Transmission of innate immune signaling by packaging of cGAMP in viral particles

See allHide authors and affiliations

Science  11 Sep 2015:
Vol. 349, Issue 6253, pp. 1232-1236
DOI: 10.1126/science.aab3628

Viruses pack antiviral mediators

Viruses often hijack host proteins for their own use, turning host cells into virion-spewing machines. However, Bridgeman et al. and Gentili et al. now report a sneaky way that the host can fight back (see the Perspective by Schoggins). Host cells that expressed the enzyme cGAS, an innate immune receptor that senses cytoplasmic DNA, packaged the cGAS-generated second messenger cGAMP into virions. Virions could then transfer cGAMP to neighboring cells, triggering an antiviral gene program in these newly infected cells. Such transfer of an antiviral mediator may help to speed up the immune response to put the brakes on viral spread.

Science, this issue pp. 1228 and 1232; see also p. 1166


Infected cells detect viruses through a variety of receptors that initiate cell-intrinsic innate defense responses. Cyclic guanosine monophosphate (GMP)–adenosine monophosphate (AMP) synthase (cGAS) is a cytosolic sensor for many DNA viruses and HIV-1. In response to cytosolic viral DNA, cGAS synthesizes the second messenger 2′3′-cyclic GMP-AMP (cGAMP), which activates antiviral signaling pathways. We show that in cells producing virus, cGAS-synthesized cGAMP can be packaged in viral particles and extracellular vesicles. Viral particles efficiently delivered cGAMP to target cells. cGAMP transfer by viral particles to dendritic cells activated innate immunity and antiviral defenses. Finally, we show that cell-free murine cytomegalovirus and Modified Vaccinia Ankara virus contained cGAMP. Thus, transfer of cGAMP by viruses may represent a defense mechanism to propagate immune responses to uninfected target cells.

To protect multicellular organisms against viruses, it is vital that infected cells trigger antiviral defense responses that can be rapidly transmitted to noninfected cells. Cells are equipped with cytosolic viral sensors that recognize viral infection and induce innate immune responses (1). The resulting innate immune responses can restrain the pathogen, repair the host, and shape an adaptive immune response (2). Cyclic guanosine monophosphate (GMP)–adenosine monophosphate (AMP) synthase (cGAS) is a cytosolic DNA sensor that synthesizes 2′3′-cyclic GMP-AMP (cGAMP), a second messenger that binds to the STING protein and activates a type I interferon (IFN) response (35). cGAS is essential to induce an antiviral response against several DNA viruses, as well as HIV (68).

Dendritic cells (DCs) link innate sensing of pathogens to induction of adaptive immune responses. In DCs, the cGAS-cGAMP pathway also activates the expression of cytokines and costimulatory molecules for T cell activation, similar to adjuvants (7, 9). In human DCs, HIV-1 normally evades sensing by cGAS, suggesting a crucial role for this pathway in host–HIV-1 interactions (6, 10).

To study cGAS function, we sought to manipulate its expression in human monocyte–derived DCs. We generated a lentivector expressing cGAS and infected monocytes before differentiating them into DCs (see supplementary materials and methods). At day 4, DCs exposed to the cGAS virus expressed CD86 and were activated, despite low transduction efficiency (Fig. 1A). In contrast, infection with a control lentivector efficiently transduced DCs but did not change basal activation of the cells (Fig. 1A). This observation confirmed that lentivectors themselves do not activate DCs (10) and suggested that an activating innate immune signal was associated with infection with a cGAS-expressing lentivector.

Fig. 1 cGAS lentiviral vector activates monocyte-derived DCs.

(A) Reporter blue fluorescent protein (BFP) and CD86 expression in DCs after infection of monocytes with a lentivirus coding BFP-2A or BFP-2A–cGAS, in the presence or absence of Vpx. (B) CD86 expression as in (A), with titrated virus without Vpx and statistical analysis on the highest dose (n = 4 independent donors combined from two experiments, paired t test; ***P < 0.001). Bars indicate average. (C) CXCL10 production as in (A), with titrated virus without Vpx and statistical analysis on the highest dose (n = 4 independent donors combined from two experiments, paired t test on log-transformed data; **P < 0.01). (D) BFP and CD86 expression in DCs after infection of monocytes with a lentivirus coding BFP-2A or BFP-2A–cGAS or with VLPs produced in the presence of a nonlentiviral plasmid encoding for cGAS (PSTCD-cGAS). (E) CD86 expression and CXCL10 production as in (D) [n = 5 independent donors combined from three experiments, one-way analysis of variance (ANOVA) with post-hoc Tukey test for CD86 expression analysis, one-way ANOVA with post-hoc Tukey test on log-transformed data for CXCL10; ***P < 0.001, **P < 0.01, *P < 0.05, ns = not significant]. Symbols reflect individual donors.

