Replacement of oxygen with sulfur on the furanose ring of cyclic dinucleotides enhances the immunostimulatory effect via STING activation

Abstract

Cyclic dinucleotides (CDNs) are secondary messengers composed of two purine nucleotides linked via two phosphodiester linkages: c-di-GMP, c-di-AMP, 3′,3′-cGAMP, and 2′,3′-cGAMP. CDNs activate the stimulator of interferon genes (STING) and trigger immune responses in mammalian species. CDNs are thus fascinating molecules as drug candidates, and chemically stable CDN analogues that act as STING agonists are highly desired at present. We herein report the practical synthesis of 4′-thiomodified c-di-AMP analogues, which have sulfur atoms at the 4′-position on the furanose ring instead of oxygen atoms, using simple phosphoramidite chemistry. The resulting 4′-thiomodified c-di-AMP analogues acted as potent STING agonists with long-term activity. Our results show that replacing O4′ on CDNs with sulfur can lead to enhanced immunostimulatory effects via STING activation.

Cyclic dinucleotide analogues that have sulfur atoms on the furanose rings act as potent and stable STING agonists.

Cyclic dinucleotides (CDNs) have been recently recognized secondary messengers found in a variety of organisms.¹ Bacteria produce molecules in which two purine nucleotides are linked via two 3′–5′ phosphodiester linkages to give symmetrical cyclophane structures: cyclic-di-GMP (c-di-GMP),² cyclic-di-AMP (c-di-AMP),³ and 3′,3′-cyclic-GMP-AMP (3′,3′-cGAMP)⁴ (Fig. 1). These bacterial CDNs play a critical role in some bacterial signaling networks involved in quorum sensing, control of biofilm formation, and cell wall homeostasis via binding to the related protein receptors and/or riboswitches.⁵

Fig. 1. Structures of naturally occurring and 4′-thiomodified CDNs.

Recently, the asymmetrical cyclic dipurine molecule: 2′,3′-cGAMP, in which guanosine and adenosine monophosphates are linked via one 3′–5′ linkage and another 2′–5′ linkage, was also discovered in mammalian cells.^6–8 2′,3′-cGAMP triggers immune responses in mammalian species. The cyclic GMP–AMP synthase (cGAS), which can identify the cytosolic DNA resulting from pathogen invasion, cyclizes ATP and GTP to produce 2′,3′-cGAMP.^9,10 The produced 2′,3′-cGAMP then acts as a secondary messenger that triggers the stimulator of interferon genes (STING) pathway,¹¹ resulting in the production of type I interferons (IFNs.)^12–14 Importantly, all known bacterial CDNs (c-di-GMP, c-di-AMP, and 3′,3′-cGAMP) also stimulate the STING.^15–19

Pharmacologic activation of STING signaling has shown promise in diverse clinically impactful applications, including broad-acting antiviral treatments, vaccine adjuvants, and immunogenic tumor clearance. Unfortunately, however, CDNs may be too chemically unstable for research and clinical use as 1) they are susceptible to phosphodiesterase (PDE)-mediated degradation, and 2) their hydrophilicity renders them impermeable to cell membranes. As these poor druglikenesses of CDNs, some non-nucleotide STING agonists have been developed.^20–22 However, synthetic CDN analogues with more favorable properties are urgently needed to elucidate the biological significance of CDNs.

To increase enzymatic stability, the nonbridging oxygen of phosphodiester linkages have been replaced with sulfur in order to create phosphorothioate linkages. The resulting compound 2′,3′-cGsAsMP is reported to be more resistant to PDE-mediated degradation than natural 2′,3′-cGAMP, prolonging its systemic half-life while maintaining a high affinity for STING.²³ Another phosphorothioate CDN analogue made of two AMP moieties cyclized via 2′–5′ and 3′–5′ phosphothioester bonds, known as ML-RR-S2-CDA, MIW815, or ADU-S100, shows better IFN-β responses than 2′,3′-cGAMP.²⁴ In addition, CDN analogues that lack negatively charged phosphodiester linkages have been developed to improve stability against PDE-mediated degradation.^25–29 However, they could not stimulate the STING and it was difficult to develop stable CDN analogues that have STING agonist activity.

We previously developed a series of 4′-thionucleic acids with sulfur atoms at the 4′-position on the furanose ring instead of oxygen atoms.^30,31 The 4′-thionucleic acids exhibited biological equivalency with natural DNA/RNA nucleotides.³² Capitalizing on these favorable properties, we reported the synthesis of c-di-4′-thioAMP (1) (Fig. 1).³³ The 4′-thiomodified analogue 1 exhibited enhanced resistance to PDE-mediated degradation compared with natural c-di-AMP. We also found that 1 acts as an artificial ligand for the bacterial c-di-AMP riboswitch, ydaO. However, how the replacement of oxygen with sulfur on the furanose ring of c-di-AMP affects the immunomodulating effect via binding to STING remains unclear, as our previous synthetic strategy for 1 had many drawbacks, including the need for repeated purification using high-performance liquid chromatography (HPLC).

We herein report the practical synthetic procedure of CDN analogues using phosphoramidite chemistry. Using this chemistry, we prepared 4′-thiomodified c-di-AMP analogues 1 and 2, and demonstrated that replacing oxygen with sulfur on the furanose ring enhances the immunostimulatory effect via STING activation.

