Probing the Catalytic Mechanism and Inhibition of SAMHD1 Using the Differential Properties of Rp- and Sp-dNTPαS Diastereomers

SAMHD1 is a fundamental regulator of cellular dNTPs that catalyzes their hydrolysis into 2′-deoxynucleoside and triphosphate, restricting the replication of viruses, including HIV-1, in CD4+ myeloid lineage and resting T-cells. SAMHD1 mutations are associated with the autoimmune disease Aicardi-Goutières syndrome (AGS) and certain cancers. More recently, SAMHD1 has been linked to anticancer drug resistance and the suppression of the interferon response to cytosolic nucleic acids after DNA damage. Here, we probe dNTP hydrolysis and inhibition of SAMHD1 using the Rp and Sp diastereomers of dNTPαS nucleotides. Our biochemical and enzymological data show that the α-phosphorothioate substitution in Sp-dNTPαS but not Rp-dNTPαS diastereomers prevents Mg2+ ion coordination at both the allosteric and catalytic sites, rendering SAMHD1 unable to form stable, catalytically active homotetramers or hydrolyze substrate dNTPs at the catalytic site. Furthermore, we find that Sp-dNTPαS diastereomers competitively inhibit dNTP hydrolysis, while Rp-dNTPαS nucleotides stabilize tetramerization and are hydrolyzed with similar kinetic parameters to cognate dNTPs. For the first time, we present a cocrystal structure of SAMHD1 with a substrate, Rp-dGTPαS, in which an Fe–Mg-bridging water species is poised for nucleophilic attack on the Pα. We conclude that it is the incompatibility of Mg2+, a hard Lewis acid, and the α-phosphorothioate thiol, a soft Lewis base, that prevents the Sp-dNTPαS nucleotides coordinating in a catalytically productive conformation. On the basis of these data, we present a model for SAMHD1 stereospecific hydrolysis of Rp-dNTPαS nucleotides and for a mode of competitive inhibition by Sp-dNTPαS nucleotides that competes with formation of the enzyme–substrate complex.

S terile alpha motif and HD domain containing protein 1 (SAMHD1) is a dNTP triphosphohydrolase enzyme that catalyzes the hydrolysis of dNTPs into triphosphate and 2′deoxynucleoside. 1,2 SAMHD1 is expressed in a variety of tissue types 3,4 and is a key regulator of cellular dNTP homeostasis. 5 In terminally differentiated myeloid lineage cells and resting Tcells, SAMHD1 activity reduces the dNTP pool to a level that inhibits the replication of HIV-1 6−8 and other retroviruses 9 as well as some DNA viruses. 10,11 In addition to the restriction of viral infection, SAMHD1 is also an important effector of innate immunity, and SAMHD1 mutations are found in patients with the autoimmune disease AGS, 12 early onset stroke, 13 along with chronic leukemia 14,15 and other cancers. 16−18 High SAMHD1 expression in acute myeloid leukemia has been associated with reduced efficacy of the nucleoside analogue anticancer drugs Chlofarabine and Cytarabine, 19−21 due to the hydrolysis of their active 5′-triphosphorylated forms by SAMHD1. More recently, SAMHD1 has been reported to have a triphosphohydrolase-independent function in genome maintenance pathways, facilitating homologous recombination 22 and functioning in DNA repair to suppress the release of single-stranded DNA fragments from stalled replication forks into the cytosol. 23 Human SAMHD1 is a 626-residue protein. It comprises an N-terminal nuclear localization signal, 24 a sterile alpha motif (SAM) domain, and an HD phosphohydrolase domain 25 containing the active site. In addition, C-terminal residues 600−626 are targeted by lentiviral Vpx accessory proteins to recruit SAMHD1 to the proteasome. 26,27 The active form of SAMHD1 is a homotetramer 28 where sequences N-and Cterminal to the HD domain stabilize intersubunit protein− protein interactions and incorporate four pairs of allosteric nucleotide-binding sites, AL1 and AL2, that regulate the enzyme through combined binding of G-based (AL1) and deoxynucleoside (AL2) triphosphates. 1,29−32 The allosteric regulation of SAMHD1 has been studied extensively. Numerous studies have shown that GTP or dGTP are the physiological ligands for the first allosteric site, AL1, 31,33 and that the second allosteric site, AL2, is specific for a dNTP with the following preference order: dATP > dGTP > TTP > dCTP. 34−37 The AL1-and AL2-coordinated nucleotides are bridged by a single Mg 2+ ion through their triphosphate moieties. SAMHD1 is also cell cycle regulated by cyclinA2/ CDK2 phosphorylation at threonine 592 38−40 through effects on tetramer stability that modulate activity, 33 and removing this regulation may enable SAMHD1 to inhibit HIV-1 in cycling cells. 38 The catalytic site of SAMHD1 can hydrolyze cognate dNTP substrates, with a preference for dCTP ≈ dGTP > TTP > dATP, 34 as well as dNTP analogues such as 5′-triphosphorylated anticancer and antiviral agents. 41−43 X-ray crystal structures of SAMHD1 in complex with substrate dNTPs and dNTP analogues have elucidated how SAMHD1 selectively binds these substrates 29,30,[33][34][35][36]43 and also utilizes the HD motif to tightly bind a Fe metal ion. 33,44 Recently, we reported structures of SAMHD1 in complex with α,β-imido-dNTP (dNMPNPP) inhibitors, which enabled us to propose a mechanism for SAMHD1 dNTP hydrolysis involving a bimetallic Fe−Mg center that is shared by some HD domain enzymes. 44 Modulation of SAMHD1 activity, for example, through inhibition by dNMPNPP nucleotide analogues, has been proposed as a therapeutic strategy for improving anticancer and antiviral therapeutic efficacy. 19,45, 46 We have now probed SAMHD1 catalysis and inhibition mechanisms using 2′-deoxynucleoside-5′-O-(1-thiotriphosphates) (dNTPαS) nucleotide analogues. Here, a nonbridging oxygen is replaced by sulfur at the α-phosphate of the dNTP, introducing a chiral center at the α-phosphorus (P α ) and resulting in two diastereomers (R p -dNTPαS and S p -dNTPαS). Our enzymological data reveal that S p -dNTPαS diastereomers only weakly support SAMHD1 tetramerization, due to the hard/soft mismatch between the P α phosphorothioate and the hard Lewis acid AL1-AL2-bridging Mg 2+ that is required for tetramer assembly. We also determined that S p -dNTPαS nucleotides are competitive inhibitors of SAMHD1 catalysis with equilibrium inhibition constants, K i , in the micromolar range, as they bind in the active site but cannot maintain the metal and water ion coordination required to support nucleophilic attack on a substrate dNTP P α . By contrast, R p -dNTPαS nucleotides are SAMHD1 allosteric activators as well as substrates with kinetic parameters comparable with natural dNTP substrates. We cocrystallized R p -dGTPαS in AL1, AL2, and the active site of a catalytically inactive SAMHD1 mutant H215A, for the first time trapping a substrate in the active site with an Fe−Mg-bridging water species in line with the P α −O 5 ′ scissile bond. On the basis of these data, we present a model for hydrolysis of R p -dNTPαS that supports a SAMHD1 catalytic mechanism that utilizes a bimetallic center and activated water molecule to hydrolyze dNTP substrates and describe a mode of inhibition by S p -dNTPαS nucleotides that competes with substrate dNTPs and prevents formation of an ES complex.

