SAMHD1 is a critical restriction factor for HIV-1 in non-cycling cells and its antiviral activity is regulated by T592 phosphorylation. Here, we show that SAMHD1 dephosphorylation at T592 is controlled during the cell cycle, occurring during M/G1 transition in proliferating cells. Using several complementary proteomics and biochemical approaches, we identify the phosphatase PP2A-B55α responsible for rendering SAMHD1 antivirally active. SAMHD1 is specifically targeted by PP2A-B55α holoenzymes during mitotic exit, in line with observations that PP2A-B55α is a key mitotic exit phosphatase in mammalian cells. Strikingly, as HeLa or activated primary CD4+ T cells enter the G1 phase, pronounced reduction of RT products is observed upon HIV-1 infection dependent on the presence of dephosphorylated SAMHD1. Moreover, PP2A controls SAMHD1 pT592 level in non-cycling monocyte-derived macrophages (MDMs). Thus, the PP2A-B55α holoenzyme is a key regulator to switch on the antiviral activity of SAMHD1.
Sterile α motif (SAM) domain and HD domain-containing protein 1 (SAMHD1) is a Mg2+-dependent triphosphohydrolase (dNTPase) converting deoxynucleoside triphosphates (dNTPs) into deoxynucleosides and inorganic triphosphates1. Besides the dNTPase function, SAMHD1 binds to single-stranded nucleic acids2,3 and is proposed to exert nuclease activity4,5,6, a function which is heavily debated3,7,8.
Mutations in SAMHD1 cause the hereditary autoimmune disease Aicardi-Goutières syndrome (AGS), associated with elevated production of interferon (IFN) α9. Moreover, SAMHD1 is frequently mutated in a variety of cancer types, such as chronic lymphocytic leukemia (CLL) and colorectal cancer10,11. Importantly, SAMHD1 restricts a diverse set of DNA and retroviruses12,13,14,15: In particular, human immunodeficiency virus (HIV)-1 is restricted at an early replication step in non-cycling myeloid cells and resting CD4+ T cells16,17,18,19. As a potent dNTPase, SAMHD1 efficiently reduces cellular dNTP levels in non-cycling cells below those required to support HIV-1 reverse transcription (RT)1,20. Furthermore, SAMHD1’s RNase activity was proposed to mediate HIV-1 restriction5; it is, however, unclear whether this additional enzymatic activity may be causative for HIV-1 inhibition3,8.
Regardless of the precise restriction mechanism, SAMHD1 expression alone is not sufficient to induce a potent block of HIV-1 replication, as activated CD4+ T and cycling THP-1 cells express high SAMHD1 levels, but are permissive for HIV-1 infection16,18. SAMHD1 is phosphorylated at threonine (T) 592 in asynchronously proliferating cells (SAMHD1 pT592), rendering it inactive against HIV-121,22,23. SAMHD1 interacts with cyclin-dependent kinase (CDK) 1 and 2/cyclin A2 in cycling cells21,24, in accordance with T592 as a target site for CDKs (consensus sequence: S/T-P-x-K/R, SAMHD1 motif: 592TPQK595). How T592 phosphorylation of SAMHD1 influences its structural and enzymatic properties, tetramerization propensity25,26,27,28 and dNTPase activity22,23, is a matter of debate. Nevertheless, only dephosphorylated SAMHD1 at T592 is able to actively restrict HIV-121,22,23,24. Remarkably, the importance of a dephosphorylated antiviral-active state of SAMHD1 has been proposed for hepatitis B virus (HBV)15 as well, suggesting this specific post-translational modification as an important regulatory mechanism. Besides the control of SAMHD1’s antiviral activity, phosphorylation at T592 has been proposed to play a novel role in promoting the resection of arrested replication forks and preventing the accumulation of single-stranded DNA (ssDNA) derived from stalled forks in the cytoplasm29. This reinforces the importance of both, phosphorylation and dephosphorylation at this specific residue, for diverse physiological functional states of SAMHD1.
In this report, two complementary proteomics approaches identified the serine/threonine protein phosphatase 2 A (PP2A) as the responsible phosphatase actively removing the phosphate at T592 in SAMHD1. Particularly, PP2A holoenzymes containing the regulatory subunit B55α, which is critical for substrate specificity, efficiently acted on T592 in vitro and in cells. Intriguingly, PP2A-B55α holoenzymes are responsible for dephosphorylation of SAMHD1 at T592 in proliferating cells during mitotic exit, an important transition between M and G1 phase of the cell cycle. Concomitantly, we observed a rapid drop in dATP levels, suggesting either a coincidental or causative relationship between dephosphorylation and dNTPase activity. Importantly, upon entry into G1 phase, HIV-1 infection led to reduction of early and late RT products in activated CD4+ T and HeLa cells, depending on the presence of dephosphorylated SAMHD1. Thus, we defined the time window of PP2A activity during which SAMHD1 is rendered antivirally active. Additionally, PP2A controls SAMHD1 T592 phosphorylation in non-cycling MDMs, important HIV-1 target cells. Furthermore, we provide evidence for PP2A involvement in the IFN-inducible dephosphorylation of SAMHD1 in MDMs.
Cell cycle-dependent regulation of SAMHD1 pT592 level
To characterize the cell cycle-dependent (de)phosphorylation of SAMHD1 at T592 in more detail, we synchronized HeLa cells at the G1/S border using a double-thymidine block. Cell cycle-progression was monitored by immunoblotting using cyclin-specific antibodies (Fig. 1a) and by flow cytometric analysis of DNA content (Fig. 1b). Interestingly, SAMHD1 protein levels remained constant in all cell cycle phases (Fig. 1a), including S phase (0–4 h post-release). SAMHD1 phosphorylation at T592 appeared high in early S phase (0–4 h post-release)—consistent with reports of initial CDK2-dependent phosphorylation at T59224,30. After maximal activity of CDKs/cyclin A2, SAMHD1 phosphorylation is maintained in G2/M phase (7–8 h post-release)—accompanied by maximal cyclin A2/B1 expression (Fig. 1a). However, when cells re-enter interphase, SAMHD1 phosphorylation at T592 is rapidly lost and paralleled by complete degradation of cyclins A2/B1 (starting at 10 h and maximal at 11.5 h post-release) (Fig. 1a). Moreover, we monitored dATP levels during cell cycle progression (Fig. 1c, Supplementary Fig. 1a, b). Interestingly, dATP levels remain consistently high from S to G2/M phase (0–8 h post-release). However, with transition of cells into G1 phase, starting at 9 h post-release (Supplementary Fig. 1b), SAMHD1 dephosphorylation (Supplementary Fig. 1a) is accompanied by a pronounced decrease in dATP levels (Fig. 1c). SAMHD1 is again phosphorylated at T592 as cells re-enter S phase (16–18 h post-release; Supplementary Fig. 1d, e) and dATP levels start to rise (13 h post-release; Fig. 1c and Supplementary Fig. 1a, b). Taken together, these results suggest the involvement of a cellular phosphatase that actively removes the T592 phosphate from SAMHD1 during M/G1 transition.
