HUSH antagonism, such as a suitable gap between TASOR, DCAF1-associated E3 ubiquitin ligase and the E2 ubiquitin transferase enzyme. Alternatively, unknown components could contribute to Vpx-mediated degradation of TASOR. Further structural analysis by Cryo-Electro Microscopy of the ubiquitin ligase complex would be necessary to understand the positioning of TASOR and DCAF1.
Importantly, the phenotype of Vpx mutants is slightly different in macrophages, in which Vpx R34A, Vpx R42A and Vpx V48A still induce TASOR degradation. Only Vpx R34A-R42A and Vpx RQV mutants, which are strongly impaired in DCAF1 binding, do not efficiently degrade TASOR in macrophages. We wonder whether this could result from a lower expression of TASOR in macrophages compared to T cells, with Vpx being the limiting factor in T cells to remove all DCAF1-bound TASOR, in line with a stoechiometric mechanism. In contrast, SAMHD1 is perfectly degraded in macrophages irrespective of whether Vpx binds efficiently or not DCAF1, in line with a catalytic mechanism. Alternatively, the different abilities of some Vpx mutants to induce TASOR degradation in macrophages, but not in T cells, may also rely on the need or use of cell-specific host factors. The ability of some Vpx mutant to induce SAMHD1 degradation, while binding to DCAF1 was severely impaired, was also very intriguing. This reminds us of SIVdeb Vpx ability to induce SAMHD1 degradation without apparent binding to DCAF1 [58]. Whether an alternative ubiquitin ligase could be used by Vpx to induce SAMHD1 degradation in some specific circumstances remains a possibility.

Studying the impact of HIV infection in myeloid cells
The use of Vpx mutants capable of degrading SAMHD1, but not HUSH, could be useful for future studies to investigate the impact of HIV infection in myeloid lineages. Indeed, to date, several studies have used Vpx to overcome the reverse transcription blockage in macrophages or dendritic cells to efficiently infect these cells and study the impact of the infection on the cellular landscape [59,60]. One may wonder whether the reported effects are actually the result of the infection or whether they could also result from HUSH degradation. This could be particularly true when studying the modification of the chromatin environment or HIV-induced innate sensing, as HUSH could interfere with both pathways. Indeed, it has been proposed that the regulation of LINE-1 by HUSH serves as a gatekeeper of type 1 interferon signaling, which, when deregulated, could lead to autoinflammatory diseases [61]. On the other hand, we have seen that TASOR protein levels return rapidly after degradation by Vpx, which may reduce the side effects of HUSH depletion. Interestingly, SAMHD1 depletion by Vpx is longlasting in macrophages compared to HUSH depletion. This difference might reflect the different outcomes resulting from the antagonism of these two restriction factors.

Restriction of retroviruses along evolution
The restriction of retroviruses by host proteins underlines the long co-evolution history between hosts and viruses. On the viral side, residues involved in the binding of the same substrate often differ between different lentiviral lineages, consistent with the molecular arms-race between hosts and viruses. In future studies, we will question whether differences in DCAF1 binding could also be demonstrated between viral proteins from different lineages, which could have some impact on HUSH, but not SAMHD1, antagonism. Understanding virus-host interaction mechanisms is important to better understand viral pathogenesis and to propose therapeutic strategies that could target restriction factors. Exploiting the activity of HUSH could be advantageous in different strategies, depending on whether the objective is to enhance or lock virus expression.

Cell culture
Cell lines were regularly tested for mycoplasma contamination: contaminated cells were discarded to perform experiments only with non-contaminated cells. Cells were cultured in media from GIBCO: DMEM (HeLa, 293FT) or RPMI (THP-1, J-Lat), containing 10% heatinactivated fetal bovine serum (FBS, Eurobio), 1,000 units ml −1 penicillin, 1,000 μ g ml −1 streptomycin and 2 mM glutamine (RPMI only) (Life Technologies). Cells were checked permanently according to morphology and functional features (SAMHD1 expression for THP-1 cells; no adherence and low GFP expression for J-Lat A1 T cells, morphology for 293FT and HeLa cells). 293FT cells, optimized from VLP production, were a gift from N. Manel. J-Lat A1 T cells were a gift from E. Verdin. All overexpression experiments were performed with the help of CaCl2 and HBSS (51558-Merck) siRNA treatment siRNA transfections were performed with DharmaFECT1 (Dharmacon, GE Lifesciences). The final concentration for all siRNA was 40nM. The following siRNA was used: siDCAF1: GGAGGGAAUUGUCGAGAAU (Dharmacon). The non-targeting control siRNAs (MIS-SION siRNA Universal Negative Control #1, SIC001) were purchased from Sigma Aldrich.

