A cGAS-Independent STING/IRF7 Pathway Mediates the Immunogenicity of DNA Vaccines

Live attenuated vaccines are the most effective vaccines because they can elicit balanced T cell and Ab responses due to their in vivo replication potentials producing Ags inside the host. However, safety remains a concern for live attenuated vaccines (1). In contrast, many traditional inactivated or subunit vaccines, although safe when delivered as exogenous Ags, are not effective in eliciting T cell responses, and they also lead to less optimal Ab responses due to the lack of strong T cell help. The discovery of DNA vaccines in the early 1990s provided a unique vaccination technology that is effective in delivering endogenous Ags to elicit both MHC class I– and MHC class II–restricted T cell responses and balanced humoral and cellular immunities (2–6). More recent progress in human studies with candidate HIV-1 and pandemic influenza vaccines has further confirmed that DNA vaccines are able to induce high-quality and long-lasting Ab responses in both animals and humans, particularly when used as a priming vaccination, followed by boosting with inactivated or recombinant protein vaccines (7–21). Studies on the underlying immunological mechanisms of DNA vaccination indicate that DNA priming is especially effective in guiding germinal center B cell development, possibly through the function of T follicular helper cells (22).

At the same time, it has been long thought that DNA vaccines also induce innate immune signals, which may further contribute to the generation of acquired immunity. It has been known since the discovery of the DNA vaccine concept >20 y ago that the unmethylated CpG motifs encoded within the DNA vaccine plasmid could mediate immune responses via the endosomal DNA receptor, TLR9 (23). However, both TLR9- and MyD88-deficient mice mount immune responses comparable to wild-type mice, indicating that Ag-specific immune responses elicited by DNA vaccines are not TLR9-dependent (24–27).

More recent reports suggest that other innate signaling pathways are also involved in regulating the function of DNA vaccines (28, 29). We reported the role that Aim2 plays in the induction of influenza hemagglutinin (HA) Ag-specific Ab responses (30). Additionally, the immunostimulatory double-stranded nature of the DNA plasmid itself, as cytosolic DNA, is a potent inducer of type I IFN (IFN-αβ) via the stimulator of IFN genes (Sting) and the noncanonical IκB kinase, TANK binding kinase-1 (Tbk1) (24, 31). Sting/Tbk1 activation triggers translocation of the IFN regulatory factor (Irf) 3 transcription factor into the nucleus, driving IFN-αβ production through a positive feedback loop. Sting/Tbk1-mediated IFN-αβ production is required for DNA vaccine immunogenicity (24, 29, 32). However, the exact requirements of this pathway remain ambiguous, as it has been reported that Irf3 deletion diminishes T cell immunity but has little impact on B cell responses (33). Therefore, it was suggested that IFN-αβ does not play a significant role in generating high-level Ab responses following DNA vaccination (33), which is in stark contradiction to two other reports (29, 32). Additionally, the upstream DNA vaccine sensor has yet to be described, although multiple reports have identified the ubiquitously expressed cyclic GMP–AMP synthase (cGas) as a robust inducer of IFN-αβ capable of directly binding cytosolic dsDNA (27). As cGas is vital for immune responses to DNA viruses and bacterial infections (27, 34), we hypothesized that it would also be essential for DNA vaccine immunogenicity.

In the present study, we investigated the role of multiple components of the DNA sensing pathway in immune responses elicited by a model DNA vaccine expressing the influenza HA Ag. We determined the deletion effects of cGas, Sting, Irf3, and the closely related Irf7 on both the innate and adaptive immune response. Our data identified the key modulator in the generation of Ag-specific immune responses downstream of Sting. Such information will provide much needed insight into the design and development of safer and more effective vaccines.

