HIV-2 and SIVmac Accessory Virulence Factor Vpx Down-regulates SAMHD1 Enzyme Catalysis Prior to Proteasome-dependent Degradation*

Background: SAMHD1 is counteracted by the HIV-2/SIV virulence factor Vpx, which directs it for degradation. Results: Vpx in complex with DDB1-DCAF1 binds to the C terminus of SAMHD1, inhibits its catalytic activity, and dissociates SAMHD1 tetramers. Conclusion: The viral countermeasures by Vpx are manifested at multiple levels. Significance: Identifying molecular mechanisms of viral countermeasures against HIV restriction factors is important for designing antiviral therapeutics. SAMHD1, a dGTP-regulated deoxyribonucleoside triphosphate (dNTP) triphosphohydrolase, down-regulates dNTP pools in terminally differentiated and quiescent cells, thereby inhibiting HIV-1 infection at the reverse transcription step. HIV-2 and simian immunodeficiency virus (SIV) counteract this restriction via a virion-associated virulence accessory factor, Vpx (Vpr in some SIVs), which loads SAMHD1 onto CRL4-DCAF1 E3 ubiquitin ligase for polyubiquitination, programming it for proteasome-dependent degradation. However, the detailed molecular mechanisms of SAMHD1 recruitment to the E3 ligase have not been defined. Further, whether divergent, orthologous Vpx proteins, encoded by distinct HIV/SIV strains, bind SAMHD1 in a similar manner, at a molecular level, is not known. We applied surface plasmon resonance analysis to assess the requirements for and kinetics of binding between various primate SAMHD1 proteins and Vpx proteins from SIV or HIV-2 strains. Our data indicate that Vpx proteins, bound to DCAF1, interface with the C terminus of primate SAMHD1 proteins with nanomolar affinity, manifested by rapid association and slow dissociation. Further, we provide evidence that Vpx binding to SAMHD1 inhibits its catalytic activity and induces disassembly of a dGTP-dependent oligomer. Our studies reveal a previously unrecognized biochemical mechanism of Vpx-mediated SAMHD1 inhibition: direct down-modulation of its catalytic activity, mediated by the same binding event that leads to SAMHD1 recruitment to the E3 ubiquitin ligase for proteasome-dependent degradation.

RING ubiquitin ligase) E3 ubiquitin ligase, and recruits SAMHD1 to the E3 ligase for proteasome-dependent degradation (1)(2)(3)20). The discovery of an interplay between the virulence factor and these host cell factors resolved a longstanding unanswered observation; Vpx facilitates transduction of dendritic cells and macrophages and relieves the inhibition of HIV-1 infection in restricting cells (21)(22)(23)(24).
Phylogenetic tree and functional analyses of Vpx and its homolog, Vpr, combined with similar analyses of SAMHD1, indicate that these virulence accessory factors and the host restriction factor have undergone an evolutionary arms race (25)(26)(27). Interestingly, two distinct regions of SAMHD1, the SAM domain, at the N terminus, and the C terminus, distal to the dNTPase domain, display strong positive selection during primate evolution. However, the detailed molecular mechanisms by which Vpx recruits SAMHD1 to CRL4-DCAF1 are yet to be specified.
Here, we show that HIV-2 and SIVmac (isolated from Macaque monkey) Vpx recruit human and several simian SAMHD1 proteins to the CRL4-DCAF1 E3 ubiquitin ligase by interacting with the highly conserved C terminus of SAMHD1. Further, we show that a highly conserved, specific sequence motif at the Vpx N terminus is essential for efficient recruitment of SAMHD1. Real-time binding assays, using surface plasmon resonance (SPR) analysis, suggest that Vpx significantly increases the association rate and decreases the dissociation rate of SAMHD1 binding to the substrate adaptorreceptor complex of the E3 ubiquitin ligase. Surprisingly, Vpx-mediated recruitment of SAMHD1 to the substrate adaptor-receptor complex deactivates its dNTPase catalysis and subsequently disassembles dGTP-dependent tetramers to dimer and monomers. These results suggest that Vpx employs an additional viral countermeasure, prior to and independently of proteasome-dependent down-regulation of SAMHD1.