■ MATERIALS AND METHODS
Protein Expression and Purification. For expression in Escherichia coli, the DNA sequences coding for human SAMHD1 residues M1−M626, SAMHD1 and Q109−M626, SAMHD1(109−626) were amplified by PCR and inserted into a pET52b expression vector (Novagen) using ligationindependent cloning (SAMHD1) or the XmaI/NotI restriction sites (SAMHD1(109−626) to produce N-terminal StrepII-tag fusion proteins. The H215A active site mutant was prepared from the parent Q109−M626 construct using the Quikchange II kit. Primer sequences for PCR and mutagenesis are provided in Table S2, and all insert sequences were verified by DNA sequencing. Strep-tagged SAMHD1 constructs were expressed in the E. coli strain Rosetta 2 (DE3) (Novagen) grown at 37°C with shaking. Protein expression was induced by addition of 0.1 mM IPTG to log phase cultures (A 600 = 0.5), and the cells were incubated for a further 20 h at 18°C. Cells were harvested by centrifugation resuspended in 50 mL of lysis buffer (50 mM Tris-HCl pH 7.8, 500 mM NaCl, 4 mM MgCl 2 , 0.5 mM TCEP, 1× EDTA-free mini complete protease inhibitors (Roche), 0.1 U/mL benzonase (Novagen)) per 10 g of cell pellet and lysed by sonication. The lysate was cleared by centrifugation for 1 h at 50000g and 4°C then applied to a 10 mL StrepTactin affinity column (IBA) followed by 300 mL of wash buffer (50 mM Tris-HCl pH 7.8, 500 mM NaCl, 4 mM MgCl 2 , 0.5 mM TCEP) at 4°C. Bound proteins were eluted from the column by circulation of 0.5 mg of 3C protease (GE) in 25 mL of wash buffer over the column in a closed circuit overnight. 3C protease was removed by affinity chromatography using a 1 mL GSTrap column (GE), and the eluent was applied to a Superdex 200 26/60 (GE) size exclusion column equilibrated with 10 mM Tris-HCl pH 7.8, 150 mM NaCl, 4 mM MgCl 2 , 0.5 mM TCEP. Peak fractions were concentrated to approximately 20 mg mL −1 and flash frozen in liquid nitrogen in small aliquots.
Nucleotides. Deoxyribonucleotide triphosphates and racemic mixtures of R p -and S p -dNTPαS nucleotides were purchased from Jena Biosciences Germany, DE, or TriLink Biotechnologies, US. Purified R p -dNTPαS and S p -dNTPαS diastereomers were from BioLog, DE.
Crystallization and Structure Determination. Prior to crystallization, H215A-SAMHD1(109−626) was diluted to 5 mg mL −1 with gel filtration buffer, supplemented with 2 mM R p -dGTPαS. Crystals of the H215A-SAMHD1(109−626)-R p -dGTPαS−Mg complex were produced by sitting drop vapor diffusion at 18°C using a mosquito crystal robot (SPT Labtech) to prepare 0.2 μL droplets containing an equal volume of the protein/nucleotide solution and mother liquor. The best crystals were obtained using a mother liquor of 0.1 M Bis-tris-HCl pH 6, 15% (w/v) PEG 3350, 0.15 M lithium sulfate. For data collection, the crystals were cryoprotected in mother liquor containing 30% (v/v) glycerol and flash frozen in liquid nitrogen. Data sets were collected on beamline I04 at the Diamond Light Source, UK, at a wavelength of 0.97949 Å. Details of the data collection, processing, and structure refinement statistics are presented in Table S1. Data were processed using the autoPROC pipeline 47 (Global Phasing LtD). Internally, indexing and integration utilized XDS; 48,49 point-group symmetry was determined with POINTLESS; 50 isotropic scaling was carried out using AIMLESS; 51 data were anisotropically scaled in autoPROC using STARANISO (http://staraniso.globalphasing.org/cgi-bin/staraniso.cgi) (Global Phasing LtD); and structure factors were generating using CTRUNCATE. 52 The crystal belonged to the P2 1 2 1 2 1 spacegroup with 8 copies of the H215A-SAMHD1(109−626) monomer and 24 copies of R p -dGTPαS in the asymmetric unit. The structure was solved by molecular replacement using the program PHASER 53  SEC-MALLS. Size exclusion chromatography coupled to multi-angle laser light scattering (SEC-MALLS) was used to determine the molar mass composition of SAMHD1 samples upon addition of R p -and S p -dNTPαS nucleotide analogues and/or activators. SAMHD1 was incubated at 4°C for 5 min after the addition of nucleotide analogues (0.5 mM) and activator (0.2 mM GTP), and then samples (100 μL) were applied to a Superdex 200 10/300 INCREASE GL column equilibrated in 20 mM Tris-HCl, 150 mM NaCl, 5 mM MgCl 2 , 0.5 mM TCEP, and 3 mM NaN 3 , pH 8.0, at a flow rate of 1.0 mL/min. The scattered light intensity and protein concentration of the column eluate were recorded using a DAWN-HELEOS laser photometer and an OPTILAB-rEX differential refractometer (dRI) (dn/dc = 0.186) respectively. The weightaveraged molecular mass of material contained in chromatographic peaks was determined using the combined data from both detectors in the ASTRA software version 6.1 (Wyatt Technology Corp., Santa Barbara, CA).