PP2A interacts with SAMHD1 in proliferating cells
To assess which phosphatase potentially binds to and acts on SAMHD1 upon entry into G1 phase of the cell cycle, we conducted two independent mass spectrometry (MS) experiments in asynchronously growing human embryonic kidney (HEK) 293T cells. We performed affinity purification of SAMHD1-associated protein complexes followed by in-gel or on-bead digestion, respectively, and subsequent MS analysis (see methods; Fig. 2a and Supplementary Fig. 2a-c). Interestingly, we identified the known phosphorylating kinase CDK1 and the PP2A Aα subunit as SAMHD1 co-purifying proteins in both MS experiments (Fig. 2a and Supplementary Fig. 2a-c), suggesting PP2A as a potential SAMHD1-interacting phosphatase. To validate the MS findings, we overexpressed FLAG-tagged SAMHD1 in HEK293T cells and identified both endogenous PP2A A and C subunits in the pull-down (Fig. 2b).
To further characterize the interaction of SAMHD1 with PP2A, we performed co-immunoprecipitations (CoIPs) with SAMHD1 mutants in HEK293T cells (Supplementary Fig. 2d). None of the tested mutants, including single exchange of amino acids essential for SAMHD1 catalytic activities, displayed an altered binding capability to PP2A A and C subunit (Supplementary Fig. 2d). Exchange of the CDK target site T592 to a phosphomimetic (T592E) or a non-phosphorylatable (T592A) residue did also not alter binding of PP2A A and C subunits, indicating that PP2A binding is also not depending on this specific motif. Moreover, a mutant of SAMHD1 (R451E) known to prevent formation of dimers and tetramers31 influenced binding to PP2A A and C subunits only to a minor extent (Supplementary Fig. 2e), suggesting that residues responsible for PP2A binding might be accessible in all structural states of SAMHD1.
SAMHD1 pT592 is dephosphorylated by PP2A in vitro
Although members of both the PP2A and the protein phosphatase 1 (PP1) families are known to counteract CDK1-mediated phosphorylations during mitosis in mammalian cells32, the MS and subsequent CoIP experiments rather suggested PP2A as the most plausible SAMHD1 phosphatase.
To clarify which phosphatase acts on SAMHD1 at T592, we performed in vitro-dephosphorylation assays using recombinant SAMHD1 purified from Sf9 cells that proved to be already significantly phosphorylated on T59233. For the dephosphorylation, we used equilibrated units of a purified PP2A dimer (PP2AD; = PP2A A and C subunit) and a purified PP1 enzyme. After in vitro-dephosphorylation, the assay mixture was resolved by 1D-SDS-PAGE and the gel pieces containing SAMHD1 were digested with trypsin. The resulting peptide mixture was subjected to targeted LC-MS/MS (tandem mass spectrometry), followed by relative quantification analysis of phosphorylated and non-phosphorylated peptides encompassing T592 (Fig. 3a). The intensities of selected reaction monitoring (SRM) transitions corresponding to phosphorylated peptides were clearly diminished upon PP2A treatment in comparison to treatment with PP1 or buffer only. Concomitantly, the intensities of SRM transitions corresponding to the non-phosphorylated counterparts were higher upon PP2A treatment in comparison to treatment with PP1 or buffer only (Fig. 3a). From these intensity values, an average T592 phosphorylation stoichiometry of 76% could be calculated for control SAMHD1, whereas after PP2A treatment the average T592 phosphorylation stoichiometry goes down to 5% (Supplementary Table 1).
Immunoblot analysis further validated these MS data, revealing significantly decreased pT592 immune-reactivity, only upon treatment with PP2AD (Fig. 3b). Additionally, SAMHD1 pT592 dephosphorylation by PP2AD was time- and concentration-dependent, and could be completely inhibited by PP2A-selective concentrations of okadaic acid (OA), a pharmacological inhibitor of PP2A-like phosphatases34 (Fig. 3c). Thus, our in vitro-dephosphorylation assays clearly exclude PP1 as a SAMHD1 phosphatase, and provide further support for PP2A as the likely enzyme dephosphorylating pT592.
PP2A-B55α interacts with and dephosphorylates SAMHD1 at T592
Active PP2A holoenzymes are heterotrimers consisting of a scaffolding A subunit (2 isoforms: Aα/β; PPP2R1A/B), a catalytic C subunit (2 isoforms: Cα/β; PPP2CA/B) and a regulatory B-type subunit, which modulates substrate specificity as well as the temporal and spatial activity of the enzyme in cells35. PP2A B-type subunits are encoded by 15 different genes—resulting in 23 isoforms, which are classified into four different families: B (B55/PR55; PPP2R2), Bʹ (B56/PR61; PPP2R5), Bʺ (PR72; PPP2R3) and B‴ (striatins; STRN)34. Thus, we set out to identify the particular PP2A holoenzyme(s) specifically interacting with and dephosphorylating SAMHD1 at the M/G1 transition in cycling cells.
First, we isolated specific PP2A holoenzymes by expressing a large set of glutathione S-transferase (GST)-tagged B-type subunits in HEK293T cells and subsequently analyzed co-purification of FLAG-tagged SAMHD1. Interestingly, SAMHD1 was found enriched after pull-down of the PP2A-B55α holoenzyme, whereas no appreciable interaction could be observed with any other B-type subunit tested (Fig. 4a). At this point, we cannot exclude binding of other subtypes that may interact with SAMHD1 with lower affinity and could possibly target other phosphorylation sites. However, the reciprocal pull-down of SAMHD1 confirmed the interaction with PP2A-B55α trimers (Fig. 4b). Moreover, FLAG-tagged SAMHD1 co-purified with endogenously expressed PP2A-B55α holoenzymes in proliferating HEK293T cells (Fig. 4c). Endogenous SAMHD1 immunoprecipitated from proliferating THP-1 cells interacted with PP2A B55α subunit (Fig. 4d), thereby validating the interaction in the native-like environment at endogenous levels of both proteins. In line with our CoIP results, loss of SAMHD1 T592 phosphorylation during cell cycle-progression resembles PRC1 pT481 dephosphorylation (Supplementary Fig. 1c, f), a known target of B55α-containing PP2A holoenzymes36.
Known PP2A-B55α substrates share a common feature, a bipartite polybasic recognition determinant surrounding the CDK site in question37. Indeed, we identified basic patches surrounding the T592 site in human SAMHD1 that share similarity to the reported recognition determinant for B55 substrates37 (Fig. 4e, highlighted in black). To directly test if SAMHD1 might be a substrate of B55 and assess the specificity of interaction, we compared basic residues in the C-terminal region of human and murine (isoform 1) SAMHD1 surrounding T592. Isoform 1 of murine SAMHD1 contains less basic residues (Fig. 4e, highlighted in gray), which correlated with an almost complete loss of interaction with PP2A (Fig. 4f; Hs vs Mm wt). Removing basic residues in human SAMHD1 reduced (Fig. 4e, f; Hs∙Mm), whereas introduction of basic residues into murine SAMHD1 markedly increased interaction with PP2A after pull-down (Fig. 4e, f; Mm∙Hs), demonstrating key sites in SAMHD1 responsible for binding to PP2A-B55α holoenzymes. Taken together, these experiments demonstrate that PP2A-B55α holoenzymes specifically interact with SAMHD1 in asynchronously proliferating cells.
Next, we isolated PP2A-B55α holoenzymes from GFP-B55α-expressing HEK293T cells by a GFP-trapping approach, and used this PP2A trimer to dephosphorylate SAMHD1 pT592 in vitro. Again, a time- and concentration-dependent pT592 dephosphorylation was observed that was completely inhibited by addition of OA (Fig. 5a). When compared with in vitro-dephosphorylation by PP2AD, dephosphorylation by PP2A-B55α trimer proved at least as efficient (Fig. 5b), consistent with the reported preference of B55-type regulatory subunits to direct PP2A activity to phosphoserine-proline or phosphothreonine-proline sites37,38.