EXPERIMENTAL PROCEDURES
Cloning and Plasmid Construction-The cDNAs encoding full-length SAMHD1 (SAMHD1-FL) or residues 113-626 (⌬N-SAMHD1) or residues 1-595 (SAMHD1-⌬C) of human, rhesus, or De Brazza's SAMHD1 were cloned into pCDNA3.1 (Invitrogen) with an HA tag at its N terminus or into pET28 (EMD Biosciences) with His 6 and T7 tags at its N terminus. The cDNAs encoding SAMHD1 from other primates, including African green monkey, mandrill, red capped monkey, and sooty mangabey were also cloned into the pET28 vector. SAMHD1 full-length or the C terminus (residue 595-626) was also cloned into pET41 (EMD Biosciences) with N-terminal GST and His 6 tags. N-terminally HA-tagged full-length DCAF1 or DCAF1 C-terminal residues 1040 -1400 (DCAF1 CA ) were also cloned into pCDNA3.1 vector. The SIVmac, HIV-2 Rod9, and HIV-2 7312a Vpx (Vpx SIVmac , Vpx Rod9 , and Vpx 7312a , respectively) were cloned into pCDNA3.1 with either an HA or V5 tag at its N terminus or into pET43 vectors with a C-terminal His 6 tag, modified to include a tobacco etch virus protease site at the C terminus of a NusA fusion protein. The cDNAs encoding Vpx SIVmac and Vpx Rod9 residues 1-102 or Vpx 7312a residues 1-101 (Vpx(⌬C) SIVmac , Vpx(⌬C) Rod9 , and Vpx(⌬C) 7312a , respectively) were also cloned into the modified pET43 vectors.
The cDNAs for monkey SAMHD1 and Vpx Rod9 and Vpx 7312a were provided by M. Emerman (Fred Hutchison Cancer Center, Seattle, WA). Site-specific mutants of Vpx and SAMHD1 were prepared using QuikChange mutagenesis kits (Agilent).
Protein Expression and Purification-The various SAMHD1 and NusA-Vpx fusion proteins were expressed in Escherichia coli Rosetta 2 (DE3) cultured in Luria-Bertani medium with 0.4 mM isopropyl 1-thio-␤-D-galactopyranoside at 18°C for 16 h. Proteins were first purified using a 5-ml nickel-nitrilotriacetic acid column (GE Healthcare), and then the aggregates were removed by gel filtration column chromatography (Hi-Load Superdex200 16/60, GE Healthcare) equilibrated with a buffer containing 25 mM sodium phosphate, pH 7.5, 150 mM NaCl, 2 mM DTT, 10% glycerol, and 0.02% sodium azide. The DDB1-DCAF1 CB (DCAF1 residues 1045-1396) complex was expressed and purified from SF21 cells co-infected with recombinant bacuolviruses at a multiplicity of infection of 2 for 40 h, as described previously (20). For preparation of multiprotein complexes, DDB1-DCAF1 CB and NusA-Vpx proteins were mixed at a molar ratio of 1:3, digested with tobacco etch virus protease, and purified over an 8-ml MONO Q column (GE Healthcare) at pH 7.5, using a 0 -1 M NaCl gradient. All other proteins were prepared as described previously (20,28).
Mammalian Cell Lines, Transfection, and Immunoblotting-Human embryonic kidney cell lines (HEK293 from ATCC) were cultured in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum with antibiotics. Cells were plated on 6-cm plates, 24 h prior to transient transfection, and grown to 90 -95% confluence. HEK293 cells were transfected with 7 g of a mixture of pCNDA3.1 plasmids encoding specific cDNAs, as indicated, using LipofectAMINE 2000 (Invitrogen), according to the manufacturer's protocol. When indicated, the transfected cells were treated with 25 M MG132 (Boston Biochem) for 6 h, 42 h after transfection. Transfected cells were harvested and treated with 300 l of lysis buffer containing 25 mM sodium phosphate, pH 7.5, 300 mM NaCl, 1 mM EDTA, 0.3% Nonidet P-40, 1% Tween, 5% glycerol, and 1 mM phenylmethylsulfonyl fluoride. Proteins in the lysate were separated by 4 -20% gradient SDS-PAGE, transferred to PVDF membrane, and subsequently identified by immunoblotting. For detection of proteins, anti-HA (Covance), anti-V5 (Sigma), and anti-actin (Sigma) antibodies were used.
Analytical Size Exclusion Column Chromatography-SAMHD1 proteins (0.5 M) or their mixtures with dGTP (200 M) in 25 mM sodium phosphate, pH 7.5, 150 mM NaCl, 5% glycerol, 1 mM DTT, and 0.02% sodium azide were injected into a 24-ml analytical Superdex200 column (10 ϫ 300 mm, GE Healthcare) at a flow rate of 0.8 ml/min, equilibrated with the same buffer lacking the reducing agent. The elution profiles were recorded by monitoring fluorescence, with excitation at 282 nm and emission at 313 nm.