NMR Analysis of SAMHD1 Catalysis. One-dimensional 1 H NMR spectroscopy was used to measure SAMHD1 hydrolysis rates of dNTPs and R p -and S p -dNTPαS analogues. Reactions were prepared in NMR buffer (20 mM Tris-HCl pH 8.0, 150 mM NaCl, 5 mM MgCl 2 , 2 mM TCEP, 5% D 2 O) containing 0.5 mM of each dNTP or dNTPαS analogue, 100 μM GTP and 1 μM SAMHD1. In inhibition studies, 10−100 μM ZnCl 2 or CdCl 2 was additionally included in assays. 1 H NMR spectra (two dummy scans, four scans) were recorded at 30 s intervals at 22°C as a pseudo 2D array using a Bruker Avance 600 MHz NMR spectrometer equipped with a 5 mm TCI cryoprobe. Solvent suppression was achieved using excitation sculpting. 60 Experiments were typically carried out for between 1 and 10 h. The integrals for clearly resolved substrate and product peaks at each time point were extracted using the Bruker Dynamics Centre software package and used to construct plots of substrate or product against time. Initial rates were extracted from the linear part of the curve in order to determine k cat values. Under these conditions, the limit of detection is ∼5% product over the span of the experiment. This equates to a minimum detectable SAMHD1 normalized hydrolysis rate of 0.00075 s −1 .
Real-Time Measurement of Triphosphohydrolase Activity. To obtain quantitative kinetic parameters for substrate hydrolysis (K M and k cat ), SAMHD1 divalent metal ion dependencies and inhibition by S p -dNTPαS analogues (K i ), a real-time continuous coupled assay was employed utilizing the biosensor MDCC-PBP 61,62 to measure phosphate release from combined SAMHD1 triphosphohydrolase and Saccharomyces cerevisiae Ppx1 exopolyphosphatase activity. 42 In a typical experiment, solutions containing wt-SAMHD1(1− 626), Ppx1, MDCC-PBP, and GTP were incubated for 5 min in assay buffer (20 mM Tris pH 8.0, 150 mM NaCl, 5 mM MgCl 2 , and 2 mM TCEP) at 25°C before the reaction was initiated by the addition of substrate nucleotides and nucleotide analogues. The final concentrations were 0.2 μM SAMHD1, 0.02 μM Ppx1, 40 μM MDCC-PBP, 0.2 mM GTP, and varying concentrations of dNTP substrates and dNTPαS analogues. In divalent metal ion titration experiments, an assay buffer without 5 mM MgCl 2 was employed, and the different metal chloride salts MgCl 2 , MnCl 2 , CoCl 2 , NiCl 2 , ZnCl 2 , and CdCl 2 were added over a concentration range of 0.1−10 mM. Throughout reactions the fluorescence intensity was recorded at 430 nm excitation and 465 nm emission wavelengths at 15− 20 s time intervals over a period of 10−30 min in a Clariostar multiwell plate reader (BMG Labtech). Steady-state rates were obtained from time courses of P i formation by linear regression of the data points in the linear phase of the reaction (<10% substrate consumed). The lower limit of detection under these conditions is ∼0.5 μM product accumulated over 20 min corresponding to a rate of 0.002 s −1 ). Rates were normalized for SAMHD1 concentration and plotted against substrate concentration. Michaelis constants (K M ) and catalytic constants (k cat ) for substrates were then determined by nonlinear least-squares fitting using either a Michaelis−Menten or Hill-function in the software package Prism 9 (Graphpad).
For inhibition studies, experiments were conducted at three constant substrate concentrations (1, 0.3, and 0.1 or 0.3, 0.1, and 0.03 mM TTP), the SAMHD1, Ppx1, MDCC-PBP and GTP concentrations were maintained as above, and the S p -dNTPαS inhibitor concentration was varied. The data from the three independent experiments were analyzed globally by nonlinear least-squares fitting using the equation for competitive inhibition (1); where V/[SAMHD1] is the steady-state rate, normalized to the SAMHD1 concentration, [S] is the (fixed) substrate concentration, [I] is the (variable) inhibitor concentration, K i is the inhibition constant, and k cat and K M are the catalytic and Michaelis−Menten constants for substrate turnover in the absence of inhibitor.
The fitting was performed with a fixed value of K M for GTPactivated TTP hydrolysis, determined previously, 42 and only k cat and K i were allowed to vary. All measurements were performed in at least triplicate.
RP-HPLC Analysis of R p -TTPαS and S p -TTPαS Hydrolysis. SAMHD1 rates of hydrolysis of equimolar mixtures of R p -TTPαS and S p -TTPαS were determined by reverse-phase chromatography analysis of reactions. Typically, 2 μM SAMHD1 was incubated with 0.2 mM GTP and 0.5 mM each of R p -TTPαS and S p -TTPαS in a reaction buffer of 20 mM Tris-HCl, 150 mM NaCl, 2 mM TCEP (pH 8.0) supplemented with either 5 mM MgCl 2 , 1 mM MnCl 2 , or 1 mM CoCl 2 . Samples were withdrawn at intervals from 0 to 60 min and quenched by 7-fold dilution into RP buffer (100 mM K 2 HPO 4 /KH 2 PO 4 pH 6.5, 10 mM tetrabutylammonium bromide, 17% acetonitrile). R p -TTPαS, S p -TTPαS, and reaction products were then separated from precipitated protein by filtration through a 0.22 μm centrifugal filter (Durapore-PVDF, Millipore). Samples (5 nmol) were applied Biochemistry pubs.acs.org/biochemistry Article to a Zorbax SB-C18 column (4.6 × 250 mm, 3.5 μm, 80 Å pore size, Agilent Technologies), maintained at 30°C, and mounted on a Jasco HPLC system controlled by Chromnav software (v1.19, Jasco). The thymidine reaction product (R t = 2.5 min), activator GTP (R t = 3.9 min), and substrates (R p -TTPαS (R t = 8.6 min) and S p -TTPαS (R t = 7.7 min) were separated under isocratic flow by application of RP buffer at 1 mL min −1 over 15 min. Absorbance data from the column eluent were continuously monitored between 200 and 650 nm (1 nm intervals) using an MD-2010 photodiode array detector (Jasco). The amount of R p -TTPαS and S p -TTPαS throughout the course of the reaction was determined by peak integration of the 260 nm absorbance data. Rates were determined by linear regression of a plot of the amount of R p -TTPαS and S p -TTPαS against reaction time.