Efficient lentivector transduction of DCs requires the protein Vpx that alleviates the restriction to HIV infection imposed by SAMHD1 (1113). Omitting Vpx prevented efficient transduction, but DC activation by the cGAS lentivirus was maintained (Fig. 1, A and B), suggesting that cGAS expression in the target cells was not required. Type I IFN–inducible cytokine CXCL10 was also produced by DCs (Fig. 1C). The resulting DCs were fully differentiated as positive for DC-SIGN and negative for CD14 (fig. S1A). The cGAS lentivector also activated DCs that were fully differentiated before infection (fig. S1, B and C). To exclude a low level of cGAS vector transduction, we produced HIV-1 virus-like particles (VLPs) lacking the lentiviral genome, and we coexpressed cGAS from a nonlentiviral plasmid (fig. S1D). The VLP-containing supernatant from cGAS-expressing cells activated DCs (Fig. 1, D and E), indicating transmission of an innate signal.

To determine the nature of this signal, we fractionated viral supernatants over a 10-kD filter. Activity was depleted from the filtrate and maintained in the retentate (fig. S2, A and B). We performed differential ultracentrifugation (14) to separate extracellular vesicles (EVs) released by cells (15) from soluble factors. Cell debris and large apoptotic bodies pellet first (2000 g), followed by medium-sized vesicles (10,000 g), and finally small EVs, including exosomes (100,000 g) (16, 17). The activity corresponded to the fraction that contained Gag, regardless of the presence of the exosome-associated proteins CD63, CD9, and CD81 and cytosolic syntenin-1 (Fig. 2A and fig. S2, C and D). This finding suggested that viral particles can transmit the innate signal.

Fig. 2 HIV-1 particles transfer an innate signal initiated by cGAS.

(A) CD86 expression and CXCL10 production in DCs after dose-response infections of monocytes with differentially fractionated supernatants containing VLPs produced from 293FT expressing WT cGAS or an inactive cGAS mutant lacking the DNA binding domain (ΔDBD). The volume of each fraction used for infection and the corresponding concentration factor compared to the initial supernatant are indicated (n = 3 independent donors combined from two experiments; mean and SEM plotted). (B) BFP and CD86 expression in DCs after exposure of monocytes to cell-free supernatants of cells transfected with combinations of plasmids expressing Gag, Pol, and VSV-G together with plasmids encoding cGAS, cGAS E225A and D227A, or control. (C) Analysis of CD86 expression and CXCL10 production as in (B) (n = 6 independent donors combined from three experiments; one-way ANOVA with post-hoc Tukey test for CD86 expression, one-way ANOVA with post-hoc Tukey test on log transformed data for CXCL10 production; ***P < 0.001, **P < 0.01, *P < 0.05, ns = not significant). (D) CD86 expression and CXCL10 production in DCs after infection of monocytes with CCR5-tropic HIV-1 viral particles produced in presence or absence of cGAS and HIV-1 entry inhibitors. Azidothymidine (AZT) was added to inhibit HIV-1 cDNA synthesis in the target cells (n = 6 independent donors combined from three experiments; one-way ANOVA with post-hoc Tukey test for CD86 expression, one-way ANOVA with post-hoc Tukey test on log transformed data for CXCL10 production; ***P < 0.001, **P < 0.01). Symbols reflect individual donors.