SPR-SPR experiments were performed using a BIAcore 2000 instrument (GE Healthcare) at 12°C. CM5 sensor chips were coated with anti-GST antibodies according to the manufacturer's protocol (GST capture kit, GE Healthcare). GST-SAMHD1 variants, at a concentration of 10 nM in running buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.005% P20, 0.02% sodium azide, 1 mM tris(2-carboxyethyl)phosphine), were injected into the flow cell at a rate of 5 l/min until the increase in response units (RU) reached 50 -60 RU. The first flow cell was used as an in-line reference with GST as ligand. The analytes, at different concentrations in the running buffer, were injected into the flow cells for 2 min at a flow rate of 30 l/min, followed by a 3-min dissociation phase. The sensor surface was regenerated by repeated injections of a regeneration buffer containing 10 mM sodium acetate, pH 5.0. For steady-state analysis, the equilibrium response of each injection at 120 s was plotted against the concentration of injected protein, using a non-linear, one-site nonspecific model (GraphPad Prism 5, GraphPad Software) to obtain the equilibrium dissociation constant (K d ) and binding response maximum (R max ). Non-equilibrium data, including the rate of association, k on , were globally fit to a predefined one-state model using BIAevaluation software (version 4.1).
Chemical Cross-linking Assays-Typically, 60 l of SAMHD1 proteins (1.6 M) were mixed with an equal volume of dGTP (100 M) in a buffer containing 25 mM sodium phosphate, pH 7.5, 50 mM NaCl, 5% glycerol, 2 mM MgCl 2 , and 0.02% sodium azide. After 60 s, the mixtures were diluted with 3.13 mM glutaraldehyde, in the same buffer, and incubated for 3 min before cross-linking reactions (50 l) were quenched with 290 mM Tris-HCl, pH 6.8 (20 l). The resulting reactions were separated by 4 -20% SDS-PAGE and visualized with Coomassie Brilliant Blue staining.
Deoxyribonucleoside Triphosphate Triphosphohydrolase (dNTPase) Assays-Assays of SAMHD1 dGTP-dependent enzymatic activities were carried out in a reaction buffer containing 20 mM Tris-HCl, pH 7.8, 50 mM NaCl, 2 mM MgCl 2 , 5% glycerol, an appropriate concentration of dNTP (0 -100 M), and 0.1 M recombinant SAMHD1. The reactions were stopped by mixing a 50-l enzymatic reaction with 20 l of 70 mM EDTA, after a specific time interval. Quenched reactions (50 l) were injected into a Capcell Pak C18 reversed-phase column (4.6 ϫ 250 mm; Phenomenex), pre-equilibrated with a buffer containing 10 mM ammonium phosphate, pH 7.8, and 4.8% methanol. Deoxyribonucleosides and dNTPs were eluted with a linear gradient (4.8 -19.2%) of methanol over 22.5 min at a flow rate of 1.5 ml/min. The amounts of products were quantified by peak integration of the absorbance trace at UV 260 nm and converted to moles based on the calibration curve of deoxyribonucleosides.
Multiangle Light Scattering-100 l of protein solution, typically at ϳ2 mg/ml, was injected into an analytical Superdex200 column (10 ϫ 300 mm; GE Healthcare) with in-line multiangle light scattering (HELEOS, Wyatt Technology), variable wavelength UV detector (Agilent 1100, Agilent Technologies), and refractive index detector (Optilab rEX, Wyatt Technology) at a flow rate of 0.5 ml/min in a buffer containing 25 mM sodium phosphate, pH 7.5, 50 -150 mM NaCl, and 0.02% sodium azide. The molecular masses of eluted protein species were determined using the ASTRA version 5.3.4 program (Wyatt Technologies).
bind the C-terminal sequences of primate SAMHD1 proteins in a similar manner.