■ RESULTS R p -and S p -dNTPαS Diastereomers. The substitution by sulfur of a nonbridging diastereotopic oxygen at the αphosphate of a dNTP introduces a chiral center at P α , with the replacement of the pro-R and pro-S oxygen atoms resulting in the formation of the R p -dNTPαS and S p -dNTPαS diastereomers containing S 1A −O 2A and O 1A −S 2A atoms respectively 63 ( Figure 1). Given that the incorporation of R p and S p diastereomers into nucleotides and nucleic acids often results in differential properties with respect to the action of stereoselective enzymes and receptors, 64−68 we sought to test the ability of R p -dNTPαS and S p -dNTPαS analogues to support SAMHD1 tetramerization through binding at AL1 and AL2 and assess their properties as substrates or inhibitors at the SAMHD1 active site. SAMHD1 Allosteric Sites Are Selective for R p -over S p -dNTPαS. We first analyzed the ability of R p -and S p -dNTPαS diastereomers to support SAMHD1 tetramerization through binding at allosteric sites AL1 and AL2, which is required for catalysis. SEC-MALLS analysis of SAMHD1 tetramerization showed that in the absence of GTP, R p -dGTPαS strongly induced SAMHD1 tetramerization, S p -dGTPαS was ineffectual, but an equimolar mixture of R p -dGTPαS and S p -dGTPαS induced a similar level of tetramerization as R p -dGTPαS alone ( Figure 2A).
These data demonstrate that R p -dGTPαS is sufficient to bind at both AL1 and AL2 to induce tetramer formation, while S p -dGTPαS is impaired in binding either at AL1 or AL2 or both. Further SEC-MALLS data that included GTP ( Figure  2B) show that the R p -diastereomers of dGTPαS, dATPαS, and TTPαS generally stabilized SAMHD1 tetramerization, through AL2-binding, more than their S p -diastereomer counterparts, and dCTPαS diastereomers did not induce significant tetramerization, as previously reported for dCTP 35 as well as dCMPNPP and α,β-methyleno-dCTP (dCMPCPP) analogues. 44 Therefore, these data demonstrate that there is a clear preference for R p -over S p -dGTPαS in AL1 and for R pover S p -dNTPαS nucleotides in AL2.
To further investigate this preference, we cocrystallized the catalytic domain, residues 109−626, of a catalytically inactive H215A SAMHD1 mutant 44 with R p -dGTPαS and magnesium ions. The structure of this H215A-SAMHD1(109−626)-R p -dGTPαS−Mg complex was determined by molecular replacement to 2.3 Å resolution and contains two SAMHD1 tetramers in the asymmetric unit ( Figure S1) with electron density for nucleotides and metal ions found in each of the allosteric and active sites ( Figures S2 and S3). Details of the data collection   Table S1. Inspection of the allosteric site of this H215-SAMHD1(109−626)-R p -dGTPαS−Mg complex ( Figure 2C) provides a structural explanation for the observed preference for R p -over S p -dNTPαS nucleotides. Here, it is apparent that AL1 selectivity for R p -dGTPαS results from the α-phosphate O 2A oxygen of the AL1-bound nucleotide that coordinates the AL1-AL2bridging magnesium ion (Mg1). This interaction would be disrupted by the pro-S thio-substitution in S p -dGTPαS due to the incompatibility of a hard Lewis acid (Mg1) and a soft Lewis base (P α -phosphorothioate). In addition, AL2 selectivity for R p -dNTPαS diastereomers results from hydrogen bonding between Lys354 (N ζ H), His376 (N ε2 H), and the O 2A oxygen in the AL2-coordinated nucleotide, which would again be perturbed by the thiol substitution in S p -dNTPαS nucleotides. These observations are further supported by a previous study where SAMHD1 was cocrystallized with a R p -and S p -dGTPαS racemic mixture. 29 There, only R p -dGTPαS was observed in the allosteric site, 29 suggesting a strong selectivity preference for the R p over the S p diastereomer. R p -but not S p -dNTPαS Are Hydrolyzed by GTP-Activated SAMHD1. In order to inform the SAMHD1 dNTP hydrolysis mechanism, the properties of R p -and S p -dNTPαS nucleotides with respect to SAMHD1 catalytic activity were assessed using a fluorescence-based coupled enzyme assay 42 and by 1 H NMR spectroscopy. Data from the coupled enzyme assay revealed that R p -dATPαS was hydrolyzed with a similar Michaelis constant (K M ) as dATP but with about a 2-fold reduction in catalytic rate constant (k cat ) in a GTP-stimulated reaction ( Figure 3A and Table 1). By contrast, no measurable hydrolysis of S p -dATPαS was observed above the limit of detection (<0.002 s −1 ) ( Figure 3A). Examination of the hydrolysis of other R p -diastereomers (R p -dGTPαS, R p -TTPαS, and R p -dCTPαS) showed a 2−3 fold variation in K M values relative to canonical nucleotides and 2−3 fold reductions in k cat ( Figure 3B,C and Table 1) but with the same rank order of turnover TTP > dATP > dCTP > dGTP. However, with both dCTP and R p -dCTPαS, significant sigmoidal behavior is apparent, likely as a result of poor AL2 binding, and so Hill coefficients (h) were applied to adequately fit the data (Table  1). Nevertheless, these data clearly demonstrate that in the presence of GTP all R p -dNTPαS nucleotides are hydrolyzed by SAMHD1 with kinetic constants comparable to the canonical nucleotides.
