Rationale,progress and development of vaccines utilizing STING-activating cyclic dinucleotide adjuvants

Introduction

How the innate immune system is engaged by targeted ligands shapes the development of an adaptive response and is central to vaccine design [Coffman et al. 2010; Dubensky and Reed, 2010; Kastenmuller et al. 2011; Reed et al. 2009]. In particular, despite a broadly held belief that an immune correlate of effective preventive or therapeutic vaccination against intracellular pathogens and cancer is cellular immunity, there is a lack of available adjuvants and formulations that initiate antigen-specific effector and long-lived memory CD4 and CD8 T-cell responses. Potent adjuvants should in addition promote humoral immunity, as functional antibodies play an essential role either by directly neutralizing extracellular pathogens, or killing pathogen-infected or malignant cells through antibody-dependent cell-mediated cytotoxicity (ADCC) mechanisms. Intracellular pathogens and cancer continue to have a major impact on morbidity and mortality at a global scale due in part to the continuing need for adjuvants that promote effective cellular and humoral immunity.

Adjuvants are compounds that serve to enhance the magnitude, breadth, quality and longevity of specific immune responses to antigens, but have minimal toxic side effects. Optimal adjuvants can address the vaccinologist’s dilemma, which in general is defined by a spectrum of vaccines ranging from those based on live-attenuated pathogens that are immunogenic but with potential safety risks, to sub-unit vaccines that, while safe, have weak potency. A potent adjuvant is characterized by affording dosage sparing (number of vaccinations required to elicit an effective immune response), dose sparing (reduction in the amount of antigen required to achieve an effective immune response), and also immune response broadening [Reed et al. 2009]. There continues to be a strong need for molecularly defined adjuvants that can be coformulated with antigens derived from intracellular pathogens or malignancies to initiate and/or activate effective cellular and humoral immune responses that are correlated with protection against intracellular pathogens and reduction of tumor burden [Rappuoli et al. 2011].

Compartmentalization of pattern recognition receptors and adjuvant design

The design and development of modern adjuvants are guided by a fundamental understanding that conserved microbial structures known as pathogen-associated molecular patterns (PAMPs) are sensed by germ-line encoded host cell pattern recognition receptors (PRRs), triggering a downstream signaling cascade resulting in the induction of cytokines and chemokines, and initiation of a specific adaptive immune response [Iwasaki and Medzhitov, 2010]. How the innate immune system is engaged by PAMPs presented from an infectious agent shapes the development of an adaptive response appropriate to combat the invading pathogen from causing disease. One objective of adjuvant design is to select defined PAMPs or synthetic molecules which activate designated PRRs and initiate a desired response. Adjuvants such as monophosphoryl lipid A (MPL A) and CpG are microbial-derived PAMPs recognized by Toll-like receptors (TLRs), a class of PRRs that signal through MyD88 and Toll-interleukin-1-resistance (TIR) domain-containing adaptor-inducing interferon-β (TRIF) adaptor molecules and mediate induction of nuclear factor-κB (NF-κB) dependent proinflammatory cytokines [Kawai and Akira, 2010]. MPL A (TLR-4 agonist) and CpG (TLR-9 agonist) are the most clinically advanced adjuvants, and are components of vaccines that are approved or pending approval by the US Food and Drug Administration (FDA) [Ahmed et al. 2011; Einstein et al. 2009]. While TLRs present on the cell surface (e.g. TLR-4) and endosomes (e.g. TLR-9) sense extracellular and vacuolar pathogens, the productive growth cycle of multiple pathogens including viruses and intracellular bacteria occurs in the cytosol. The compartmentalization of extracellular, vacuolar and cytosolic PRRs has led to the hypothesis that the innate immune system can sense productively replicating pathogenic microbes by monitoring the cytosol [Vance et al. 2009]. This provides a rationale for the use of agonists that activate PRRs comprising the cytosolic surveillance pathway (CSP) and may be an effective strategy for the design of effective vaccines for eliciting cellular immunity, a presumptive immune correlate of protection against intracellular pathogens.

