The SAM domain of mouse SAMHD1 is critical for its activation and regulation

Human SAMHD1 (hSAMHD1) is a retroviral restriction factor that blocks HIV-1 infection by depleting the cellular nucleotides required for viral reverse transcription. SAMHD1 is allosterically activated by nucleotides that induce assembly of the active tetramer. Although the catalytic core of hSAMHD1 has been studied extensively, previous structures have not captured the regulatory SAM domain. Here we report the crystal structure of full-length SAMHD1 by capturing mouse SAMHD1 (mSAMHD1) structures in three different nucleotide bound states. Although mSAMHD1 and hSAMHD1 are highly similar in sequence and function, we find that mSAMHD1 possesses a more complex nucleotide-induced activation process, highlighting the regulatory role of the SAM domain. Our results provide insights into the regulation of SAMHD1 activity, thereby facilitating the improvement of HIV mouse models and the development of new therapies for certain cancers and autoimmune diseases.


ABSTRACT
Human SAMHD1 (hSAMHD1) is a retroviral restriction factor that blocks HIV-1 infection by depleting the cellular nucleotides required for viral reverse transcription. SAMHD1 is allosterically activated by nucleotides that induce assembly of the active tetramer. Although the catalytic core of hSAMHD1 has been studied extensively, previous structures have not captured the regulatory SAM domain. In this study, we determined the first crystal structure of full-length SAMHD1 by capturing mouse SAMHD1 (mSAMHD1) structures in three different nucleotide bound states. Although mSAMHD1 and hSAMHD1 are highly similar in sequence and function, we found that mSAMHD1 possesses a more complex nucleotide-induced activation process, highlighting the regulatory role of the SAM domain. Our results provide new insights into the regulation of SAMHD1 activity, thereby will facilitate the improvement of HIV mouse models and the development of new therapies for certain cancers and autoimmune diseases.
The sterile alpha-motif (SAM) and histidine-aspartate (HD) domain-containing protein 1 (SAMHD1) is a dNTP phosphohydrolase that restricts viral replication by limiting the cellular dNTP pool [1][2][3][4][5][6][7] . Without an adequate supply of dNTPs, retroviruses like HIV-1 cannot complete reverse transcription. In addition to its role in the antiviral response, SAMHD1 is also implicated in the autoimmune disease Aicardi-Goutieres syndrome (AGS). Homozygous mutations in the SAMHD1 gene lead to the accumulation of nucleotides in the cell and result in symptoms that mimic a congenital viral infection 8,9 . This highlights the importance of SAMHD1 activity in the human immune system and dNTP metabolism 6,10,11 .
Interestingly, the HD domain of hSAMHD1 was found to have higher enzymatic and antiviral activities than the full-length enzyme 15 , suggesting a potential regulatory role for the SAM domain. Without a structure of full-length SAMHD1, it has previously been difficult to determine the role of the SAM domain. In addition, despite past investigations on hSAMHD1, mechanisms for mSAMHD1 activation and viral restriction remain unclear.
Two isoforms of mSAMHD1 resulting from alternative splicing share about 72-74% sequence identities with hSAMHD1 18 . These isoforms are identical until residue 593 where the C-termini differ. While isoform 1 (iso1) has a phosphorylation site at residue T634 that regulates its antiviral activity, isoform 2 (iso2) lacks this site.
Phosphorylation of the corresponding residue in hSAMHD1 (T592) has been shown to negatively regulate SAMHD1's ability to restrict HIV by affecting the formation of an active tetramer [19][20][21][22][23][24] . The effect of phosphorylation of mSAMHD1 iso1 is still unclear, although studies suggest that the T634D phosphomimetic mutation leads to loss of HIV-1 restriction in non-dividing cells, it has no effect on murine leukemia virus infection in dividing cells 18 . This suggests that different mechanisms of regulating retroviral restriction exist between hSAMHD1 and mSAMHD1.
In this study, we explore the structural and functional differences between human and mouse SAMHD1. While both mSAMHD1 and hSAMHD1 are allosterically activated by nucleotides, we find that the two enzymes have different assembly processes for the active tetramers. Furthermore, our results show that the role of the SAM domain differs between mSAMHD1 and hSAMHD1. The SAM domain is not required for dNTP hydrolysis and viral restriction by hSAMHD1, but is essential for these activities of mSAMHD1. To better understand the regulatory role of the SAM domain, we crystallized full-length mSAMHD1 in three different states (0, 1 and 2 allosteric nucleotides bound). In addition to capturing important interactions between the SAM and HD domains, these structures delineate the mSAMHD1 allosteric activation process that governs SAMHD1 enzymatic activities. These results are important for the assessment of SAMHD1 as a potential therapeutic target for HIV-1 infection and autoimmune diseases such as AGS.