To gain mechanistic insight into how Vpx facilitates SAMHD1 binding to DDB1-DCAF1, we developed an SPRbased assay to evaluate the rates of association and dissociation between SAMHD1 and DDB1-DCAF1 CB (residues 1045-1396) in complex with Vpx. Specifically, full-length human SAMHD1 was expressed as a GST fusion protein (GST-Hu SAMHD1-FL) and immobilized onto a CM5 chip coated with anti-GST antibody. Then increasing concentrations of DDB1-DCAF1 CB -Vpx(⌬C) SIVmac were applied, and sensorgrams were recorded. Surprisingly, DDB1-DCAF1 CB -Vpx(⌬C) SIVmac dissociation from human SAMHD1 was negligible for 180 s after the injection was complete ( Fig. 2A). Similar observations were made with immobilized rhesus (Fig. 2B) and De Brazza's SAMHD1 (data not shown). In fact, the rate of association (k on ) and the dissociation constant (K d ) were essentially the same for all three SAMHD1 proteins, ranging from 18 to 34 ϫ 10 3 M Ϫ1 s Ϫ1 and from 213 to 521 nM, respectively (Table 1 and Fig. 2C). Similar SPR experiments showed that SAMHD1 did not bind to the DDB1-DCAF1 CB complex in the absence of viral protein (supplemental Fig. 2A), confirming previous observations that Vpx mediates recruitment of SAMHD1 to the DDB1-DCAF1 substrate adaptor-receptor complex (20). Because human, rhesus, and De Brazza's SAMHD1 showed binding kinetics similar to those of DDB1-DCAF1-Vpx complexes and have strict sequence conservation at the C terminus (supplemental Fig. 1), we speculated that the C terminus of SAMHD1 is sufficient for efficient interaction with DDB1-DCAF1-Vpx. To test this possibility, the C terminus (residues 595-626, SAMHD1-CTD) of human SAMHD1 was immobilized onto a CM5 chip, as a GST fusion protein, and sensorgrams were recorded for increasing concentrations of DDB1-DCAF1 CB -Vpx(⌬C) SIVmac (Fig. 2D). The binding kinetics for the SAMHD1 C-terminal fragment  Table 1. C, binding curves of DDB1-DCAF1 CB -Vpx(⌬C) SIVmac to human (Hu; solid circles), rhesus (Rh; solid squares), and De Brazza's (DB; solid triangles) SAMHD1 were generated by plotting the response levels (RU) at 120 s (marked as dissociation in A and B). The average K d values for the interactions between human, rhesus, or De Brazza's SAMHD1 protein and the DDB1-DCAF1 CB -Vpx(⌬C) SIVmac complex, for two independent experiments, are summarized in Table 1. D, SPR sensorgrams of DDB1-DCAF1 CB -Vpx(⌬C) SIVmac binding to GST-Hu SAMHD1-CTD (residues 595-626). E, SPR sensorgrams of DDB1-DCAF1 CB -Vpx(⌬C) SIVmac , with Ala substitution at Glu-15 of Vpx (E15A), binding to GST-Hu SAMHD1-CTD. F, binding curve of DDB1-DCAF1 CB -Vpx(⌬C) SIVmac to human SAMHD1-CTD is shown. The data were analyzed as described in C. The average k on and K d values are shown in Table 1. Response levels (RU) of DDB1-DCAF1 CB -Vpx(⌬C) SIVmac , with Vpx alanine substitutions at Asn-12, Glu-15, Glu-16, or Thr-17 binding to human SAMHD1-CTD are also shown.
were essentially the same as those of the full-length protein, with a similar binding affinity (Table 1). Similar results were obtained when DDB1-DCAF1 in complex with Vpx 7312a was subjected to binding to the C terminus of human SAMHD1 (supplemental Fig. 2, B and C, and Table 1). Vpx SIVmac cannot enhance SIV and HIV-1 infection of macrophages and differentiated THP-1 cells when specific N-terminal residues (Asn-12, Glu-15, Glu-16, or Thr-17) are mutated (29). Because these mutants retain their ability to interact with DCAF1, these residues were hypothesized to be at the binding interface for a cellular restriction factor (29).
Indeed, we showed that these Vpx mutants fail to down-regulate SAMHD1 in a proteasome-dependent manner (20). For these reasons, we monitored the binding kinetics for immobilized human SAMHD1-CTD and DDB1-DCAF1-Vpx SIVmac , after mutations were introduced into the N terminus of Vpx (N12A, E15A, E16A, or T17A). The results (Fig. 2, E and F) indicate that N-terminal Vpx mutations significantly increase the rate of DDB1-DCAF1-Vpx dissociation from SAMHD1. Conversely, mutation of SAMHD1, at C-terminal residues, completely abolished DDB1-DCAF1-Vpx binding (supplemental Fig. 2, C and D). Taken together, these results suggest that   Vpx not only specifically increases the rate of association between SAMHD1 and DDB1-DCAF1-Vpx but also substantially decreases their dissociation rate, which may account for subsequent polyubiquitination of SAMHD1 by the CRL4 E3 ubiquitin ligase.

The C Terminus of SAMHD1 and the N Terminus of Vpx Are Required for Vpx-dependent SAMHD1 Ubiquitination and
Down-regulation-To examine the functional importance of the SAMHD1 C terminus as it relates to Vpx-mediated recruitment to the CRL4-DCAF1 ligase, in vitro ubiquitination assays were performed. In the absence of Vpx, wild-type SAMHD1 proteins were not ubiquitinated by CRL4-DCAF1 E3 ubiquitin ligase (Fig. 3A, left). However, when Vpx 7312a was complexed with the E3 ubiquitin ligase, both human and monkey SAMHD1 proteins were robustly polyubiquitinated (Fig. 3A,  right). Further, the C-terminal region of SAMHD1 was required for polyubiquitination, because SAMHD1 proteins lacking this region did not show evidence of polyubiquitination. Similar results were obtained when in vitro ubiquitination assays were performed with Vpx SIVmac and Vpx Rod9 (data not shown).