The CSP is integral to immunity against intracellular pathogens

Nucleic acids from bacterial, viral, protozoan and fungal pathogens are sensed by several distinct cytosolic signaling pathways. When activated, these individual pathways induce a characteristic cytokine profile, which in turn shapes the antigen-specific immune response. For example, the nucleotide binding oligomerization domain (NOD)-like receptor (NLR) family, such as ‘absent in melanoma 2′ (AIM2), senses cytosolic double-stranded (ds)DNA, triggering activation of the inflammasome and caspase-1 dependent production of interleukin-1β (IL-1β) [Strowig et al. 2012]. The signaling cascade resulting from activation of the inflammasome stimulates priming of Th17-biased CD4⁺ T-cell immunity, associated with protection against diverse pathogens, such as Streptococcus pneumoniae [Olliver et al. 2011].

Type I interferons (IFN-α, IFN-β) are the signature cytokines induced by two distinct TLR-independent cytosolic signaling pathways. In the first pathway, various forms of single-stranded (ss)RNA and dsRNA are sensed by RNA helicases, including retinoic acid-inducible gene I (RIG-I) and melanoma differentiation-associated gene 5 (MDA5), and through the IFN-β promoter stimulator 1 (IPS-1) adaptor protein mediate phosphorylation of the interferon regulatory factor 3 (IRF-3) transcription factor, leading to induction of IFN-β [Ireton and Gale, 2011]. IPS-1^-/- deficient mice have increased susceptibility to infection with RNA viruses. Sensors that signal through the IPS-1 pathway are directly targeted for inactivation by various viral proteins, demonstrating a requirement of this cytosolic host defense pathway to control productive virus infection. Synthetic dsRNA, such as polyinosinic:polycytidylic acid (poly(I:C)) and poly ICLC, an analog that is formulated with poly-l-lysine to resist RNase digestion, is an agonist for both TLR3 and MDA5 pathways, is a powerful inducer of IFN-β, and is currently being evaluated in several diverse clinical settings [Caskey et al. 2011].

STING (stimulator of interferon genes) is the central mediator for the second cytosolic pathway that triggers type 1 interferon, in response to sensing cytosolic dsDNA from infectious pathogens or aberrant host cells (danger associated molecular patterns, DAMPS) [Barber, 2011]. Several recent landmark discoveries provide the rationale for the development of vaccine adjuvants which activate the STING signaling pathway. Alternatively known as TMEM173, MITA, ERIS and MPYS, STING was discovered by Glen Barber and colleagues using cDNA expression cloning methods as a MyD88-independent host cell defense factor expressed in macrophages, dendritic cells (DCs) and fibroblasts was found to induce expression of IFN-β and NF-κB dependent pro-inflammatory cytokines in response to sensing cytoplasmic DNA (Figure 1) [Ishikawa and Barber, 2008].

OPEN IN VIEWER

Figure 1.

STING (stimulator of interferon genes) is a central sensor of cytosolic nucleic acids. STING is a MyD88-independent host cell defense factor that senses cytosolic nucleic acids, and in response activates and TBK-1/IRF-3 signaling cascades, inducing the expression of pro-inflammatory cytokines and IFN-β. The activating ligands for STING are CDNs, second messengers that are produced by bacteria or by cellular cGAS in response to binding cytosolic dsDNA. CDNs produced by bacteria and cGAS are structurally distinct. Less well studied, activation of STING by CDNs has also been shown to induce the expression of STAT6-dependent Th2 cytokines.

STING-activating second messengers

Cyclic dinucleotides (CDNs) have long been studied as ubiquitous small molecule second messengers synthesized by bacteria which regulate diverse processes including motility and formation of biofilms [Romling et al. 2013]. Recently, Vance and colleagues discovered that CDNs are the principal ligand for STING [Burdette and Vance, 2011; McWhirter et al. 2009]. In response to binding CDNs, STING activates signaling through the TANK-binding kinase 1 (TBK-1)/IRF-3 axis and induces the expression of IFN-β [Burdette and Vance, 2013]. Portnoy and colleagues showed that cyclic di-adenosine monophosphate (c-di-AMP) is the critical signaling molecule secreted by multidrug resistance efflux pumps from the intracellular bacterium Listeria monocytogenes (Lm) into the cytosol of infected host antigen presenting cells, and is correlated with CD4 and CD8 T-cell mediated protection in the mouse listeriosis model [Crimmins et al. 2008; Woodward et al. 2010]. When cells are infected with Lm mutants that are confined to the vacuole, IFN-β is not induced and protective immunity is not induced [Bahjat et al. 2006; Crimmins et al. 2008]. Induction of IFN-β in Lm-infected macrophages is dependent upon activation of the STING signaling pathway, and the levels of type I IFN induced by c-di-AMP in macrophages from MyD88^-/- Trif^-/- and C57BL/6 parental mice are indistinguishable [Leber et al. 2008; Witte et al. 2012]. In contrast, IFN-β is not induced by CDNs in macrophages derived from goldenticket (gt) mice encoding a nonfunctional mutant STING protein [Sauer et al. 2011]. Interestingly, the extracellular bacterium, Vibrio cholera, was shown by the John Mekalanos laboratory to produce a hybrid cyclic guanosine monophosphate (GMP-AMP (cGAMP) molecule, which also induces the STING pathway [Davies et al. 2012]. The activation of innate immunity with these ubiquitous second messengers suggests that sensing CDNs may be integral to host defense against bacterial infection.

