Mutations in SAMHD1 cause Aicardi-Goutières syndrome (AGS), which is a genetic autoimmune disorder that mimics a congenital viral infection^9,10. SAMHD1 is a myeloid cell-specific HIV-1 restriction factor that is counteracted by Vpx^1,2. SAMHD1 possesses a dGTP-dependent dNTPase activity^11,12, and studies have proposed that SAMHD1 blocks HIV-1 infection by depleting the intracellular dNTP pool^3-5. Contrary to the proposed role for dNTPase activity in SAMHD1-mediated restriction of HIV-1, the phosphorylation of SAMHD1 regulates the ability of this enzyme to restrict HIV-1 infection but does not alter the cellular dNTP levels^6-8. In addition, SAMHD1 is a nucleic acid binding protein^13-15 and exhibits a nuclease activity in vitro against diverse nucleic acid substrates¹⁶. Although the physiological relevance of SAMHD1 nuclease activity remains unknown, the aforementioned characteristics suggest that SAMHD1 might bind to HIV-1 RNA and degrade this RNA or reverse transcription intermediates.

In contrast to Beloglazova et al.¹⁶, other studies have not observed SAMHD1 nuclease activity^11-13; therefore, we tested whether SAMHD1 possesses nuclease activity. We incubated full-length, wild-type human SAMHD1 (SAMHD1_WT) (Supplementary Fig. 1) with various types of nucleic acid substrates. SAMHD1_WT hydrolyzed both 20- and 30-mer single-stranded RNAs (ssRNA) processively (Fig. 1a and Supplementary Fig. 2a). No nuclease activity was detected on double-stranded RNA (dsRNA), single-stranded DNA (ssDNA), double-stranded DNA (dsDNA), and RNA-DNA hybrid substrates, whereas these substrates were digested by positive control nucleases (Supplementary Fig. 2b–e) (substrate sequences are shown in Supplementary Table 1). A previous study¹⁶ observed that SAMHD1 degrades ssRNA, ssDNA, and RNA in DNA-RNA hybrids; however, we observed that only ssRNAs, including in vitro-HIV-1 transcripts (Supplementary Fig. 3), were sensitive to SAMHD1 digestion when the experiments were repeated using the various types of 40-mer nucleic acid substrates described by Beloglazova et al.¹⁶ (Supplementary Fig. 4). These discrepancies might be due to differences in assay conditions or protein purification procedures; however, our results are consistent with the observation of Beloglazova et al.¹⁶ that SAMHD1 cleaves ssRNAs in vitro. The degradation of ssRNA by SAMHD1 required Mg²⁺ (Supplementary Fig. 5) and correlated with the presence of the SAMHD1_WT protein across chromatographic fractions (Supplementary Fig. 6). Considering the preferential binding of SAMHD1 to ssRNA compared with ssDNA^13,15 and no interaction between endogenous SAMHD1 and dsRNA, dsDNA or RNA-DNA hybrids¹⁵, we concluded that SAMHD1 is an RNase that digests ssRNA.

Open in a separate window

Figure 1

The RNase but not the dNTPase function of SAMHD1 is required for HIV-1 restriction. (a) Purified GST-SAMHD1_WT (150 nM) was incubated for 30 min at 37 °C with 5′-end ³²P labeled nucleic acid substrates (20-mer, 50 nM). The RNA and DNA are indicated by red and blue lines, respectively. (b) An RNase activity assay performed for SAMHD1 mutants as described in Fig. 1a using 20-mer ssRNA substrates. (c) A dGTP-triphosphohydrolase assay was performed for GST-SAMHD1 variants as described previously³. (d) The HA-tagged SAMHD1_WT and SAMHD1_Q548A proteins were purified from the PMA-differentiated U937 cells using HA-antibody. An RNase and a dGTP triphosphohydrolase assays were performed as described above. (e) Inhibition of the SAMHD1 RNase activity by dGTP. The assay was performed as in a in the presence of dGTP. (f) The intracellular dNTP pools were measured in U937 cells stably expressing SAMHD1 variants as described²⁶. (g) U937 cells stably expressing SAMHD1mutants were treated with PMA and infected by pLaiΔenvGFP3, a HIV-1-GFP reporter virus with the LAI backbone²⁷. After 48 h, the percentages of GFP-positive cells were analyzed using flow cytometry. The percentage of GFP-positive cells was calculated relative to the number of GFP-positive mock-transfected cells. In f and g, the data are presented as the mean ± s.d. of triplicate independent experiments. (* and ** indicate significant differences compared with the mock-transfected control at P < 0.05 and P < 0.001, respectively, using the two-tailed Student's t test).

