Nucleic acid recognition orchestrates the anti-viral response to retroviruses

SUMMARY

Intrinsic restriction factors and viral nucleic-acid sensors are important for the anti-viral response. Here, we show how upstream sensing of retroviral reverse transcripts integrates with the downstream effector APOBEC3, an IFN-induced cytidine deaminase that introduces lethal mutations during retroviral reverse transcription. Using a Murine Leukemia Virus (MLV) variant with an unstable capsid that induces a strong IFNβ antiviral response, we identify three sensors, IFI203, DDX41 and cGAS, required for MLV nucleic-acid recognition. These sensors then signal using the adaptor STING, leading to increased production of IFNβ and other targets downstream of the transcription factor IRF3. Using knockout and mutant mice, we show that APOBEC3 limits the levels of reverse transcripts that trigger cytosolic sensing, and that nucleic-acid sensing in vivo increases expression of IFN-regulated restriction factors like APOBEC3 that in turn reduce viral load. These studies underscore the importance of the multiple layers of protection afforded by host factors.

INTRODUCTION

The genomes of mammals and other species contain many genes that restrict pathogenic infectious retroviruses (Goff, 2004b; Harris et al., 2012; Malim and Bieniasz, 2012; Okeoma and Ross, 2010). Among these are the Apolipoprotein B Editing Complex 3 (APOBEC3) genes, which have undergone positive selection and are present in different copy numbers in different species (Compton et al., 2012; OhAinle et al., 2006; Sawyer et al., 2004). APOBEC3 proteins, which are packaged into retroviral virions, are cytidine deaminases that act during reverse transcription. Deamination of cytosine residues in reverse-transcribed DNA results in G-to-A mutations and lethal mutations in virus coding regions or to degradation of reverse transcripts (Harris et al., 2003). APOBEC3 proteins also inhibit reverse transcription and block viral cDNA synthesis (Bishop et al., 2008; Iwatani et al., 2007; MacMillan et al., 2013).

Recent work has suggested that retroviral reverse transcripts are sensed by cellular DNA sensors such as cyclic GMP-AMP synthase (cGAS), DExD/H-box helicase 41 (DDX41) and Interferon (IFN)-Induced 16 (IFI16) (Gao et al., 2013; Jakobsen et al., 2013; Monroe et al., 2014; Zhang et al., 2011). cGAS produces cyclic GMP-AMP upon DNA binding, which in turn activates Stimulator of IFN Genes (STING; see below) (Sun et al., 2013; Wu et al., 2013). DDX41 plays a role in the IFN response to B-form DNA and to DNA but not RNA viruses (Zhang et al., 2011). IFI16 belongs to the Absent in Melanoma 2 (AIM2)-like receptor (ALR) family, characterized by the presence of a N terminal pyrin and 1 to 2 C-terminal Hin domains, termed Hina and Hinb, which encode DNA binding activity (Brunette et al., 2012; Cridland et al., 2012). The ALRs are also under positive selection; while the human genome contains 4 ALRs, the mouse genome contains 13–14 genes, encoded at a single locus (Brunette et al., 2012; Cridland et al., 2012). Several groups have suggested that the mouse Ifi204 gene is the functional equivalent of IFI16, because it contains two C-terminal Hin domains and its knockdown abrogates the response to transfected DNA (Lee et al., 2013; Paludan and Bowie, 2013; Unterholzner et al., 2010). The specific role(s) of the different mouse ALRs in host responses to pathogen nucleic acids has not been determined.

IFI16, DDX41 and cGAS interact with and signal through STING, which activates the kinase TBK1, resulting in phosphorylation of the transcription factor IFN regulatory factor 3 (IRF3) and its trafficking to the nucleus where it activates transcription of genes including type 1 IFNs (Bhat and Fitzgerald, 2014). While the loss of any one of these factors abrogates the STING-mediated IFN response to cytosolic DNA, the inter-relationship between the actions of these factors has not been determined. Moreover, although cGAS- and IFI16-mediated activation induce the type I IFN in tissue culture cells and affect HIV and MLV infection levels, how IFNs decrease infection and whether the STING-mediated pathway is important to the in vivo anti-retroviral response is not known (Jakobsen et al., 2013).

We showed recently that retroviral capsid stability affected the ability of cytosolic sensors to detect reverse transcripts. We found that a variant of murine leukemia virus (MLV) that lacks expression of the viral glycosylated-Gag protein (MLV^gGag) had a destabilized capsid and induced higher levels of IFNβ RNA in macrophages compared to wild type virus (MLV^WT) (Stavrou et al., 2013). The IFNβ response was further increased by knockdown of the cytosolic DNA exonuclease Three Prime Repair Exonuclease 1 (Trex1), which degrades cytosolic retroviral reverse transcribed DNA (Yan et al., 2010). While MLV^gGag was attenuated for infection in wild type mice, it replicated to the same extent as MLV^WT in APOBEC3 null mice suggesting that exclusion of restriction factors like APOBEC3 and DNA sensors by capsid is critical for virus replication in vivo (Kolokithas et al., 2010; Stavrou et al., 2013).

Here we show that the ligand sensed during MLV infection is reverse-transcribed DNA and that three factors, IFI203, DDX41, and cGAS, as well as the downstream effector STING, are required for the Trex1-dependent IFNβ response to both MLV and HIV. IFI203 and STING are also critical to in vivo control of MLV, since mice with a STING mutation (STING^mut) or with a polymorphism in Ifi203 gene (DBA/2) both showed a diminished IFNβ response and were significantly more infected with virus compared to C57BL/6 mice. Importantly, we found that there was a hierarchy of restriction in vivo with APOBEC3 as the primary and cytosolic sensing the second line of antiviral defense. Thus, the major function of in vivo cytosolic sensing appears to be the induction of IFN-stimulated anti-retroviral genes like Apobec3.

