The C-terminal domain of feline and bovine SAMHD1 proteins has a crucial role in lentiviral restriction

SAM and HD domain-containing protein 1 (SAMHD1) is a host factor that restricts reverse transcription of lentiviruses such as HIV in myeloid cells and resting T cells through its dNTP triphosphohydrolase (dNTPase) activity. Lentiviruses counteract this restriction by expressing the accessory protein Vpx or Vpr, which targets SAMHD1 for proteasomal degradation. SAMHD1 is conserved among mammals, and the feline and bovine SAMHD1 proteins (fSAM and bSAM) restrict lentiviruses by reducing cellular dNTP concentrations. However, the functional regions of fSAM and bSAM that are required for their biological functions are not well-characterized. Here, to establish alternative models to investigate SAMHD1 in vivo, we studied the restriction profile of fSAM and bSAM against different primate lentiviruses. We found that both fSAM and bSAM strongly restrict primate lentiviruses and that Vpx induces the proteasomal degradation of both fSAM and bSAM. Further investigation identified one and five amino acid sites in the C-terminal domain (CTD) of fSAM and bSAM, respectively, that are required for Vpx-mediated degradation. We also found that the CTD of bSAM is directly involved in mediating bSAM's antiviral activity by regulating dNTPase activity, whereas the CTD of fSAM is not. Our results suggest that the CTDs of fSAM and bSAM have important roles in their antiviral functions. These findings advance our understanding of the mechanism of fSAM- and bSAM-mediated viral restriction and might inform strategies for improving HIV animal models.

Sterile alpha motif and histidine-aspartate-domain-containing protein 1 (SAMHD1) is an antiviral triphosphohydrolase that can reduce the concentration of intracellular deoxynucleo-side triphosphates (dNTPs) 2 below a usable threshold for efficient HIV type 1 (HIV-1) replication (1)(2)(3)(4)(5). Activation of this catalytic activity relies on SAMHD1 tetramerization initiated by GTP/dGTP binding (6 -10). Recently, a RNase activity has been described as the second enzymatic antiviral mechanism of SAMHD1, but conflicting results were reported (11)(12)(13). In addition to HIV-1, SAMHD1 also has been shown to restrict a variety of other retroviruses, such as HIV-2 and simian immunodeficiency virus (SIV) (14,15). As counter mechanisms, the viral protein X (Vpx) of HIV-2 and some strains of SIV antagonize the antiviral function of SAMHD1 by recruiting SAMHD1 to the E3 ubiquitin ligase complex CRL4 DCAF1 consisting of DCAF1, DDB1, CUL4, and RBX1 (16 -20). This recruitment leads to proteasomal degradation of SAMHD1 through which Vpx relieves SAMHD1-mediated restriction of lentiviral replication in myeloid-lineage cells and resting T cells (18 -24). The Vpx expressed by each of distinct HIV-2 and SIV strains is unique in terms of sequence, and structural and functional studies of SAMHD1 have identified unique features of the Vpx-SAMHD1 interaction depending on the source/sequence of Vpx (19,(25)(26)(27)(28). SAMHD1 consists of a conserved N-terminal sterile alpha motif (SAM) domain containing a nuclear localization signal that determines its intracellular distribution (29,30), a catalytic histidine-aspartate (HD) domain that is proposed to have the dNTP triphosphohydrolase (dNTPase) and nuclease activities (1,7,10,31), and a C-terminal domain (CTD) that is recognized by Vpx of HIV-2/SIVmac (macaque)/SIVsm (sooty mangabey) lineage for degradation and stabilizes the tetramerization of SAMHD1 (20,26,27,32,33). SAMHD1 is conserved among primate and nonprimate mammals such as mice, cats, and cows. To elucidate the in vivo physiological functions of SAMHD1, small animal models can be used as a tool to dissect the contribution of SAMHD1 to antiviral immunity in vivo. SAMHD1-knockout mice are readily established and the studies show that dNTP levels are increased and the restriction of VSV-G-pseudotyped HIV-1 is relieved in SAMHD1-knockout cells and mice (34,35). However, mice are not the natural hosts of HIV-1 and not subject to lentiviruses, and loss of SAMHD1 does not result in increased viral loads in mice infected with murine leukemia virus (35). Thus, the role of SAMHD1 in mice is unclear. By contrast, feline and bovine SAMHD1 (fSAM and bSAM) could restrict the lentiviruses of their own host species (feline and bovine immunodeficiency viruses, FIV and BIV, respectively) by using their dNTPase activities (36). FIV and BIV also infect nondividing myeloid cells during their pathogenesis (37)(38)(39)(40) and can cause immune dysfunction like HIV (41,42). Therefore, SAMHD1-knockout cats, and potentially cows, may be better models to study HIV pathogenesis and the biological function of SAMHD1, and the molecular mechanisms underlying feline and bovine SAMHD1-mediated restriction of HIV are worth being investigated.
Here, we investigate the antiviral ability of fSAM and bSAM against different primate lentiviruses and the crucial regions in fSAM and bSAM that are required for their biological functions. We identified key sites in the C-terminal regions of fSAM and bSAM which are required for Vpx-mediated degradation and revealed the role of bovine SAMHD1 CTD in regulating the catalytic function and antiviral activity. Our findings will provide insights into the establishment of alternative models to investigate SAMHD1 in vivo.