SAM domain of mSAMHD1 is required for tetramerization
The oligomerization states of purified mouse and human SAMHD1 are different in the absence of added nucleotides. hSAMHD1 exists as a monomer in a nucleotide-free purification buffer, as monitored by both size exclusion chromatography (SEC) and analytical ultracentrifugation sedimentation velocity (AUC). However, under the same conditions, mSAMHD1 is predominantly a dimer with some population of tetramers ( Fig. 1a-c). After addition of the non-hydrolyzable nucleotide dGTP-α-S, hSAMHD1 readily forms tetramers, but a significant portion of mSAMHD1 remains as dimer ( Fig.   1a-c). In fact, complete tetramer formation of mSAMHD1 requires up to 24 hours of assembly with dGTP-α-S (Supplementary Fig. 1a), suggesting that mSAMHD1 has a more complex and slower process of tetramer assembly.
To determine whether the SAM domain plays a role in the tetramer formation of mSAMHD1, we repeated these assays with truncated proteins that do not contain the SAM domains, namely mouse SAMHD1 HD domain (mHD, residues 145-658 and 145-651 for iso 1 and iso 2, respectively). mHD exists in an equilibrium of monomers and dimers in the absence of allosteric nucleotides (Fig. 1f), similar to what has been observed with human SAMHD1 HD (hHD, residues 113-626) 12,15 . While only monomers of mHD were observed by SEC (Fig. 1e), a substantial amount of mHD dimers are detected by AUC (Fig. 1f). This may be due to sample dilution or the stripping of potential nucleotides co-purified with the mHD dimer during the SEC run.
Importantly, this contrasts with the behavior of full-length mSAMHD1, which readily purifies as dimers in the absence of additional nucleotides ( Fig. 1b and 1c). This suggests that in the absence of nucleotides the SAM domain is important for mSAMHD1 oligomerization. Surprisingly, upon addition of dGTP-α-S, mHD forms significantly less tetramers than hHD does ( Fig. 1e and 1f). These experiments demonstrate that mHD is deficient in forming a stable tetramer, highlighting important differences in the functions of the SAM domains of mouse and human SAMHD1. Although full-length mSAMHD1 inherently tends to oligomerize in the absence of nucleotides (Fig. 1b), this tendency is modestly enhanced in the presence of nucleotides. In this regard, mSAMHD1 is less sensitive to nucleotides for oligomerization than hSAMHD1.

The HD domain of mSAMHD1 is not sufficient for its dNTPase activity
Like hSAMHD1, mSAMHD1 is a dNTPase that reduces the dNTP level in mouse cells and restricts retroviral infection 25 . To confirm the catalytic activity of mSAMHD1, we analyzed nucleotide hydrolysis by both isoforms of the enzyme in vitro. We incubated highly purified recombinant mSAMHD1 protein with dGTP and monitored subsequent reaction products using high performance liquid chromatography (HPLC) (Fig. 1g) and a malachite green (MG) colorimetric assay (Supplementary Fig. 1b). These two independent assays show that both isoforms of full-length mSAMHD1 and hSAMHD1 exhibit similar rates of dNTP hydrolysis ( Fig. 1g and Supplementary Fig. 1b).
Surprisingly, our activity assay revealed that mHD is not sufficient for catalytic activity ( Fig. 1g and Supplementary Fig. 1b). We found that although mHD displays residual dNTPase activity over time, the hydrolysis rate is drastically reduced from that of the full-length enzyme (Fig. 1h). This is in stark contrast to hHD, which has higher dNTPase activity than full-length hSAMHD1 2,12,26-28 (Fig. 1g). Since removing the SAM domain leads to opposite effects on the activities of human and mouse SAMHD1, the regulatory role of the SAM domain likely differs between the two enzymes. Consistently, mHD was also deficient in tetramer formation ( Fig. 1e and 1f), suggesting that the SAM domain contributes to the allosteric activation of mSAMHD1.
We further assessed the HIV-1 restriction activity of mHD. While the full-length mSAMHD1 was able to restrict HIV-1 infection, mHD increased viral infection by 1.6fold compared to vector control cells (Fig. 1i). These data suggest that, unlike the HD domain of hSAMHD1 that has strong antiviral activities 26 , the HD domain of mSAMHD1 is not sufficient for its anti-HIV function.