RESULTS

MLV reverse transcripts activate a rapid IFN response

MLV^gGag virions have relatively unstable capsids compared to MLV^WT and induce a large and rapid Trex1-dependent IFNβ response (Stavrou et al., 2013). In NR-9456 mouse macrophages, the IFNβ RNA response peaked at 2 hr post-infection and returned to basal levels by 6 hr post-infection (Fig. 1A); IFNβ protein levels peaked at 4 hr post-infection (Fig. S1A). Similarly, IRF3-responsive IFIT1 RNA levels rapidly increased in response to infection and expression of the IFN-inducible genes ISG15 and CXCL10, as well as IFIT1, increased at 8 hr post-infection (Fig. S1B), indicating that an early step in infection induced IFNβ transcription. MLV^WT also induced a response, albeit at lower levels than MLV^gGag (Fig. 1A). For most of the study we used MLV^gGag because of the higher level IFNβ induction. We also used Trex1 knockdown for the in vitro and ex vivo experiments to enhance the IFN response. Knockdown of Trex1 was verified by RT-qPCR (Fig. S1C).

Figure 1.

MLV infection induces a rapid IFNβ response. A) siRNA-treated NR-9456 cells were infected with MLV^WT or MLV^gGag. RNA isolated at the indicated times was analyzed by RT-qPCR; values are presented as the level of IFNβ relative to β-actin normalized to mock-infected cells. Open bars, MLV^WT and control siRNA; gray bars, MLV^gGag and control siRNA; diagonal bars, MLV^WT and siTrex1; horizontal bars, MLV^gGag and siTrex1. Average of 2 experiments (3 technical replicates/experiment); bars indicate standard error (S.E.). B) NR-9456 cells treated with Trex1 or control siRNA were infected with UV-inactivated MLV^gGag or in the presence of 100μg/ml AZT. Average of 2 experiments (3 technical replicates/experiment); bars indicate S.E. C) and D) NR-9456 cells infected with HIV pseudoviruses bearing amphotropic (C) or ecotropic- (D) MLV Envs. Average of 3 independent experiments; bars indicate standard deviation (S.D.). *, p ≤ 0.05; **, p ≤ 0.001; **, p ≤ .01 (two-tailed T test). See Fig. S1.

To confirm that the response was due to reverse-transcribed DNA, we infected NR-9456 cells with MLV^gGag in the presence of the reverse transcriptase inhibitor AZT or with UV-inactivated virions. Both AZT treatment and UV-inactivation dramatically reduced the Trex1-dependent increase in IFNβ RNA levels (Fig. 1B), confirming that MLV reverse-transcribed DNA induced the IFN response.

Interestingly, the Trex1-sensitive response to MLV was more rapid than that reported for HIV-1, which was detected at ~24 hr post-infection (Gao et al., 2013; Jakobsen et al., 2013; Yan et al., 2010). To determine if this was due to differences in cell lines or to viral uncoating or reverse transcription kinetics, we pseudotyped HIV cores with the ecotropic (Fig. 1C) or amphotropic (Fig. 1D) MLV envelope (Env) proteins and carried out a time course of IFNβ induction in NR-9456 cells. Interestingly, the time course of IFNβ induction was much slower with HIV than with MLV cores.

IFI203, DDX41, cGAS and STING are required for the IFN response

A number of molecules are involved in cytosolic sensing of nucleic acids, including IFI16 which has been implicated in the induction of type I IFN in response to infection by DNA viruses and HIV (Jakobsen et al., 2013; Monroe et al., 2014; Unterholzner et al., 2010). Additionally, MLV-induced IFNβ RNA induction was diminished in a mouse fibroblast cGAS knockout cell line, albeit at 20 hr post-infection (Gao et al., 2013). To determine whether ALRs were required for the rapid IFN response to MLV reverse transcripts, we first determined which genes were expressed in NR-9456 cells and found that with the exception of pyhinB, the ALRs and other molecules were expressed at the RNA level (Figs. S2A and S2B). We then carried out a targeted siRNA screen directed against the mouse ALRs as well as AIM2, DDX41, DAI, STING, and cGAS in conjunction with Trex1-knockdown to determine if any were involved in the response to MLV^gGag. Knockdown of IFI203, DDX41 or cGAS reduced the virus- and Trex1-dependent increase in IFNβ RNA (Fig. 2A) and protein (Fig. S1A), as well as IFIT1, ISG15 and CXCL10 RNAs (Fig. S1B). IFNβ RNA levels were also reduced in cells receiving IFI203, DDX41 and cGAS siRNAs without Trex1 knockdown (Fig. 2B). Knockdown specificity was verified at the RNA (Fig. S2B) and protein (Fig. S2C) levels and the effect on IFNβ levels was confirmed with additional independent siRNAs (Fig. S1C). All of the siRNAs used in these and subsequent analyses decreased expression of only the targeted gene and not the other cytosolic sensors (Fig. S2D).

Figure 2.