To further validate the specificity of the PP2A-B55α holoenzyme, recombinant SAMHD1 was incubated with respective holoenzymes of the PP2A B56/Bʹ and PR72/Bʺ subunit families. PP2A-B56α, PP2A-B56β, and PP2A-PR72 holoenzymes which did not bind to SAMHD1 in cells (Fig. 4a) were also not able to remove T592 phosphorylation in vitro (Fig. 5c; Supplementary Fig. 4e). Taken together, our results not only support a specific interaction of PP2A-B55α holoenzymes with SAMHD1 in cells, but also emphasize specific SAMHD1 pT592 dephosphorylation by this PP2A trimer.
Silencing of PP2A-B55α increases SAMHD1 T592 phosphorylation
To underscore the involvement of PP2A in SAMHD1 pT592 dephosphorylation, we treated HeLa cells with low, PP2A-selective concentrations (nM range) of OA34 for 8 h. Increased SAMHD1 phosphorylation at T592 was observed (Fig. 5d), suggesting that the inhibition of PP2A by OA leads to accumulation of phosphorylated SAMHD1.
To provide further support, we transfected several siRNAs that were previously shown to be on-target for single or all subunits of PP2A-B55α trimers39 into HeLa cells. 52 h post-transfection, protein levels of PP2A subunits were successfully decreased as confirmed by immunoblotting (Fig. 5e). As reported before, downregulation of PP2A Aα or Cα subunit leads to co-depletion of other subunits, including B55α39. Upon PP2A single subunit or triple knock-down, phosphorylation of endogenous SAMHD1 was increased compared to control siRNA. Intriguingly, already single knockdown of the PP2A B55α subunit led to an increase of SAMHD1 T592 phosphorylation (Fig. 5e), indicating that PP2A-B55α holoenzymes are mediating SAMHD1 dephosphorylation at this specific residue in cells. Surprisingly, we observed overall SAMHD1 protein levels decreasing after PP2A knockdown (Fig. 5e), suggesting that hyperphosphorylated SAMHD1 might be degraded by an unknown mechanism.
PP2A-B55α controls SAMHD1 pT592 level in non-cycling MDMs
We next analyzed whether phosphorylation of SAMHD1 at T592 was also controlled by PP2A in non-proliferating primary human HIV-1 target cells, namely MDMs (Supplementary Fig. 3a). As a consequence of PP2A inhibition with a PP2A-selective concentration of OA (2 nM), SAMHD1 phosphorylation at T592 increased over 8 h in different donors (Fig. 5f). This indicates that PP2A is involved in keeping T592 phosphorylation in MDMs in balance as well. Interestingly, we observed strong donor-dependent differences in steady-state SAMHD1 phosphorylation levels at T592, which correlated with minichromosome maintenance protein 2 (MCM2) expression (Fig. 5f and Supplementary Fig. 3b). As MCM2 is not expressed in G040, it can be used as a marker to track conversion of MDMs from the quiescent state (G0) to G1 phase. Interestingly, these varying levels of SAMHD1 T592 phosphorylation in different donors correlated with susceptibility to HIV-1 infection (Supplementary Fig. 3c, d). Nevertheless, in all observed donors, SAMHD1 phosphorylation at T592 increased upon inhibition of PP2A (Fig. 5f; Supplementary Fig. 3e). To prove that indeed PP2A-B55α trimers are responsible for dephosphorylating SAMHD1 in MDMs, same as in dividing cells (Fig. 5e), we silenced the PP2A B55α subunit in MDMs (Fig. 5g; Supplementary Fig. 3f). Highly efficient silencing of B55α mRNA (Fig. 5g, right panel) translates in apparent, but modest downregulation of B55α subunit protein expression (Fig. 5g, left panel), but remarkably, T592 phosphorylation of endogenous SAMHD1 was increased compared to control siRNA in MDMs (Fig. 5g, left panel; Supplementary Fig. 3f).
To summarize, employing the phosphatase inhibitor OA and siRNA-mediated silencing of PP2A, we observed an increase in SAMHD1 phosphorylation at T592 in cycling HeLa cells (Fig. 5d, e) as well as in differentiated MDMs (Fig. 5f, g; Supplementary Fig. 3e, f). Taken together, our results demonstrate that PP2A is important for controlling SAMHD1 phosphorylation at T592, in both cycling and differentiated cells.
Upregulation of B55α subunit in MDMs upon IFN treatment
Previous reports have shown that SAMHD1’s antiviral activity is regulated by type I IFN. SAMHD1 phosphorylation at T592 is decreasing in IFN-treated MDMs and monocyte-derived dendritic cells21. It was speculated that IFNs may induce an unknown phosphatase activity21. We hypothesized that specific subunits of PP2A may be differentially induced by IFNs. Therefore, we analyzed mRNA levels of various regulatory B-type subunits after treatment of MDMs with type I and II IFNs or upon infection with Sendai virus as a viral IFN inducer (Supplementary Fig. 4a, c). Primers used for quantitative real-time reverse transcription PCR (RT-qPCR) were isoform-specific, enabling us to discriminate between different PP2A B-type subunits and isoforms (Supplementary Fig. 4d). Tested B-type subunits were not detectable (B56δ) or not inducible by IFNs or Sendai virus (B56γ and ε) in MDMs, except PP2A B55α and to lower extent B56β subunit (Supplementary Fig. 4a, c). The in vitro-dephosphorylation assay clearly excluded B56β as a relevant B-type subunit for dephosphorylating T592 (Supplementary Fig. 4e), validating our previous findings for B55α as the only relevant subunit (Fig. 4 and Fig. 5). Moreover, mRNA levels of the PP2A B55α subunit were strongly induced by IFNs and Sendai virus 8 h and 24 h post-treatment, with maximal induction (up to 15.5-fold) through IFNβ (Supplementary Fig. 4a). In concordance, PP2A B55α subunit protein levels were upregulated after 24 h upon type I or II IFN induction (Supplementary Fig. 4b). Strikingly, B55α upregulation correlated with significant SAMHD1 dephosphorylation at T592 in MDMs (Supplementary Fig. 4b, left panel). In contrast to differentiated MDMs, IFN treatment of cycling HEK293T cells resulted in an upregulation of total SAMHD1 levels, which was paralleled by an increase in T592 phosphorylation and correlated with decreasing PP2A B55α subunit levels (Supplementary Fig. 4f), suggesting a cell type-specific regulation of SAMHD1 T592 phosphorylation through IFN. Taken together, our findings on IFN inducibility of B55α in MDMs provide further support for a physiological role of the PP2A-B55α holoenzyme as the responsible SAMHD1 pT592 phosphatase.