Crystal structure of activated full-length mSAMHD1 tetramer
We determined the crystal structures of the full-length mSAMHD1 protein, which furnishes our understanding of the fully functional state of the enzyme. The structure of full-length hSAMHD1 has remained elusive, despite extensive effort by many groups; all published structures of hSAMHD1 only contain the catalytic HD domain 2,15,17 . Although a full-length hSAMHD1 protein was recently crystallized, the SAM domain could not be visualized, indicating inherent flexibility or disorder 29 . A structural understanding of the full-length SAMHD1 is important as the SAM domain is required for mSAMHD1 catalysis and is likely important for regulating hSAMHD1 activity.
The structure of the full-length mSAMHD1 iso1 protein with both allosteric sites occupied (GTP in Allo-site1 and dGTP in Allo-site2) was determined to a resolution of 3.5 Å, and hereafter is referred to as 2-Allo ( Fig. 2a and Table 1). The 2-Allo structure was determined as a tetramer in I222 space group with one copy of the protein in each asymmetric unit of the crystal. The majority of the SAM and HD domains (residues 71 to 624) were built into the structure, although the C-terminal region including the reported phosphorylation site (T634) 15,16,18,20 was not observed. Analysis of the 2-Allo structure in comparison to hHD structure reveals that the tetramer interface is fairly disordered with more than 25 residues unresolved (Supplementary Fig. 2a). This is the first full-length SAMHD1 structure that captures the relationship between the SAM and HD domains, and shows that each SAM domain bridges three HD domains at the tetramer interface ( Fig. 2a and 2b).  Fig. 2b). The mHD tetramer has an RMSD of less than 0.6 Å when compared to hHD tetramer (PDBID 4BZB) 16 ( Supplementary Fig. 2c). Although the overall architecture and fold of the individual HD domains are conserved between human and mouse, there are large differences in their relative orientations which affect the overall compactness of the tetramers (discussed in latter sections). The most significant deviation between mSAMHD1 and hSAMHD1 subunit structures occurs at the linker region between the mHD and mSAM domains, which is the last observed region of the hSAMHD1 structure crystallized in the fulllength background 29 (black arrow, Supplementary Fig. 2d). This is consistent with the flexibility between the SAM and HD domains of hSAMHD1 and may explain why crystallization attempts have been unsuccessful with full-length hSAMHD1. Since mSAMHD1 and hSAMHD1 proteins share 72% sequence identity 18 and the overall folds of both SAM and HD domains are highly conserved, we believe our mSAMHD1 structure provides an excellent structural model for full-length hSAMHD1.

SAM domain of mSAMHD1 is essential for capping allosteric sites
Our 2-Allo mSAMHD1 structure shows that the SAM domain contributes to the formation of the allosteric site pocket by interacting with the HD domains to stabilize the allosteric nucleotides (Fig. 2c). Each allosteric site is formed as a deep cleft at the interface of HD domains from three subunits ( Fig. 2b and 2c). Each SAM domain interacts intra-and inter-molecularly to contact the three HD domains, capping the nucleotides bound in the allosteric site pocket (Fig. 2c). These interactions explain the important role of the SAM domain in stabilizing the active tetramer ( Fig. 1b and 1c) and hence the subsequent effects on mSAMHD1 enzymatic and antiviral activities ( Fig. 1g and 1i). mSAMHD1's dependence on the SAM domain for allosteric activation is a striking feature that differentiates it from hSAMHD1. It is possible that the SAM domain has a different structural or regulatory role in the context of hSAMHD1.