These findings were supported by in vitro cellular assays; wild-type human SAMHD1 or deletion constructs were expressed along with DCAF1 and Vpx in HEK293 cells by cotransfection, and Vpx-dependent depletion of SAMHD1 was assessed (Fig. 3B). Consistent with the previous report (20), the SAMHD1 lacking C-terminal residues 596 -626 was resistant to down-regulation by Vpx SIVmac (Fig. 3B) and Vpx Rod and Vpx 7312a (data not shown), whereas N-terminal deletion had no effect on Vpx-dependent SAMHD1 down-regulation. Similar results were obtained with rhesus or De Brazza's SAMHD1 in combination with various Vpx proteins (data not shown). Notably, the C-terminal region of DCAF1 (DCAF1 CA ; residues 1040 -1400) was sufficient to mediate Vpx-dependent modulation of SAMHD1 level by co-transfection (Fig. 3C). This decrease in SAMHD1 level by Vpx was alleviated by treatment with MG132, indicating proteasome-dependent down-regulation of SAMHD1 (1)(2)(3)20).
In an analogous manner, we explored whether the N-terminal region of Vpx resides at the critical recruitment interface for both human and monkey SAMHD1, as suggested by our in vitro kinetic studies. Mutation of two consecutive acidic Glu residues (15 and 16) to Ala in Vpx SIVmac resulted in significant abrogation of Vpx-dependent SAMHD1 down-regulation (Fig. 3D,  lanes 4 and 5). Similar results were observed for both Vpx Rod9 and Vpx 7312a (Fig. 3D, lanes 6 -9). Of note, the degree of human as well as monkey SAMHD1 down-regulation (data not shown) mediated by Vpx Rod was relatively low compared with the other Vpx proteins. Vpx 7312a contains a third acidic residue (Asp-13), and mutation of this residue to Gly, the same amino acid type for Vpx SIVmac and Vpx Rod (Fig. 1B), did not affect its ability to down-regulate SAMHD1 (Fig. 3E). Taken together, these results suggest that SIVmac and HIV-2 Vpx degrade SAMHD1 in a proteasome-dependent manner by using a common molecular interface.
SAMHD1 Forms dGTP-dependent Tetramers-Previous reports established that the dNTPase activity of human SAMHD1 is positively regulated by dGTP, the binding of which, at allosteric sites, stabilizes dimerization and enhances its catalytic activity (8). However, our biochemical and biological analyses of full-length human SAMHD1 suggested that this protein is inactive as a monomer or dimer and that binding of dGTP induces formation of a tetramer, which is the catalytically active form of the protein (19). To confirm that monkey SAMHD1 behaves in a manner similar to the human protein (i.e. transits from an inactive monomer/dimer to an active tetramer in a dGTP-dependent manner) we carried out several biochemical assays. Chemical cross-linking analyses indicated that the six examined monkey SAMHD1 proteins do, indeed, form tetramers in a dGTP-dependent manner (Fig. 4A)   tures of SAMHD1 and dGTP support tetramer formation upon the addition dGTP. In particular, the elution volumes of rhesus and De Brazza's SAMHD1 in the presence of dGTP (Fig. 4B, top panels) were similar to that of human SAMHD1 (11.0 ml; reported previously (19) and in Fig. 6B). In the absence of dGTP, both monkey SAMHD1 proteins eluted at 14.1 ml (Fig.  4B, bottom panels), which is also similar to human SAMHD1 (13.6 ml; reported previously (19) and in Fig. 6A). These comparisons suggest a high degree of similarity among primate SAMHD1 proteins in terms of oligomerization behavior. To complement these comparisons, we examined the in vitro dGTP-dependent catalytic activity of all three SAMHD1 proteins. Both rhesus and De Brazza's SAMHD1 exhibited dNTPase activity when dGTP was present in the reaction mixture; no activity was observed with other dNTPs (data not shown), similar to the human protein. Further, dNTPase assays defined the preference for dNTP substrates as dGTP Ͼ dCTP Ͼ TTP Ͼ dATP, which was consistent across all three primate SAMHD1 proteins (Fig. 4C).