Hydrolysis of R p -and S p -dNTPαS nucleotides by SAMHD1 was also investigated using 1 H NMR spectroscopy that readily distinguishes R p -and S p -dNTPαS diastereomers by their 1 H NMR spectrum. The spectra of R p -dATPαS and S p -dATPαS ( Figure 4A) contain two singlet peaks in the downfield nucleobase region from the C8H and C2H protons of the adenine base. The chemical shifts of the R p -dATPαS C8H and C2H protons are 8.431 and 8.140 ppm, and the S p -dATPαS C8H and C2H have chemical shifts of 8.463 and 8.145 ppm. Other dNTPαS diastereomers are also distinguishable by the unique resonances of base protons. Therefore, 1 H NMR was  Biochemistry pubs.acs.org/biochemistry Article used to measure GTP-stimulated SAMHD1 hydrolysis of each R p -and S p -dNTPαS diastereomer. These data ( Figure 4B,C) clearly demonstrate that, while R p -dNTPαS diastereomers are SAMHD1 substrates, the S p -dNTPαS diastereomers are refractory to hydrolysis, in good agreement with observations from the coupled enzyme assay (Figure 3). Moreover, the apparent k cat values measured for R p -dNTPαS substrates were 2−4 fold lower than those of the canonical dNTPs (Table 2) with a rank order of hydrolysis of R p -TTPαS > R p -dATPαS > R p -dGTPαS ≈ R p -dCTPαS, mirroring that of the canonical dNTPs (TTP > dATP > dGTP > dCTP) in a 1 H NMR assay 44 and close to that observed in the coupled enzyme assay (Table  1). S p -dNTPαS Diastereomers Inhibit SAMHD1 Catalysis. Having demonstrated that R p -dNTPαS diastereomers can stabilize SAMHD1 tetramers through AL2-binding and that they are hydrolyzed by SAMHD1 with catalytic parameters similar to their cognate canonical dNTP, we next wanted to investigate the refractory S p -dNTPαS diastereomers in the context of SAMHD1 catalysis. SEC-MALLS experiments showed that the stabilization of SAMHD1 tetramers through AL2 binding of S p -dNTPαS was much less than that by R p -dNTPαS ( Figure 2B). Therefore, the lack of hydrolysis in 1 H NMR and coupled enzyme fluorescence experiments (Figures 3 and 4B−C) may either be a result of using a poor AL2 activator or that S p -dNTPαS diastereomers are directly refractory to hydrolysis by the SAMHD1 active site. To test these ideas and promote tetramerization of SAMHD1 in 1 H NMR assays measuring S p -dNTPαS hydrolysis, we combined GTP and a 1:1 mix of each R p -and S p -dNTPαS pair and simultaneously monitored both R p -and S p -dNTPαS as substrates ( Figure 4D). Analysis of these experiments reveals three key observations. First, all R p -dNTPαS diastereomers are hydrolyzed, confirming tetramerization of SAMHD1 through AL2 binding. Second, all the S p -dNTPαS diastereomers are still refractory to hydrolysis, indicating that, although SAMHD1 is activated through AL2 binding by R p -dNTPαS, S p -dNTPαS diastereomers are not hydrolyzed at the active site. Third, although the R p -dNTPαS diastereomers are still hydrolyzed, they are hydrolyzed at significantly reduced rates, 2−8-fold slower than in the absence of S p -dNTPαS (Table 3). Thus, we concluded that not only are S p -dNTPαS diastereomers refractory to hydrolysis they are competitive inhibitors of SAMHD1 nucleotide hydrolysis through binding at the active site.
These data provide semiquantitative measurements of competitive inhibition by S p -dNTPαS diastereomers. Therefore, further studies using enzyme-coupled inhibition assays  Table 1; error bars represent the standard error of the mean (SEM) of at least three independent measurements.    Table 2 and Table 3. In C and D, no significant reduction in the S p -dNTPαS peak intensities is apparent, indicating that S p -dNTPαS analogues are refractory to SAMHD1 hydrolysis.  Biochemistry pubs.acs.org/biochemistry Article substrate by SAMHD1. These data fit well to a competitive inhibition model, demonstrating that all S p -dNTPαS diastereomers competitively inhibit SAMHD1 triphosphohydrolase activity ( Figure 5) with K i ranging from 117 μM for S p -dATPαS to 0.82 μM for S p -dGTPαS with a rank order of K i of S p -dATPαS > S p -TTPαS > S p -dCTPαS > S p -dGTPαS ( Table  4) that mirrors the same nucleobase rank order as observed previously with the dNMPNPP inhibitors. 44 Conformation of R p -and S p -dNTPαS Diastereomers in the SAMHD1 Active Site. The H215A-SAMHD1(109− 626)-R p -dGTPαS crystal structure contains a R p -dGTPαS substrate bound at the active site ( Figure 6 and Figure S3), as well as in allosteric sites AL1 and AL2. In this structure, the active site R p -dGTPαS coordinates the His/Asp-bound Fe, two Mg ions (Mg2 and Mg3), and hydrating water molecules. Several amino acids also interact with or pack against the guanine base, 2′-deoxyribose and thio-substituted triphosphate, including Gln149, Arg164, His210, Lys312, Tyr315, Arg366, and Tyr374. Although Ala215, that replaces histidine in the H215A mutant, cannot provide the general acid required for catalysis of the substrate R p -dGTPαS, an Fe/Mg3-bridged water, W0, that could act as a nucleophile for catalysis is positioned in line with the scissile P α −O 5 ′ phosphoester bond of the substrate R p -dGTPαS ( Figure 6). This suggests that the R p -dGTPαS substrate conformation in the active site is representative of the precatalytic state and is consistent with our enzymological data, which demonstrates that the R p -dNTPαS diastereomers are substrates, albeit with a small reduction in k cat relative to canonical dNTPs (Figures 3 and 4).
Comparison of the configuration of R p -dGTPαS bound in the H215A active site with that of dGMPNPP and dAMPNPP inhibitors in wild-type active sites ( Figure 7A−C) reveals that the nucleotide coordination, together with the positioning of metal ions and water molecules, is highly conserved. Specifically, the Fe, Mg2, and Mg3 active site metal ions, are equivalently coordinated by side chains from the HD motif residues His167, His206, Asp207, and Asp311, and by the side chain of His233, as well as α, β, and γ-phosphate oxygens and active site water molecules.