C57BL/6 mice were obtained from Taconic Laboratories. Sting^−/− mice were a gift from G. Barber (University of Miami) and were fully backcrossed to C57BL/6. Irf3^−/− and Irf7^−/− mice were from T. Taniguchi (University of Tokyo, Tokyo, Japan). Irf3^−/−/Irf7^−/− double-knockout (DKO) mice and cGas^−/− mice were generated in-house by the K. Fitzgerald laboratory at the University of Massachusetts Medical School. All mice were maintained in the Department of Animal Medicine at University of Massachusetts Medical School according to Institutional Animal Care and Use Committee–approved protocols. The H1HA DNA vaccine used in this study encodes a codon-optimized gene of full-length HA protein from influenza A/Texas/04/09 as previously reported (30). Mice received 100 μg H1HA DNA vaccine by i.m. injection at weeks 0 and 2 to study HA-specific Ab responses. A third boosting vaccination was delivered 1 wk prior to sacrifice for the analysis of HA-specific T cell immunity. Serum was collected prior to vaccination and 6 h after vaccination for cytokine assays and at week 4 for Ab analysis. After collection, serum samples were stored at −80°C until the conclusion of the study. For in vivo immunogenicity studies involving Sting^−/−, Irf3^−/−, Irf7^−/−, and DKO mice, an n = 10 is reported. cGas^−/− mouse studies are reported as an n = 8. Immunogenicity data are representative of the average of two independently performed studies, where the N value equals the total number of mice per phenotype. Tissue-punch biopsies were harvested at 0, 6, and 12 h after vaccination for real-time PCR (RT-PCR). Splenocytes were harvested at time of sacrifice. RT-PCR studies used an n = 5.

Cells were treated under several conditions. LPS stimulation was performed at a concentration of 200 ng/ml. Poly(deoxyadenylic-deoxythymidylic) acid [poly(dA:dT)] (Sigma-Aldrich) DNA and HA-expressing DNA vaccine plasmid were added to the cells in the presence of Lipofectamine 2000 at a concentration of 1.5 μg/ml. Cultures were incubated 16–18 h at 37^oC and supernatants were harvested. Murine IFN-β sandwich ELISA was used as previously described (35). IFN-α and IFN-β RT-PCR was performed on RNA isolated from immortalized bone marrow–derived macrophages (BMDM). RT-PCR reactions were performed on a Bio-Rad CFX-96 cycler. Primer sequences are designed based on GenBank information.

Cytokines were quantified in sera collected from individual mice prior to vaccination at hour 0 and 6 h after the first DNA vaccination using a mouse Th1/Th2/Th17 cytometric bead array kit (BD Biosciences) according to the manufacturer’s instructions. The panel of cytokines included IFN-γ, IL-2, IL-4, IL-6, TNF, IL-10, and IL-17. All serum samples from each time point of interest were run in a single cytometric bead array experiment. Prior to assay, serum samples were diluted 1:2 in sample diluent. Samples were read on an LSR II flow cytometer and analyzed with FCAP Array software version 3.0 (BD Biosciences).

Microtiter plates were coated with standard H1HA Ag at 1.0 μg/ml in PBS for 1 h at room temperature and assayed as previously described (36). For sodium thiocyanate, displacement was performed at a serum dilution of 1:100. After washing of serum samples, sodium thiocyanate was added at various (0, 0.5, 1.0, 1.5, 2.0, 2.5, and 3.0 M) concentrations in PBS for 15 min followed by five washes in ELISA wash buffer. The assay was then completed as above.

Splenocyte T and B cell ELISPOT reagents were obtained from Mabtech. H1HA-specific T cells were quantified per the manufacturer’s instructions. Positive controls were stimulated with 20 ng/ml PMA (Sigma-Aldrich) and 500 ng/ml ionomycin (Sigma-Aldrich). The H1HA-relevant peptide used was a CD8⁺ T cell epitope (IYSTVASSL). H1HA-specific Ab-secreting cells were detected by coating of MAIPSWU (Millipore) plates with the standard H1HA Ag used for ELISA (1.0 μg/ml). Positive spots were visualized on a CTL imager and counting was performed with Immunospot software (Cellular Technology)

C57BL/6 and various knockout mice were shaved and vaccinated with 100 μg H1HA DNA vaccine plasmid i.m. in the hind quad muscle. Punch biopsies were performed from the injection site at 6 and 12 h after vaccination and snap frozen. RNA was isolated from tissue biopsies using TRIzol reagent (Life Technologies), and cDNA was generated using the Bio-Rad iScript cDNA synthesis kit. cDNA was then used for RT-PCR reactions on a Bio-Rad CFX-96 cycler. Primer sequences were designed based on GenBank HA sequences.