SAM-to-HD interactions are important for mSAMHD1 oligomerization and activity
We further determined that the interactions between the SAM and HD domains are critical for mSAMHD1 enzymatic and antiviral activities, consistent with our biochemical and structural data that the SAM domain is important for both dimerization and tetramerization of the molecule. Located at the interface of three HD domains, the SAM domain engages in various interactions with HD domains from the same (intrasubunit) and adjacent subunits (inter-subunit) in the tetramer (Fig. 2b). These interactions help position the SAM domain in the correct orientation to cap the allosteric sites ( Fig.   2b and 2c). The residues involved in these interactions are not conserved between the otherwise highly homologous mouse and human SAMHD1 proteins ( Supplementary   Fig. 3). The SAM domain is known to be dispensable for hSAMHD1 enzymatic and antiviral activities 2,12,26-28 . The functional integrity of hHD suggests that the critical SAM-to-HD interactions observed in mSAMHD1 are not required in hSAMHD1. This notion is supported by modeling the corresponding hSAMHD1 residues into the SAM-to-HD interfaces of mSAMHD1. This model predicts the disruption of the observed tetramer-stabilizing interactions ( Fig. 3 and Supplementary Fig. 3

and 4a).
To validate the importance of the observed SAM-to-HD interfaces, we first assessed the inter-subunit interaction's contribution to stabilizing the mSAMHD1 tetramer and maintaining its function. We tested the effect of mutating the key mSAMHD1 SAM-to-HD residues to their hSAMHD1 counterparts (Supplementary Fig.   3). At the inter-subunit interface, residues S566 and N567 of one HD domain interact across subunits with S142 and Q137 of the SAM domain from an adjacent mSAMHD1 molecule (Fig. 3a). Modeling a S142I/S566T/N567A triple mutant (ITA) onto our 2-Allo structure showed that S142I and S566T result in a steric clash and N567A disrupts a salt bridge with Q137 (Fig. 3a). As expected, our experimental results showed that the ITA mutant was defective in forming a tetramer and displayed a substantial reduction in dNTPase activity (Fig. 3b). Similarly, this ITA mutant did not confer HIV-1 restriction compared to full-length wild-type mSAMHD1 (Fig. 3c), demonstrating that these SAM-HD inter-subunit interactions are important for mSAMHD1 function in vivo.
Next, we assessed the importance of the intra-subunit SAM-to-HD interactions in oligomerization and enzymatic function (Fig. 2b). At the mSAMHD1 intra-subunit interface, SAM domain residue R143 forms a salt bridge with HD domain residue E225 and the aromatic residues F109 and F112 from the SAM domain form stacking interactions with H161 and Y289 from the HD domain, respectively (Fig. 3d). We swapped these residues with their human counterparts to make the triple mutant F109L/F112C/R143H (LCH) (Supplementary Fig. 3). Modeling of these mutations shows that they disrupt the electrostatic and stacking interactions that position the SAM domain to stabilize the tetramer (Fig. 3d). Indeed, the LCH mutation completely abolished mSAMHD1 tetramer formation and dNTPase activity (Fig. 3b). This mutation also significantly reduced the HIV-1 restriction function of mSAMHD1, albeit the mutant expressed lower than WT protein (Fig. 3c).
In a complimentary study, we created the L77F/C80F/H111R (FFR) mutations in hSAMHD1 (Supplementary Fig. 4a) to introduce the SAM-to-HD interactions observed in mSAMHD1. Because these interactions are important for stabilizing the mSAMHD1 tetramer, we expected their introduction into hSAMHD1 to have a similar tetramerstabilizing effect. Indeed, our results show that this triple mutant displayed higher dNTPase activity and formed a more stable tetramer than WT hSAMHD1 (Fig. 3e).
Correspondingly, we observed a 2-fold increase in HIV-1 restriction by the FFR mutant compared to full-length WT hSAMHD1 (Fig. 3f). These data demonstrate that enhancing interactions between the SAM and HD domains of hSAMHD1 lead to a more stable tetramer with higher enzymatic and HIV-1 restriction activities.