Next, we examined which components were required. Activity was abrogated when Glu225→Ala225 (E225A) and Asp227→Ala227 (D227A) mutations were introduced to the cGAS catalytic active site (18) (Fig. 2, B and C). VLPs are produced by expressing the viral proteins Gag and Pol and the fusogenic viral envelope protein VSV-G (vesicular stomatitis virus glycoprotein). Omitting expression of Gag and Pol, VSV-G, or all three proteins decreased induction of CD86 and CXCL10 in DCs (Fig. 2, B and C, and fig. S3A). The absence of VSV-G led to a strong decrease in DC activation (Fig. 2, B and C, and fig. S3A), indicating that fusogenic viral protein–containing extracellular material is the major DC-activating factor. We next examined the effect of tetherin expression, an inhibitor of viral particle release (19). Tetherin inhibited viral particle release and significantly decreased transfer of the innate signal (fig. S3, B to D). To rule out the possibility that transmission of the innate signal was unique to VSV-G, which can lead to production of tubulovesicular structures (20), we produced VLPs carrying the influenza virus envelope proteins H1N1 and H5N1 instead of VSV-G. When produced in the presence of cGAS, the VLPs activated DCs (fig. S4, A and E). VSV-G–pseudotyped gammaretroviral murine leukemia virus particles and HIV-1 particles expressing the wild-type (WT) CCR5-tropic envelope protein BaL also transmitted the innate signal (fig. S4, B to E). Incubation of target cells with HIV-1 entry inhibitors disrupted transfer of the innate signal without impairing the response to transfected cGAMP (Fig. 2D and fig. S4F). These data demonstrate that fusogenic viral particles transfer the innate signal from cGAS-expressing cells to DCs.

We could not detect cGAS protein in the pelleted supernatants (fig. S1D and fig. S2C). We hypothesized that cGAMP, a small molecule (mass = 675 daltons) produced in the cytosol, could be packaged in the viral particles and EVs because these structures contain cytosol from the producing cells (16).

If cGAMP was transferred by viral particles, the latter should activate a type I IFN response in a STING-dependent but cGAS-independent manner. We transfected an IFN reporter construct with or without a STING plasmid in 293FT cells that lack detectable cGAS expression (fig. S1D), and we validated the assay by transfected synthetic cGAMP or a cGAS expression plasmid (Fig. 3A). VLPs produced from cGAS-expressing cells activated the reporter only in the presence of STING (Fig. 3A and fig. S5A). Supernatants from cGAS-expressing cells that did not produce VLPs were much less effective (Fig. 3A and fig. S5A). To further demonstrate that cGAMP was present in the viral particles, we used a bioassay based on permeabilized THP-1 cells (a monocytic cell line) and an IFN reporter cell line (21, 22) (Fig. 3B). We extracted small molecules from virus-producing cells and pelleted VLPs. cGAMP activity was detected in cells transfected with cGAS but not in control cells. cGAMP activity was also detected in the pelleted VLPs and was lost when we introduced the cGAS E225A and D227A mutations (Fig. 3B). To confirm the presence of cGAMP in viral particles, we quantified the amount of cGAMP in the fractionated supernatants of cells expressing cGAS with or without viral particles. Expression of viral particles increased the quantity of pelleted cGAMP, whereas levels of the exosomal protein syntenin-1 were not affected (fig. S5, B to E). We also confirmed the presence of cGAMP by performing mass spectrometry analysis on the extract of pelleted viral particles, using synthetic cGAMP for reference (Fig. 3C).

Fig. 3 Viral particles package and transfer cGAMP.

(A) 293FT cells transfected with a luciferase reporter plasmid under control of the IFN-β promoter with or without a STING coding plasmid. The cells were stimulated with titrated amounts of supernatants from cells producing viral particles in the presence (cGAS viral particles) or absence (control viral particles) of murine cGAS and supernatants from cells expressing murine cGAS (cGAS no Gag/Pol no VSV-G), stimulated with synthetic cGAMP using lipofectamine, or transfected with a plasmid coding for cGAS (cGAS transfection) (n = 3 independent experiments; one-way ANOVA with post-hoc Tukey test on log-transformed data; ****P < 0.0001. Graph shows average ± SEM). The dilution factor of the wedges is threefold per bar. (B) cGAMP quantification in extracts coming from 293FT transfected cells and pelleted viral particles. 293FT cells were transfected with a lentiviral packaging plasmid in the presence of cGAS alone or catalytically inactive cGAS with E225A and D227A mutations. Type I IFN activity was measured after permeabilized PMA-differentiated THP-1 cells were exposed to synthetic 2′3′-cGAMP (left) or the benzonase-resistant extracts (right) (n = 3 independent experiments, except n = 2 for empty control. Graph shows average ± SEM). The dilution factor of the wedges is threefold per bar. (C) Liquid chromatography–tandem mass spectrometry analysis was used to identify cGAMP in extracts of pelleted cell-free EVs from cells producing VLPs in the presence of cGAS. (Top) Extracted ion chromatogram (XIC) of synthetic cGAMP [mass/charge ratio (m/z) = 675.05 to 675.15], full mass spectrum (MS1) at 12.6 min, and tandem mass spectrum (MS2) cumulated (12.6 to 12.9 min) after fragmentation of the precursor ion (m/z = 675.1) shown in the MS1 scan. (Bottom) XIC of cGAS viral particle extract (m/z = 675.05 to 675.15), full mass spectrum (MS1) at 12.6 min, and tandem mass spectrum (MS2) cumulated (12.6 to 12.8 min) after fragmentation of the precursor ion (m/z = 675.1) shown in the MS1 scan (with subtracted MS2 scan at 12.57 and 12.82 min). Arrows indicate the m/z values specific to cGAMP.