Unlike dNTPase activity, RNase activity does not require the cofactor dGTP (Fig. 1a); therefore, we proposed that these two catalytic functions of SAMHD1 might be physically separable. We characterized AGS-associated SAMHD1 mutations and mutations near the allosteric sites for defects in RNase or dNTPase activity and identified several point mutations that caused loss of function. Mutations at residues 207 and 311 in the catalytic site abolished both RNase and dNTPase activities (Figs. 1b,c). The allosteric mutant SAMHD1_D137N had no effect on the RNase activity but did abolish the dNTPase activity (Figs. 1b,c). The AGS mutant SAMHD1_Q548A showed the opposite phenotype. Unlike the processive degradation seen in SAMHD1_WT and SAMHD1_D137N, a mutation at Q548 caused the enzyme to digest ssRNA in a distributive manner featured by accumulation of intermediate size fragments (16~19-mer) (Fig. 1d). Considering that a shift from a processive to distributive degradation often occurs under the conditions that decrease the affinity of exonuclease to the substrates¹⁷, this switch in mode may be because of a lowered affinity of SAMHD1_Q548A to ssRNA. Indeed, recent studies have shown that SAMHD1_Q548A exhibits impaired ssRNA-binding^15,16. Notably, dGTP inhibited the RNase activity of SAMHD1 on ssRNA substrates, and this inhibition correlated with a shift from a processive to a distributive hydrolysis (Fig. 1e), suggesting that dGTP interferes with the binding of ssRNA substrates to SAMHD1. The identification of the SAMHD1 mutants that retain only one of two enzymatic activities enabled us to determine the discrete contribution of SAMHD1 RNase activity to HIV-1 restriction. For this analysis, we stably expressed SAMHD1 mutants in U937 pro-monocytic cells, which do not express endogenous SAMHD1^1,3,6,7 (Supplementary Fig. 7). Consistent with in vitro data (Fig. 1c,d), SAMHD1_D207N, SAMHD1_D311A, and SAMHD1_D137N did not reduce the intracellular levels of dNTPs, whereas SAMHD1_Q548A decreased the cellular levels of dNTPs (Fig. 1f). As expected, SAMHD1_D207N and SAMHD1_D311A did not restrict HIV-1 infection. Notably, SAMHD1_D137N restricted HIV-1 infection to nearly the same extent as SAMHD1_WT. SAMHD1_Q548A was unable to block HIV-1 restriction despite its ability to decrease the cellular levels of dNTPs (Fig. 1g). These data suggest that the RNase but not the dNTPase function of SAMHD1 is essential for HIV-1 restriction.

The quantification of viral cDNA intermediates showed that SAMHD1_D137N inhibited the synthesis of both the early and late viral cDNA products (Supplementary Fig. 8), indicating that SAMHD1 RNase functions at the early step of HIV-1 reverse transcription. Thus, we tested whether SAMHD1 RNase activity affects HIV-1 RNA levels. HIV-1 RNA was significantly reduced in cells expressing SAMHD1_WT or SAMHD1_D137N 3 and 6 h post-infection (hpi) (Fig. 2a), but SAMHD1_D207N, SAMHD1_D311A, and SAMHD1_Q548A did not reduce the HIV-1 RNA level; these results were consistent with FACS analyses (Fig. 1g). To exclude primer-specificity bias, 12 pairs of primers spanning the entire HIV-1 genomic RNA (Supplementary Table 2) were used to amplify fragments from HIV-1-infected cells. The HIV-1 RNA level in SAMHD1_WT-expressing cells was distinctly lower compared with the mock-transfected cells (Supplementary Fig. 9). To exclude the effect of the viral reverse transcriptase RNase H activity on the HIV-1 RNA level, we used RNase H-defective HIV-1 (HIV-1_D443N) in which the Asp-443 of reverse transcriptase was mutated to Asn (D443N)¹⁸. The HIV-1_D443N RNA level was lower in SAMHD1_WT-expressing cells compared with mock-transfected cells (Fig. 2b), indicating that SAMHD1 can degrade HIV-1 RNA in the absence of reverse transcription.