Until recently, how STING sensed cytoplasmic DNA remained elusive. Unlike the AIM2 cytosolic sensor which directly binds dsDNA, STING lacks any obvious DNA-binding domains. The evidence that other candidate cytosolic DNA sensors such as DDX41, DNA-PK and DAI kinase were the primary mediators of dsDNA signaling through STING was inconclusive. This conundrum was solved with the discovery by Chen and colleagues of cyclic GMP-AMP synthase (cGAS), a host cell nucleotidyl transferase that directly binds dsDNA, and in response synthesizes a second messenger, cGAMP, which activates the STING pathway and induces IFN-β expression [Kranzusch et al. 2013; Sun et al. 2013; Wu et al. 2013]. Cells without a functional cGAS are unable to express IFN-β in response to stimulation with cytosolic DNA. Although the helicase DDX41was shown to sense both CDNs and cytosolic dsDNA, and activate IFN-β expression through the STING pathway [Parvatiyar et al. 2012], Chen and colleagues showed subsequently that induction of IFN-β in response to foreign DNA in cells expressing both cGAS and STING was not diminished by knockdown of helicase DDX41 expression [Sun et al. 2013]. Furthermore, the magnitude of induced IFN-β was orders of magnitude higher in cells transfected with cGAS as compared with DDX41, demonstrating that cGAS is the critical cytosolic DNA sensor.

However, there was an additional twist of cGAS activity that remained to be discovered. Cells expressing a particular STING allele were nonresponsive to stimulation by CDNs, but responsive to stimulation with dsDNA in a cGAS-dependent and TLR9 (MyD88)-independent manner. This observation was incompatible with a mechanism defined by cGAS synthesizing STING-activating CDN ligands in response to sensing cytosolic dsDNA. This apparent paradox was resolved very recently by several independent investigators, including the laboratories of Patel, Vance, Hornung, Hopfner and Chen, who showed that cGAS produces a noncanonical CDN that activates STING alleles that are nonresponsive to canonical CDNs [Ablasser et al. 2013; Civril et al. 2013; Diner et al. 2013; Gao et al. 2013; Kranzusch et al. 2013; Zhang et al. 2013]. Unlike the CDN second messengers produced by bacteria, in which the two purine nucleosides are joined by a phosphate bridge with bis-(3′,5′) linkages, the internucleotide phosphate bridge in the cGAMP synthesized by cGAS is joined by noncanonical 2′,5′ and 3′,5′ linkages, represented as c[G(2′,5′)pA(3′,5′)p] (Figure 2).

OPEN IN VIEWER

Figure 2.

STING activating ligands and CDN synthesis: (A) ‘canonical’ cyclic-di-GMP (cyclic[G(3′,5′)pG(3′,5′)p]) produced by bacteria; (B) ‘noncanonical’ cyclic-GMP-AMP cyclic[G(2′,5′)pA(3′,5′)p] produced by cellular cGAS; (C) DMXAA, a small molecule structurally unrelated to CDNs; and (D) schematic representation for synthesis of dioxo- and dithio-CDNs using phosphoramidite and H-phosphonate methods for the dimerization and cyclization steps, respectively. Full sequence requires 10 steps from commercially available starting material.