Next, we sought to determine whether transfer of cGAMP by viral particles would occur at a natural level of cGAS expression. HeLa cells express the cGAS protein and produced cGAMP after DNA stimulation in a cGAS-dependent manner (fig. S6, A and B). We pelleted the supernatants of unstimulated HeLa, DNA-stimulated HeLa, or HeLa transfected with VLPs coding plasmids that also provide a DNA stimulus. cGAMP was detected in the supernatants of all DNA-stimulated HeLa cells, consistent with its packaging in both EVs and viral particles (fig. S6C). However, although HeLa-derived VLPs induced CXCL10 production in phorbol 12-myristate 13-acetate (PMA)–treated THP-1 cells, EVs from control HeLa or DNA-stimulated HeLa cells did not (fig. S6D). To ascertain whether cGAMP was transferred, we tested the supernatants in the luciferase reporter assay. HeLa-derived VLPs, but not EVs from DNA-stimulated HeLa, activated the reporter in a STING-dependent manner (fig. S6, E and F). Overall, these data demonstrate that viral particles and EVs package cGAMP produced by endogenous cGAS, but only viral particles can efficiently transfer the second messenger cGAMP.

Transfected DNA provides a strong stimulation for cGAS. We sought to test whether HIV-1 infection itself, which generates cDNA after reverse transcription, would lead to sufficient cGAMP production for packaging in virus progeny. Upon infection of HeLa cells, cGAMP accumulation was observed in the cells and inhibited by reverse transcriptase inhibitors, indicating that cGAMP was produced in response to the infection (fig. S7, A and B). cGAMP was also detected in the supernatants of both transfected and infected cells (fig. S7, C to E). Thus, when HIV-1 infection activates cGAS, cGAMP is incorporated in the extracellular material containing the viral progeny. Additionally, we examined the presence of cGAMP in two DNA viruses: murine cytomegalovirus (mCMV), a member of the herpesviridae family of which herpes simplex virus (HSV) was shown to stimulate cGAS (4), and Modified Vaccinia Ankara virus (MVA), an attenuated poxvirus that requires cGAS for sensing by murine DCs (23). MVA and mCMV were produced by infection of cGAS-expressing cells (table S1). We found that cGAMP was present in the extracts of sucrose-purified stocks of both viruses (Fig. 4A and fig. S8, A and B).

Fig. 4 cGAMP is packaged in enveloped viruses produced by infection, and its transfer by viral particles induces an antiviral response in target cells.

(A) Detection of cGAMP in extracts of MVA–green fluorescent protein (GFP) [4 × 106 plaque-forming units (pfu)] and mCMV (2.7 × 104 pfu). XICs were measured by ultrahigh-performance liquid chromatography coupled with triple quadrupole mass spectrometry in multiple-reaction monitoring mode of the target ion with m/z = 476.1. Gray traces show the blank before the sample. Red dashed lines indicate the retention time of cGAMP. A representative analysis is shown, and additional replicates are shown in fig. S8. (B) Outline of the experiment in (C). (C) Monocytes were infected with the indicated supernatants or treated with type I IFN. The resulting DCs were infected with a dose-response of HSV-1 encoding GFP. GFP expression was analyzed after 24 hours (n = 4 independent donors combined from two experiments; one-way ANOVA with post-hoc Tukey test; ***P < 0.001, ns = not significant). Symbols reflect individual donors. MOI, multiplicity of infection.

Finally, we tested whether eliciting cGAMP transfer to DCs by HIV-1 particles could induce a functional antiviral response. We exposed monocytes to cGAMP-containing VLPs, control VLPs, or supernatants from cGAS-expressing cells and challenged them with a replication-competent HSV-1 (24) and a replication-defective vaccinia strain (Fig. 4B and fig. S8C). IFN treatment was used as a positive control (Fig. 4C and fig. S8D). cGAMP-containing VLPs, but not control VLPs or supernatants from cGAS-expressing cells, conferred protection against viral infections (Fig. 4C and fig. S8D). Collectively, our results provide evidence that cGAMP can be transferred between cells by virtue of packaging within viral particles, defining a mechanism of innate immune signal transmission (fig. S9).