Antiviral activity of fSAM and bSAM against primate lentiviruses
fSAM and bSAM are able to restrict FIV and BIV in SAMHD1-knockout THP-1 cells (36). To facilitate the in vivo study of the anti-HIV activity of SAMHD1, we extended the analysis to their antiviral abilities against HIV-1, HIV-2, and SIV in vitro. We first constructed HA-tagged fSAM and bSAM expression vectors and tested their expression in TZM-bl cells. Human SAMHD1 (hSAM) was also expressed as a positive control, and cellular GAPDH protein was used as a loading control ( Fig. 1A). Next, we developed a quick detection method to determine the antiviral activity of fSAM and bSAM by using TZM-bl cells. This cell line contains an HIV-1 Tat-regulated ␤-gal reporter gene that can be expressed after a single round of infection with not only HIV-1, but also HIV-2 and SIV (43). First, TZM-bl cells were transfected with human, feline or bovine SAMHD1 expression vector or empty vector. Next, the cells were detached at 48 h post transfection and re-seeded into 24-well plates with viral infection solutions containing HIV-1 (NL4 -3), SIVmac239, or HIV-2 ROD. 48 h later, the infected cells were stained with the substrate of ␤-gal (X-Gal) and the number of positive blue cells was counted, by which the infectivity of the three viruses could be calculated. Consistent with previous findings (29), hSAM showed strong antiviral activities against HIV-1, SIV, and HIV-2, which confirms the validity of this method (Fig. 1, B-D). fSAM and bSAM could also restrict the three viruses as efficiently as hSAM.
To confirm this result, we generated hSAM-, fSAM-, and bSAM-expressing stable U937 cell lines which do not express endogenous SAMHD1 and stable U937 cell lines expressing catalytically inactive human, feline and bovine SAMHD1 proteins with alanine mutations in the HD domain (hSAM HD/AA , fSAM HD/AA , and bSAM HD/AA ) (Fig. S1). After differentiated with PMA, the cells were infected with HIV-1, SIVmac239, or HIV-2 ROD. p24 antigen or reverse transcriptase (RT) levels in the culture supernatants were monitored for 5 days to determine the dynamic changes of viral replication. The results showed that hSAM, fSAM, and bSAM could significantly restrict the viral replication of these three viruses, whereas the catalytically inactive mutants could not (Fig. 2, A-C). Because Vpx expressed from SIVmac239 and HIV-2 ROD can induce the degradation of hSAM, the restriction efficiency of hSAM against SIVmac239 and HIV-2 ROD was lower than that against HIV-1 (Fig. 2, A-C). Similar phenomena were observed in fSAM and bSAM, indicating that Vpx may also mediate the degradation of these two SAMHD1 proteins. Immunoblot 48 h later, the infected TZM-bl cells were stained with X-Gal and the positive blue cells were counted under optical microscope. Viral infectivity was determined by the number of blue cells. The viral infectivity in the empty vector-transfected cells was set to 100% (positive control, PC). Error bars represent the S.D. calculated from three independent infections. Statistical analysis was performed between the indicated groups using Student's t test.

Function of C-terminal domains of feline and bovine SAMHD1
analysis of the U937 cells showed that the three pairs of WT SAMHD1 proteins and their catalytically inactive mutants were expressed at similar levels, respectively (Fig. 2D). We also constructed another catalytically inactive mutant for each of human, feline, and bovine WT SAMHD1 proteins (hSAM H233A , fSAM H233A , and bSAM H221A ) (Fig. S1) and tested their antiviral activities against HIV-1, SIVmac239, and HIV-2 ROD using TZM-bl cells. As expected, the antiviral activity of these mutants was largely reduced (Fig. 2, E-G). We then tested the dNTPase activity of hSAM, fSAM, bSAM, and their catalytically inactive mutants in vitro by a malachite green-coupled dGTP-pyrophosphatase hydrolysis assay (44). In this assay, the final product inorganic phosphate (P i ) of the dGTP substrate hydrolysis by the SAMHD1 proteins was quantified to deter- Error bars represent the S.D. calculated from three independent infections. Statistical analysis was performed between each group of the WT proteins and their mutants and between each mutant and PC using Student's t test. H, in vitro detection of SAMHD1-catalyzed inorganic phosphate (P i ) release. HA-tagged SAMHD1 proteins were isolated from transfected HEK293 cells by immunoprecipitation. An aliquot of the immunoprecipitated SAMHD1 proteins was analyzed by Western blotting to ascertain comparable protein using anti-HA antibody. The levels of P i released after in vitro dGTP-pyrophosphatase hydrolysis reactions were detected by malachite green. Error bars represent the S.D. calculated from three independent reactions.

Function of C-terminal domains of feline and bovine SAMHD1
mine the dNTPase activity. The results confirmed that only few dGTP could be hydrolyzed by the catalytically inactive mutants (Fig. 2H). Together, these results demonstrated that feline and bovine SAMHD1 could restrict the infection and replication of primate lentiviruses by using their dNTPase activities.