In the R p -dGTPαS structure, the Fe is coordinated by the αphosphorothioate sulfur rather than the phosphate oxygen present in a canonical dNTP substrate. Our enzymological data demonstrate that although the S p -dNTPαS diastereomers are refractory to hydrolysis they still act as competitive inhibitors of SAMHD1. This indicates that S p -dNTPαS diastereomers can still bind the active site, likely through the same electrostatic interactions with the basic side chains of Arg164, Lys312, and Arg366, hydrogen bonds with Gln149 and Tyr315, and π−π stacking with Tyr374 that are observed in the R p -dGTPαS structure ( Figure 6). Therefore, to assess how the S p -diastereomer alters the catalytic competence of the active site, we modeled an S p -dGTPαS nucleotide into the R p -dGTPαS 2Fo-Fc difference density in our H215A SAMHD1 structure. In the modeled S p -dGTPαS structure ( Figure 7D), the α-phosphorothioate sulfur and nonbridging oxygen atoms have switched positions. As a result, the α-phosphorothioate nonbridging oxygen coordinates Fe, and the sulfur is now   Biochemistry pubs.acs.org/biochemistry Article positioned so as to coordinate Mg3 to maintain the octahedral geometry of the coordination sphere. It is apparent that a sulfur-Mg 2+ configuration of this kind does not satisfy the pairing-selectivity principle of hard Lewis acid Mg 2+ cation with a hard Lewis base. Furthermore, analyses of the PDB database reveal that, although coordination of Fe by thiol groups is prevalent in proteins, sulfur coordination of Mg 2+ does not occur. 69−71 Therefore, we hypothesize that the loss of coordination between the α-phosphate nonbridging S 2A sulfur and Mg3 prevents the formation of a catalytically competent configuration of an S p -dNTPαS diastereomer in the active site. One consequence of the absence of this coordination is a diminished electron-withdrawing environment around the α-phosphate, resulting in a reduction of electrophilicity and therefore reactivity of P α . In addition, and perhaps more importantly, the hard/soft mismatch between a Mg 2+ ion and the phosphorothioate thiol moiety could both distort nucleotide binding and result in the loss of Mg3 from the active site.
Regardless of which of these effects dominates, it is unlikely that W0, the hydroxide nucleophile bridged by Fe−Mg3 in the R p -dGTPαS structure, could be positioned by an S p -dNTPαS nucleotide in line with the P α −O 5 ′ bond to initiate catalysis. Therefore, overall, our observations support the hypothesis that the hydrolyzable R p -dNTPαS nucleotides maintain coordination with the active site Fe and Mg3 through the αphosphorothioate group and, together with other active site residues, support hydroxide-mediated nucleophilic attack of P α to initiate the P α −O 5 ′ bond cleavage. By contrast, although S p -dNTPαS diastereomers are able to bind at the active site they act as competitive inhibitors, as they cannot maintain the metal and water ion coordination required to support nucleophilic attack on the P α .
Metal Ion Dependencies of dNTP, R p -, and S p -dNTPαS diastereomers. In order to test our Lewis acid−Lewis base hard/soft mismatch hypothesis, we examined the SAMHD1 metal ion dependency of GTP-stimulated hydrolysis of TTP, R p -TTPαS, and S p -TTPαS. We first employed a range of divalent metal cations (Mg 2+ , Mn 2+ , Co 2+ , Ni 2+ , Zn 2+ , and Cd 2+ ) that constitute hard and softer Lewis acids in the SAMHD1-Ppx1 coupled enzyme assay. However, in control experiments, we found Zn 2+ and Cd 2+ did not support triphosphate hydrolysis by Ppx1( Figure S4) and strongly inhibited Ppx1 in the presence of Mg 2+ , so these ions were excluded from further analysis using the coupled enzyme assay. Nevertheless, Zn 2+ and Cd 2+ were amenable to 1 H NMR experiments. These direct assays of TTP hydrolysis showed that Zn 2+ and Cd 2+ were also potent inhibitors of SAMHD1 activity, each reducing the SAMHD1 TTP hydrolysis rate >10fold at 10 μM and >100-fold at 100 μM in the presence of 5 mM Mg 2+ ( Figure S5). These data support previous observations of SAMHD1 inhibition by Zn 2+33,72 and now show Cd 2+ is similarly effective.
Of the remaining divalent metal ions, using the coupled enzyme assay, we found that Ni 2+ supported very slow hydrolysis of TTP, R p -TTPαS, and even S p -TTPαS at the lowest Ni 2+ concentrations employed (0.2−0.4 mM). By contrast, Mg 2+ , Mn 2+ , and Co 2+ all stimulated hydrolysis to very different degrees depending on the substrate and also with significantly different concentration dependencies ( Figure 8A− C). It is apparent that TTP hydrolysis is strongly Mg 2+ dependent with a maximum stimulation above 1 mM. TTP is also hydrolyzed effectively with Mn 2+ and Co 2+ , but here the maximum rate is achieved with 0.2−1 mM metal ion, and increased concentration is actually inhibitory to catalysis. This is especially apparent with Co 2+ ( Figure 8A). Hydrolysis of R p -TTPαS is also stimulated by Mg 2+ above 1 mM, but here Mn 2+ supports faster rates. Similar to the observation with TTP, Co 2+ also supports hydrolysis at sub-millimolar concentrations but is inhibitory at a higher concentration ( Figure 8B). The hydrolysis of S p -TTPαS in the presence of Mg 2+ is below the detection limit, consistent with the notion of the hard/soft mismatch of the Mg 2+ ion and the phosphorothioate thiol moiety. By contrast, the softer Mn 2+ and Co 2+ that can coordinate the phosphorothioate do support hydrolysis of S p -TTPαS but also with Co 2+ being inhibitory at a higher millimolar concentration ( Figure 8C). To test if mixtures of metal ions might better support hydrolysis, as there are three divalent metal ion binding sites in each SAMHD1 monomer with potentially different metal ion binding requirements, we determined the rates of hydrolysis with pairs of metal ions at 1.25 mM each. These data (Figure 8D−F and Table 5) largely recapitulate the observations with single metals in that TTP is hydrolyzed effectively by Mg 2+ , and Mn 2+ and that although Co 2+ supports hydrolysis it is inhibitory at millimolar concentration even in the presence of Mg 2+ ( Figure 8D).  Hydrolysis of R p -TTPαS is supported by Mg 2+ , Mn 2+ , and Co 2+ but is most strongly stimulated by Mn 2+ that in the background of Mg 2+ in an ion mixture increases the k cat 4-fold rate from 0.1 to 0.4 s −1 ( Figure 8E and Table 5). Mg 2+stimulated hydrolysis of S p -TTPαS is not measurable above the limit of detection of the assay (0.002 s −1 ). However, upon addition of Mn 2+ and Co 2+ either alone or combined with Mg 2+ , the S p -TTPαS hydrolysis rate is increased at least 10-fold by Mn 2+ and 20-fold by Co 2+ (Figure 8F and Table 5). Taken together, these data show even though there is a complex relationship between metal ion type, concentration and SAMHD1 substrate, the hydrolysis of S p -TTPαS is not supported by the hard Lewis acid Mg 2+ but can be by softer Mn 2+ and to a greater extent Co 2+ ions.