Isolation of primary cells
PBMCs from the blood of anonymous donors (obtained in accordance with the ethical guidelines of the Institut Cochin, Paris and Etablissement Français du Sang) were isolated by Ficoll (GE Healthcare) density-gradient separation. Monocytes were isolated by positive selection with CD14 magnetic MicroBeads (Miltenyi Biotec). Monocyte-derived macrophages (MDMs) were obtained by 7 days stimulation with 20 ng.ml −1 macrophage colony-stimulating factor (M-CSF) and 10 ng.ml −1 granulocyte-macrophage colony-stimulating factor (GM-CSF) (Miltenyi Biotec).

Virus-Like Particle (VLP) production and transduction
VLPs were produced in 293FT cells by cotransfection of envelope and packaging vectors by the calcium-phosphate precipitation method. 3.10 6 cells were plated the day prior transfection in 10 cm culture dishes. 3μg of VSV-G plasmid, 8μg of SIV3+ ΔVprΔVpx packaging vector (a gift from N. Landau described in [62]) and 8μg of pAS1B-HA-Vpx (WT or mutants) or pAS1b-HA (for empty VLP) or pAS1b-HA-VprSIVagm.ver9063 were then transfected. Cell culture medium was collected 72h after transfection and filtered through 0.45 μm pore filters. VLPs were concentrated by sucrose gradient and ultracentrifugation (1h30 at 100 000g). The quality of VLP production and Vpx incorporation was analyzed and quantified by revelation of HA-Vpx and HIV-2 capsid (p27) levels by Western blot. VLP volumes were adjusted in order to transduce the same quantity of Vpx WT and mutants onto cells. J-Lat A1 suspension cells were transduced with VLPs 6h in reduced medium prior to mock or overnight TNF-α (1 ng ml −1 ) treatment.

Flow cytometry analysis
Cells were collected and resuspended in PBS-EDTA (0.5mM). Data were collected and analyzed with a BD Accuri C6 cytometer and software v100.264.21. At least 10,000 events in P1

PLOS PATHOGENS
Mechanism of TASOR degradation by HIV-2 Vpx were collected, the GFP-positive population was determined using untreated J-Lat A1 T cells according to the low percentage of GFP expressing cells. The same gate was maintained for all conditions. Analysis was performed on the whole GFP-positive population.

Mechanism of TASOR degradation by HIV-2 Vpx
WT was co-expressed with HA-TASOR short isoform in HeLa cells, then an anti-Flag immunoprecipitation was performed. (TIF)

S2 Fig. Interactions between TASOR and DCAF1 and between TASOR and Vpx. (A)
TASOR interacts with the two isoforms of DCAF1 in an overexpression system. HA-TASOR short isoform was co-expressed with Myc-DCAF1 isoform 1 (iso1) or isoform 3 (iso3) in HeLa cells, then an anti-HA immunoprecipitation was performed. ( � ) The Myc-DCAF1 band is detected in the HA-TASOR panel. (B) Endogenous TASOR protein level is not affected following DCAF1 depletion. HeLa cells were transfected with 40nM of siCTL (-) or siDCAF1 (+) and cells were harvested at 48h and 72h. (C) Flag-Vpx interacts with HA-TASOR in absence of DCAF1. HeLa cells were treated with siRNA CTL or siRNA DCAF1. After 24h, Flag-Vpx WT was co-expressed with HA-TASOR for 48h, then an anti-Flag immunoprecipitation was performed. In each panel, the indicated proteins were revealed by western blot.