PP2A-B55α removes pT592 in SAMHD1 during mitotic exit
Interestingly, PP2A-B55α holoenzymes are established regulators of mitotic exit in cycling cells39, entirely in line with the observation that SAMHD1 loses T592 phosphorylation during M/G1 transition (Fig. 1a). To characterize SAMHD1 dephosphorylation at T592 at M/G1 transition in yet more detail, we chemically induced exit from mitosis in HeLa cells39. Using this assay, we were able to monitor dephosphorylation of CDK substrates (including SAMHD1), which is taking place during a brief time window from onset of anaphase until reformation of the nuclear envelope. HeLa cells were arrested in metaphase using nocodazole and MG-132 (proteasome inhibitor) and forced to exit mitosis by adding flavopiridol (CDK1 inhibitor) in presence of MG-132. Dephosphorylation of CDK substrates was detected at different time points after induction using an antibody specific to phosphorylated threonine (pT) within the context of the CDK target sequence pTPXK (Supplementary Fig. 5a). As expected, overall CDK substrate phosphorylation dropped to 9.3 ± SD of 1.2% (n = 3) within 18 min after flavopiridol addition (Supplementary Fig. 5a). Phosphorylation of SAMHD1 T592 decreased similarly to 27.3 ± 8.5% (n = 3) within 18 min (Supplementary Fig. 5b), demonstrating that SAMHD1 is specifically dephosphorylated during mitotic exit. Mitotic exit also resulted in a decrease in dATP levels within only 18 min (Supplementary Fig. 5c, d), which is in concordance with the drop in dATP in cells transitioning into G1 phase as observed in Fig. 1c. Notably, the assay is performed while MG-132 is constantly present (as indicated by constant expression of cyclin B1; Supplementary Fig. 5a, b, d), therefore degradation of other proteins involved in dNTP metabolism are most likely not responsible for this decrease in dATP levels. In line with the requirement for PP2A-B55α for timely CDK1 substrate dephosphorylation39, siRNA-mediated depletion of PP2A B55α subunit, delayed dephosphorylation of overall CDK substrates compared to control cells (reduction of pT signal to 28 ± 14.1% for siCtr compared to 40 ± 15.6% for siB55α; n = 2) (Supplementary Fig. 5e). Interestingly, also dephosphorylation of SAMHD1 at T592 was delayed compared to control cells (reduction of SAMHD1 pT592 signal to 42.5 ± 2.1% for siCtr compared to 66 ± 5.7% for siB55α; n = 2) (Fig. 6). Therefore, we conclude that SAMHD1 is specifically targeted by PP2A-B55α holoenzymes in cycling cells during a short time window at the end of mitosis ( = mitotic exit). We propose that SAMHD1 is converted into an active restriction factor during mitotic exit, thereby providing an antiviral-active state during G1 phase in cycling cells.
Reduction in HIV-1 RT products in G1 phase
Our results demonstrate that PP2A-B55α holoenzymes are responsible for dephosphorylating SAMHD1 at T592 during mitotic exit in cycling cells (Fig. 6). Additionally, entering G1 phase is accompanied by a notable and fast decrease in dATP levels (Fig. 1c and Supplementary Fig. 5c, d). In fact, we observed a significant decline for all four dNTPs upon G1 entry compared to S ( = 2 h post-release) and G2/M phase ( = 7 h post-release) (Supplementary Fig. 6a–c). Specifically, dCTP level are particularly low in G1 phase (4.22 ± 0.33 pmol/106 cells) compared to the low levels observed for the other dNTPs (dATP (17.97 ± 4.83 pmol/106 cells), dGTP (16.32 ± 1.96 pmol/106 cells) and dTTP (25.6 ± 5.03 pmol/106 cells); all dNTPs: n = 3, Supplementary Fig. 6a). However, dNTP levels in G1 phase are still higher when compared to the absolute levels in primary MDMs (Supplementary Fig. 7a, c) and resting CD4+ T cells (Supplementary Fig. 7b, c). Moreover, when taking into account the differences in cell volumes of the various cell types, HeLa cells in G1 phase contain ~21–35-fold and ~7–11-fold more dCTP than macrophages and resting CD4+ T cells, respectively (Supplementary Fig. 7d), questioning whether the observed dNTP levels in G1 phase may influence the critical threshold required for RT.
As a consequence, we were interested whether even in cycling cells dephosphorylated SAMHD1 would be able to reduce HIV-1 infection. Therefore, we synchronized HeLa cells using a double-thymidine block and infected at different time points post-release (Supplementary Fig. 6d). Notably, the points of infection were specifically chosen to ensure that onset of RT coincided with different cell cycle-phases and SAMHD1 phosphorylation states. Total DNA was harvested 4 h post-infection to prevent cells from progressing further in the cell cycle. During infection from S to G2/M phase (“inf 1”, 2–6 h post-release), SAMHD1 remains highly phosphorylated at T592 (Supplementary Fig. 6f) and cells are readily infected by HIV-1 (Supplementary Fig. 6e). In contrast, we detected less early RT products and a significant drop in late RT products (Supplementary Fig. 6e), when cells enter G1 phase during the infection period (“inf 2”). The reduction in RT products is mirrored by a decrease in SAMHD1 phosphorylation at T592 (“inf 2”, 7–11 h post-release; Supplementary Fig. 6f). Strikingly, infection with HIV-1 in the presence of Vpx (Supplementary Fig. 6g), which leads to degradation of SAMHD1 (Supplementary Fig. 6h), does not change the levels of RT products; the infection rates are comparable between G2/M and G1 phase (Supplementary Fig. 6g). In conclusion, reduction of RT products in cycling HeLa cells depends on the presence of SAMHD1 dephosphorylated at T592 (Supplementary Fig. 6i).
To validate that SAMHD1 could be antivirally active in the G1 phase of primary target cells for HIV-1, we infected activated CD4+ T cells with HIV-1 and subsequently arrested the cells in mitosis using nocodazole (Fig. 7a). Subsequently, nocodazole was either removed after 16 h, allowing CD4+ T cells to exit mitosis/enter G1 phase (“inf”), or nocodazole treatment was continued until 24 h post-infection (“inf + N”; Fig. 7a) to keep cells in mitosis. Indeed, continued nocodazole treatment leads to increased SAMHD1 phosphorylation at T592 in all donors analyzed compared to non-synchronized, activated CD4+ T cells (Fig. 7f, comparing “ + N” to condition “a”). In contrast, SAMHD1 phosphorylation at T592 is reduced after nocodazole wash-out, as CD4+ T cells were able to enter G1 phase, which is also indicated by decreasing cyclin B1-levels (Fig. 7f, “-N”). Interestingly, we observed higher early and late HIV-1 RT products upon HIV-1 infection in absence of Vpx (-Vpx) in mitotic cells, which displayed high SAMHD1 pT592-levels (“inf + N”), than compared to CD4+ T cells that entered G1 phase characterized by a reduced pT592 signal (“inf”) (Fig. 7b, d, “-Vpx”). Again, as seen in HeLa cells, infection with HIV-1 in the presence of Vpx leads to SAMHD1 degradation (Fig. 7f) and a significantly less pronounced difference in HIV-1 products (Fig. 7b, d, “ + Vpx”; Fig. 7c, e). Therefore, we conclude that the observed reduction of HIV-1 RT products upon G1 entry in cycling CD4+ T cells is depending on the presence of dephosphorylated, active SAMHD1.
Since SAMHD1 dephosphorylation at T592 by PP2A-B55α holoenzymes reduces HIV-1 infection efficiency, we tested whether HIV accessory proteins actively alter B55α subunit expression. No significant modulation of endogenous B55α levels could be detected (Supplementary Fig. 8), whereas B56δ subunit expression was counteracted by Vif as described recently41.
Taken together, our results demonstrate that PP2A is important for controlling SAMHD1 dephosphorylation at T592, in both cycling and non-cycling cells, and in consequence for its restrictive activity against HIV-1.