DDX41, IFI203, cGAS and STING are required for the IFNβ response. A) NR-9456 cells were treated with the indicated siRNAs for 24 hr and incubated with MLV^gGag for 2 hr. RT-qPCR was used to determine IFNβ RNA levels. The screen was performed in triplicate. B) DDX41, IFI203, cGAS and STING knockdown without Trex-1 knockdown. C) Knockdown was carried out as in A), except that RNA was isolated 24 hrs post-infection with amphotropic MLV Env pseudotyped HIV-1. Average of 3 experiments are presented in (B) and (C) (3 technical replicates/experiment); bars indicate S.D. * p≤ 0.05, ** p≤ 0.0001 (one-way ANOVA). See Figs. S1 and S2.

We also tested whether HIV reverse transcripts induced IFI203-, cGAS- or DDX41-dependent IFNβ induction in mouse cells using amphotropic MLV Env-pseudotyped HIV cores. Knockdown of IFI203, cGAS or DDX41 reduced the IFNβ response at 24 hr post-infection (Fig. 2C). Interestingly, while IFI203, cGAS or STING knockdown abrogated the response to HIV reverse transcripts, DDX41 knockdown had a more modest effect, in accord with previous work (Zhang et al., 2011). As was seen with MLV, knockdown of IFI204 had no effect on IFN induction.

DBA/2 mice have a genomic deletion of part of the ALR locus which encompasses the Mndal gene; Ifi203 is also expressed at very low levels in mesenteric lymph nodes (Zhang et al., 2009) or bone marrow-derived macrophages (BMMs) (Fig. S3A). To determine if the lack of IFI203 affected the IFNβ response, MLV^gGag was used to infect DBA/2 and BL/6 BMMs; BL/6 mice have an intact ALR locus and express high levels of IFI203 (Fig. S3A). The Trex1-dependent response to virus was reduced more than 5-fold in DBA/2 BMMs (Fig. 3A), although the response to lipopolysaccharide (LPS) was equivalent to BL/6 (Fig. S3B). Since Ifi204 is intact in DBA/2 mice and is expressed at high levels (Fig. S3A), these data support a role for IFI203 and not IFI204 in the sensing of MLV reverse transcripts.

Figure 3.

IFI203 mutant and STING^mut BMMs show little IFNβ response to MLV. A) BL/6 and DBA/2 BMMS were treated with control or Trex1 siRNAs, infected with MLV^gGag and 2 hr post-infection, RNA was analyzed by RT-PCR. Average of 3 independent experiments (3 technical replicates/experiment) with S.D. *, p ≤ 0.05; **, p ≤ 0.005 (two-tailed T-test). B) BL/6 and STING^mut BMMs treated with the indicated siRNAs were infected with MLV^gGag and 2 hr post-infection, analyzed for IFNβ RNA levels. Mock, uninfected cells. Average of 3 independent experiments (3 technical replicates/experiment) with S.D. *, p ≤ 0.001 (one-way ANOVA). See Fig. S3.

IFI16, DDX41 and cGAS sensing converges on STING. To determine if STING was required for the response to MLV, we first showed that STING knockdown diminished the response to MLV and HIV reverse transcripts in NR-9648 cells (Fig. 2). We next tested BMMs from STING^mut mice (Sauer et al., 2011) for their MLV response. The Trex1-dependent response to MLV^gGag in STING^mut BMMs was similar to that seen in BL/6 BMMs transfected with IFI203, DDX41 and STING siRNAs (Fig. 3B). Thus, the IFN response mediated by cGAS, DDX41- or IFI203-sensing of MLV reverse transcripts required STING pathway signaling.

DDX41 and IFI203 bind MLV reverse-transcripts and each other

These results suggested that IFI203 and DDX41 might be sensors of MLV reverse transcripts. To determine if IFI203 and DDX41 bound viral DNA, 293T cells were transfected with wild type IFI203 and DDX41 expression plasmids, as well as an IFI203 molecule with mutations in the Hinb DNA binding domain (Fig. 4A). At 24 hr post-infection, cell lysates were prepared and immunoprecipitated with anti-HA (IFI203) or –myc (DDX41) beads. The protein-bound beads were incubated with MLV DNA from Endogenous Reverse Transcriptase (EnRT) reactions prepared with purified MLV virions. The DNA was eluted and subjected to RT-qPCR using primers for MLV early reverse-transcribed (strong stop) ssDNA. Both wild type IFI203 and DDX41 bound ssDNA, whereas extracts prepared from control plasmid-transfected cells or those transfected with the IFI203 mutant did not (Fig. 4B). Similarly, when IFI203-, IFI203 mutant-, DDX41- and cGAS-transfected cells were infected with MLV^gGag, only wild type IFI203, DDX41 and cGAS immunoprecipitated MLV reverse transcripts (Fig. S4A).

Figure 4.

IFI203 and DDX41 bind MLV reverse transcripts and each other. A) Map of IFI203 and the mutations engineered into the HINb DNA binding domain. B) DNA pulldown assays with extracts from cells transfected with the indicated constructs. Average of 3-6 independent experiments with S.D. *, p ≤ 0.05; **, p ≤ 0.001; ***, p ≤ 0.003; ‡, p ≤ 0.07 (two-tailed T test). C) Extracts from 293T cells transiently transfected with the indicated expression vectors were immunoprecipitated with anti-IFI203 (HA), anti-DDX41 (myc) or anti-STING (FLAG) antibodies and analyzed on western blots. Shown are the results of a single experiment (representative of 6 independent experiments). See Fig. S4.