Proteosomal degradation of feline and bovine SAMHD1 by Vpx
Vpx encoded by SIVmac and HIV-2 target the C-terminal domain of hSAM for proteosomal degradation to antagonize the antiviral function of hSAM (18 -20, 45). HIV-1 does not encode a Vpx protein, and its viral protein R (Vpr) does not counteract hSAM (25). It has been reported that FIV and BIV fail to proteosomally degrade the SAMHD1 proteins of their hosts (36), and whether their genomes contain a certain accessory protein that functions as Vpx is unclear. To understand whether Vpx from primate lentivirus could mediate the degradation of fSAM and bSAM, we first examined the degradation of hSAM by SIVmac239 with or without Vpx as the control of degradation assay. HEK293 cells were co-transfected with the hSAM expression vector and SIVmac239 protein-expressing plasmid either with or without Vpx (pSIV or pSIV⌬Vpx). The expression levels of SAMHD1 and SIVmac239 were determined with anti-HA and anti-SIV p27 antibodies. As shown in Fig. 3A, WT SIVmac239 reduced the expression level of hSAM to 31.6% and the level was retrieved to 88.3% without the Vpx expression from SIVmac239 (lanes 3 and 4). Then, we tested the degradation of fSAM and bSAM by the WT or Vpx-deleted SIVmac239. Both of the proteins could be degraded by WT SIVmac239 but only about 15 and 30% of the total fSAM and bSAM were degraded (Fig. 3, B and C). To confirm the degradation result of fSAM, we repeated the immunoblot analysis in a feline kidney epithelial cell line, CRFK. We found that the degradation percentage of fSAM (about 45%) in the cell line of its own species was much higher than that in HEK293 cells (Fig.  3D). We also repeated the degradation assay of bSAM in Madin-Darby bovine kidney (MDBK) cells. However, because of the low transfection efficiency of this cell line, we did not obtain a clear and convincing result.
Next, we investigated whether fSAM and bSAM were degraded via the proteosomal pathway. HA-tagged human, feline, or bovine SAMHD1 proteins were co-expressed with HA-tagged Vpx from the SIVmac239 (Vpx mac ) or HIV-2 ROD (Vpx ROD ) strain or empty vector in HEK293 cells in the presence of the proteasome inhibitor MG132 or DMSO as a negative control. The results showed that MG132 could block the Vpx mac -and Vpx ROD -induced degradation of fSAM and bSAM (Fig. 3, E-G and Fig. S2), indicating that fSAM and bSAM are both degraded proteasomally in HEK293 cells. We also infected were co-transfected with 600 ng of (A) pVR1012-homo-SAMHD1-HA, (B) pVR1012-feline-SAMHD1-HA, or (C) pVR1012-bovine-SAMHD1-HA and 1.5 g of pSIVmac239 or pSIVmac239⌬Vpx or empty vector. D, CRFK cells (1 ϫ 10 6 ) were co-transfected with 600 ng of pVR1012-feline-SAMHD1-HA and 1.5 g of pSIVmac239 or pSIVmac239⌬Vpx or empty vector. Cells were harvested at 48 h post transfection and then analyzed by Western blotting using anti-HA, anti-SIV p27, and anti-GAPDH antibodies. The percentage of SAMHD1 in the presence of pSIVmac239 or pSIVmac239⌬Vpx was calculated relative to that of the corresponding SAMHD1 in the absence of them (set to 100%). E-G, HEK293 cells (1 ϫ 10 6 ) were co-transfected with 600 ng of (E) pVR1012-homo-SAMHD1-HA, (F) pVR1012feline-SAMHD1-HA, or (G) pVR1012-bovine-SAMHD1-HA and 500 ng of pCG-Vpx mac -HA or empty vector. The transfected cells were treated with the proteasome inhibitor MG132 at 20 M or DMSO as a negative control at 36 h after transfection. Cells were harvested 12 h later (48 h post transfection) and then analyzed by Western blotting using anti-HA and anti-GAPDH antibodies. The percentage of SAMHD1 in the presence of Vpx mac with DMSO or MG132 treatment was calculated relative to that of the corresponding SAMHD1 in the absence of Vpx mac (set to 100%). H, empty U937 cells (5 ϫ 10 5 ) or U937 cells stably expressing each WT SAMHD1 protein were differentiated with PMA for 20 h and infected with SIVmac239 or SIVmac239⌬Vpx (1 ng of RT). The viral replication curves were measured by the concentration of RT in the culture supernatants collected on the day of infection (day 0) and at day 1, 3, and 5 post infection. Error bars represent the S.D. calculated from three independent infections.

Function of C-terminal domains of feline and bovine SAMHD1
the U937 cell lines stably expressing hSAM, fSAM, and bSAM with SIVmac239 and Vpx-deleted SIVmac239 (SIV⌬Vpx) viruses and quantified the RT levels in the culture supernatants for 5 days. The restriction efficiency of the three SAMHD1 proteins against Vpx-deleted SIVmac239 was higher than that against WT SIVmac239 (Fig. 3H), indicating that feline and bovine SAMHD1 were partially degraded by SIVmac239 Vpx in U937 cells. Collectively, these results demonstrated that Vpx is able to mediate the proteosomal degradation of feline and bovine SAMHD1, although this counteraction is weaker than that between Vpx and human SAMHD1.