SAMHD1-mediated restriction of HIV-1 replication in non-dividing cells is part of the antiretroviral defense program. This restriction is strictly regulated by dephosphorylation of T592 in SAMHD121,22,23. The dephosphorylation at this specific residue might also impact SAMHD1 functions, such as the ability to degrade RNA6, tetramerize and hydrolyze dNTPs25,26,27, although this has been controversially discussed8,22,23,28,42,43. Phosphomimetic mutants of SAMHD1 are able to deplete the cellular levels of dNTPs22,23,42, suggesting that phosphorylation of this specific residue has no influence on the dNTPase function of SAMHD1. Nevertheless, these mutants abrogate viral inhibition, furthermore supporting the importance of post-translational regulation of SAMHD1 at T592 in HIV-121,22,23,42 and also HBV replication15. Moreover, phospho-regulation at T592 impacts replication fork progression, degradation of nascent DNA at arrested forks and chronic inflammation29.
SAMHD1 dephosphorylation at T592 is regulated during cell cycle-progression in proliferating cells. We show that in synchronized HeLa cells, phosphorylation at T592 starts building up at the G1/S border and remains constant until mitosis, which is in line with the requirements for phosphorylated SAMHD1 as a promoter for fork resection in S phase29. Upon re-entry of interphase, SAMHD1 is rapidly dephosphorylated at T592 (Fig. 1 and Supplementary Fig. 1), which is in concordance with a previous report by Yan et al. detecting a loss of phosphorylated T592 in G1-sorted cell populations of proliferating cells27. Moreover, we show a concomitant drop of all four dNTPs upon entering G1 phase (Fig. 1c; Supplementary Fig. 6a), reinforcing that dNTP levels are tightly regulated at the transition from M to G1. As dephosphorylation of SAMHD1 at T592 happens concurrently, this suggests either a coincidental or causative relationship between dephosphorylation and dNTPase activity. Tight cell cycle-regulation is in line with the cellular requirements for balancing dNTP levels during cell cycle-progression44. It could be hypothesized that SAMHD1 as the counterpart to the ribonucleotide reductase (RNR), which synthesizes dNTPs de novo45, becomes an active dNTPase in the G1 phase (Fig. 1c), whereas the R2 subunit of RNR is absent in G0/G1 upon its degradation in mitosis. Furthermore, we extended this observation by analyzing the exact temporal transition between M and G1 phase. We chemically induced mitotic exit events in synchronized cells by inactivation of CDKs in absence of proteasome-mediated degradation. SAMHD1 dephosphorylation at T592 happens during mitotic exit (Supplementary Fig. 5b)—a complex, irreversible and unidirectional process that starts with central spindle assembly and chromosome segregation. This process is highly regulated (i) through degradation of mitotic factors and (ii) by phosphatases that remove phosphorylations from mitotic substrates32. Intriguingly, concurrently with SAMHD1 dephosphorylation at T592, dATP levels dropped rapidly within 18 min upon forced exit out of mitosis (Supplementary Fig. 5c, d). As this assay was performed in the presence of a proteasomal inhibitor, the degradation of the R2 subunit of the RNR or possibly other not yet discovered factors responsible for regulating cellular dNTP pools might not be held accountable for this rapid decline in dNTPs. But at this point, we cannot rule out that not yet identified metabolic enzymes or (co)-factors regulating the decline in dNTP pools might get activated during mitotic exit. Since SAMHD1 expression levels did not change within different cell cycle-phases (Fig. 1), consistent with reported observations in sorted, unsynchronized human primary CD4+ T cells and monocytic cell lines27,46, we suggest that SAMHD1 function is not regulated by fluctuating expression levels (as it was proposed by45,47). Instead, we propose that SAMHD1’s diverse functions are regulated throughout the cell cycle by post-translational modifications. Further studies are needed to determine how T592 dephosphorylation mechanistically modulates SAMHD1’s enzymatic activities and functions in vivo, specifically the antiviral mechanism. We cannot rule out that exit from mitosis itself or the dephosphorylation event at T592 might trigger cell cycle-specific protein interactions, possibly in interplay with additional post-translational modifications46, or nucleic acid binding that would influence the enzymatic functions of SAMHD1 during G1 phase.
Intriguingly, our studies revealed the responsible phosphatase PP2A that renders SAMHD1 restrictive against HIV-1. As PP2A substrate specificity is defined by the regulatory B-type subunits48, we revealed that PP2A holoenzymes containing the specific subunit B55α dephosphorylate SAMHD1 at T592 during mitotic exit (Fig. 6). In PP2A-B55α-depleted cells, the removal of phosphorylation at CDK substrates and SAMHD1 phosphorylation at T592 was delayed in comparison to control cells (Fig. 6 and Supplementary Fig. 5). The results support the conclusion that T592 in SAMHD1 is a mitotic substrate of PP2A-B55α trimers, and are in line with the fact that this particular holoenzyme represents a key mitotic exit phosphatase in mammalian cells39. Importantly, the discovery of PP2A-B55α as the responsible phosphatase acting on T592 in SAMHD1 during mitotic exit is in line with current knowledge on the timely coordinated and conserved program that occurs in mammalian cells during mitotic exit49 (see Supplementary Note 1).
We identified key basic residues flanking T592, a unique recognition site shared only by B55 substrates, responsible for interaction between SAMHD1 and the holoenzyme PP2A-B55α (Fig. 4e, f). Interestingly, mitotic exit substrates are recognized in a temporal order encoded in this site. The phosphatase decodes this signal through a complementary negatively charged surface on the regulatory subunit B5537. Additionally, it was shown that the phospho-amino acid threonine over serine determines the turn-over rate during mitotic exit37,50, suggesting that threonine at residue 592 determines human SAMHD1 as an ideal strong B55 substrate. An interesting side aspect comparing human to murine SAMHD1 (Fig. 4e, f) suggests that murine isoform 151,52 may (i) be a low affinity substrate of PP2A-B55 and (ii) may display slower dephosphorylation kinetics compared to human SAMHD1 as it contains much less basic residues. Of note, human and murine B55α subunit display 100% sequence identity. This would argue for a different regulation of murine compared to human SAMHD1. The identified key residues critical for binding of human SAMHD1 to PP2A are situated in the C-terminus, a platform for interaction with cellular and viral partners53,54,55, and are highly accessible based on the structural model (see Supplementary Note 1).
Importantly, as soon as proliferating cells enter the G1 cell cycle-stage, HIV-1 infection resulted in significant reduction of RT products depending on the presence of SAMHD1 dephosphorylated at T592 (Fig. 7 and Supplementary Fig. 6d-i), while concomitantly dNTP levels are low (Fig. 1c; Supplementary Fig. 5c; Supplementary Fig. 6a-c). We propose the following model (Fig. 8a): even in proliferating cells, the G1 stage of the cell cycle represents a short window where SAMHD1 could be antivirally active and delay RT kinetics. This could explain previous observations on restrictive SAMHD1 in proliferating cells20,56. Our results suggest that SAMHD1 is dephosphorylated by PP2A-B55α trimers in cycling cells every time cells exit mitosis and enter interphase. Active SAMHD1 in G1 then could delay RT of HIV-1. Whether the reduced cellular pool of dNTPs is responsible for HIV-1 inhibition in G1 phase, the proposed controversial nuclease activity4,5 or nucleic acid-binding properties2,3 triggered upon mitotic exit, will need to be determined. In fact, the calculated dNTP concentrations in cells (mainly) refractory to HIV-1 infection, macrophages and resting CD4+ T cells, are lower than the measured dNTP levels in the G1 phase in HeLa cells (Supplementary Fig. 7d). When taking into account the range of Km and Kd values of HIV-1 RT57, this suggests that the lowest measured level in G1 phase that would be rate-limiting (1.09–1.8 µM dCTP) lies still above the general Km values of HIV-1 RT and might not significantly affect the replication kinetics. Still, the limited substrate concentrations could influence dNTP binding affinity. Particularly, highly structured RNA templates such as retroviral RNA are known to interrupt processive DNA synthesis resulting in stalling of RT58. This mechanism is influenced by restricted substrate availability59 and could contribute to the observed reduction of RT in G1 phase. On the other hand, it is likely that additional mechanism of restriction by SAMHD1 play a role that are triggered upon mitotic exit and correlate with dephosphorylated SAMHD1 at T592. The latter hypothesis would agree with Welbourn et al.42 who provide evidence for a dNTPase-independent function of SAMHD1. This function particularly contributes to the lentiviral restriction observed in HeLa cells where dNTP levels were artificially lowered by inhibiting the RNR42, paralleling our experiment in synchronized cells providing a state of low dNTP levels in G1 phase. In contrast, SAMHD1 is phosphorylated from S to M phase (Fig. 1a) and rendered inactive against HIV-1. We speculate that HIV-1-infected cells are reprogrammed to avoid progression to the end of mitosis, where PP2A-B55α holoenzymes would be active and could remove CDK-mediated phosphorylations from targets like SAMHD1. Thereby, HIV-1 actively evades dephosphorylated, restrictive SAMHD1 present in G1 phase in cycling cells. As HIV-1 capsid uncoating is triggered only after first strand transfer of RT60 and as SAMHD1 restriction is a reversible process61, it might be likely that HIV-1 capsids are able to “wait around” for the S phase to continue the infection process.
