DNA tumor virus oncogenes antagonize the cGAS-STING DNA-sensing pathway

See allHide authors and affiliations

Science  30 Oct 2015:
Vol. 350, Issue 6260, pp. 568-571
DOI: 10.1126/science.aab3291

Viral oncogenes remove the host's STING

Cancer-causing viruses, such as the human papilloma virus (HPV) that causes cervical cancer, account for 12% of human cancers. One way they can cause cancer is by targeting tumor suppressor proteins in the host. Now Lau et al. report that DNA tumor viruses can also thwart the host's immune system. Oncogenes from HPV and human adenovirus bound to the protein STING, a key component of the cGAS-STING pathway that senses and defends against intracellular DNA. In this way, the viruses subvert the host's antiviral immunity and set up shop, which, for an unlucky few, eventually causes cancer.

Science, this issue p. 568


Cyclic guanosine monophosphate–adenosine monophosphate synthase (cGAS) detects intracellular DNA and signals through the adapter protein STING to initiate the antiviral response to DNA viruses. Whether DNA viruses can prevent activation of the cGAS-STING pathway remains largely unknown. Here, we identify the oncogenes of the DNA tumor viruses, including E7 from human papillomavirus (HPV) and E1A from adenovirus, as potent and specific inhibitors of the cGAS-STING pathway. We show that the LXCXE motif of these oncoproteins, which is essential for blockade of the retinoblastoma tumor suppressor, is also important for antagonizing DNA sensing. E1A and E7 bind to STING, and silencing of these oncogenes in human tumor cells restores the cGAS-STING pathway. Our findings reveal a host-virus conflict that may have shaped the evolution of viral oncogenes.

Many viruses encode antagonists that enable evasion of innate immune detection of viral RNA and DNA (1). These antagonists can be broadly grouped into two classes. The first class consists of virus-encoded proteins that disrupt components of the antiviral interferon (IFN) response that are shared among RNA and DNA sensors. Examples of this class include several herpesvirus-encoded proteins that block IFN signaling (2), and the Vpu accessory factor of human immunodeficiency virus (HIV), which degrades the interferon regulatory factor 3 (IRF3) transcription factor (3). A second class of virus-encoded antagonists targets the most proximal components of RNA and DNA sensing and thus mediates specific inhibition of one pathway while leaving the other intact. For example, the nonstructural protein 3/4A (NS3/4A) protease of hepatitis C virus cleaves the mitochondrial antiviral-signaling (MAVS) adapter protein that is essential for RNA-activated innate immune signaling (4), and the NS1 protein of influenza A virus binds to and blocks activation of the retinoic acid–inducible gene I (RIG-I) RNA sensor (5).

We hypothesized that DNA viruses must encode dedicated antagonists of the proximal components of the DNA-sensing pathway that signal through cyclic guanosine monophosphate–adenosine monophosphate synthase (cGAS) (6) and the adapter protein STING/TMEM173 (7). Our exploration of cGAS-STING pathway antagonists began with our previous observation that most immortalized and tumor cell lines fail to respond to intracellular DNA, whereas primary cells mount a vigorous DNA-activated antiviral response (8). To explore this phenomenon in detail, we studied human embryonic kidney 293 (HEK 293) cells and HeLa cells, two of the most widely used human cell lines in biology. We transfected these cells with calf thymus (CT) DNA, a specific activator of the cGAS-STING pathway (9), or with a triphosphate RNA ligand that activates RIG-I (10). Whereas primary human fibroblasts mounted robust type I IFN responses to both DNA and RIG-I ligand (Fig. 1A), HEK 293 cells and HeLa cells responded only to RIG-I ligand and not to DNA (Fig. 1A). Similarly, primary mouse embryonic fibroblasts (MEFs) and bone marrow–derived macrophages responded to both DNA and RNA ligands, but immortalized mouse fibroblasts responded only to RNA (Fig. 1B). These data demonstrate a specific loss of the cGAS-STING pathway in multiple immortalized cell lines.

Fig. 1 Viral oncogenes antagonize the cGAS-STING pathway.