The function of fSAM and bSAM CTD related to Vpx-induced degradation
Alignment of the C-terminal amino acid sequences of human, simian, feline, and bovine SAMHD1 proteins revealed sequence divergence in the CTD of fSAM and bSAM compared with that of primate SAMHD1 (Fig. 4A), especially in bSAM that lacks a typical C-terminal sequence. To determine how these domains mediate Vpx-induced degradation of fSAM and bSAM, we first constructed an hSAM variant truncated at the C terminus (hSAM 1-600 , truncated from amino acid 601 to 626) as a control. As shown previously (45), SIVmac239 Vpx was unable to degrade C-terminally truncated human SAMHD1 (Fig. 4B, lane 4). Next, we constructed a CTD-truncated fSAM variant (fSAM 1-601 , truncated from amino acid 602 to 627) and about 28% of fSAM in the HEK293 cells was retrieved (Fig. 4C, lane 4), suggesting that the CTD of fSAM contributed to the degradation induced by Vpx. To investigate whether hSAM CTD (hCTD, residue 601 to 626) containing the Vpx recognition sequence can increase the sensitivity of fSAM and bSAM to Vpx-mediated degradation, we then constructed a fSAM-hSAM chimeric protein by swapping the fSAM 1-601 with hCTD (fSAM 1-601 ϩhCTD) and a bSAM-hSAM chimeric protein by adding the hCTD to the C-terminal end of bSAM (bSAMϩhCTD). The co-transfection results in HEK293 cells demonstrated that the fSAM 1-601 ϩhCTD chimera was more sensitive to Vpx-mediated degradation than its WT SAMHD1 protein but the bSAMϩhCTD chimera was not (Fig. 4C, lanes 6 and 10), indicating that bSAM has its own sequence other than the typical CTD sequence for mediating Vpx-induced degradation. To find this sequence, we constructed C-terminally truncated bSAM variants (bSAM 1-572 and bSAM 1-580 , truncated off the last 17 and 9 amino acids, respectively) of which the region showing the lowest similarity to the SAMHD1 proteins from other species (Fig. 4A) was segmentally deleted. Compared with WT bSAM, the level of bSAM  in HEK293 cells was about 31% higher in the presence of SIVmac239 Vpx, whereas no difference was observed between WT bSAM and bSAM   (Fig. 4C, lanes 12 and 14), suggesting that the amino acid sequence from residue 573 to 580 is crucial to bSAM degradation mediated by Vpx.
To identify which sites might be necessary for fSAM and bSAM CTD-Vpx binding and subsequent degradation, we performed computational homology modeling studies based on the hSAM CTD-SIVsm Vpx-DCAF1 CTD X-ray crystal structure (26) using Discovery Studio to predict the binding models of fSAM and bSAM CTD with SIVmac239 Vpx and human DCAF1 CTD. The modeling results showed one possible site in fSAM CTD interacting with Vpx mac , which is a hydrogen bond between Thr-619 of fSAM and Tyr-65 of Vpx mac helix 3 (Fig.  4D). For bSAM, although it lacks more than 20 amino acids at the C terminus compared with hSAM and fSAM (Fig. 4A), its C-terminal tail also forms a structure (two short helices and a loop in the middle) similar to that of hSAM (26) and fSAM (Fig.  4E). Lys-579 of bSAM was predicted to be hydrogen bonded to Tyr-65 of Vpx mac , and there is also a possible hydrogen bond between Arg-578 of bSAM and Asn-1090 of DCAF1. Therefore, we first tested whether fSAM with Thr-619 to alanine mutation (fSAM T619A , marked with red triangle in the sequence of fSAM in Fig. 4A) would impact its degradation mediated by Vpx. As expected, fSAM T619A was more resistant to SIVmac239 Vpx-induced degradation than its WT protein (Fig. 4F). Next, to further investigate which sites are crucial to Vpx-induced bSAM degradation, each amino acid except Pro-576 in the range of residues 573 to 580 was mutated to alanine individually (marked with green diamonds in Fig. 4A). We found that the bSAM mutants with Arg-578, Lys-579 or each of the three leucine (Leu-575, -577, and -580) mutations could resist SIVmac239 Vpx-mediated degradation in HEK293 cells (Fig. 4G, lanes 8, 10, 12, 14, and 16), indicating that these sites are required in the formation of bSAM-Vpx-CRL4 DCAF1 E3 ubiquitin ligase complex. This is consistent with the homology modeling result and the hydrophobic side chains of Leu-575, Leu-577, and Leu-580 from bSAM form a hydrophobic interface between the N termini of Vpx mac helix 1 and 3 (Fig. 4E).
Next, we tested whether the fSAM and bSAM mutants that could not be degraded in the presence of Vpx are not able to interact with Vpx. To do this, HA-tagged hSAM or hSAM 1-600 were co-expressed with myc-tagged SIVmac239 Vpx or FLAGtagged human DCAF1 in HEK293 cells to validate the further experiments by co-immunoprecipitation using anti-HA antibody-conjugated agarose beads (Fig. 5A). Previous studies have found that Vpx binds DCAF1 using conserved motifs in helix 1 and helix 3, which in turn recruits other components of the CRL4 DCAF1 E3 ubiquitin ligase (16 -20) to facilitate hSAM ubiquitination and subsequent degradation through recognition of the C-terminal sequence of hSAM (20,26,45), whereas hSAM alone was found to be unable to interact with DCAF1 in co-immunoprecipitation assays (20,45). Consistent with the previous results, hSAM  was unable to bind SIVmac239 Vpx (Fig. 5A, lane 6) and both hSAM and hSAM 1-600 were unable to bind DCAF1 (lanes 2 and 5. No band of DCAF1 was detected in the co-immunoprecipitated complex, so the data were not shown.). DCAF1 could only be co-precipitated in the presence of Vpx (Fig. S3). The co-immunoprecipitation assays between DCAF1 and fSAM or bSAM showed that fSAM and bSAM also could not immunoprecipitate DCAF1 (Fig. 5B, lanes 2 and 6, the blank data of DCAF1 were not shown). Among the mutants of fSAM and bSAM, fSAM T619A and bSAM L577A were unable to bind Vpx, and bSAM L575A , bSAM K579A and bSAM L580A demonstrated clearly decreased ability to interact with Vpx (Fig. 5B), confirming that mutations at these sites disrupted the interaction and hydrophobic interface of fSAM and bSAM with Vpx. Thus, it could be indicated that the abolished and decreased ability of Vpx to recruit these mutants to the E3 ubiquitin ligase complex leads to the inability of Vpx to degrade them.