Open in a separate window

Figure 2

SAMHD1 directly degrades HIV-1 RNA in human monocytic cells. (a) The total cellular RNA was extracted from SAMHD1-expressing, HIV-1-GFP-infected U937 cells. The HIV-1 RNA content was quantified by qRT-PCR using HIV-1 gag-specific primers. The data were normalized to an internal β-actin. (b) Mock- and SAMHD1_WT-expressing U937 cells were infected with HIV-1-GFP or HIV-1_D443N-GFP. The HIV-1 RNA content was quantified by qRT-PCR using HIV-1 gfp-specific primers. (c) Relative HIV-1 genome abundance in reads per kilobase per million mapped reads (RPKM) calculated from RNA-Seq data. (d) Coverage of the RNA-Seq data across the HIV-1 genome in U937 cells expressing SAMHD1_WT, SAMHD1_D207N and SAMHD1_D137N after HIV-1-GFP infection. Data are representative of two independent experiments (c,d). (e) shRNA-mediated silencing of SAMHD1 (left) or Vpx-mediated degradation of SAMHD1 (right) in THP-1 cells increases HIV-1 RNA stability. THP-1 cells expressing shRNAs, Vpx or negative control were treated with PMA for 24 h followed by infection with HIV-1-GFP. The total viral RNA was quantified by qRT-PCR at the indicated times post-infection. Knockdown efficiency was determined by immunoblot analysis. (f) Enrichment of HIV-1 genomic RNA in the SAMHD1 immunoprecipitates. U937 cells expressing HA-tagged SAMHD1 variants were infected for 90 min with HIV-1-GFP. The purified RNAs in the anti-HA immunoprecipitates were quantified by qRT-PCR using HIV-1 gag-specific primers. In a, b, e, and f, data are presented as the mean ± s.d. of triplicate independent experiments. (*P < 0.05 and **P < 0.001 versus Mock or shControl, two-tailed Student's t test).

RNA-Seq analysis also confirmed the degradation of HIV-1 RNA by SAMHD1. At 3 hpi, the relative read count of HIV-1 RNA in SAMHD1_WT and SAMHD1_D137N-expressing cells was lower than that in mock-transfected or SAMHD1_D207N-expressing cells despite the considerable reduction of HIV-1 RNA levels in all cell types compared with the levels observed at 1 hpi (Figs. 2c,d). shRNA-mediated knockdown of SAMHD1 or Vpx-mediated degradation of endogenous SAMHD1 in THP-1 cells also caused an approximately four-fold increase in the HIV-1 RNA level (Fig. 2e). RNA immunoprecipitation (RIP) analysis¹⁹ revealed that a significant enrichment of HIV-1 RNA was observed in the immunoprecipitates from cells expressing SAMHD1_WT-HA. The HIV-1 RNA enrichment was more pronounced for the SAMHD1_D207N-HA catalytic mutant with an approximately eight-fold increase (Fig. 2f), suggesting that SAMHD1 associates with HIV-1 genomic RNA. In contrast, interaction of SAMHD1_Q548A-HA with HIV-1 RNA was significantly reduced when compared with the SAMHD1_WT-HA (Fig. 2f), supporting that impaired binding of SAMHD1_Q548A-HA to HIV-1 RNA accounts for its inability to cleave short ssRNA and HIV-1 RNA (Figs. 1d and ​and2a2a).