Recruitment of SAMHD1 to DDB1-DCAF1 via Vpx Reduces Its dGTP-dependent Catalytic Activity-The C terminus of SAMHD1 plays an important role in the stability of the dGTPinduced tetramer (19); we therefore reasoned that DDB1-DCAF1-Vpx binding would negatively influence its catalytic activity. To explore this hypothesis, the dNTPase activity of SAMHD1 was determined, first in the absence of any additional proteins, to establish a base line/total in vitro dNTPase activity for SAMHD1, and then in the presence of increasing concen-trations of DDB1-DCAF1 CB -Vpx(⌬C) SIVmac (Fig. 5A). At a 1:1 stoichiometric ratio of DDB1-DCAF1 CB -Vpx(⌬C) SIVmac to SAMHD1 (1ϫ), ϳ13% of the total in vitro dNTPase activity of SAMHD1 was observed. This inhibition was specific, because neither DDB1-DCAF1 CB alone nor DDB1-DCAF1 CB -Vpr significantly influenced the catalytic activity of SAMHD1. Further, the inhibition was dose-dependent and observed for both rhesus and De Brazza's SAMHD1 proteins as well (Fig. 5B). Taken together, these results suggest that SAMHD1, when recruited to the substrate adaptor-receptor complex by Vpx, is not only directed for degradation but is also deactivated. Interestingly, the interaction between SAMHD1 and Vpx alone is not sufficient to effectively inhibit SAMHD1 dNTPase activity (Fig. 5C). This result is not surprising, because Vpx alone does not form a stable complex with SAMHD1 (data not shown), and the ternary complex comprising SAMHD1, DCAF1, and Vpx is the minimum unit for a stable interaction (20).
Our data indicate an essential role for the SAMHD1 C terminus and Vpx N terminus in the functional interaction between these proteins (Figs. 2 and 3). To further confirm the importance of these regions, the dNTPase activity of SAMHD1-⌬C was measured with increasing concentrations of DDB1-DCAF1 CB -Vpx(⌬C) SIVmac The catalytic activity of SAMHD1 alone was essentially unaltered by loss of its C-terminal residues (19), and the addition of DDB1-DCAF1 CB -Vpx(⌬C) SIVmac did not reduce this activity (Fig. 5D). Further, a site-specific SAMHD1 mutant protein (triple mutation at L620A/F621A/ K622A) was not inhibited by DDB1-DCAF1 CB -Vpx(⌬C) SIVmac ( Fig. 5D). This observation is consistent with our previous findings that Leu-620 and Phe-621 are at the interface where the N terminus of Vpx, in complex with DDB1-DCAF1, binds to SAMHD1 (20).

Recruitment of SAMHD1 to DDB1-DCAF1 via Vpx Induces Disassembly of dGTP-dependent SAMHD1 Tetramers into
Dimer and Monomers-To understand the molecular mechanisms by which DDB1-DCAF1-Vpx inhibits SAMHD1 catalysis, light scattering experiments were performed. First, SAMHD1 and DDB1-DCAF1 CB -Vpx(⌬C) SIVmac were individually subjected to analytical size exclusion column chromatography, and the molecular mass of proteins in each eluting peak was determined by multiangle light scattering. SAMHD1 alone, without dGTP, eluted between 12.5 and 14.5 ml, and the molecular mass of the protein was determined to be 125 kDa, representing a mixture of dimer and monomer populations (Fig. 6A, green trace; the theoretical molecular mass of SAMHD1 is 75 kDa). The molecular mass of DDB1-DCAF1 CB -Vpx(⌬C) SIVmac alone was estimated to be 189 kDa (data not shown). A mixture of DDB1-DCAF1 CB -Vpx(⌬C) SIVmac and SAMHD1 at a molar ratio of 1.5:1, in the absence of dGTP, yielded two major peaks, at 10.8 and 12 ml, with the molar mass of protein in each peak estimated to be 256 and 185 kDa, respectively (Fig. 6A, red  trace). The mass of the first peak corresponds to a molecular complex of SAMHD1 and DDB1-DCAF1 CB -Vpx(⌬C) SIVmac , whereas the second peak may correspond to unbound DDB1-DCAF1 CB -Vpx(⌬C) SIVmac , present in slight excess of SAMHD1 in the initial