The spread of innate responses is generally attributed to the production of cytokines, including IFNs. Effectors such as APOBEC3G can be packaged into viral particles and EVs (2527), but they do not signal in the target cells. cGAMP can diffuse between physically connected cells via gap junctions (28). Viral transfer of cGAMP does not require a direct contact between the cells, which may allow distant transmission of an innate signaling molecule within the organism or between hosts. This process could maximize the rapid induction of effector responses in target cells and bystander uninfected cells through local IFN production. Similarly, immunostimulatory cyclic dinucleotides naturally produced by bacteria can be delivered into the target cell (3).

Our results establish that HIV-1 may generate a cGAMP-containing progeny in cells that express cGAS. However, HIV-1 escapes cytosolic innate sensing by cGAS and thus is unlikely to contain cGAMP upon natural infection (6, 10, 29, 30). Eliciting cGAMP transfer by HIV-1 leads to activation of antiviral defenses. This parallels antiviral restriction factors that are inactive against the WT virus (31). In contrast, we find that purified stocks of mCMV and MVA contain cGAMP. Thus, cGAMP packaging is likely relevant to a large number of viruses and vaccines that naturally stimulate cGAS. The contribution of virally packaged cGAMP as compared to other innate signals will probably vary according to the functionality of cGAS in target cells.

Our results indicate that cell-derived EVs that include exosomes can package cGAMP as well. EVs also package and transmit cellular RNA between cells (32, 33). However, transmission of cGAMP by cellular EVs is not efficient. Viral envelope proteins increase transmission, indicating that fusion of EVs from noninfected cells with target cells is limiting. Nonetheless, low-efficiency transmission of cGAMP by host EVs may contribute to the tonic IFN response (8, 34).

Packaging of cGAMP within viral particles can be interpreted as an immune tagging process, allowing infected cells to further signify progeny viruses as nonself, or danger, to alert subsequent target cells. We speculate that other signaling molecules are also packaged and disseminated by viral particles. Subverting viral particles with cGAMP constitutes an attractive approach for therapeutics and vaccines.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S9

Table S1

References (3546)

References and Notes

  1. Acknowledgments: We declare no conflicts of interest. We thank M. Sitbon and J.-L. Battini for providing supernatant from the R187 hybridoma; A. García-Sastre for providing H1N1 and H5N1 envelope coding plasmids; O. Schwartz for providing the IFN-β–pGL3 plasmid and the Tetherin plasmid; P. Zimmermann for providing the anti–syntenin-1 antibody; P. Desai and L. Lorenzo for providing HSV-1 K26GFP; H. Raux and D. Blondel for help with vaccinia virus; P. Benaroch, S. Amigorena, and D. Rookhuizen for critical reading of the manuscript; and J. Rehwinkel and M. Benkirane for discussions. Some reagents were provided by the NIH AIDS Reagent Program. MVA-HIV and MVA-GFP are available from Agence Nationale de la Recherche sur le Sida (ANRS) and Transgene, respectively, under a material transfer agreement with ANRS. The data presented in this manuscript are tabulated in the main paper and in the supplementary materials. N.M. and Institut Curie have filed a provisional patent application (62051016) that relates to methods of preparation of viral particles with cyclic dinucleotides and their use for inducing immune responses. This work was supported by the ATIP-Avenir program, ANRS (France Recherche Nord and Sud SIDA-HIV Hépatites), the Ville de Paris Emergence program, European FP7 Marie Curie Actions, Labex VRI (ANR-10-LABX-77), Labex DCBIOL (ANR-10-IDEX-0001-02 PSL* and ANR-11-LABX-0043), ACTERIA Foundation, the Fondation Schlumberger pour l’Education et la Recherche European Research Council grant 309848 HIVINNATE (for N.M.), Labex DCBIOL and Association pour la Recherche sur le Cancer (ARC) SL220120605293 (for C.T.), and Région Ile-de-France and Institut Thématique Multi-Organisme Cancer 2014 for D.L. M.T. is a fellow from ARC.
View Abstract

Stay Connected to Science

Navigate This Article