Intriguingly, HIV-1 actively manipulates cell cycle-progression by inducing a G2 arrest mediated by the accessory protein Vpr62. First indications suggested an involvement of PP2A as (i) okadaic acid and (ii) regulatory and catalytic subunits of PP2A were shown to influence the Vpr-induced G2 arrest63,64. Interestingly, the HIV-1 accessory protein Vif has been reported to affect specific PP2A holoenzymes by directing the B56 regulatory subunits to CUL5-dependent proteasomal degradation41. Indeed, our results confirm Vif-dependent influence on B56 subunit expression (Supplementary Fig. 8); however, our proofs indicate no striking influence on PP2A-B55α subunit expression by any of the accessory proteins (Supplementary Fig. 8). Intriguingly, in fission yeast, it was demonstrated that the events during mitotic exit are driven by a coordinated and sequential activation of various phosphatases, the so-called phosphatase relay, where PP1 activation is required for the reactivation of both PP2A-B55 and PP2A-B56 and, in turn, full activation of the phosphatase relay65. Given the conservation of PP1 docking sites in both subunits across species66, proper control of mitotic exit events may indeed be dependent on a phosphatase relay. This would suggest that antagonism of PP2A-B56 by HIV-1 Vif could possibly influence PP2A-B55α action during mitotic exit and possibly represent another level of active manipulation of the cell cycle besides the mentioned G2/M arrest by Vpr, in order to avoid antivirally active SAMHD1 in the G1 stage.
Moreover, we provide first evidence that PP2A is involved in controlling the level of SAMHD1 T592 phosphorylation in differentiated MDMs (Fig. 5f, g). We propose the following model (Fig. 8b): MDMs contain active kinases that are constantly kept in check by the action of PP2A to hold a balance of dephosphorylated targets, such as SAMHD1. This is supported by the fact that silencing of various CDKs in MDMs has been reported to reduce phosphorylation of T59230. Intriguingly, we could show that the PP2A B55α subunit was specifically upregulated at mRNA and protein level by IFNs in MDMs (Supplementary Fig. 4a, b; see Supplementary Note 1), indicating that PP2A-B55α holoenzymes are preferentially involved in dephosphorylating SAMHD1 after IFN stimulation in MDMs. This would explain the reported observation that a phosphatase activity might be inducible by IFNs in MDMs21, which reflects another layer of control over SAMHD1’s restrictive capacity against HIV-1.
Taken together, our findings suggest PP2A-B55α as the key regulator to dephosphorylate SAMHD1 at T592 in MDMs and during mitotic exit in proliferating cells, thereby converting SAMHD1 into an active HIV-1 restriction factor. Understanding the activation process of SAMHD1 will facilitate the development of new therapeutic strategies to control virus replication and to regulate the innate immune system.
Human buffy coats of anonymous blood donors were purchased from the German Red Cross Blood Donor Service Baden-Württemberg Hessen.
Cell lines and human primary cells
Human HeLa ‘Kyoto’39 (resource identification: RRID:CVCL_1922) and HEK 293 T (ATCC No.: CRL-3216) or 293 T/17 (ATCC No.: CRL-11268) cells were grown in DMEM (Lonza) supplemented with 10% (v/v) fetal bovine serum (FBS; Sigma-Aldrich) and 2 mM l-glutamine (Biochrom) at 37 °C and 5% CO2. Human monocytic THP-1 (ATCC No.: TIB-202) cells were grown in RPMI-1640 (Biowest) supplemented with 10% (v/v) FBS and 2 mM l-glutamine at 37 °C and 5% CO2. All the cell lines were regularly tested for the absence of mycoplasma using the MycoAlert Mycoplasma Detection Kit (Lonza). Authentication of cell lines was confirmed by examining genetic characteristics through PCR-single-locus-technology (Eurofins Medigenomix Forensik).
Human peripheral blood mononuclear cells (PBMCs) were isolated by density gradient centrifugation (980g, 30 min, RT) using Histopaque-1077 (Sigma-Aldrich). PBMCs were diluted in PBS, centrifuged (550g, 10 min, RT) and cell pellets were resuspended in 10 mL 0.87% ammonium chloride for lysis of erythrocytes (10 min, 37 °C). PBMCs were washed twice with PBS and the amount of isolated cells was determined. Monocytes were purified from PBMCs using CD14 MicroBeads (Miltenyi Biotec) according to the manufacturer’s instructions. CD4+ T cells were purified from PBMCs using CD4 MicroBeads (Miltenyi Biotec) according to the manufacturer’s instructions. Separation from unlabeled cells was carried out using an autoMACS Pro Separator (Miltenyi Biotec).
For the generation of monocyte-derived macrophages (MDMs), isolated monocytes were cultured in RPMI-1640 supplemented with 10% FBS, 2 mM l-glutamine, 10 mM HEPES, 1 mM sodium pyruvate and 100 U/mL granulocyte-macrophage colony-stimulating factor (GM-CSF; PeproTech). After 72 h, fresh medium containing the respective cytokines was added. On day 5, all cells were harvested and the amount of differentiated cells determined using the Cellometer Auto T4 (Nexcelom Bioscience).
For activation of CD4+ T cells, isolated cells were cultured at a density of 3 × 106 cells/mL in RPMI-1640 supplemented with 10% FBS, 2 mM l-glutamine, 1% Pen-Strep and 25 mM HEPES, 500 U/mL Proleukin S (Novartis) and 5 µg/mL phytohemagglutinin (PHA-P; Sigma-Aldrich) for 5 days, including renewal of the culture medium/supplements 72 h after start of activation.
IFN treatment and RT-qPCRs
For IFN stimulation, 7.5 × 105 cells/12-well (RNA samples for quantitative real-time reverse transcription PCR (RT-qPCR)) or > 5 × 105 cells/12-well (whole-cell lysates for immunoblotting) were seeded. MDMs were left untreated or treated with 1000 U/mL IFNα2a (PeproTech), IFNβ1a (PBL Interferon Source) or IFNγ (PeproTech) for 8/24 h (RNA samples) or 24 h (whole-cell lysates) at 37 °C. As a control for IFN induction, MDMs were infected with Sendai virus (final dilution = 1:200).