Function of C-terminal domains of feline and bovine SAMHD1 The impact of essential amino acids of fSAM and bSAM CTD on antiviral ability
To test whether the mutations in the C terminus of fSAM and bSAM that compromised Vpx-induced degradation impact their antiviral activities, HIV-1, SIVmac239, and HIV-2 ROD were used to infect TZM-bl cells that were transfected with human, feline, or bovine WT SAMHD1 expression vector or each of their mutants or empty vector. 48 h later, the number of positive blue cells was counted to calculate the infectivity of the three viruses. Unlike WT hSAM, fSAM, and bSAM that could restrict all of the three viruses, respectively, we observed varied activity of each SAMHD1 mutant in restriction of the viruses (Fig. 6, A-C). hSAM 1-600 and fSAM 1-601 had no differences in restriction of the three viruses compared with their WT SAMHD1 proteins, but bSAM 1-572 abolished the antiviral activity. fSAM T619A could slightly restrict the three viruses. bSAM K579A and bSAM L580A also could restrict all of the three viruses but the antiviral ability of bSAM K579A was weakened, whereas bSAM L575A and bSAM R578A only displayed mild antiviral activity against HIV-1 compared with WT bSAM. bSAM K577A could not restrict any of the three viruses. We also examined the expression level of all these SAMHD1 proteins in TZM-bl cells on the day of infection and found that the intracellular SAMHD1 levels were almost equal (Fig. 6D). These results demonstrated that CTD of hSAM and fSAM and Leu-580 of bSAM are not required for the antiviral function of these WT SAMHD1 proteins, and Thr-619 of fSAM and Leu-575, Lys-577, Arg-578 and Lys-579 of bSAM are partially or fully necessary for feline and bovine SAMHD1 in restriction of different viral strains.
To further understand why the antiviral activity of these CTD mutants were varied, we analyzed their dNTPase activities in vitro. The deletion of residues 601-626 of hSAM caused only a less than 2-fold reduction in the dNTPase activity (Fig. 7,  A and D). Similarly, the deletion of fSAM CTD also caused a modest reduction in the dNTPase activity, and so did the T619A mutation of fSAM (Fig. 7, B and E). In contrast, the deletion of the CTD of bSAM largely impacted its dNTP hydrolysis function (Fig. 7, C and F). bSAM L577A and bSAM R578A showed a more than 3-fold reduction in the dNTPase activity and bSAM R579A showed a half-reduced activity compared with the WT bSAM, whereas bSAM L575A and bSAM R580A still had the same dNTP hydrolysis capacity as the WT bSAM (Fig. 7, C and F). Because significant differences in the dNTPase activity were observed between many mutants (hSAM 1-600 , fSAM 1-601 , fSAM T619A , bSAM 1-572 , bSAM L577A , bSAM R578A , and bSAM K579A ) and their WT SAMHD1 proteins but the antiviral activity of two of the mutants (hSAM 1-600 and fSAM 1-601 ) were not impacted (Fig. 6, A-C), we proposed that a 2-fold reduction in the dNTPase activity in vitro (Fig. 7, D-F) was the threshold of unaffected antiviral activity in our assays, suggesting that the antiviral activity of fSAM and bSAM CTD mutants largely depends on their dNTPase activity and the CTD of SAMHD1, especially of bovine SAMHD1, contributes to the maintenance of dNTPase activity of SAMHD1.

Discussion
Generation of SAMHD1-knockout animal models can facilitate the investigation of restriction of HIV-1 vectors and endogenous retroviruses in vivo and has been carried out in mice (46,47). In the current study, we found that both fSAM and bSAM could restrict HIV-1, HIV-2, and SIV. In addition, they could be degraded via a proteasomal pathway mediated by Vpx, which is different from mouse SAMHD1 (mSAM) that cannot be degraded by Vpx because of the altered amino acid sequence in the CTD (20). Because both fSAM and bSAM are able to restrict their endogenous retroviruses FIV and BIV, respectively, by reducing intracellular dNTP pool (36), these findings indicate that the physiological functions of fSAM and bSAM in vivo might be similar to that of SAMHD1 in human.