Next, we analyzed the role of SAMHD1 RNase in degrading HIV-1 RNA in primary human monocyte-derived macrophages (MDMs) and CD4⁺ T cells. The differentiation of MDM and activation state of CD4⁺ T cells were analyzed by monitoring their surface markers (Supplementary Figs. 10 and 11). SAMHD1 silencing in MDMs markedly reduced the cellular SAMHD1 levels (Fig. 3a, left) and rendered MDMs more permissive to HIV-1 infection (Fig. 3a and Supplementary Fig. 12), a result that is correlated with the increased stability of HIV-1 RNA. The enhancement of HIV-1 RNA stability ranged from approximately two to four-fold, depending on the donor (Fig. 3a, middle). Similar to the results obtained using MDMs, SAMHD1 silencing in resting CD4⁺ T cells increased the infectivity of HIV-1 (Fig. 3b and Supplementary Fig. 12). In these cells, increased HIV-1 RNA stability was observed 3 hpi (Fig. 3b, middle). The degradation of SAMHD1 by Vpx could also stabilize the viral RNAs of HIV-1 and HIV-1_D443N in resting CD4⁺ T cells (Fig. 3c). These results show that the antiviral activity of SAMHD1 correlates with its RNase activity in primary MDMs and resting CD4⁺ T cells. Biochemical fractionation revealed that SAMHD1 is localized in both the cytoplasm and nucleus of MDMs and monocytes (Supplementary Fig. 13), which is consistent with previous observations^4,13. Considering that cytoplasmic SAMHD1 can block HIV-1 infection^20,21, these data suggest that after virus uncoating, SAMHD1 associates with and degrades HIV-1 RNA in the cytoplasm.

Open in a separate window

Figure 3

SAMHD1 degrades HIV-1 RNA in primary human MDMs and CD4⁺ T cells. (a) CD14⁺ monocytes isolated from three donors were differentiated into MDMs. MDMs were transfected with siRNA to SAMHD1 or control siRNA and then infected with 100 ng of HIV-1-GFP. The efficiency of SAMHD1 knockdown was evaluated by qRT-PCR (left). Total viral RNA was quantified at the indicated times post-infection by qRT-PCR using gfp-specific primers. The data were normalized to an internal β-actin (middle). Quantification of GFP-positive cells by flow cytometry of MDMs (right). (b) Post-activated resting CD4⁺ T cells from three donors were transfected with siRNA to SAMHD1 or control siRNA and then infected with 100 ng of HIV-1-GFP. The efficiency of SAMHD1 knockdown (left), HIV-1 RNA levels (middle) and HIV-1 infectivity (right) were analyzed as in a. (c) Effect of Vpx treatment on the HIV-1 RNA stability in resting CD4⁺ T cells. Resting CD4⁺ T cells were challenged with HIV-1-GFP or HIV-1-GFP_D443N in the presence of Vpx VLP. HIV-1 RNA levels were quantified as in a. In a-c, data are presented as the mean ± s.d. of three experiments. (* and ** indicate significant differences compared with siControl or –Vpx VLP control at P < 0.05 and P < 0.001, respectively, using the two-tailed Student's t test).

The phosphorylation of SAMHD1 at T592 causes the loss of HIV-1 restriction without decreasing cellular dNTP levels^6-8. We tested whether the phosphorylation of SAMHD1 affects its RNase activity in HIV-1 restriction. In PMA-treated U937 cells where SAMHD1_WT is not phosphorylated on T592^6,7, SAMHD1_WT blocked HIV-1 infection. In contrast, SAMHD1_T592D containing the phosphomimetic D residue was unable to inhibit HIV-1 infection (Fig. 4a), confirming previous results⁶. Notably, the stability of HIV-1 RNA was enhanced by three to four-fold in cells expressing SAMHD1_T592D compared with cells expressing SAMHD1_WT (Fig. 4b). In contrast to this in vivo observation, the recombinant SAMHD1_T592D protein was as effective as the SAMHD1_WT and the nonphosphorable SAMHD1_T592V in degrading ssRNA substrates in vitro (Fig. 4c). RIP analysis revealed that SAMHD1_WT and SAMHD1_T592D bound to HIV-1 genomic RNA with similar affinities, indicating that phosphorylation of SAMHD1 T592 does not interfere with substrate binding (Fig. 4d). The discrepancy between the in vivo and in vitro RNase activities suggests that the phosphorylation of SAMHD1 might promote the recruitment of cellular factors that negatively regulate its RNase activity or interfere with the interaction of SAMHD1 with unidentified cofactors. Our results reveal that the phosphorylation of SAMHD1 at T592 is a mechanism that negatively regulates its RNase activity in vivo and impedes HIV-1 restriction.