Human STING has known polymorphisms, including alleles encoding histidine at position 232, which are refractory to canonical CDNs, but not noncanonical CDNs [Diner et al. 2013; Jin et al. 2011]. Interestingly, 5,6-dimethylxanthenone-4-acetic acid (DMXAA) (Figure 2), a small molecule that is structurally unrelated to CDNs, stimulates STING signaling in mice, but binds without activating human STING [Conlon et al. 2013; Prantner et al. 2012; Roberts et al. 2007]. DMXAA has been shown to induce an innate immune-mediated antitumor response in mice [Fridlender et al. 2013], but ultimately failed when combined with chemotherapy in a large placebo-controlled phase III clinical efficacy trial in patients with nonsmall cell lung cancer [Lara et al. 2011]. It is tempting to speculate that the disappointing results of this clinical trial may have been due to the inability of DMXAA to activate the STING signaling cascade in humans, but how STING signaling participates in shaping the adaptive immune response in humans in response to infection with bacteria, which produce CDNs with canonical linkages, is a fundamental question that remains to be answered.

Immunologic potency of CDNs in mice

The immunostimulatory and vaccine adjuvant properties of CDNs were first published in 2007 [Karaolis et al. 2007a], long before STING was shown to be a direct sensor of c-di-GMP [Burdette et al. 2011]. The first report to demonstrate the presence of CDNs in eukaryotic cells showed that plants express a c-di-GMP dependent cellulose synthase similar to the bacterium Acetobacter xylinum [Amor et al. 1991; Ross et al. 1990]. Several years later, the observation that c-di-GMP increased CD4 T-cell receptor expression in Jurkat cells provided early insight that CDNs affect gene expression in immune cells [Steinberger et al. 1999].

CDN synthesis and design of compounds with increased potency

The availability of CDN-STING crystal structures along with recent results describing human STING allele-CDN dependent signaling relationships will facilitate structure-based studies to develop effective CDN compounds that activate innate immunity across genetically diverse human populations [Chin et al. 2013; Huang et al. 2012; Ouyang et al. 2012; Shang et al. 2012; Shu et al. 2012; Su et al. 2012]. Hydrogen bonding between CDN oxygen and phosphate atoms with amino acids at the interface of the STING homo-dimer binding cleft promotes its conformational transition and activate downstream signaling [Burdette and Vance, 2013]. The recent report from the Vance group showing that particular human STING alleles were refractory to CDNs with canonical 3′–5′ linkages suggests that noncanonical CDNs may be the preferred structure of compounds for advancement to human trials [Diner et al. 2013]. The design and screening of novel compounds will be facilitated by well-established processes for synthesizing CDNs. As shown in Figure 2, a wide diversity of ligands can bind to STING and activate signaling, including c-di-AMP, c-di-GMP, c-di-IMP and c-GAMP as well as DMXAA, a planar aromatic small molecule.

The synthesis of CDNs is achieved by the coupling of suitably protected and activated purine nucleosides to form a linear dimer followed by a ring closure reaction. A range of chemistries have been employed for the coupling and ring closure reactions including phosphate triester, H-phosphonate and phosphoramidite [Gaffney et al. 2010; Grajkowski et al. 2010; Hayakawa, 2003; Hyodo et al. 2006; Ross et al. 1990; Yan and Aguilar, 2007]. Figure 2 schematically depicts the conversion of a 3′-O-phosphoramidite (1) to linear dimer (3) followed by ring closure via the H-phosphonate to give cyclic dinucleotide (4). The phosphorous III intermediates generated at these two steps can either be oxidized or sulfurized (Gaffney et al. 2010). Both mono- and di-thio modified c-di-GMPs retain immunostimulatory capacity in vivo that is comparable with unmodified c-di-GMP [Yan et al. 2008]. After deprotection and purification, the product CDN is characterized by both one- and two-dimensional NMR spectroscopy in conjunction with liquid chromatography-mass spectrometry (LC/MS) and high resolution mass spectroscopy [Gao et al. 2013]. ¹H–³¹P heteronuclear multiple bond correlation (HMBC) experiments can be especially useful in establishing the regiochemistry of the phosphodiester bond in noncanonical CDNs [Diner et al. 2013].