mixture. Preincubation of SAMHD1 with dGTP resulted in a new peak, at ϳ11.0 ml, which corresponds to SAMHD1 tetramer (306 kDa) (green trace in Fig. 6B), which is consistent with our previous results (19). The elution volume of DDB1-DCAF1 CB -Vpx(⌬C) SIVmac , when preincubated with dGTP, was essentially the same (data not shown). The addition of DDB1-DCAF1 CB -Vpx(⌬C) SIVmac to the SAMHD1-dGTP mixture (at an ϳ1.25:1 stoichiometric ratio; Fig. 6B, red trace) produced a new peak at ϳ8.8 ml. The average molecular mass of proteins in this peak was estimated to be ϳ830 -940 kDa, close to the mass of a supramolecular protein complex comprising four SAMHD1 (75 kDa each) and four DDB1-DCAF1 CB -Vpx(⌬C) SIVmac (181 kDa each). Interestingly, upon increasing the concentration of DDB1-DCAF1 CB -Vpx(⌬C) SIVmac (to an ϳ2.5:1 stoichiometric ratio; Fig. 6B, blue trace), the UV 280 nm absorbance height of the 8.8 ml peak was reduced, and a second peak at ϳ9.6 ml emerged. The average molecular mass of protein in the second peak was estimated to be 470 -570 kDa, which is close to the mass of a protein complex comprising two SAMHD1 and two DDB1-DCAF1 CB -Vpx(⌬C) SIVmac proteins. A peak eluting at ϳ10.8 ml that had a molar mass of 260 -310 kDa and probably represents one SAMHD1 bound to DDB1-DCAF1 CB -Vpx(⌬C) SIVmac showed a concomitant increase in its UV 280 nm absorbance height, upon increasing the amount of DDB1-DCAF1 CB -Vpx(⌬C) SIVmac in the mixture. The three protein peaks, eluting at three distinctive elution volumes (8.8, 9.6, and 10.8 ml), were composed of nearly stoichiometric amounts of SAMHD1, DDB1, DCAF1 CB , and Vpx(⌬C) SIVmac , as judged by SDS-PAGE analysis (Fig. 6B, bottom, compare  lanes 3, 7, and 11). Taken together, these data suggest that DDB1-DCAF1 CB -Vpx(⌬C) SIVmac binding to dGTP-induced SAMHD1 tetramer results in disassembly of the supramolecular complex into dimeric and monomeric SAMHD1-DDB1-DCAF1 CB -Vpx(⌬C) SIVmac protein complexes. At the same time, the recruitment of active forms of SAMHD1 to the virushost protein complex also results in inhibition of its catalysis (Fig. 5). Because we did not observe further reduction of SAMHD1 dNTPase activity upon increasing the molar ratio, from 1 to 2, of DDB1-DCAF1 CB -Vpx(⌬C) SIVmac over SAMHD1 (Fig. 5A), inhibition of SAMHD1 catalysis may precede the disassembly of dGTP-SAMHD1 tetramer.

DISCUSSION
The discovery of SAMHD1 as an antiviral restriction factor shed light on the molecular mechanisms underlying the inability of HIV-1 to infect monocytes and dendritic cells (30 -34). SAMHD1 exerts its antiviral function by maintaining cellular dNTP pools at a level insufficient for HIV-1 reverse transcription (6,10,11). HIV-2, on the other hand, encodes a virulence accessory factor, Vpx, which overcomes the SAMHD1-mediated antiviral response (1)(2)(3). Vpx has been demonstrated to reprogram the CRL4 E3 ubiquitin ligase by binding DCAF1, a substrate receptor of the ligase (35)(36)(37), and recruiting SAMHD1 for proteasome-dependent degradation (2, 3, 20). However, the precise molecular mechanism of SAMHD1 acti-vation and the interplay between Vpx and host cellular factors, with respect to SAMHD1 down-regulation, are yet to be defined. In this report, we provide extensive biochemical analyses showing 1) a common mode of human and monkey SAMHD1 recruitment by SIVmac and HIV-2 Vpx, in complex with the substrate adaptor-receptor component of CRL4 E3 ubiquitin ligase, 2) real-time binding kinetics of SAMHD1 recruitment in vitro, 3) dGTP-induced tetramerization and allosteric activation of human and monkey SAMHD1 proteins, and 4) regulation of SAMHD1 catalysis by Vpx during the course of recruitment to the substrate adaptor-receptor complex.