IFI16 and DDX41 interact with STING (Unterholzner et al., 2010; Zhang et al., 2011). To determine if IFI203 also bound STING or DDX41, we transfected cells with IFI203, DDX41 and STING expression vectors, alone and in combination. IFI203 and DDX41 co-immunoprecipitated when transfected together or in combination with STING and IFI203 and STING co-immunoprecipitated alone or in combination with DDX41 (Fig. 4C). DDX41 also co-immunoprecipitated with STING (Fig. 4B). As expected, since the proteins interacted in the absence of infection, mutation of the IFI 203 DNA binding site did not affect the interaction with DDX41 or STING (Fig. S4B). Neither IFI203 nor DDX41 co-immunoprecipitated with cGAS (Fig. S4C).

Sensing of MLV occurs in the cytoplasm

Trex1 is a cytosolic exonuclease and STING is associated with the endoplasmic reticulum; it moves to the perinuclear region upon stimulation. Similarly, cGAS and DDX41 sense DNA in the cytoplasm (Cai et al., 2014; Stein and Falck-Pedersen, 2012; Zhang et al., 2011). IFI203 is largely localized in the nucleus, although small amounts are found in the perinuclear cytoplasm when over-expressed with STING (Brunette et al., 2012; Zhang et al., 2008b). First, we examined the subcellular localization of IFI203, DDX41 and STING. We transiently transfected 293-MCAT cells expressing the MLV receptor with IFI203, DDX41 and STING expression vectors, alone and in combination, and then infected them with MLV^gGag. Whole cell, nuclear and cytoplasmic fractions were analyzed by western blots. DDX41 and STING were found both in the nuclear and the cytoplasmic fraction in all cases. IFI203 was found in the nuclear fraction in the presence or absence of DDX41 or STING (Fig. 5A). A small amount of IFI203 did appear in the cytoplasm of triply-transfected cells infected with virus. While western blots with antibodies to laminB (nucleus) and tubulin (cytoplasm) showed that the fractions were pure (Fig. 5A), we cannot rule out that STING and DDX41 were associated with membranes found in the perinuclear region or that this was an artifact of over-expression.

Figure 5.

MLV reverse transcribed DNA is sensed in the cytoplasm. A) Subcellular localization of IFI203, DDX41 and STING. 293-MCAT cells were transfected with IFI203, DDX41 and STING expression plasmids, alone and in combination and infected with MLV^gGag. Cells were fractionated into nuclear (N) and cytoplasmic (C) fractions and analyzed by western blots along with whole cell extracts (W). Anti-tubulin and anti-lamin antibodies were used to verify the purity of the C and N fractions, respectively. Representative of 3 independent experiments. B) NR-9456 cells were transfected with the indicated siRNAs, serum-starved for 48 hr. and infected with MLV^gGag. Average of 3 independent experiments (3 technical replicates/experiment) with S.D. NS, no statistical difference (two tailed T-test). See Fig. S5.

The MLV reverse transcription complex (RTC) depends on cell division for nuclear entry and inhibition of cell division by growth in low serum blocks infection (Harel et al., 1981; Roe et al., 1993). We tested whether sensing of viral DNA required nuclear entry by the MLV RTC. Trex1- or control-siRNA transfected NR-9456 cells were grown in media supplemented with 0.03% or 10% serum for 48 hr and then infected with MLV^gGag. While cell division decreased 4-fold in cells grown in low serum (Fig. S5), there was no difference in IFNβ RNA induction between cells grown in low or high serum (Fig. 5B). Taken together with the movement of IFI203 into the cytoplasm upon infection, this suggests that sensing of MLV reverse transcripts occurs in the cytoplasm.

APOBEC3 is the 1^st and DNA sensing the 2^nd line of defense against MLV in vivo

We showed previously that MLV^gGag was more susceptible to APOBEC3-mediated inhibition of reverse transcription than was MLV^WT and that reverse transcripts were more abundant in cells lacking APOBEC3 (Stavrou et al., 2013). This suggested that APOBEC3 might reduce the level of DNA ligands sensed by cytosolic sensors. To test this, virions were isolated from wild type and APOBEC3 knockout (A3 KO) splenocytes and used to infect Trex1 siRNA-treated primary BMMs from A3 KO mice. At 2 and 4 hrs post-infection, substantially less IFNβ RNA was induced with virions containing APOBEC3 than with those isolated from KO mice (Fig. 6A). Thus, when APOBEC3 failed to block the production of reverse transcripts after MLV infection, there were more viral DNA ligands available for sensing, leading to an increased IFNβ response.

Figure 6.

APOBEC3 and STING are important for virus control in vivo. A) Primary A3 KO BMMs transfected with either siControl or siTrex1, were infected with MLV^gGag isolated from A3 KO (-A3) or wild type (+A3) splenocytes. Average of 2 experiments, 3 technical replicates/experiment; bars indicate S.E. B) Newborn mice of the indicated genotypes were inoculated with MLV^WT or MLV^gGag. Each point represents the virus titer in individual mice. Bars indicate S.E. *, p≤.0001, ** p≤0.02 (non-parametric t-test). n, number of mice per group. C) MLV^gGag reverts in STING^mut mice. Mice of the indicated genotype were infected with MLV^gGag and at 6 week post-infection, virus was sequenced. WT M-MLV has a TAT(Y) codon. *Data taken from Stavrou et al., 2013.

To determine whether MLV reverse transcript-induced signaling through the STING axis was important for the control of infection in vivo, we inoculated newborn wild type, A3 KO, STING^mut and A3 KO/STING^mut double mutant mice with MLV^WT or MLV^gGag. At 18 days post-infection, virus titers in spleens were measured. As shown previously, MLV^gGag infection was greatly attenuated in wild type but not A3 KO mice (Fig. 6B) (Kolokithas et al., 2010; Stavrou et al., 2013). Moreover, MLV^WT levels were almost a log greater and MLV^gGag levels almost 2 logs greater in STING^mut than in wild type mice (Fig. 6B). Thus, loss of STING-mediated signaling increased the susceptibility of mice to MLV infection.