To test this notion further and to assess if allosteric binding of R p -TTPαS might further enhance S p -TTPαS hydrolysis, we examined the divalent metal ion dependency of hydrolysis reactions containing both R p -TTPαS and S p -TTPαS nucleotides. As 1 H NMR detection of nucleotide base protons was not possible with the paramagnetic Mn 2+ and Co 2+ ions present, to discriminate between hydrolysis of the two substrates in the same reaction, we took advantage of the fact that the diastereomers are separable using ion-pair reversephase HPLC. Analysis of a GTP-Mg 2+ -stimulated reaction containing equimolar R p -TTPαS and S p -TTPαS showed that R p -TTPαS is hydrolyzed in the presence of S p -TTPαS, while S p -TTPαS remains refractory ( Figure 9A,D). Nonetheless, the rate of R p -TTPαS hydrolysis ( Figure 9G) is reduced compared to that of a GTP-Mg 2+ stimulated reaction of R p -TTPαS alone  Figure 8E). Therefore, these data support our conclusions from both NMR and coupled enzyme assays demonstrating that Mg 2+ cannot support hydrolysis of S p -dNTPαS nucleotides and that they are competitive inhibitors of R p -dNTPαS nucleotide hydrolysis. In GTP-Mn 2+ -and GTP-Co 2+ -stimulated reactions, some hydrolysis of S p -TTPαS along with that of R p -TTPαS is observed ( Figure 9B,E & C,F) but with no significant increase of the rate ( Figure 9G) compared to Mn 2+or Co 2+ -stimulated hydrolysis of S p -TTPαS alone ( Figure 8F). Therefore, these data indicate that while the softer Mn 2+ and Co 2+ ions do support SAMHD1 hydrolysis of S p -dNTPαS nucleotides the presence of R p -dNTPαS nucleotides at allosteric sites does not enhance S p -dNTPαS nucleotide hydrolysis further.

■ DISCUSSION
Despite the importance of SAMHD1-mediated dNTP regulation of cell proliferation and viral restriction, a proposed catalytic mechanism for dNTP triphosphohydrolysis by SAMHD1 was only recently reported. 44 Thio-substituted nucleotide analogues are often inhibitory or are poorly hydrolyzed by enzymes, making them useful for structural analysis 73,74 and have been exploited in a number of mechanistic studies of phospho-hydrolytic enzymes. 65,75,76 Therefore, in this study, we employed α-thio-substituted R pand S p -dNTPαS diastereomers ( Figure 1) to probe SAMHD1 catalysis and allostery. Depending on the diastereomer, some SAMHD1 protein-nucleotide interactions are disrupted, while others are maintained, resulting in differences in tetramerization/allosteric activation and in catalysis. Our X-ray crystallographic, enzymological, and biochemical studies using R p -and S p -dNTPαS diastereomers now provide insight into the specificity of SAMHD1−nucleotide−metal ion interactions at the allosteric and active sites. Moreover, the R p -dGTPαS structure provides a model for the enzyme−substrate [ES] complex, while our S p -dNTPαS data reveal a new class of SAMHD1 inhibitors that compete for the apo-active site. R p and S p Stereoselectivity at the SAMHD1 Allosteric Site. Previous studies have demonstrated the importance of Mg for SAMHD1 activity. 29 Our present study now highlights the functional importance of these nucleotide−Mg interactions at the allosteric site as demonstrated by the observation that only R p -dNTPαS and not S p -dNTPαS diastereomers are able to coordinate Mg at AL1 and AL2 to support tetramerization.
At AL1, which is specific for a guanine-based nucleotide, only R p -dGTPαS supports tetramerization. Inspection of AL1 in the H215A-SAMHD1(109−626)-R p -dGTPαS crystal structure reveals that R p -dGTPαS maintains coordination of the AL1-AL2 bridging Mg ion through an α-phosphate oxygen in the same way as a canonical nucleotide. By contrast, with S p -dGTPαS the incompatibility of soft Lewis base α-phosphorothioate sulfur and hard Lewis acid Mg disallows this nucleotide-Mg coordination at AL1-AL2 and so is refractory to the subunit packing required for tetramer assembly.
Our biochemical studies reveal less discrimination at AL2 than AL1 but nevertheless do demonstrate that AL2 binding of R p -dNTPαS diastereomers stabilizes SAMHD1 tetramerization to a greater extent than S p -dNTPαS diastereomers. Here, our structural analysis reveals that R p or S p thio-substitution to the AL2-bound nucleotide has little direct effect on the Mgcoordination. Instead, where a canonical deoxynucleotide or the R p -dNTPαS the α-phosphate makes hydrogen bonds with the basic side chains of Lys354 and His376 AL2-interacting residues, the geometry demands that in an Sp-dNTPαS the S pphosphorothioate is required to make these interactions. Given Error is the SD of three independent measurements. c value below the reliable limit of detection (0.002 s −1 ).  Biochemistry pubs.acs.org/biochemistry Article the reduced electronegativity of sulfur relative to oxygen and that it is a very poor hydrogen bond acceptor, 77 a loss of this hydrogen bonding likely explains the reduced capacity of S p -dNTPαS deoxynucleotides to support SAMHD1 tetramerization through binding at AL2. Therefore, taken together, it is apparent that both allosteric sites discriminate R p over S p , but the selection is mediated in different ways. At AL1, it is through the loss of a direct interaction with the Mg ion and at AL2 it is through the lack of capacity for a S p -phosphorothioate to make hydrogen bonding interactions with the key residues that support tetramerization upon nucleotide binding. R p -dNTPαS Hydrolysis and S p -dNTPαS Inhibition of SAMHD1. Our enzymological and biochemical data clearly show that the structural differences arising from the stereochemistry of R p -and S p -dNTPαS analogues have significant effects on SAMHD1 activity. R p -dNTPαS nucleotides are substrates of SAMHD1 with catalytic constants comparable with those of canonical nucleotides. In contrast, S p -dNTPαS nucleotides are inhibitors of SAMHD1 triphosphohydrolase activity, likely through binding competitively at the active site.