(A) Primary human foreskin fibroblasts (HFFs), HEK 293 cells, and HeLa cells were treated with the indicated ligands for 8 hours, followed by measurement of type I IFN activity in culture supernatants. (B) Primary murine embryonic fibroblasts (MEFs), bone marrow–derived macrophages (BMDM), or Jackson MEFs (Jax MEFs) were treated and measured as in (A). (C) Primary MEFs were transduced with retroviruses encoding the indicated oncogenes, and the expression of each oncoprotein was confirmed by Western blot of cell extracts. (D) Cells transduced with each oncogene were treated with DNA or RNA, and Ifnb mRNA responses were measured by quantitative reverse transcription–polymerase chain reaction (QRT-PCR) and compared to those of cells transduced with control retrovirus. Data are the combined measurements of three independent sets of transductions, with three independent measurements at consecutive cell passages per transduction. Means and standard deviations are shown, and statistical significance was determined by using unpaired t test with equal SD with two-tailed P-values (A and B) or two-way analysis of variance (D): **P < 0.01; ****P < 0.0001. ns, not significant.

We considered the manner in which these cells were immortalized. HEK 293 cells were transformed by introduction of human adenovirus 5 (hAd5) DNA into fetal kidney cells (11). HeLa cells derive from a cervical carcinoma caused by infection with human papillomavirus 18 (HPV18) (12, 13). Our immortalized mouse fibroblasts were generated by introduction of the large T antigen of simian virus 40 (SV40). hAd5, HPV18, and SV40 are known as “DNA tumor viruses” (14), owing to their ability to cause cancer in experimental animals and (in the case of HPV18) in humans (15). These viruses encode dominant oncogenes that bind to and inactivate the two principal tumor suppressors in cells: p53 and retinoblastoma (Rb) (14, 16). Additionally, these viral oncogenes interfere with IRF3- or IFN-dependent transcription that is shared by both the cGAS-STING and RIG-I–like receptor pathways, thus mediating generic inhibition of the antiviral response (1720). Our findings led us to investigate whether these oncogenes might also specifically disrupt the unique proximal signaling components of the cGAS-STING pathway.

We generated retroviral expression vectors containing the oncogenes derived from these viruses: HPV18 E6, HPV18 E7, hAd5 E1A, and SV40 large T antigen. As a control, we included a constitutively activated form of the cellular proto-oncogene H-Ras (G12V) (21). We transduced primary MEFs with each of these retroviruses (Fig. 1C) and monitored transcription of IFN-β in response to transfection with DNA or RIG-I ligand. We found that transduction of MEFs with H-Ras G12V mildly affected their response to both RNA and DNA ligands (Fig. 1D). Similarly, HPV18 E6 transduction resulted in a mild impairment of both pathways (Fig. 1D). Notably, transduction with HPV18 E7 or hAd5 E1A inhibited DNA-activated signaling, whereas the RIG-I pathway remained intact (Fig. 1D). Transduction with SV40 large T antigen led to a reduction in both cGAS-STING and RIG-I responses (Fig. 1D), suggesting that large T antigen may block a common component of the inducible antiviral response. These data reveal that E1A and E7 specifically block the cGAS-STING pathway when introduced into primary cells.

We next examined the mechanism by which E1A and E7 antagonize cGAS-STING pathway signaling. Both of these oncoproteins bind to and inhibit the function of the Rb tumor suppressor pathway, using a Leu-X-Cys-X-Glu (LXCXE) protein motif (22). To test whether the LXCXE motif was also important for cGAS-STING pathway inhibition, we introduced conservative mutations into this motif (VXSXD) that are known to disrupt Rb binding (Fig. 2A) (23). We then transduced primary MEFs with retroviruses encoding these mutants and measured the responses of the transduced cells to DNA and RNA ligands. Whereas the wild-type E1A and E7 potently blocked the cGAS-STING response, the VXSXD mutants of these oncoproteins did not (Fig. 2, B and C). Neither the wild-type nor the mutant oncoproteins had an inhibitory effect on RIG-I–mediated signaling (Fig. 2, B and C). These data demonstrate that the mechanism of cGAS-STING pathway antagonism by E1A and E7 is similar to the mechanism by which they disrupt the Rb pathway.

Fig. 2 The LXCXE motif of viral oncoproteins is important for cGAS-STING pathway blockade.

(A) Amino acid sequences of the regions of E1A and E7 containing the LXCXE motifs, with the locations of the conservative mutations introduced to disrupt the LXCXE motifs. Single-letter abbreviations for the amino acid residues are as follows: A, Ala; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; L, Leu; N, Asn; P, Pro; Q, Gln; S, Ser; T, Thr; and V, Val. (B and C) Primary MEFs were transduced with retroviruses encoding the wild-type or LXCXE-mutant oncogenes, and Ifnb mRNA induction in response to the indicated ligands was measured by QRT-PCR and normalized to Hprt mRNA expression within each sample. Results are representative of three independent transductions, with three independent measurements per transduction. Means and standard deviations are shown, and statistical significance was determined by using unpaired t test with equal SD with two-tailed P-values: *P < 0.02; **P < 0.01. No Tx denotes “No treatment.”