Crystal structures of allosteric intermediates indicate a multi-step activation process
In addition to the 2-Allo structure, we also captured crystal structures of fulllength mSAMHD1 in two more activation intermediate states: no allosteric nucleotides bound (No-Allo) and one allosteric nucleotide bound to Allo-site 1 only (1-Allo) (Fig.   4a). The No-Allo crystal was obtained without the addition of nucleotides in the crystallization mixture and diffracted to 3.5 Å resolution. The 1-Allo crystal was crystallized with dGTP-α-S in the crystallization mixture and diffracted to 3.4 Å resolution. The No-Allo and 1-Allo mSAMHD1 structures also crystallized as tetramers in different space groups (C2 and P2 1 2 1 2) with 4 or 2 copies of mSAMHD1 in the asymmetric unit, respectively.
To confirm that these intermediate states, which lack the necessary nucleotides at Allo-site 2, are inactive forms of the enzyme, we tested whether the residues that interact with the Allo-site 2 nucleotides are critical for catalysis (Supplementary Fig. 4b and   4c). These residues at the Allo-site 2 pocket do not interact with the Allo-site 1 nucleotides nor directly contribute to tetramerization. Mutating these residues in both isoforms of mSAMHD1 abolished dNTPase activity (Supplementary Fig. 4c). These results demonstrate the importance of nucleotide-recognition residues, and hence nucleotide binding at Allo-site 2, for catalytic activity. Although mSAMHD1 is capable of forming tetramers without nucleotides bound to Allo-site 2, these states are inactive and likely represent activation intermediates of the enzyme.
To understand the differences between the three tetramer states, we systemically analyzed the conformations of the mSAMHD1 subunits that assemble the tetramer (Supplementary Fig. 5). First, we aligned the subunits within each structure and between the three states and found that their RMSDs are very low (0.5-1.0 Å) ( Supplementary Fig. 5a-c). This indicates that the overall conformations of the individual monomeric subunits that make up the tetramer do not change much between the states. Both the No-Allo and 2-Allo tetramers are symmetrical and obey the 222 point-group symmetry with three perpendicular 2-fold axes. However, this symmetry is broken in the 1-Allo tetramer, which is a dimer of dimers, with only one overall 2-fold axis. Interestingly, the dimeric building block (designated as dimer A) remains rigid with minor conformational variations (RMSDs 0.7-1.3 Å) among the three different tetramer states ( Fig. 4b and 4c, and Supplementary Fig. 5d and 5e). This suggests that the dimer may serve as an intermediate assembly unit for the formation of the active mSAMHD1 tetramer.
Despite relatively rigid monomer and dimer building blocks, the three mSAMHD1 tetramers (No-Allo, 1-Allo, and 2-Allo) (Fig. 4a) display large conformational differences (RMSDs 2.4-3 Å) with varying compactness (Fig. 4d and e and Supplementary Fig. 5f). Aligning the three structures by the rigid dimer A highlights the relative movement of the other half of the tetramer (designated as dimer Fig. 4d). Starting from the most open No-Allo state, dimer A' swings progressively towards dimer A via the 1-Allo-state to close the tetramer interface in the 2-Allo state ( Fig. 4d and Supplementary Video 1). The consequence of the swinging motions is a formation of a compact tetramer highlighted by the distance between the SAM domains of dimers A and A' (Fig. 4e and Supplementary Video 2). These significant rearrangements between the two dimers occur as nucleotides sequentially bind the allosteric sites to activate the enzyme.