All data are presented as the mean of individual mice ± SEM. Statistical analysis was performed using a Student t test, a one-way ANOVA followed by a Tukey posttest, or a two-way ANOVA followed by a Bonferonni posttest.

It was recently reported, using a model OVA-encoding vaccine, that Sting-deficient mice exhibited severe defects in the adaptive immune responses generated by DNA vaccination (24). To confirm these results with a DNA vaccine expressing a viral immunogen, we vaccinated mice with DNA vaccine pH1HA expressing the HA Ag of the influenza A virus, which was responsible for the H1N1 pandemic in 2009. As HA is highly immunogenic and is the major protective Ag in clinically licensed inactivated and live-attenuated influenza vaccines, HA DNA vaccines provide a sensitive study model (15, 16, 18, 30, 36–38) to detect the impact of innate immunity on a clinically relevant immunogen.

To gain a more complete understanding of the requirement for Sting signaling in innate immune responses, wild-type (WT) C57BL/6 mice and Sting^−/− mice were vaccinated with the DNA vaccine pH1HA and serum cytokines were measured 6 h after vaccination. Quantification of cytokine levels revealed that Sting^−/− mice had a marked decrease in TNF-α and IL-6 production (Fig. 1A, 1B). Sting deletion also negatively impacted the generation of Ag-specific adaptive immunity. Whereas WT mice yielded high levels of HA-specific IFN-γ–secreting CD8⁺ T cells following stimulation with an MHC class I peptide encoded within the pH1HA vaccine, Sting^−/− splenocytes failed to respond to peptide stimulation (Fig. 1C). Furthermore, humoral responses were also impaired because vaccinated WT, but not Sting^−/−, mice elicited robust anti-HA IgG responses (Fig. 2A, 2B). Correspondingly, the HA-specific B cell population was decreased in both the circulating (spleen) and memory (bone marrow) compartments of Sting^−/− mice, as was the avidity of anti-HA–binding Abs (Fig. 2C–E).

Previous attempts to identify the upstream sensor for DNA vaccination have been unsuccessful (28, 29). It is widely thought that a cytosolic DNA sensor regulates IFN-αβ production in response to DNA vaccination. Studies during the last few years have identified cytosolic DNA-binding proteins important in the immune response to intracellular dsDNA. Among these, cGas is a powerful activator of the Sting/Tbk1 pathway, providing a strong candidate for the unknown DNA vaccine sensor. cGas is a nucleotidyl transferase enzyme that catalyzes the conversion of ATP and GTP into a novel cyclic dinucleotide, cGMP-AMP, which in turn binds and activates Sting. Hence, we evaluated the role of cGas in DNA vaccination. Contrary to expectations, no significant change in the magnitude of the innate or adaptive responses were seen in cGas^−/− mice, as they had comparable levels of serum cytokines, anti-HA Abs, and cytotoxic T cell responses to WT mice (Fig. 3, Supplemental Fig. 1). The ability to generate high-level adaptive immune responses following DNA vaccination suggested that IFN-αβ production was not dependent on cGas. This result was surprising to us, given the importance of cGas in sensing cytosolic dsDNA. We next quantified local IFN-αβ production at the site of injection by taking punch biopsies at 6 and 12 h after vaccination. This ensured that we only measured pH1HA-induced IFN-αβ locally. IFN-αβ production was attenuated in cGas^−/− mice at the initial 6 h time point. However, the levels of IFN-αβ in vaccinated cGas^−/− mice recovered by the 12 h time point, approaching those seen in WT mouse controls (Fig. 4). These findings indicate that although cGas contributes to the initial innate immune response upon detection of the DNA vaccine plasmid itself, this effect was transient and dispensable for sustained immunity. There may be an alternative DNA-sensing, IFN-αβ–inducing pathway in cGas^−/− mice for DNA vaccine–induced immune responses.