MLV^WT, which largely counteracts the antiviral affects of APOBEC3 because of its stable capsid, infected STING^mut mice at very high levels (Fig. 6B). Interestingly, whereas MLV^gGag was dramatically inhibited in wild type mice, infection by this virus was modestly inhibited in STING^mut mice, suggesting that the basal levels of APOBEC3 in these mice only partially controlled infection (Fig. 6B). Indeed, when MLV^gGag was used to infect A3 KO/STING^mut mice, infection was restored to the level of MLV^WT, demonstrating that both APOBEC3 and STING were important for virus control.

A single nucleotide change converts a tyrosine residue in MLV^WT to a stop codon in MLV^gGag (TAT to TAG). MLV^gGag reverts to MLV^WT in wild type but not A3 KO mice (Stavrou et al., 2013). Since MLV^gGag was also inhibited by STING, we tested whether it would be maintained in STING^mut mice. In contrast to what is seen in A3 KO mice, MLV^gGag reverted to MLV^WT in all of the STING^mut mice (Fig. 6C). Thus, while the looser capsid of MLV^gGag makes it more accessible to both APOBEC3 and DNA cytosolic sensors, only APOBEC3 imposes strong selection on the capsid.

Apobec3 is an IFN-stimulated gene. We speculated that in the absence of STING- or IFI203-mediated IFN induction in STING^mut or DBA/2 mice, the level of APOBEC3 would be low and thus unable to fully restrict virus. To test this, we subcutaneously injected wild type BL/6, DBA/2, A3 KO, STING^mut and A3 KO/STING^mut mice with MLV^gGag and 24 hrs post-infection, examined IFNβ, APOBEC3 and MLV RNA levels in the draining lymph nodes. A3 KO mice had the highest levels of IFNβ RNA, consistent with the presence of increased reverse transcript levels that activate the IFI203/DDX41/STING pathway (Fig. 7A). STING^mut mice, whether alone or in combination with the A3 KO allele, showed no IFNβ induction, while wild type BL/6 mice showed intermediate levels (Fig. 7A). DBA/2 mice also showed little IFNβ response to virus (Fig. S3C). STING^mut mice showed only basal levels of APOBEC3 RNA, equivalent to cells isolated from uninfected lymph nodes, while wild type mice had approximately 3-fold higher APOBEC3 RNA levels (Fig. 7A). This basal level of APOBEC3 controlled MLV infection in the wild type and STING^mut mice, while the lymph nodes from all mice lacking APOBEC3 were highly infected with MLV (Fig. 7A). DBA/2 mice exhibited about 2-fold higher infection levels than BL/6 mice (Fig. S3D).

Figure 7.

APOBEC3 limits cytosolic sensing but is induced by IFNβ. A) Four mice of each genotype received MLV^gGag subcutaneously. At 14 hr post-infection RNA from the draining lymph nodes was analyzed by RT-qPCR for IFNβ, APOBEC3 or MLV. Each diamond represents an individual mouse and the horizontal bar the averages. For IFNβ and APOBEC3, all values are normalized to uninfected BL/6. For MLV, a standard curve was generated. P values are indicated (two-tailed T test). B) MLV reverse transcripts induce expression of APOBEC3 and BST2/tetherin. NR-9456 cells were treated with control or Trex1 siRNAs and infected with MLV^gGag. RNA was subjected to RT-PCR with primers specific for APOBEC3 or BST2/tetherin (left axes) and IFNβ RNA (right axis). Average of 3 independent experiments. Area under the Curve analysis showed that the Trex1 siRNA treated macrophages had significantly higher BST2 and APOBEC3 (p≤0.05) levels compared to the control siRNA treated. At 8hpi the BST2 and APOBEC3 levels are significantly higher from both control- (p=0.0004 for BST2 and p=0.0013 for APOBEC3) and Trex1-siRNA treated cells (p≤.0001 for BST2 and p=0.0001 for APOBEC3) when compared to all other time points (one-way ANOVA with a Newman Keuls post-test). Error bars represent S.D.

To determine if APOBEC3 induction was the result of retroviral DNA sensing, we tested whether this required Trex1. We infected mouse macrophages with MLV^gGag in conjunction with Trex1 knockdown and examined the levels of APOBEC3 RNA, as well as another IFN-inducible retroviral restriction factor, BST2/tetherin. Both were induced and this induction lagged behind the IFNβ induction, suggesting that it was a secondary response (Fig. 7B). Taken together, these experiments demonstrate that APOBEC3 prevents the generation of reverse transcripts needed to efficiently activate the STING-mediated IFN response. However, when APOBEC3 fails to block reverse transcription, DNA sensing results in IFNβ production which in a feedback loop, up-regulates APOBEC3 and perhaps other intrinsic anti-viral factors like BST2/tetherin, resulting in the repression of virus spread in vivo.