Next steps: prospects for advancement of CDN adjuvants to evaluation in humans

The rapidly advancing mechanistic understanding of the central role STING plays in host defense against cytosolic pathogens will likely intensify efforts to develop novel vaccine adjuvants which target this signaling pathway. Several hurdles, including selection and formulation of a molecule that can signal in a diverse human population, remain to be overcome prior to evaluation of CDN-based adjuvants in humans. While noncanonical linkages enable cGAMP to mediate signaling through particular STING alleles that are nonresponsive to canonical cGAMP, whether other CDNs such as c-di-GMP having a comparatively higher STING binding avidity also require 2′,5′-3′,5′internucleotide phosphate bridges to activate nonresponsive human alleles remains to be determined. The combined results from published studies evaluating the relative potency of c-di-AMP and c-di-GMP adjuvanted vaccines have been in general inconclusive. With the exception of the H5N1 virosomes, previous work evaluating the impact of CDNs on vaccine immunogenicity in mice relies on admixing adjuvant and soluble protein antigen prior to injection. How defined adjuvants are formulated with antigen and administered affects the profile of the induced immune response and is central to vaccine design [Baldwin et al. 2012; Moon et al. 2011; Reed et al. 2009; Wille-Reece et al. 2005]. Codelivery of antigen and immunostimulatory factors to antigen-presenting cells is a necessary step to elicit functional antigen-specific Th1 CD4 and CD8 T cells and protective immunity.

Formulations that provide a sustained localized release of adjuvant and antigen sufficient to prime or boost effective immune responses without systemic exposure and induction of pro-inflammatory cytokines are an essential safety requirement for vaccine development. As the systemic cytokine profile was not evaluated in previous studies with CDN-based vaccines, the adjuvant dose-dependent safety profile determined in acceptable animal toxicology models remains a major unanswered question for this class of adjuvants that needs to be addressed prior to clinical evaluation. Since STING is a cytosolic PRR, it is likely that formulations which facilitate trafficking of vaccine compositions to this cellular compartment may increase the overall potency of CDN-based adjuvants.

With the urgent need for vaccines that elicit protection against intracellular pathogens, we have focused our efforts on developing CDN compounds and formulations that promote priming of antigen-specific cellular immunity. When formulated together with antigen using Addavax (InvivoGen, San Diego, CA) a 2% squalene-in-water emulsion similar to MF-59, a component of Fluad^© inactivated influenza virus vaccines marketed by Novartis [Seubert et al. 2011], we observed that c-di-GMP promoted priming of robust OVA-specific CD4⁺ and CD8⁺ T-cell immunity at levels that were substantially higher than those reported previously (Figure 3) [Ebensen et al. 2011]. We are exploring a variety of formulations and CDN compounds that have been modified to facilitate either coformulation/encapsulation efficiency with antigen or cellular uptake. Such an approach has been shown to increase the safety profile by limiting the systemic exposure of other adjuvants, including imidazoquinolines, a class of TLR7/8 agonists [Smirnov et al. 2011].

OPEN IN VIEWER

Figure 3.

CDNs promote priming of robust antigen-specific CD8⁺ T-cell immunity. C57BL/6 mice (n = 3) were vaccinated twice at a 3-week interval with 50 µg cyclic di-GMP coformulated with 10 µg OVA protein and 2% squalene-in water emulsion (Addavax). Mice were injected by subcutaneous (base of the tail) or intramuscular routes. OVA_257-264-specific CD8⁺ T-cell responses were measured in the spleen at the peak of the secondary response at 6 days post boost by intracellular cytokine staining for IFN-γ and TNF-α cytokines. Shown are representative FACS plots, each containing the mean and standard deviation of responses for the group of mice.

Conclusion

While the immunostimulatory properties of CDNs have been known since 2007, this class of adjuvants has not received extensive attention. However, with the multiple recent high visibility publications describing STING as a central player in a pathway to sense cytosolic nucleic acids, either indirectly from CDNs produced by cGAS in response to binding cytosolic DNA or directly from CDNs secreted by bacterial pathogens, this will more than likely change. Several steps in development will be required before the first vaccine candidates using CDN-based adjuvants are acceptable for evaluation in humans. These steps include selecting CDN molecules that can activate signaling across the diversity of human STING alleles, development of formulations that facilitate trafficking of vaccines to the cytosol, and conducting toxicology studies in animal models to define the CDN immunogenicity–toxicity therapeutic index to support eventual adjuvant dose levels for testing in humans. Since numerous intracellular pathogens still await the development of effective vaccines, there is both a strong scientific rationale and unmet medical need to develop CDN-based adjuvants for advancement to humans.