DISCUSSION
SAMHD1 plays an important role in controlling cellular dNTP pool levels and has been identified as an HIV-1 restriction factor in non-dividing cells 1,3,31 . Additionally, mutations in SAMHD1 are associated with autoimmune disease states, such as AGS 8 . In this study, we provide a rigorous biochemical analysis and crystal structures of fulllength mouse SAMHD1 that for the first time capture the interactions between the SAM and HD domains. This study establishes a mechanistic framework for understanding the key differences between human and mouse SAMHD1 activities, allosteric activation and retroviral restriction properties of mSAMHD1.
Our results show that a key difference between the human and mouse enzyme is that mSAMHD1, but not hSAMHD1, requires the SAM domain for dNTPase and antiviral activities. This is likely due to the SAM domain's contribution to nucleotide binding at the allosteric site. By capping the allosteric nucleotides in the pocket, the SAM domain is essential to the formation of active mSAMHD1 tetramers. Our analysis shows that SAM-to-HD interactions are crucial for mSAMHD1 oligomerization and function.
Mutating key residues involved in SAM-to-HD interactions substantially decreased the ability of mSAMHD1 to form a tetramer, which results in significantly reduced dNTPase and antiviral activities. While mSAMHD1 depends on its SAM domain to stabilize the active tetramer, residues involved in these interactions are not conserved in hSAMHD1.
Consistent with their tetramer-enhancing role, we were able to engineer these SAM-to-HD interactions into hSAMHD1 to create a more stable tetramer with higher dNTPase and antiviral activities. This hyper-stable tetramer hSAMHD1 construct may be useful for future biochemical and structural studies that will provide insights into the effect of the SAM domain in regulating the activities of hSAMHD1.
The SAM domains of hSAMHD1 and mSAMHD1 may play different roles in the regulation of enzymatic and other cellular functions. In contrast to mSAMHD1, hSAMHD1 does not require the SAM domain for stable formation of the tetrameric catalytic core. In fact, the hHD domain is more active than the full-length protein, so it is possible that the SAM domain imposes a negative regulatory effect on the catalytic activity of hSAMHD1. These opposing effects of SAM domain on mSAMHD1 and hSAMHD1 activities might reflect differences in the sensitivity to nucleotides. It has been shown that the nucleotides levels in human cells are lower than those in mouse cells 32 . As a result of being exposed to higher concentrations of nucleotides, mHD may have evolved lower affinities for nucleotide activators whose binding are enhanced by the SAM domain. Conversely, the hHD itself may have developed a higher affinity for nucleotides at relatively lower nucleotides concentrations. It also remains possible that the SAM domain plays other important roles in regulating these enzymes in the cell or processes yet undiscovered functions.
The existence of mSAMHD1 activation intermediates may also explain the previous biochemical observations that mSAMHD1 is active in the absence of additional nucleotide activators 33 . Bloch et al. showed that mSAMHD1 tetramer has dNTPase activity in the absence of additional dGTP and suggested that this activity is from the nucleotides bound to SAMHD1 during purification 33 . We also observed dimer and tetramer species during our purification of mSAMHD1, thus we tested whether these states of mSAMHD1 exhibit dNTPase activity in the absence of added GTP/dGTP. We carried out a dNTPase activity assay in the presence of dATP, which cannot bind to Allosite 1 to assemble an active tetramer. Thus, activity would only be detected if mSAMHD1 is purified with Allo-site 1 GTP or dGTP bound. We found that the dimer of mSAMHD1 did not exhibit hydrolysis activity similarly to the inactive RN control (Supplementary Fig. 1c), indicating that mSAMHD1 assembles dimers in the absence of any bound nucleotides. In contrast, the purified tetramer did hydrolyze dATP, albeit at a very slow rate, requiring incubation times up to 2 hours (Supplementary Fig. 1c). Since our experiment did not include GTP or dGTP, which are the only allosteric activators to bind Allo-site 1 15-17 , we conclude that some populations of mSAMHD1 protein were purified with GTP or dGTP bound to the allosteric site 1, consistent with the previous observations 33 . These tetramers likely represent activation intermediates similar to our 1-Allo structure, and they are not fully active states of the enzyme.
Our study indicates that the allosteric activation of mSAMHD1 and hSAMHD1 is considerably different. While hSAMHD1 transitions from a monomer or dimer to a tetramer with the addition of allosteric nucleotides, mSAMHD1 seems to have a more complex landscape of oligomerzation states. In the absence of nucleotides, mSAMHD1 exists in an equilibrium between stable dimers and inactive tetramers. Three tetramer states exist with 0 (No-Allo), 1 (1-Allo), or 2 (2-Allo) nucleotide(s) bound at each allosteric site. Our No-Allo and 1-Allo structures show that these inactive tetramer intermediates are composed of two relatively rigid dimers with loose tetramer interfaces.
In the presence of similar concentrations of nucleotides, mSAMHD1 forms fewer tetramers in solution than hSAMHD1. Our 2-Allo mSAMHD1 structure shows that even this activated form of mSAMHD1 is still less compact than hSAMHD1, indicating an overall less stable tetramer.
A model of the mSAMHD1 activation process emerges from our structural and biochemical results, which includes a series of inactive tetramer intermediates with increasing numbers of allosteric nucleotides bound (Fig. 5). Initially, inactive mSAMHD1 exists in solution predominantly as dimers that can associate into loose tetramers. Nucleotide binding to Allo-site 1 causes the dimer components to swing around a pivot point and close in towards each other. Subsequent nucleotide binding to Allo-site 2 brings the dimers even closer together and positions the SAM domains to cap the allosteric sites for stable binding of the allosteric nucleotides. This activation model differs from the one proposed for hSAMHD1, as there is no evidence to suggest that hSAMHD1 can form tetramers in the absence of nucleotides. In the case of hSAMHD1, nucleotides are required for oligomerization whereas for mSAMHD1, nucleotides are required for conformational change and activation of the preformed tetramers.
Although mSAMHD1 appears to form less tetramers than hSAMHD1 at the same nucleotide concentration, their dNTPase activities are similar, indicating a higher activity for activated mSAMHD1. In fact, it has been previously shown that hSAMHD1 has a lower K cat than that of mSAMHD1 34 . This could be advantageous as mouse cells have a higher concentration of nucleotides than that of human cells 1,35 . Thus, mSAMHD1's relative insensitivity to nucleotides, and its slow and complex activation, ensure that mSAMHD1 is not continuously hydrolyzing dNTPs. These differences in oligomerization and enzymatic activity may help to explain the differences between human and mouse SAMHD1 and inform the improvement of a mouse model for the study of HIV-1, certain cancers and autoimmune diseases.