Although the effects of Sting deletion on DNA vaccine immunogenicity were previously reported, the mechanism by which Sting promotes IFN-αβ production is not. Therefore, we also dissected the downstream signaling molecules to elucidate the IFN-αβ pathway. Irf3 is a critical downstream mediator of Sting-driven type I IFN responses. As Irf3 is endogenously expressed at high levels and is required for initiating the IFN-αβ cascade, we analyzed its effects on DNA vaccine immunogenicity. In vitro stimulation of immortalized BMDM cultures showed a clear negative effect of Irf3 deletion on IFN-β production (Fig. 5). As expected, the synthetic B-form dsDNA poly(dA:dT) and pH1HA induced robust IFN-β levels in Irf^+/+ BMDM as measured by both RT-PCR and cell culture ELISA. Conversely, IFN-β production was limited in Irf3^−/− and, to a greater extent, BMDM lacking both Irf3 and Irf7 (DKO). Surprisingly, however, vaccination of Irf3^−/− mice did not result in impaired adaptive immunity. WT and Irf3^−/− mice had equivalent levels of IFN-γ–producing CD8⁺ T cells, in contrast to previously reported results (33) (Fig. 6A). Furthermore, Irf3^−/− mice exhibited high levels of anti-HA IgG titers, a result not seen in DKO mice (Fig. 6B–D). Overall, whereas IFN-αβ is required for DNA vaccine immunogenicity, Irf3 alone does not play a substantial role in impacting the acquired immune response.

We next investigated the role of Irf7 in the context of DNA vaccination. As shown in Fig. 7, both WT and Irf3^−/− mice yielded commensurate levels of serum cytokines 6 h after vaccination. However, Irf7^−/− and DKO mice showed clear defects in TNF-α and IL-6 production, at levels comparable to Sting^−/− mice. The similar cytokine profiles seen in Sting^−/− and Irf7^−/− mice indicated that both are required for innate signaling following DNA vaccination. Furthermore, both molecules have been identified as key regulators of IFN-αβ production following DNA virus infection (24, 39). As DNA vaccine immunogenicity is IFN-αβ–dependent (29), we performed a more detailed analysis on the effect of Sting and Irf7 deletion on IFN-αβ expression. To ensure we only measured pH1HA-induced IFN-αβ, we limited our measurements to the site of vaccination. WT, Sting^−/−, Irf7^−/−, and DKO mice were vaccinated with the pH1HA vaccine, and punch biopsies of the injection site were harvested 6 or 12 h later. RT-PCR analysis of mRNA illustrates that Sting^−/− and DKO mice lack significant IFN-α and IFN-β expression compared with WT controls (Fig. 8A, 8B). Interestingly, Irf7^−/− mice exhibited a similar decrease in IFN-αβ production, although to a lesser degree. Consistent with a decrease in IFN-αβ, both Sting^−/− and Irf7^−/− mice had a corresponding decrease in the IFN-stimulated gene IP10, illustrating the wide-ranging effect of Sting and Irf7 deletion on the innate immune response (Fig. 8C).

We next tested whether Irf7 deletion would inhibit development of Ag-specific immunity. pH1HA vaccination of WT mice elicited strong Ag-specific humoral responses to the encoded HA Ag, whereas Irf7^−/− mice failed to generate the same levels of anti-HA IgG after two vaccinations (Fig. 9A, 9B). Likewise, Ab-binding avidity was decreased ∼3-fold compared with WT controls (Fig. 9C), consistent with the underdevelopment of HA-specific B cells in both the spleen and bone marrow (Fig. 9D, 9E). Vaccination of Irf7^−/− mice also failed to induce significant numbers of IFN-γ⁺CD8⁺ splenocytes (Fig. 9F). Collectively, these results strongly support the view that Irf7 plays a broad role in the activation of both T and B cell subsets following DNA vaccination.

The complicated interplay between DNA vaccination and the innate immune system is just beginning to be elucidated, as the canonical TLR pathways have little influence on DNA vaccine immunogenicity. Instead, previous reports have shown that sensing of intracytoplasmic DNA plasmid governs DNA vaccine immunogenicity via the Sting/Tbk1/IFN-αβ pathway. However, the processes involved are still unclear. In particular, the requirement for IFN-αβ in generating high-level Ab responses has yielded contradictory results (29, 32, 33). Similarly, the necessary transcription factors downstream of Sting and Tbk1 remain uncertain. The present study investigates these factors and provides an updated understanding of the requirement for such factors in DNA vaccine immunogenicity.