Function of C-terminal domains of feline and bovine SAMHD1
Therefore, these two species of mammals, especially cats, would be better models to study the in vivo function of SAMHD1. In fact, FIV-infected cat models have been widely used to study HIV pathogenesis (48). The BIV/cow model is also a good means for studying lentiviruses because there are no safety issues with BIV (40). Thus, it is necessary to study the features of feline and bovine SAMHD1 to help with the establishment of the animal models.
fSAM and bSAM were less sensitive to Vpx-mediated degradation than hSAM in HEK293 cells, however, more fSAM could be degraded in CRFK cells (Fig. 3, A-D), suggesting that Vpx from primate lentiviruses have evolved to counteract their host SAMHD1 proteins more efficiently and the feline E3 ubiquitin ligase complex may be more effective to its own targeted protein. Although overexpression of Vpx (Fig. 3, E-G), compared with co-transfection with SIVmac239 protein-expressing plasmid (Fig. 3, A-C), would result in more degradation of SAMHD1, our results clearly demonstrated that both fSAM and bSAM could be degraded in the presence of Vpx. This is not consistent with a previous study which found that Vpx from SIVmac239 did not mediate the degradation of feline and bovine SAMHD1 in HEK293T cells (36). We are curious about the reason for the different results obtained by others and ourselves, so we repeated the degradation assay in HEK293T cells by co-expression of human, feline, or bovine SAMHD1 proteins with Vpx mac or Vpx ROD . Interestingly, fSAM and bSAM were almost not degraded in the presence of Vpx mac , and the degradation of hSAM was also weakened in our HEK293T cell line (Fig. S4A), whereas the levels of degradation of these three SAMHD1 proteins mediated by Vpx ROD were similar to what was observed in HEK293 cells (Figs. S2 and S4B ). This suggested that in some circumstances, different cell lines may have different results, and the HEK293 cell line is more suitable for our investigations between Vpx and SAMHD1 in this study. By homology modeling based on the hSAM CTD-SIVsm Vpx-DCAF1 CTD X-ray crystal structure, two (Thr-619 and Ly-579) possible sites interacting with Vpx mac were predicted in fSAM and bSAM CTD, respectively, and Arg-578 of bSAM is a potential site interacting with DCAF1 (Fig. 4, D and E). All of these three sites were required for Vpx-induced degradation (Fig. 4, F  and G), but only mutation at Thr-619 of fSAM and Lys-579 of bSAM abolished or decreased the interaction with Vpx (Fig.  5B). The interaction between Arg-578 of bSAM and DCAF1 could not be detected in our experiments because the WT bSAM was unable to interact with DCAF1, which is consistent with previous results obtained between human SAMHD1 and DCAF1 by co-immunoprecipitation (20,45) although the X-ray crystal structure showed that Lys-622 of hSAM has a direct

Function of C-terminal domains of feline and bovine SAMHD1
interaction with Asp-1092 of DCAF1 (26). Because our result showed that mutation at Arg-578 of bSAM did not impact the interaction between bSAM and Vpx but diminished Vpx-induced degradation, this site might be necessary to the stability of the structure of bSAM-Vpx-DCAF1 complex required for the subsequent degradation.
Human SAMHD1 is highly expressed in HIV-1 nonpermissive cells such as THP-1, monocytes, and monocyte-derived dendritic cells, whereas it is absent from HIV-1-sensitive T cell lines such as Jurkat and SupT1, and myeloid cell line U937 (19). Among these cell lines, THP-1 and U937 are widely used in studying SAMHD1-mediated restriction and related events (4,14,19,29,30,49,50). Although TZM-bl is not a T cell line, this cell line was artificially modified with surface CD4 and CCR5 and is susceptible to infection by both R5 and X4 HIV-1 isolates (43). The widely adopted protocols for evaluating viral infectivity, 50% inhibitory concentration (IC 50 ) of antiviral reagent and 50% inhibitory dose (ID 50 ) of neutralizing antibody are built on this cell line (51)(52)(53). Therefore, it would be convenient to use SAMHD1 transiently transfected TZM-bl cells to detect the antiviral activity of SAMHD1 in vitro. We obtained consistent results from TZM-bl and stable U937 cells for the antiviral ability of human, feline, and bovine SAMHD1 (Figs. 1 and 2). Because SAMHD1 was overexpressed in these two cell lines before infection and the quantified amounts of viruses used for infection were relatively low in our experiments, the effect of degradation after SIV and HIV-2 infection should be limited and the antiviral activity of SAMHD1 was predominant to the results. Our results showed that the C-terminal residues 601 to 626 are not required for the antiviral activity of human SAMHD1 (Fig. 6, A-C). Consistent results were also found by Schwefel et al. (26). This result is supported by the finding that removal of residues 582 to 626 of hSAM had only a modest effect on the dNTPase activity and had essentially no effect on the capacity of hSAM to oligomerize (54). However, we found that deletion of the CTD of bSAM abolished its antiviral activity (Fig. 6, A-C), whereas the antiviral activity of CTD-deleted fSAM was not affected, suggesting that the function of CTD may be specific among different species. Consistent with this possibility, previous study found that CTD is important for the dNTPase activity of mSAM and this activity is regulated through tetramer stabilization by the CTD (55). Mutations that disrupt the allosteric site or the tetramer interface of the protein prevent dNTPase activity (33,55). Thus, it is possible that the bSAM mutants altered the conformation of the C-terminal tail in such a way that they disrupted tetramerization, consequently inactivating the enzyme. The half-reduced dNTPase activity in vitro might reflect the threshold of intracellular dNTP levels for efficient viral transcription. These findings highlighted the role of CTD in regulating the catalytic function of bovine SAMHD1, and it will be important to investigate how these mutations influence the antiviral ability of fSAM and bSAM against FIV and BIV in future analysis.