On the basis of these data, we explored the cGAS-STING pathway–specific target of viral oncogenes. We first tested whether the three members of the Rb gene family (Rb, p107, p130) (24) were important for cGAS-STING pathway activation. However, primary mouse macrophages with lentiviral short hairpin RNA–mediated depletion of each of these Rb family members responded normally to DNA transfection (fig. S1, A and B). Moreover, triple-knockout MEFs lacking Rb, p107, and p130 similarly mounted a potent IFN-β response to DNA, despite their documented growth abnormalities (fig. S1C) (25). This demonstrated that the host target of viral oncogenes in the cGAS-STING pathway is distinct from their targets in cell cycle regulation. We therefore tested whether E1A and E7 targeted a proximal signaling component of the cGAS-STING pathway. We hypothesized that, similar to the means by which E1A and E7 disrupt Rb, these oncoproteins would block the cGAS-STING pathway through direct binding to a cellular protein. We did not detect an interaction between viral oncoproteins and cGAS. Notably, when we immunoprecipitated epitope-tagged human STING from HEK 293 cells, we observed a robust interaction between STING and hAd5 E1A that is endogenously expressed in these cells (Fig. 3A). Moreover, we noticed that 13S E1A interacted more robustly with STING compared to 12S E1A, especially at lower concentrations of STING expression vector (Fig. 3A). Indeed, a direct comparison of STING binding to FLAG epitope–tagged 13S E1A or 12S E1A revealed a specific interaction with 13S E1A and a much weaker interaction with 12S E1A (Fig. 3B), demonstrating that the unique zinc finger motif present in 13S E1A is important for robust STING binding.

Fig. 3 Viral oncoproteins bind to STING using the LXCXE motif.

(A) HEK 293 cells were transfected with the indicated concentrations of human STING-HA expression vector, followed by immunoprecipitation of STING-HA and Western blotting for endogenous E1A protein or STING-HA. Inputs are shown in the left panels, and immunoprecipitations (IP) are in the right panels. (B) HEK 293 cells were transfected with 5 μg of human STING-HA expression vector, together with either 13S E1A-FLAG or 12S E1A-FLAG, followed by immunoprecipitation of STING-HA and Western blotting for E1A-FLAG or STING-HA. (C) HEK 293 cells were transfected with the indicated STING-HA expression vectors and processed for immunoprecipitations and Western blot, as in (A). (D) HeLa cells were transfected with full-length (FL) STING-HA, followed by immunoprecipitation of STING-HA and Western blotting for endogenous E7 protein or STING. The asterisk indicates an unknown E7 antibody-immunoreactive protein migrating at ~35 kD. HC, immunoglobulin heavy chain; LC, immunoglobulin light chain. (E) HEK 293 cells were transfected with plasmids encoding wild-type or VXSXD-mutant E1A-FLAG, together with STING-HA expression vector, followed by immunoprecipitation of STING-HA. All results are representative of two (B) or three (A, C, D, E) independent experiments.

We next tested whether specific domains of STING were important for interaction with E1A. The N-terminal 148 amino acids of STING constitute the transmembrane domains that anchor it to the endoplasmic reticulum (ER) membrane, whereas amino acids 149 to 379 encode a cytosolic signaling domain that mediates STING dimerization and binding to cyclic dinucleotides and IRF3 (26, 27). We were unable to express the N terminus of STING to appreciable levels in HEK 293 cells, suggesting instability of this domain in live cells. However, using epitope-tagged constructs encoding STING amino acids 149 to 379, or 149 to 379 anchored to the ER membrane by a heterologous transmembrane domain derived from cytochrome p450 (28), we found that the isolated C terminus of STING is insufficient to mediate interaction with E1A (Fig. 3C), revealing that additional elements of STING are required for robust interaction with E1A. We then overexpressed STING in HeLa cells and detected an interaction with the endogenous HPV18 E7 protein (Fig. 3D). We reproducibly detected a ~35-kD protein with the E7 antibody that also specifically interacted with STING (Fig. 3D, asterisk), the nature of which is unknown.

We examined whether the LXCXE motif in E1A that is essential for blockade of the cGAS-STING pathway (Fig. 2) is also important for E1A binding to STING. To do this, we generated FLAG epitope–tagged 13S E1A constructs containing either the native LXCXE motif or the mutant VXSXD motif. We found that the VXSXD mutant of E1A was severely compromised for interaction with STING–hemagglutinin A (HA) in HEK 293 cells (Fig. 3E). Together, these data identify STING as a binding partner of E1A and E7 and suggest a potential mechanism of cGAS-STING pathway blockade by viral oncogenes. The contribution of the LXCXE motif of E1A to STING binding precisely mirrors the role of this same motif in Rb binding, revealing a mechanistic parallel between antagonism of the DNA-activated antiviral response and inactivation of a tumor suppressor.