DISCUSSION

Retroviruses undergo reverse transcription in the cytoplasm and thereby generate potential DNA ligands for cytosolic sensors. Indeed, Yan and colleagues showed several years ago that Trex1 prevented the induction of type I IFNs in response to HIV infection in cell culture and showed that DNA was the ligand (Yan et al., 2010). Cytosolic DNA is recognized by various sensors such as IFI16, DDX41 and cGAS, belonging to the ALR, DEAD-box helicase and nucleotidyltransferase families, respectively, which play roles in both the innate immune and inflammatory response to viruses in cell culture (Gao et al., 2013; Jakobsen et al., 2013; Lahaye et al., 2013; Monroe et al., 2014; Thompson et al., 2014b; Unterholzner et al., 2010; Zhang et al., 2011). The IFN response to pathogens may require multiple sensors; similar to our findings, the IFN response to DNA in THP-1 macrophages was impaired by knockdown of IFI16 or DDX41 (Jakobsen et al., 2013; Zhang et al., 2011) and the response to Listeria depends on at least two sensors, IFI16 and cGAS and perhaps DDX41 as well (Hansen et al., 2014; Parvatiyar et al., 2012). However, what role these sensors play in controlling retroviral infection in vivo has not been demonstrated. Here, we show that cytosolic sensing of reverse-transcribed retroviral DNA is likely important for the induction of type 1 IFNs, which in turn activates the expression of host restriction factors like APOBEC3 that are potent, critical inhibitors of infection.

Mouse APOBEC3 restricts infection by mouse retroviruses largely by blocking reverse transcription (MacMillan et al., 2013; Nair et al., 2014; Okeoma et al., 2009a; Stavrou et al., 2014; Stavrou et al., 2013). Here we show that by decreasing the DNA ligands of cytosolic sensors, APOBEC3-mediated inhibition of reverse transcription reduces IFNβ induction in response to MLV infection. Human APOBEC3 proteins also restrict HIV by blocking reverse transcription (Berger et al., 2011; Bishop et al., 2008; Gillick et al., 2012; Holmes et al., 2007) and heavy cytidine deamination has been reported to lead to degradation of reverse transcripts (Yang et al., 2007). Thus, it is likely that APOBEC3 plays a similar role in HIV infection of human cells.

MLV causes T cell leukemia. However, like HIV, it initially infects sentinel cells of the immune system in vivo, such as macrophages and dendritic cells, before transmission to lymphocytes (Low et al., 2009; Okimoto and Fan, 1999). Previously, Gao and colleagues showed that cGAS played a role in the response to infection by vesicular stomatitis virus G-pseudotyped MLV in Trex1-deficient murine fibroblasts (Gao et al., 2013). Here we show that in addition to cGAS, two sensors, IFI203 and DDX41 are critical to the IFN response to newly reverse-transcribed MLV DNA in macrophages. Importantly, IFI203 levels are highest in vivo in spleen and thymus, and in particular in T and B cells and BMMs, consistent with its role in restricting MLV and perhaps other murine retroviruses that target these cell types (Cridland et al., 2012). DDX41 and cGAS are also highly expressed in macrophages and dendritic cells (Wu et al., 2009).

A number of studies have implicated DNA sensors that signal through STING, including IFI16 (Bhat and Fitzgerald, 2014; Hansen et al., 2014; Thompson et al., 2014b; Zhang et al., 2011). We show that IFI203, which has only a single Hinb domain, is the murine sensor of reverse-transcribed DNA, since its knockdown abrogated the IFN response to both MLV and HIV. Moreover, DBA/2 mice which naturally express very low levels of IFI203 but normal levels of IFI204, showed a diminished IFNβ response to MLV. Interestingly, IFI204 was predicted to be the mouse homologue of IFI16 because of similarities in the structure (2 Hin domains) and sequence (Lee et al., 2013). In contrast, Ifi203 has a single Hinb domain (Li et al., 2013; Zhang et al., 2011). Although both the Hina and Hinb domains have been implicated in IFI16 binding to DNA, it may be that binding to reverse-transcribed DNA requires only its Hinb domain; IFI16 is the only human pyhin family member with a Hinb domain (Cridland et al., 2012). Given that IFI204 senses transfected and herpesvirus-1 (HSV-1) DNA, this suggests that it may be important for the response to intracellular pathogens other than retroviruses (Unterholzner et al., 2010). Moreover, there may be cell type specificity to sensing; it was recently suggested that IFI204 is a redundant DNA cytosolic sensor, because its knockdown did not abrogate the IFN response to DNA in BMMs and embryonic fibroblasts (Brunette et al., 2012). Finally, that IFI203 is required for IFN induction upon retrovirus infection and IFI204 upon HSV infection indicates that different ALRs act as sensors for different viruses and might explain why mice have 13 ALR genes, reflecting the diversity of pathogens to which this species was evolutionarily subjected.

Recent work has also suggested that IFI16 functions as a transcription factor in the nucleus, either negatively regulating HSV-1 transcription (Li et al., 2012; Orzalli et al., 2012) or positively regulating IFNβ expression in response to RNA-mediated stimulation of the RIG-I pathway (Thompson et al., 2014a). IFI203 is also predominantly found in the nuclear fraction and thus could function as a transcriptional regulator of gene expression. Because it is also found in the nucleus as well as the cytoplasm, it is formally possible that DDX41 functions as a transcription factor. This is clearly not the case with reverse transcribed MLV DNA, since quiescent cells, in which MLV DNA remains cytoplasmic, still showed Trex1-dependent stimulation of IFNβ (Fig. 5B). Although we showed that IFI203 and DDX41 bind to native reverse transcription products (Fig. 4B) and that a fraction of IFI203 relocates to the cytoplasm during retroviral infection (Fig. 5A), suggesting that both are directly interacting with viral DNA, we cannot exclude the possibility that IFI203 and DDX41 are activated by an as-of-yet unidentified pathway and act as transcription factors to induce IFNβ production upon virus infection. However, the lack of a requirement for nuclear localization of MLV DNA, combined with our finding that some IFI203 is found in the cytoplasm of MLV-infected cells when expressed with DDX41 and STING, suggest that IFI203 and DDX41 are part of the cytoplasmic DNA sensing machinery. Whether IFI203 and DDX41 function upstream, downstream or in parallel with cGAS, remains to be determined.