The necessity for Irf3 in cellular-mediated immune responses was previously demonstrated (33). Supporting these findings, multiple reports have shown that Irf3-dependent IFN-β can augment the production of Th1 and Th2 cytokines both in vitro and in vivo (40, 41). However, our results show a more limited impact for Irf3, with reductions in IFN-γ⁺ T cell numbers seen only in Irf3^−/−/Irf7^−/− DKO mice. Still, in agreement with previous reports (33), we did not see a substantial effect on humoral immunity, confirming that Irf3 is not required for B cell activation. The discrepancy in CD8⁺ T cell responses may be attributed to the inclusion of more than one vaccination in our study. Another possible explanation may be our choice of a more clinically relevant Ag, which is highly immunogenic, providing a more sensitive system to detect the impact of innate immunity. In this study system, Irf3 alone plays a limited role in DNA vaccine–induced Ag-specific immune responses.

In contrast to Irf3, our results unexpectedly identified the important role played by Irf7 in DNA vaccine immunogenicity. Irf7^−/− mice exhibited a similar immune phenotype to Sting^−/− mice in that Irf7 deficiency resulted in significantly diminished T and B cell responses, indicating a broad contribution by Irf7 to the induction of adaptive immunity. Moreover, Irf7^−/− mice had impaired TNF-α and IL-6 production, characteristic of Sting-dependent signaling. The lack of IFN-αβ production in Irf7^−/− mice further suggests that the defects in DNA vaccination seen in Sting^−/− mice may be due to a failure to initiate the Irf7-dependent IFN-αβ feedback loop. This is in accordance with previous reports that Irf7 is the main regulator of immunity with regard to DNA virus infection (39, 42). Altogether, our results indicate that Irf7 is the key driving force behind sustained DNA vaccine–induced IFN-αβ production, implying that the temporary defect in immune priming provided by Irf3 deletion is overcome by the induction of Irf7, allowing for rescue of DNA vaccine immunogenicity.

Our attempt to identify the upstream DNA vaccine sensor yielded surprising results, as cGas deletion did not significantly limit DNA vaccine immunogenicity. In particular, the ability of cGas^−/− mice to generate WT IFN-αβ levels in vivo was somewhat unexpected. Our results suggest that although cGas is engaged at the site of vaccination and required for the early induction of IFN-αβ, it is not necessary for sustained IFN-αβ production in response to DNA vaccines. Previous studies have suggested that cGas is a nonredundant cytosolic DNA sensor (27, 43), but our data imply that at least one additional sensor functions in this context. Several potential candidates have been defined in the literature, including Ifi16 and Ddx41, among others. The engagement of one or more of these sensors and their signaling via Sting provides a possible explanation for the ability of cGas^−/− mice to generate adaptive immune responses following DNA vaccination.

In summary, our results further extend previous knowledge on the importance of the Sting pathway by adding new information both up- and downstream of the Sting molecule. Deletion of cGas did not dramatically impact the immune response, pointing to the need for searching additional upstream mechanisms or alternate cytosolic DNA receptors that control the Sting pathway. Realizing that Irf3 is not essential for DNA vaccine immunogenicity led to the finding that Irf7 is the key signaling molecule downstream of Sting in DNA vaccination. Of note, the necessity of both the Irf3 and Irf7 pathways appears to be independent of route of vaccination, as gene gun delivery did not affect data trends (results not shown).

In addition to obtaining a clearer understanding of the molecular mechanisms governing DNA vaccine immunogenicity, our findings may be relevant to clinical applications as well. Much attention has been focused recently on the inclusion of molecular adjuvants in DNA vaccine formulations (44, 45). The identification of the necessary signaling components for DNA vaccine immunogenicity may provide for additional adjuvant targets that can prime the immune system more efficiently. Overall, our results provide further insight on the cellular mechanisms through which DNA vaccines link both the innate and adaptive immune pathways to maximize protective, Ag-specific immune responses.

The authors have no financial conflicts of interest.