We returned to the transformed human cells that are specifically unresponsive to transfection with DNA. We hypothesized that the constitutive expression of E1A (HEK 293) or E7 (HeLa) renders these cells unable to activate the cGAS-STING pathway. To test this, we used a lenti-CRISPR (clustered regularly interspaced short palindromic repeat) approach (29) to disrupt these oncogenes. Notably, we found that the E1A-targeted HEK 293 cells were able to mount an IFN response to DNA transfection, but cells transduced with control guide RNAs remained as unresponsive as the parental cells (Fig. 4, A and B). We next used the same lenti-CRISPR approach to target the ~12 copies of E7 in HeLa cells (Fig. 4C), where we also observed robust recovery of the cGAS-STING pathway (Fig. 4D). Moreover, we found that HeLa cells transduced with a retrovirus encoding the bovine papillomavirus E2 protein, which silences the promoter encoding the bicistronic HPV18 E6/E7 mRNA (30), also recovered a potent IFN response to DNA transfection (fig. S2). Together, our data demonstrate that HEK 293 cells and HeLa cells both retain a functional cGAS-STING pathway, but this pathway is inhibited by the constitutive expression of the viral oncogenes that transformed these cells.

Fig. 4 Restoration of the cGAS-STING pathway in human tumor cells.

(A and B) HEK 293 cells were transduced with lenti-CRISPR constructs containing the indicated guide RNAs (gRNA), selected for 3 days in puromycin, and then harvested for evaluation of E1A protein expression by Western blot (A), or for plating prior to stimulation with the indicated ligands (B). Type I IFN activity was measured in culture supernatants 24 hours after stimulation. (C and D) HeLa cells were transduced with Lenti-CRISPR constructs containing the indicated gRNAs, selected for 3 days in puromycin, and then harvested for evaluation of E7 protein expression by Western blot (C), or for plating prior to stimulation with the indicated ligands (D). Type I IFN activity was measured in culture supernatants 24 hours after stimulation. Data are representative of three independent experiments. Means and standard deviations are shown, and statistical significance was determined by using unpaired t test with equal SD with two-tailed P-values: **P < 0.01.

Our data reveal that DNA tumor virus oncoproteins are potent and specific antagonists of the DNA-activated antiviral response. Recently, an unrelated herpesvirus-encoded protein was also found to bind to STING and block its activation (31), suggesting that targeting of STING may represent a general mechanism by which DNA viruses manipulate innate immune signaling. Notable advances over the last four decades have revealed how viral oncogenes cause cancer through inhibition of p53 and Rb (14, 16, 24). Our findings may shed new light on the long-standing question of why DNA tumor viruses evolved these oncogenes in the first place. There are two hypotheses that rationalize the origins of these oncogenes, neither of which is mutually exclusive. The “limited resources hypothesis” (16) postulates that oncogenes evolved to stimulate cell cycle progression in order to facilitate a cellular environment conducive to viral replication: abundant deoxynucleoside triphosphates for viral DNA synthesis, dissolution of the nuclear envelope to allow for viral egress, and new cellular niches to support transmission. In contrast, the “anti-antivirus hypothesis” (14) holds that the principal function of viral oncogenes is to antagonize innate immune signaling. Our data support the latter of these two hypotheses and reveal a host-virus conflict that may have shaped the origins of the DNA viruses that cause cancer in humans. Interestingly, most DNA viruses that infect humans encode proteins with LXCXE motifs (32), but only a minority of these viruses are known to cause cancer. We propose that study of these LXCXE-containing proteins will illuminate a broad class of STING-pathway antagonists.

Supplementary Materials

Materials and Methods

Figs. S1 and S2

References and Notes

  1. Acknowledgments: We are grateful to D. DiMaio for the BPV-E2 construct; to J. Kagan for the p450 ER-targeting motif; to J. Sage for providing the Rb/p107/p130 triple-knockout MEFs; and to members of the Stetson lab for discussions. The data presented in this paper are tabulated in the main paper and in the supplementary materials. D.B.S. is a scholar of the Rita Allen Foundation and a Burroughs Wellcome Fund Investigator in the Pathogenesis of Infectious Disease. E.E.G. is supported by a Cancer Research Institute Irvington postdoctoral fellowship.
View Abstract

Navigate This Article