The rapid induction of IFN expression in both macrophage cell lines and primary macrophages coincided with the peak of MLV reverse transcription, in contrast to the cGAS-dependent induction reported for MLV-infected murine fibroblasts, which occurred 20 hr post-infection, long after the initial reverse transcription of incoming virus should be completed (Pfeiffer et al., 1999; Gao et al., 2013). The difference in the kinetics of the response could be due to cell type; macrophages may have different/additional cytosolic sensors than do fibroblasts which affect the kinetics. Alternatively, the use of VSV G MLV pseudotypes to infect cGAS-depleted fibroblasts may direct virus entry to a different cellular compartment that affects sensing; MLV enters cells at the plasma membrane or through a neutral compartment, whereas VSV G directs entry through a low pH compartment (Ross et al., 2002). In contrast to the rapid IFNβ response to MLV infection, the response to ecotropic MLV or amphotropic-MLV pseudotyped HIV cores was much slower than was seen with MLV cores. This may be due to slower uncoating or the generation of reverse transcripts that serve as ligands with HIV than with MLV when HIV infects mouse cells, which present multiple blocks to infection (Goff, 2004a; Lores et al., 1992; Rehwinkel, 2014; Zhang et al., 2008a). Whether this is also the case with bona fide HIV infection of human macrophages remains to be determined.

Because APOBEC3 and other inhibitors of retrovirus infection, such as BST2 and SAMHD1 as well as Trex1 and the ALRs are ISGs, DNA sensor-mediated IFN production increases expression of these direct inhibitors of infection. Previous studies with BST2/tetherin and SAMHD1 KO mice showed little or no differences in MLV infection compared to wild type mice (Behrendt et al., 2013; Liberatore and Bieniasz, 2011). This may be because inhibition of reverse transcription in these APOBEC3-competent mice limits the number of DNA ligands needed to activate the IFI203/DDX41/cGAS/STING pathway and thereby the IFN-mediated induction of BST2/tetherin and SAMHD1 expression; we found high level IFNβ induction occurred only in A3 KO cells and mice. The tighter capsid structure of MLV^WT used in the BST2 and SAMHD1 KO mice may also prevent significant reverse transcript sensing and hence the high level IFNβ production needed to induce BST2 or SAMDHD1 expression. Retroviral capsid stability is believed to play a role in protecting the reverse transcription complex from recognition by host anti-viral factors (Rasaiyaah et al., 2013) and increased IFN production has been seen with HIV-1 and HIV-2 bearing mutant capsids in dendritic cells (Lahaye et al., 2013).

Importantly, both APOBEC3 and signaling through the sensor-STING axis are critical to the control of retrovirus infection in vivo. APOBEC3 is likely the major anti-viral restriction factor in vivo. This is supported by the observation that while capsid stability affected both cytosolic sensing and APOBEC3-mediated restriction, glyco-Gag mutant viruses reverted in APOBEC3-containing but not STING^mut mice. Thus, selection for “tight” capsid structure was imposed only by APOBEC3. However, mice lacking STING were more highly infected by either MLV^WT or MLV^gGag and this was most likely due to abrogation of the IFNβ response needed to increase the expression of IFN-inducible restriction factors like APOBEC3. Since both human and mouse APOBEC3 restrict infection by blocking reverse transcription of HIV and MLV in addition to using their deaminase activity, fewer DNA ligands for sensors such as cGAS, DDX41 and IFI203, in the case of MLV or IFI16 in the case of HIV, will be generated in APOBEC3-expressing cells such as macrophages. Because APOBEC3 acts at an earlier step than does the STING-mediated pathway, this suggests that DNA sensing via the STING pathway is a “back-up” mechanism to catch reverse transcripts that elude APOBEC3-mediated restriction.

EXPERIMENTAL PROCEDURES

Mice

A3 KO mice (BL/6 background) were bred at the University of Pennsylvania as previously described (Okeoma et al., 2007). BL/6 and DBA/2 mice were purchased from the NCI and STING^mut (C57BL/6J-Tmem173gt/J) mice from the Jackson Laboratory. The STING mutation was crossed onto the A3 KO background to generate STING^mut/A3 KO mice. All mice were housed according to the policies of the Institutional Animal Care and Use Committee of the University of Pennsylvania. The experiments performed with mice in this study were approved by this committee (IACUC protocol #801594).

siRNA Knockdown and Expression Assays

NR-9456 cells (immortalized BL/6 macrophages (Hornung et al., 2008)) were transfected with siRNAs using Lipofectamine RNAiMAX reagent (Invitrogen). RNA was isolated using the RNeasy Mini Kit (Qiagen). cDNA was made using the SuperScript III First Strand Synthesis System for RT-PCR (Invitrogen). RT-PCR was performed using the Power SYBR Green PCR master mix kit (Applied Biosystems). All knockdowns were verified by RT-PCR. Primers for RT-PCR and siRNAs are described in the supplement.