Figure 7. In vitro dNTPase activity of feline and bovine SAMHD1 and their CTD mutants.
A-C, in vitro detection of SAMHD1-catalyzed P i release. HAtagged SAMHD1 proteins were isolated from transfected HEK293 cells by immunoprecipitation. An aliquot of the immunoprecipitated SAMHD1 proteins was analyzed by Western blotting to ascertain comparable protein using anti-HA antibody. The levels of P i released after in vitro dGTP-PPase hydrolyzation reaction were detected by malachite green. Error bars represent the S.D. calculated from three independent reactions. Statistical analysis was performed between each CTD mutant and its WT SAMHD1 protein. D-F, the reduction fold was calculated by dividing the average value of P i concentration released from dGTP-PPase hydrolyzation reaction of the WT SAMHD1 protein by each of the triplicate values from each mutant. The threshold value of P i concentration reduction fold is 2.

Function of C-terminal domains of feline and bovine SAMHD1
Taken together, our study demonstrated that SAMHD1 from the three species have many common features in antiviral ability against different viral strains and degradation pathway mediated by Vpx, but have distinct differences in the functional sites of CTD with regard to Vpx-induced degradation and antiviral activities. Thr-619 of fSAM was identified as one of the only two sites in the C-terminal region that were under positive selection during SAMHD1 evolution in the Carnivora clade, including cat, and was one of the five residues that directly contact Vpx in all mammals, whereas more than 10 sites were identified being under positive selection in the C-terminal region (from residue 586 to the end, based on the human sequence numbering) of SAMHD1 from the Cetartiodactyla clade, including cow (56). This is consistent with our homology modeling results and suggests that bovine SAMHD1 may be under more intensive selection during evolution and that factors expressed by BIV or other viruses which target the CTD of bSAM may exist. The findings in this study will help to understand the SAMHD1 activities responsible for its viral restriction and will facilitate the improvement of SAMHD1/HIV animal models.

Plasmids
Plasmids pVR1012-homo-SAMHD1-HA, pCG-Vpx mac -HA, and pCG-DCAF1-FLAG were gifts from Dr. Wenyan Zhang. The DNA sequences coding for feline-SAMHD1 (GenBank accession no. XM_003983547.3) and bovine-SAMHD1 (Gen-Bank accession no. NM_001075861.1) were amplified from feline and bovine genomic cDNA (reverse-transcripted from total mRNA isolated from CRFK and MDBK cells, respectively) by PCR, sequence confirmed, and inserted into a pVR1012 vector (57) at the XbaI and BamH I restriction sites. A HA tag was added to the C terminus of each SAMHD1 sequence. To generate the sequence-truncated, or point-mutated SAMHD1 fragments, pVR1012-homo-SAMHD1-HA, pVR1012-feline-SAMHD1-HA, and pVR1012-bovine-SAMHD1-HA were used as PCR templates. Chimeric fSAM 1-601 ϩhCTD and bSAMϩhCTD fragments were constructed by overlapping PCR. All PCR products were digested and cloned into the XbaI and BamH I sites of pVR1012 vector with C-terminal HA tags. An myc tag was also added to the C terminus of SIVmac239 Vpx gene amplified from pCG-Vpx mac -HA for the co-immunoprecipitation assay. The PCR product was digested and cloned into the XbaI and BamH I sites of pVR1012 vector (pVR1012-Vpx mac -myc). Codon-optimized DNA sequence coding for HIV-2 ROD Vpx (UniProtKB accession no. P06939) were synthesized (GenScript, Piscataway, NJ) and inserted into the pVR1012 vector with a myc tag (pVR1012-Vpx ROD -myc). For generation of stable U937 cell lines, pLVX-puro vectors expressing SAMHD1 were constructed by inserting the HAtagged SAMHD1 fragments into the EcoR I and BamH I restriction sites of pLVX-puro.

Western blotting
DNA transfection was carried out using Lipofectamine 3000 (Invitrogen) according to the manufacturer's instructions. Cells were harvested at 48 h after transfection, centrifuged at 3000 rpm for 5 min and lysed. The prepared protein samples were separated by electrophoresis in SDS-PAGE, followed by transferring onto nitrocellulose-membranes. After blocking in 5% nonfat milk, the membranes were probed with primary and secondary antibodies. Staining was carried out with 0.66% 5-bromo-4-chloro-3-indolyl phosphate and 0.33% nitro blue tetrazolium solutions in 0.1 M Tris-HCl, pH 9.5. Protein band intensities were determined by Adobe Photoshop CC 2017 software (San Jose, CA) and normalized by the level of GAPDH. The remaining percentage of SAMHD1 after degradation was calculated by dividing the normalized band intensity of SAMHD1 protein in the presence of Vpx by that of the corresponding SAMHD1 protein in the absence of Vpx.

Viral infectivity assay
Viral infectivity was measured by infecting TZM-bl indicator cells which contain an HIV-1 LTR promoter-␤-gal expression cassette (43). Viruses were generated by transfection into HEK293T cells. The infectious molecular clones for generation of HIV-1 (NL4 -3), SIVmac239, and HIV-2 ROD were provided by NIH-ARP. At 72 h post transfection, culture supernatants were harvested and centrifuged at 3000 rpm for 10 min at 4°C to remove cell debris. The viral concentration was determined by measuring the concentration of RT using a Lenti RT Activity Kit (Cavidi, Uppsala, Sweden). TZM-bl cells transfected with or without SAMHD1 were detached with trypsin at 48 h post transfection and were re-seeded into 24-well plates (2 ϫ 10 5 cells per well) in 300 l viral infection solutions (containing HIV-1 of 0.2 ng RT, SIVmac239 of 1 ng RT or HIV-2 ROD of 1 ng RT, 40 nM diethylaminoethyl (DEAE)-dextran and DMEM). At 48 h after infection, the cells were fixed and stained with X-Gal, the substrate for ␤-gal, as described previously (58). The viral infectivity was determined by the number of positive blue cells. Each experiment was performed in triplicate.