HIV-2 Vpx and SIVmac Vpx show strong sequence conservation, implying a common binding mode to DDB1-DCAF1  (44) and the HD domain of SAMHD1 (3UN1) (8) and was rendered faintly to emphasize residues and dGTP molecules. B, SAMHD1 monomer, dimer, and tetramer are recruited to CRL4-DCAF1 by Vpx. Vpx, while binding to DCAF1 via its middle region (16), interacts with the C terminus of SAMHD1 via its N terminus. DDB1-DCAF1-Vpx binding to the SAMHD1 tetramer deactivates its dNTPase activity and subsequently disassembles it to dimers and monomers and induces proteasome-dependent degradation. and SAMHD1. Especially, the N terminus of these Vpx proteins contain Asn-12, Glu-15, Glu-16, and Thr-17 residues (Asn-11, Glu-14, Glu-15, and Thr-16 for HIV2 7312a), which are critical for binding to human and monkey SAMHD1 proteins by their conserved C terminus. Each of these residues at the N terminus appears to contribute equally to SAMHD1 binding (Fig. 2F), and mutation of a single residue in the region is sufficient to abrogate Vpx-dependent SAMHD1 down-regulation (1,20,38). Interestingly, HIV-2 Vpx 7312a shows a sequence divergence in this region, at position 13 (Fig. 1B); however, this residue is not important for SAMHD1 down-regulation (Fig. 3E). Notably, the N-terminal region of Vpx is not involved in the direct binary protein interaction with DCAF1 (29). However, recruitment of SAMHD1 requires both Vpx and DCAF1 in complex with DDB1 (20). These results suggest that the N-terminal region of these Vpx proteins creates a new interface in complex with DCAF1. This mode of interaction is distinct from the Vprmediated UNG2 recruitment to CRL4-DCAF1. Vpr directly interacts with UNG2 and forms a stable binary complex (28,39). Interestingly, other SIV Vpr and Vpx proteins interact with the SAM domain of their cognate SAMHD1 proteins (26), and the interfaces mediating these interactions are yet to be identified. All of these Vpr and Vpx proteins contain a common sequence motif that interacts with DCAF1 (16). Taken together, these observations suggest that HIV Vpr and Vpx, although bound to DCAF1 using a common sequence motif, create unique interfaces at separate regions for two distinct host cellular factors (UNG2 and SAMHD1).
We confirmed and extended our previous observation that the C terminus of SAMHD1 is required for efficient recruitment to the CRL4-DCAF1 E3 ubiquitin ligase in complex with Vpx, interacting with DCAF1 for transfer of multiple ubiquitins to SAMHD1 (Figs. 2 and 3). Kinetic studies of substrate polyubiquitination, with SCF E3 ligase, indicate that a relatively slow monoubiquitination (at a rate of 0.03 s Ϫ1 ) is followed by fast polyubiquitination (at a rate of ϳ3-5 s Ϫ1 ) (40). Our SPR data indicate that the rate of SAMHD1 dissociation from DDB1-DCAF1-Vpx is extremely slow (Fig. 2). This would allow sufficient time for initial monoubiquitination and subsequent multiple rounds of ubiquitin transfer from E2 to SAMHD1 bound to the Vpx-modified substrate receptor of the E3 ubiquitin ligase.
Our current findings along with our previous report (19) suggest that human and monkey SAMHD1 undergo dGTP-induced tetramerization and activation of catalysis. We propose that primate SAMHD1, in the absence of dGTP, interconverts between monomer and dimer, which are catalytically inactive forms. Binding of dGTP at the allosteric site induces formation of tetramer, which is a catalytically active dNTPase (Fig. 7A).
Our biochemical characterization of SAMHD1 catalysis upon binding to the DDB1-DCAF1-Vpx complex suggests that Vpx counteracts SAMDH1 restriction at an enzymatic level in addition to proteasome-dependent degradation (1)(2)(3)20). Vpx, bound to DDB1-DCAF1, inhibits its dNTPase activity and induces disassembly of the dGTP-induced tetramer by interacting with the C terminus, distal to the oligomeric HD domain (Fig. 7B). Light-scattering analyses indicate that DDB1-DCAF1-Vpx binding to the dGTP-induced tetramer results in several distinctive quaternary states of protein complexes (Fig.  6B). Although a small region of Vpx is required for SAMHD1 binding, a large molecular interface would probably be created by both DCAF1 and Vpx, and this extended interface may negatively influence the conformation of the catalytically active form. This notion is supported by our finding that SAMHD1-Vpx interaction is not sufficiently strong to inhibit dNTPase catalysis (Fig. 5C). Structural studies of SAMHD1 in complex with Vpx as well as components of the CRL4-DCAF1 E3 ubiquitin ligase would shed light on these important interfaces.
The direct inhibition of SAMHD1 catalysis by Vpx bound to DDB1-DCAF1 is analogous to the direct inhibition of APOBEC3G catalysis by Vif. In particular, the cytidine deaminase activity of APOBEC3G was inhibited by Vif binding, in the absence of degradation, when investigated in bacteria and in a cell-free system (41)(42)(43). Given that our observations were made using a cell-free system, our model of Vpx-mediated inhibition of SAMHD1 catalysis needs to be validated in cells.