Virus

MLV^WT and MLV^gGag were harvested from stably infected NIH3T3 fibroblasts as described (Fan et al., 1983). Viruses were titered on NIH3T3 cells using a focal immunofluorescence assay (units given as infectious centers (ICs)) (Low et al., 2009), analyzed by RT-qPCR for viral RNA levels and subjected to western blot analysis using anti-MLV antisera (Stavrou et al., 2013).

Infection of macrophages

Macrophages were siRNA-transfected and 48 hrs later infected with MLV^WT or MLV^gGag (MOI of 2 ICs) and harvested at the indicated times; for some experiments, UV-inactivated virus was used. For the AZT experiments, the cells were treated with 1μM AZT for 1 hr prior to infection with MLV^gGag. Primers used for detection of APOBEC3, actin and IFNβ RNA were previously described (Okeoma et al., 2009b; Stavrou et al., 2013).

HIV pseudoviruses

Retroviral vectors bearing the Moloney MLV Env and HIV (pNL4-3) cores were produced by transient transfection into 293T cells. Pseudoviruses were harvested at 48 hr post-infection and the pseudoviruses were treated with DNaseI (20u/ml for 45min at 37°C) (Roche) and ultracentrifugation at 180K × g for 1 hr. Pseudoviruses bearing the amphotropic MLV Env and the pNL4-3 core were provided by Zahra Parker.

BMM cultures

BMMs were isolated from hind limbs of 10- to 12-week-old mice (Caamano et al., 1999). Macrophages were cultured in DMEM supplemented with 10% FBS, 10ng/ml Macrophage Colony Stimulating Factor (Invitrogen), 1 mM sodium pyruvate, 100 U/ml penicillin, 100 μg/ml streptomycin. Cells were harvested 7 days after plating and seeded in 96-well plates for infection assays.

Plasmids

The IFI203-HA and STING-FLAG plasmids were provided by Dan Stetson. Murine DDX41 and cGAS plasmids were purchased from Origene and Open Biosystems, respectively. A myc/his tag was introduced at the DDX41 C’terminus by subcloning into pcDNA3.1myc/his (Invitrogen). A V5 tag was introduced at the cGAS C’terminus by PCR. The IFI203 DNA Hinb mutations (F217 and Y219; residues critical for DNA binding (Albrecht et al., 2005)) were introduced using the QuikChange II kit (Agilent).

In vivo infections

Two day old mice (BL/6, A3 KO, STING^mut and A3 KO/ STING^mut) were infected intraperitoneally with 2x10⁴ ICs and harvested at 18 days post-infection. Splenic virus titers were determined as previously described (Stavrou et al., 2013). For acute infections, 4 week old mice were inoculated with 10⁵ ICs in their footpad. At 14 hr post-inoculation, the draining lymph node was harvested and RNA isolated from lymphocytes. The contralateral non-draining lymph nodes were pooled from mice of each genotype and served as negative controls. Mice of each genotype were also injected subcutaneously with media; there was no response to media alone (data not shown).

Revertant Analysis

Two-day-old BL/6 and STING^mut mice were infected with MLV^gGag and sacrificed at 6 wk. DNA was isolated from spleens and the glyco-Gag region PCR-amplified and sequenced (Stavrou et al., 2013).

Co-immunoprecipitation and cell fractionation

MCAT-293T cells were transfected with the indicated plasmids and at 24 hrs post-infection, infected with MLV^gGag (2 IC/cell). Two hr post-infection, the cells were washed, lysed in 50 mM Tris pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% Triton-X-100, 1% deoxycholate, 0.1% SDS and protease inhibitor cocktail), sonicated and centrifuged. The supernatant was incubated with the indicated antibodies and protein A/G agarose and then subjected to SDS-PAGE and western blot analysis. The REAP method was used for the cell fractionation studies (Suzuki et al., 2010).

Western Blots

Protein extracts and immunoprecipitates were run on 10% SDS-polyacrylamide gels and transferred to PVDF Immobulon membranes (Thermo). The following antibodies were used: mouse anti-HA (Roche), mouse monoclonal anti-DDX41 and goat anti-LaminB (SantaCruz Biotechnology), rabbit anti-STING and anti-FLAG and HRP-conjugated anti-rabbit (Cell Signaling), mouse anti-V5 (Invitrogen), rabbit anti-tubulin (Sigma), HRP-conjugated anti-goat and -mouse antibody (Sigma). ECL kits (GE Healthcare Life Sciences) or Supersignal West Femto Chemiluminescent substrate (Thermo Scientific) were used for detection.

DNA Pull-downs

Reverse-transcribed MLV DNA was generated by EnRT (Stavrou et al., 2013), added to sonicated salmon sperm DNA and isolated using the Qiagen DNeasy Blood and Tissue Kit (Qiagen). 293T cells were transfected with the indicated plasmids and at 24 hrs post-transfection, lysates were prepared and incubated at 4°C overnight with anti-c-Myc- or anti-HA-agarose (Sigma). The IFI203-HA- or DDX41myc-bound beads were incubated with EnRT DNA. The bead-bound DNA was purified using the DNeasy Kit (Qiagen) and analyzed by RT-PCR using MLV strong stop primers (Stavrou et al., 2013).

Statistical Analysis

Statistical analysis was performed using the GraphPad/PRIZM software.

Highlights.

• IFI203, cGAS and DDX41 are sensors of MLV and HIV reverse transcripts

• APOBEC3 blocks the generation of the retroviral ligands of cytosolic sensors

• APOBEC3 and STING are both critical for in vivo control of MLV infection

• A major function of nucleic acid sensing is the induction of anti-retroviral genes