Function of C-terminal domains of feline and bovine SAMHD1 Generation and infection of stable U937 cell lines
Lentiviral stocks for overexpression of HA-tagged SAMHD1 mutants were generated by transfection of HEK293T cells with SAMHD1-expressing pLVX-puro vectors, psPAX2 packaging plasmid, and pVSV-G (Addgene, Watertown, MA). 72 h post transfection, lentiviral stocks were harvested, filtered, and used to infect U937 cells in the presence of DEAE-dextran (20 nM), after which cells were cultured and selected in RPMI 1640 with 1 g/ml of puromycin as described previously (30). The expression of SAMHD1 in PMA-differentiated cells was confirmed by Western blotting. SAMHD1-expressing stable U937 cells (5 ϫ 10 5 ) were differentiated with PMA for 20 h, infected with HIV-1 of 0.1 ng RT, SIVmac239 of 1 ng RT, SIVmac239⌬Vpx of 1 ng RT or HIV-2 ROD of 1 ng RT for 4 h, and then washed three times with RPMI 1640 medium and cultured in 6-well plates with 2 ml of RPMI 1640 containing 10% FBS. The culture supernatant (200 l) in each well was harvested regularly and the same volume of fresh medium was replenished. The viral replication was monitored by measuring the concentration of p24 (Alliance HIV-1 P24 ANTIGEN ELISA Kit, PerkinElmer) for HIV-1 or RT for SIV and HIV-2 in the culture supernatant. All infections were performed in triplicate.

In vitro dNTP hydrolysis assay
HEK293 cells in 12-well plates transfected with 600 ng HAtagged SAMHD1 constructs were harvested at 48 h after transfection and washed twice with cold reaction buffer (pH 7.4, containing 50 mM Tris-HCl, 50 mM KCl, and 5 mM MgCl 2 ), and then lysed in 250 l lysis buffer (pH 7.4, containing 50 mM Tris-HCl, 150 mM NaCl, and 1% Triton X-100) supplemented with complete Mini Protease Inhibitor Mixture Tablets (Roche, South San Francisco, CA) at 4°C for 40 min, followed by centrifugation at 16,000 ϫ g for 15 min at 4°C. Cleared cell lysates were mixed with anti-HA antibody-conjugated agarose beads (Roche) and incubated at 4°C for 3 h. The samples were then washed three times with washing buffer (pH 7.4, containing 20 mM Tris-HCl, 100 mM NaCl and 0.05% Tween 20) and once with the reaction buffer. 40 l of reaction buffer was then added to the bead pellet. 30% of the bead slurry was reserved as an input control, and the rest was diluted at a 1: 10 dilution with the reaction buffer and mixed with 1 mM dGTP and 0.01 unit pyrophosphatase (New England Biolabs, Ipswich, MA) to a final volume of 40 l in triplicate reactions. The reactions were incubated at 37°C for 2 h with occasional mixing and stopped by heating to 70°C for 5 min. The reaction products were then diluted by 25-fold with reaction buffer and the P i release was measured with a Malachite Green Detection Kit (R&D Systems, Minneapolis, MN) according to the manufacturer's instructions.

Co-immunoprecipitation assay
Cells in 6-well plates were harvested at 48 h after transfection, then washed twice with cold PBS and lysed in 250 l lysis buffer supplemented with complete Mini Protease Inhibitor Mixture Tablets at 4°C for 40 min. The cell lysates were centrifuged at 16,000 ϫ g for 15 min and the supernatants were incubated with anti-HA beads at 4°C for 3 h. Subsequently, the beads were washed three times with the washing buffer, resus-pended with 30 l 2ϫ SDS loading buffer, boiled at 97°C for 10 min, and then subjected to SDS-PAGE and immunoblotting.

Homology modeling
Homology modeling of target proteins was performed by using the protein modeling module of Discovery Studio 2.1 software package (Omaha, NE). The PDB file of the hSAM CTD-SIVsm Vpx-DCAF1 CTD X-ray crystal structure (ID: 4CC9) was used as a template. The original hSAM CTD and SIVsm Vpx sequences were aligned with the feline (residue 607 to 625) or bovine (residue 568 to 586) SAMHD1 CTD and SIV-mac239 Vpx sequences. Modeling was performed at the medium optimization level with refined loop parameters, and no additional restraints were used. The best model of 10 models built was selected based on the quality estimation score and overall structure similarity. Because two (SAMHD1 and Vpx) of the three molecules in the crystal structure were remodeled, three modeling strategies: SAMHD1 first, Vpx first and both together, were tested independently. Generally, the interacting sites displayed in all three tests were further analyzed by Western blotting and co-immunoprecipitation. The images of remodeled structures were then modified by using PyMOL v1.8 (Schrödinger, LLC, New York, NY).

Statistical analysis
Data are shown as mean Ϯ S.D. Significance is calculated by using unpaired Student's t test or repeated-measure analysis of two-way ANOVA with PRISM v6 (GraphPad Software, Inc., La Jolla, CA). * indicates p Յ 0.05, ** p Յ 0.01, and *** p Յ 0.001. ns indicates no significance.