To understand these observed differences, we employed a SAMHD1 mutant, H215A, which retains nucleotide binding but is catalytically deficient 44 to determine the structure of SAMHD1 in complex with a substrate R p -dGTPαS at the active site ( Figure 6). The use of this mutant in combination with substrate R p -dGTPαS has now enabled us to visualize a substrate precatalysis in the SAMHD1 active site for the first time and so provides an excellent structural tool for studying other SAMHD1 substrates, such as canonical dNTPs and nucleotide-based anticancer and antiviral drugs. These data demonstrate how a substrate R p -dNTPαS is positioned in the SAMHD1 active site. Unlike in previous structures, the H215A-SAMHD1(109−626)-R p -dGTPαS−Mg complex reveals how the substrate R p -dGTPαS is poised for nucleophilic attack by an Fe−Mg-bridged water species, W0, likely a hydroxide ion (Figures 6 and 7 and Supplementary Figure S3). Moreover, the substrate R p -dGTPαS binding conformation is highly similar to that of a dNMPNPP inhibitor, which, we previously proposed, mimics the precatalytic state. 44 This is despite the substitution of an α-phosphate nonbridging oxygen with the phosphorothioate in R p -dGTPαS, which nevertheless still supports Fe coordination and nucleotide hydrolysis.
Modeling of the S p -dNTPαS diastereomers at the SAMHD1 active site shows there is a similar incompatibility between the S p -α-phosphorothioate and Mg3 as with the Mg1 and S p -dGTPαS in the allosteric site. Here the S p -thio moiety would have to approach Mg3 in the active site, but due to the sulfur/ magnesium mismatch, this likely distorts nucleotide binding in the catalytic site to the extent that the attacking hydroxide nucleophile, W0, and substrate nucleotide are not aligned for catalysis. We have previously demonstrated the importance of Mg3 by alanine substitution of the Mg3-coordinating residue His233 that resulted in a 300-fold reduced k cat for GTPactivated dATP hydrolysis. 44 Thus, our observation here that S p -dNTPαS diastereomers bind in the active site, are competitive inhibitors and are not hydrolyzed by SAMHD1 further supports the notion that the Fe-proximal Mg is crucial for catalysis.
The idea of the hard−soft mismatch between the αphosphorothioate of S p -dNTPαS diastereomers with Mg3 is further supported by our metal ion dependency experiments. We employed a range of hard to soft metal ions to ascertain whether hydrolysis of S p -dNTPαS diastereomers could be rescued by employing softer metal ions that support interaction with the S p α-phosphorothioate. These data showed convincingly that, while hard Mg 2+ did not support hydrolysis of S p -dNTPαS, the S p diastereomer was hydrolyzed in the presence of the softer Mn 2+ and Co 2+ metal ions.
Mechanisms of Inhibition. Given the notion that S p -dNTPαS diastereomers bind at the active site and act as competitive inhibitors but cannot engage with the catalytic metal ions to enable the catalytic geometry means that they represent a different class of SAMHD1 inhibitor from those reported previously. 41,44,78 Figure 10 shows a comparison of the reaction mechanism schemes for canonical dNTP and R p -dNTPαS and also for inhibition by dNMPNPP and S p -dNTPαS nucleotides. In these proposed reaction mechanisms, dNTP and R p -dNTPαS ( Figure 10A,B) follow the same profile with the α-phosphates in the canonical nucleotide or αphosphorothioate and α-phosphate in the R p -dNTPαS nucleotide first coordinating the active site Fe and Mg3 respectively. The reaction then proceeds through adduction of the hydroxyl nucleophile at the α-phosphate of the ES complex to a trigonal bipyramidal transition state. Inversion of P α and breakage of the P α −O 5 ′ bond, catalyzed by His215 acting as a general acid, then results in incorporation of W0 into the newly formed triphosphate product and concomitant release of the 2′-deoxynucleoside. The proposed mechanism of inhibition by dNMPNPP nucleotides ( Figure 10C) is through increased stability of an EI complex by a Asp311 Oδ and H imido hydrogen bond. So, although the EI complex mimics the ES complex with all metal ions in place as well as the catalytic hydroxide molecule, the increased stability of the EI complex prevents formation of the transition state and bond inversion. For the S p -dNTPαS nucleotides, we now propose an alternative mechanism of inhibition ( Figure 10D). Here, although S p -dNTPαS nucleotides can bind in the active site through interactions with Fe as well as with surrounding active site side chains, they may adopt a configuration that is unable to coordinate the Mg3 metal ion and hydroxyl nucleophile. Accordingly, they represent a nonproductive EI complex that cannot assemble further into an ES complex and support catalysis.
Our results with SAMHD1 reiterate many previous observations regarding the exquisite stereoselectivity of enzymes. They show on one hand how the analysis of the differential effects of diastereomer pairs of substrate, activator, and inhibitor molecules is a powerful tool to inform on the enzyme mechanism and protein structure. Using this approach, we have uncovered two modes of competitive inhibition of SAMHD1 by nucleotide-based compounds at the active site. Type-I is exemplified by dNMPNPP nucleotides that inhibit through competition with the ES complex. Type-II, exemplified by the S p -dNTPαS nucleotides, represents a new mode of inhibition that works through competition with the initial binding of substrate nucleotides to form a transient EI complex with a conformation that does not engage the hydroxyl nucleophile. Given the need to modulate SAMHD1 activity to better understand its cellular functions, both of these modes of inhibition now provide starting points for the discovery of tool compounds that can be used to understand SAMHD1 function in HIV-1 restriction, DNA repair, and innate immune sensing. The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.biochem.0c00944. Table of X-ray data collection and refinement statistics,  table of primers used for cloning and mutagenesis, and figures describing the crystallographic AU, electron density for nucleotides bound in the allosteric and active sites, metal dependancy of Ppx1 triphosphate hydrolysis and zinc.cadmium inhibition of SAMHD1-(PDF)

Accession Codes
The atomic coordinates and structure factors of the H215A-SAMHD1(109-626)-R p -GTPαS complex have been deposited in the Protein Data Bank under accession number 7A5Y.