Open in a separate window

Figure 4

Phosphorylation of SAMHD1 regulates RNase activity

(a–c) U937 cells stably expressing an empty vector (mock) or Flag-tagged SAMHD1 variants were treated with PMA and infected with 100 ng of pLaiΔenvGFP3. (a) Flow cytometric analysis 48 hpi to measure the percentage of GFP-positive cells. The percentage of GFP-positive cells was calculated relative to the number of GFP-positive mock-transfected cells (right). Western blot analysis of cell extracts using the anti–Flag antibody. GAPDH was used as a loading control. (b) HIV-1 genomic RNA content was quantified by qRT-PCR using HIV-1 gfp-specific primers. The data were normalized to an internal β-actin. (c) Recombinant SAMHD1_T592V and SAMHD1_T592D proteins were purified from E. coli. The RNase activity assay was performed as described in Fig. 1 using 5′-end ³²P labeled 20-mer ssRNA substrates. (d) Effect of SAMHD1 T592 phosphorylation on substrate binding. RIP analysis was performed as described in Fig. 2f, except that cell lysates were immunoprecipitated using an anti-Flag antibody. In a, b, and d, data are presented as the mean ± s.d. of triplicate experiments. (* and ** indicate significant differences compared with the mock-transfected control at P < 0.05 and P < 0.001, respectively, using the two-tailed Student's t test). e. Model for anti-HIV-1 activity mediated by SAMHD1 RNase function.

The crystal structure of SAMHD1 reveals that this protein is dimeric and indicates a structural basis for allosteric activation of catalytic activity against dNTPs by dGTP binding¹¹. By contrast, recent studies have demonstrated that SAMHD1 oligomerizes in a dGTP-dependent manner up to a tetramer and that the tetramerization of SAMHD1 is required for its dNTPase and anti-HIV-1 activities^22-24. Notably, SAMHD1 with a mutation at the allosteric dGTP-binding site D137 did not undergo oligomerization and was mostly monomeric in the presence of dGTP²². The mutation of D137 (SAMHD1_D137N) abolishes SAMHD1 dNTPase but not RNase activity (Fig. 1), indicating that unlike the dNTPase activity, the tetramerization of SAMHD1 is not necessary for its RNase activity. Based on previous data and our results, we propose that SAMHD1 is a dual-function enzyme whose mutually exclusive and predominant activities depend on the dGTP pool. At low dGTP levels, SAMHD1 exists as monomers or dimers that harbor active RNase but not dNTPase activity, enabling HIV-1 restriction through the cleavage of HIV-1 RNA. At high dGTP levels, SAMHD1 tetramerizes, yielding an enzyme with an inactive RNase but active dNTPase function (Fig. 4e). An alternative but not mutually exclusive possibility is that the ssRNA and dNTP substrates might compete with each other for a common active site on the enzyme. dGTP is both a substrate and an activator of the SAMHD1 dNTPase function^11,12; this model is consistent with the observations that high levels of dGTP impair the binding of SAMHD1 to nucleic acid substrates in vitro^6,16, that the switch from a processive to distributive degradation of ssRNA in dGTP-dependent manner (Fig. 1e) is indicative of competitive inhibition of ssRNA-binding by dGTP, and that mutations in the catalytic site abolish both RNase and dNTPase activities (Fig. 1). In this respect, the partial alleviation of SAMHD1-mediated HIV-1 restriction by the addition of exogenous dNs³ might be an indirect consequence of impaired RNA substrate binding by increased dGTP levels. These two hypotheses may also explain why SAMHD1 blocks HIV-1 infection only in non-cycling cells, such as DCs and macrophages with low dNTP levels¹, but not in cycling cells with high dNTP levels^4,6,25. Another explanation might be that in non-cycling cells, dNTP limitation would slow down reverse transcription, and this could allow a sufficient time window for the action of SAMHD1 RNase against HIV-1 RNA. The identification of SAMHD1 as an RNase suggests that SAMHD1 is involved in nucleic acid metabolism related to the prevention of an innate immune response; therefore, our results provide insight into the mechanisms underlying HIV-1 restriction and the pathogenesis of AGS.