Total RNA from MDMs was isolated using the RNeasy Plus Mini Kit (QIAGEN) or NucleoSpin RNA Kit (Macherey-Nagel). Expression of mRNAs was determined in a 384-well format using QuantiTect SYBR Green RT-PCR Kit (QIAGEN) on an ABI7900 cycler (Applied Biosystems). Isoform-specific primers are listed in the Supplementary Methods.
HIV-1 infection and viral DNA quantitation by qPCR
3 × 104 HeLa cells/12-well were seeded and synchronized using a double-thymidine block. After the 2nd release, cells were counted at each time point to ensure the use of equal amounts of virus. Subsequently, cells were infected with a VSV-G-pseudotyped, full-length HIV-1 reporter viruses mutated in p6 allowing for production of virions that package SIVmac239 Vpx67 and lacking Vpr (pNL4.3 E-R- luc3 chp6_pcDNA, MOI 3; pNL4.3 E-R- luc3 chp6_Vpx, MOI 1.5). Prior to infection, HIV-1 virus stocks were incubated with 10 U/mL DNaseI (NEB) (30 min, 37 °C) to reduce contamination by plasmid DNA in subsequent qPCR assays. For control purposes, the used virus was heat-inactivated (65 °C, 20 min). Cells were spin occulated (30 min, 245g, 32 °C) and the virus removed 1 h post-infection. In order to monitor cell cycle-phases/SAMHD1 phosphorylation states, samples for immunoblotting were harvested at the time of infection and DNA harvest.
2 × 105 activated CD4+ T cells/96-well were seeded and infected with a VSV-G-pseudotyped, full-length HIV-1 reporter viruses -/+Vpx lacking Vpr (pNL4.3 E-R- luc chp6_pcDNA and pNL4.3 E-R- luc chp6_Vpx, MOI 7.5 each). Prior to infection, HIV-1 virus stocks were incubated with 10 U/mL DNaseI (NEB) (30 min, 37 °C) to reduce contamination by plasmid DNA in subsequent qPCR assays. For control purposes, the used virus was heat-inactivated (95 °C, 10 min). Cells were spin occulated (1.5 h, 800g, 32 °C) and the virus removed 2 h post-infection. With removal of virus, cells were arrested using 100 ng/mL nocodazole (Sigma-Aldrich) for 16 h ( = inf) or 24 h ( = inf + N) at 37 °C. In order to allow cell cycle-progression from mitosis into G1 phase, arrested CD4+ T cells were washed twice with warm medium to remove nocodazole (after 16 h) and again incubated for 6 h at 37 °C ( = inf). In order to monitor cell cycle-phases/SAMHD1 phosphorylation states, samples for immunoblotting were harvested prior to infection and 24 h post-infection ( = DNA harvest).
For both cell types, virus was titrated in a time-course experiment to detect the onset of RT products upon 3 h after infection in HeLa cells and upon 16 h after infection in activated CD4 + T cells, respectively.
Total DNA was isolated 4 h (HeLa) or 24 h post-infection (CD4+ T cells) post-infection using the DNeasy Blood & Tissue Kit (QIAGEN). DNA was subjected to qPCR to specifically quantify HIV-1 early and late RT products. To normalize the amount of input DNA, the copy number of the cellular gene porphobilinogen deaminase (PBGD) was quantified using a standard curve generated in the respective experiment.
A full list of plasmids used in this study and detailed protocols describing respective cloning procedures can be found in the Supplementary Methods.
For synchronization, 0.5 × 106 HeLa cells were seeded per 10-cm-cell-culture dish. 24 h after seeding, the medium was replaced with DMEM + 2 mM thymidine and cells were incubated for 16 h at 37 °C ( = 1st block). Cells were washed and released for 8 h at 37 °C into DMEM + 25 µM 2’-deoxycytidine ( = 1st release). The procedure was repeated ( = 2nd block/2nd release) and cells were harvested at different time points post-release. The respective samples were split for immunoblot analysis and determination of cell cycle-phases by propidium iodide (PI) staining/flow cytometry.
Chemically induced mitotic exit assay
For the chemically induced mitotic exit assay39, 4–5 × 105 HeLa cells were seeded in 5 mL DMEM per 6-cm-cell-culture dish. 17 h before chemical induction of mitotic exit ( = day 2), HeLa cells were arrested using nocodazole (100 ng/mL). On day 3, arrested cells were incubated for 30 min in DMEM containing 30 µM MG-132 and collected by mitotic shake-off. Cells were washed, resuspended in PBS (containing 30 µM MG-132) and divided into aliquots. Subsequently, cell aliquots were incubated at 37 °C and forced to exit mitosis by addition of flavopiridol (final concentration: 20 µM). Single cell aliquots were lysed every 3 min (over a total time period of 18 min) using radio-immunoprecipitation assay (RIPA) buffer (supplemented with protease and phosphatase inhibitors) and directly frozen at -20 °C.
RNA interference (RNAi)
For immunoblot analysis, 2 × 104 HeLa cells were seeded in 500 µL DMEM per 24-well and incubated overnight at 37 °C. HeLa cells were transfected with isoform-specific small interfering RNAs (siRNAs) for different PP2A subunits or scrambled controls (see Supplementary Methods) using Lipofectamine RNAiMAX Reagent (Invitrogen) according to the manufacturer’s instructions (siRNA amount/well: 5 pmol). 52 h post-transfection, cells were harvested for immunoblot analysis.
For the chemically induced mitotic exit assay, 5 × 105 HeLa cells were seeded in 5 mL DMEM per 6-cm-cell-culture dish and incubated overnight at 37 °C. 52 h before chemical induction of mitotic exit, HeLa cells were transfected as described above (siRNA amount/dish: 50 pmol).
For siRNA-transfection of MDMs, 3 × 105 MDMs/24-well were seeded in 500 µL RPMI-1640 (supplemented with 10% FBS, 2 mM l-glutamine, 10 mM HEPES, 1 mM sodium pyruvate) and allowed to re-attach for 2 h at 37 °C. MDMs were transfected with siRNAs specific to PP2A B55α subunit, PP2A Cα subunit or a scrambled control using Stemfect RNA Transfection Kit (Stemgent; siRNA amount/well: 10 pmol). After 24 h, the medium was changed and MDMs were again transfected with siRNAs as described above. 48 h or up to 5 days post-transfection, MDMs were harvested for immunoblot and RT-qPCR analysis.
Mass spectrometry to identify SAMHD1-interacting proteins
Two independent mass spectrometry (MS) approaches were performed (Fig. 2a: GFP-traps subjected to on-bead trypsin digestion and Supplementary Fig. 2: in-gel digestion of tandem-affinity purified CBP-SBP-SAMHD1, see Supplementary Methods for the latter approach). For GFP-trapping, 3 × 106 HEK293T cells/10-cm-cell-culture dish were seeded. After 24 h, HEK293T cells were transfected with 6 µg (pEGFP-C1-SAMHD1 or pEGFP-C1) of total plasmid DNA per dish using 18 mM PEI reagent. 48 h post-transfection, cells were harvested, pooled and lysed in 1 mL NET lysis buffer/sample.