Protein expression and purification. N-terminal 6×His-tagged mSAMHD1 constructs
were expressed in Escherichia coli and purified using Ni-NTA affinity and size-exclusion chromatography, as previously described 15 . Crystallization and data collection mSAMHD1 protein in SAMHD1 buffer was mixed with various combinations of nucleotides at 0-7 mM concentrations (Table 1) and incubated at 4 °C for 15 min before crystallization. The apo WT structure was crystallized in the absence of nucleotides, the 1-Allo WT structure was co-crystallized in the presence of 4 mM dGTP-α-S, with the crystals optimized after rounds of streak seeding. The 2-Allo RN structure was cocrystallized with 1mM GTP and 6 mM dGTP. All crystals were grown at 25 °C using the microbatch under-oil method by mixing 1 µL protein (5 mg/mL) with 1 µL crystallization buffer (100 mM SPG (Qiagen) buffer, pH 7.4, 25% PEG 1500). Crystals were cryoprotected by crystallization buffer supplemented with 25% (Vol/Vol) glycerol before freezing in liquid nitrogen. Diffraction data were collected at the Advanced Photon Source beamline 24-ID C at wavelength 0.91640 Å (2-Allo), 24-ID E at wavelength 0.97920 Å (1-Allo) and 24-ID E at wavelength 0.97918 Å (0-Allo) and temperature 100 K. The data statistics are summarized in Table 1.

Structure determination and refinement
The structures were solved by molecular replacement using PHASER 37

Malachite green colorimetric assay of SAMHD1 activity
All assays were performed at 25°C in SAMHD1 buffer. Each 40 uL reaction, containing 10 uM pyrophosphatase, 0.5 uM SAMHD1, and 125 uM substrate or allosteric activator was quenched with 40 uL 20 mM EDTA after 15 minutes. Subsequently, 20 uL Malachite Green reagent was added to the solution and developed for 15 minutes before the absorbance at 650 nm was measured 41 .

HPLC assay of SAMHD1 activity
Standard SAMHD1 dNTPase activity assays were performed in SAMHD1 buffer containing 0.5 mM GTP in a 500 µl reaction volume, with substrate dNTP at various concentrations. Reactions were incubated at 37 °C and initiated by the addition of purified SAMHD1 samples to a final concentration of 500 nM. Aliquots of reactions were removed at various time points and terminated by 5 X dilution into ice cold buffer containing 10 mM EDTA, followed by deproteinization by spinning through an Amicon Ultra 0.5-ml 10-kDa filter (Millipore) at 16,000 X g for 20 min. Deproteinized samples were analyzed by HPLC using a Synergi C18 column 150 X 4.6 mm (Phenomenex), preequilibrated in 20 mM ammonium acetate, pH 4.5 (buffer A). Samples were eluted with a methanol (buffer B) gradient over 14 min at a flow rate of 1 ml/min. The elution profiles (UV absorption at 260 nm) were recorded as described previously 20,42 .

HIV-1 restriction assay
Stable U937 cells expressing full-length or mutant SAMHD1were generated and cultured as described 43

COMPETING FINANCIAL INTERESTS:
The authors declare no competing financial interest.