Anti-GFP trapping of the same amount of lysates from GFP or GFP-SAMHD1 overexpressing cells was executed; notably, GFP-trapped complexes were washed stringently for four times using NENT300 (20 mM Tris [pH 7.4], 300 mM NaCl, 1 mM EDTA [pH 7.4], 0.1% NP40, 25% glycerol). GFP-traps were subjected to on-bead trypsin digestion and the resulting peptide mixture was analyzed by LC-MS/MS on a nano-LC hybrid quadrupole-orbitrap mass spectrometer (Q Exactive, Thermo Fisher Scientific). Relative quantification of proteins in different conditions was executed with Progenesis software (Nonlinear Dynamics) incorporating protein identifications obtained by MASCOT (Matrix Science) search engine using Swiss-Prot (Homo sapiens, 20,202 entries) as a database. Only peptides with 95% peptide identification probability, resulting in a 1% peptide FDR (Scaffold), were taken into account during the analysis. As a result, a list of proteins present in the GFP-SAMHD1 trap and not present in the GFP trap could be established (280 hits). CDK1 and the PP2A Aα ( = 65 kDa) subunit were present in this list.
With transfected nFLAG-SAMHD1 in HEK293T cells. For CoIPs, 3 × 106 HEK293T cells were seeded per 10-cm-cell-culture dish. After 24 h, HEK293T cells were transfected with 6 µg (pcDNA3.1( + )-nFLAG-SAMHD1 only) or 11 µg (5.5 µg pcDNA3.1( + )-nFLAG-SAMHD1 + 5.5 µg PP2A B-type subunits in pEGFP ( = GFP-tagged)) of total plasmid DNA per dish using 18 mM PEI reagent. 48 h post-transfection, cells were harvested and lysed in 200 µL NET lysis buffer (containing protease and phosphatase inhibitors) per dish for 30 min on ice. Lysates were pre-cleared and subsequently incubated with 25 µL ANTI-FLAG M2 Affinity Gel (Sigma-Aldrich) for 1 h at 4 °C. After washing, bound immune complexes released in 25 µL LDS Sample Buffer through boiling (95 °C, 5 min).
With transfected GST-tagged PP2A B subunits in HEK293T cells. For CoIPs, 3 × 106 HEK293T cells were seeded per 10-cm-cell-culture dish. After 24 h, HEK293T cells were transfected with 11 µg of total plasmid DNA (3.5 µg pcDNA3.1( + )-nFLAG-SAMHD1 + 7.5 µg PP2A B-type subunits in pGMEX-T1 ( = GST-tagged)) per dish using 18 mM PEI reagent. 48 h post-transfection, cells were harvested and lysed in 200 µL NET lysis buffer (containing protease and phosphatase inhibitor) per dish for 30 min on ice. Lysates were incubated with 25 µL GST-Trap_M beads (Chromotek; Nano-Traps consisting of a second type of antibody of the Camelidae species, called heavy chain antibodies (hcAbs), that bind their antigen through a single variable domain (VHH) devoid of light chains (so-called nanobody), coupled to magnetic microparticles) for 1.5 h at 4 °C. Subsequently, beads were washed and bound immune complexes released in 25 µL LDS Sample Buffer through boiling (95 °C, 5 min).
In vitro-dephosphorylation assays and targeted LC-MS/MS
De novo purified PP2AD and PP1 were obtained from a commercial source (Millipore). For comparative MS analysis, 1 µg of recombinant SAMHD1 was incubated in enzyme dilution buffer (0.15 M NaCl, 20 mM MOPS, pH 7.5, 60 mM 2-mercaptoethanol, 0.1 mM MnCl2, 1 mM MgCl2, 1 mM EGTA, 10% glycerol and 0.1 mg/mL serum albumin) with 0.032 units/µL PP2A, 0.032 units/µL PP1 or buffer only, for 1 h at 30 °C, in a reaction volume of 10 µL. 1 µL of complete protease inhibitor mix (Roche) was also included. Reactions were stopped by adding 5 × SDS-PAGE sample buffer and 10 min of boiling. Following SDS-PAGE of the samples, the 1D-gel piece containing SAMHD1 was subjected to in-gel trypsin digestion followed by C18 peptide cleanup (ZipTip, Millipore). The resulting peptide mixture was subjected to targeted LC-MS/MS on a nano-LC hybrid quadrupole-orbitrap mass spectrometer (Q Exactive, Thermo Fisher Scientific). Data analysis was executed with PinpointTM 1.4.0 (Thermo Fisher Scientific) software. Normalization of data in the different conditions was done based on a non-phosphorylatable peptide of SAMHD1 (GGFEEPVLLK) (Supplementary Table 1). Calculation of phosphorylation stoichiometry was executed as described in Schreurs et al.68, and based on a method described by Olsen et al.69. For immunoblot analysis with anti-SAMHD1 pT592 antibodies, the amount of SAMHD1 was reduced to 0.1 µg per dephosphorylation reaction. For in vitro-dephosphorylation with PP2A trimers, the relevant GFP-tagged B-type subunits (B55α, B56α, B56β, PR72) were first expressed in HEK293T cells, and the respective catalytically competent PP2A trimers harboring a specific B-type subunit were retrieved from the transfected cells by GFP-trapping, as previously described70. Whenever OA (Calbiochem) was used in the assays, the phosphatases were first pre-incubated with 50 nM OA for 10 min at 30 °C, before their addition to the substrate. All dephosphorylation reactions were stopped by adding 5 × SDS-PAGE sample buffer and boiling.
Uncropped scans of immunoblots are supplied in the Supplementary Information.
Statistical analysis was performed using the Graph Pad Prism software (Version 7.03). Data are represented as mean ± standard deviation (SD). To assess statistical significance, a paired/unpaired, two-tailed Student’s t-test was applied (ns: p ≥ 0.05; *: p < 0.05; **: p < 0.01; ***: p < 0.001) or one-way ANOVA with multiple comparisons according to Dunnett/Sidak (ns: p ≥ 0.05; *: p < 0.05; **: p < 0.01; ***: p < 0.001; P values were adjusted to account for multiple comparisons).
All data are available from the corresponding author upon request.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The authors thank Anna-Maria Engin, Nicole Gorgs, and Michaela Neuenkirch for technical assistance as well as Maximilian Rieß, Manja Burggraf, and Andreas Sommer for experimental support/critical discussion. Additionally, The authors thank Oliver T. Keppler (LMU Munich, Germany; for anti-SAMHD1 pT592 antibody), Sumit K. Chanda (Sanford Burnham Prebys Medical Discovery Institute, USA; for siRNAs), Nevan J. Krogan (University of California San Francisco (UCSF), USA; for plasmids encoding HIV-1 accessory proteins), Nathaniel R. Landau (NYU School of Medicine, USA; for pNL4.3 plasmids), Georg Kochs (University of Freiburg, Germany; for Sendai virus), Daniel Gerlich (University of Vienna, Austria; for HeLa ‘Kyoto’ cells), and Stephen Dilworth (Middlesex University London, UK; for monoclonal PP2A C antibodies) for reagents. This work is supported by Deutsche Forschungsgemeinschaft (DFG): CRC1292 project TP04 and SPP1923 project KO 4573/1-1 to R.K., BMBF 01KI1307A and German Center for Infection Research (DZIF) HZI2010Z10, DZIF TTU 01.802 to R.K. V.J. and R.D. were supported by the IAP Program of the Belgian federal government (P7/13) and the KU Leuven Research Council (OT/13/094). A.B.N. and F.D.G. were supported by an NIH R01 GM123540 grant to F.D.G. B.K., B.M. and C.S. were supported by R01 GM104198 and R01 AI136581 to B.K.
About this article
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.