Phosphoramidite chemistry has been widely utilized in the synthesis of naturally occurring and chemically modified oligonucleotides.³⁴ Most synthetic nucleosides, including 4′-thionucleosides, accept the reaction conditions of the phosphoramidite oligonucleotide synthesis.^30,31 Therefore, adopting a synthetic approach to generating CDN analogues based on phosphoramidite chemistry should expand the structural diversity of CDN analogues.

Thus far, a number of synthetic approaches for CDNs have been reported.³⁵ In most cases, the CDNs and their analogues are prepared by dimerizing nucleoside/nucleotide monomers and then subjecting the resulting dimer molecules to the cyclocondensation reaction. However, of the many synthetic reports of CDNs, there have been few using phosphoramidite chemistry in both dimerization and cyclocondensation reactions to construct the CDN skeleton.^36–38 In these reports, the major challenge has been the application of 3′,5′-hydroxyl-protecting groups that can chemoselectively be removed prior to performing the cyclocondensation of its synthetic precursor, and tedious procedures are required to prepare appropriately and specially protected nucleoside monomers. Therefore, we first screened conditions for preparing CDNs using phosphoramidite chemistry with an easily prepared 3′,5′-unprotected nucleoside monomer.

The coupling reaction between a commercially available N⁶-benzoyl (Bz)-adenosine phosphoramidite derivative 3 (ref. 39) and a cognate 3′,5′-unprotected nucleoside monomer 4 (ref. 40) with the aid of a generally used promoter for phosphoramidite coupling, 1H-tetrazole, in acetonitrile with 3 Å molecular sieves (MS) and subsequent oxidation of an internucleotide phosphite(iii) linkage with tert-butyl hydroperoxide (TBHP) afforded a complicated mixture including the desired linear dimer 5 (Table 1, entry 1 and Scheme 1). However, it was difficult to remove impurities even by performing the subsequent detritylation reaction to yield 6. We therefore sought a suitable promoter for efficient phosphoramidite coupling between 3 and 3′,5′-unprotected monomer 4 to give nucleotide dimer 5. The reaction using other tetrazole-type coupling reagents, such as a 5-(ethylthio)-1H-tetrazole⁴¹ and 5-(benzylthio)-1H-tetrazole,⁴² with subsequent oxidation afforded the pure nucleotide dimer 5 in moderate yields (2 steps) (entries 2 and 3), with the side reactions suppressed due to the enhanced reaction efficiencies compared with that using 1H-tetrazole. To further improve the reaction efficiency, we also explored the use of the acid/azole complex promoters reported by Hayakawa et al.⁴³ Whereas the use of imidazolium triflate decreased the reaction yield and elongated the coupling time (entry 4), a similar reaction using N-phenylimidazolium triflate (N-PhIMT) with subsequent oxidation showed excellent activity, and the desired nucleotide dimer 5 was isolated in 97% yield (2 steps) (entry 5).

Formation of nucleotide dimer 5.

Entry	Reagents	Coupling time (3 → 5)	Yieldsa
1	1H-Tetrazole	24 h	60%b
2	5-(Ethylthio)-1H-tetrazole	9 h	68%
3	5-(Benzylthio)-1H-tetrazole	7 h	62%
4	Imidazolium triflate	24 h	23%
5	N-Phenylimidazolium triflate (N-PhIMT)	1.5 h	97%

^a2 steps.

^bIncluding inseparable impurities. A^Bz = N⁶-benzoyladenine.

Scheme 1. Synthesis of c-di-AMP, c-di-4′-thioAMP (1) and 2 using phosphoramidite chemistry. A^Bz = N⁶-benzoyladenine.

After detritylation of 5, the resulting 3′,5′-unprotected dimer 6 was treated with strictly 1.1 mol equivalent of 2-cyanoethyl tetraisopropylphosphoramidite and N-PhIMT in acetonitrile with 3 Å MS (Scheme 1). However, no obvious reaction progress was observed on thin-layer chromatography (TLC). When the solvent system was changed to the combination of an acidic solvent CH₂Cl₂ with 4 Å MS (a suitable drying agent for CH₂Cl₂), the phosphitylation of 6 smoothly progressed to form 7. The subsequent intramolecular cyclocondensation of 7 was then performed by adding two molar equivalents of N-PhIMT to the initial phosphitylation reaction mixture. The cyclocondensation reaction of 7 was allowed to proceed within a few minutes by stepwise phosphoramidite coupling without undesired polymerization. Subsequent oxidation with TBHP gave a fully protected c-di-AMP derivative 8 (ref. 44) in 3 steps 63% yield based on the linear dimer 6. Finally, full-deprotection of 8 was performed in a manner similar to that reported by Gaffney et al.⁴⁵ The Bz and cyanoethyl (CE) protecting groups on 8 were thus removed by treatment with saturated ammonia in methanol at room temperature for 24 h. After concentration to dryness, the residue was dissolved in minimal volume of pyridine and treated with Et₃N·3HF at 65 °C for 2.5 h. The final product, c-di-AMP, was then obtained as a triethylammonium salt.

Since simple phosphoramidite chemistry for CDN synthesis had now been optimized, we next prepared 4′-thiomodified CDN analogues. To synthesize c-di-4′-thioAMP (1), the 4′-thioadenosine phosphoramidite derivative 9 (ref. 31) and the corresponding 3′,5′-unprotected monomer 10 (ref. 32) were coupled in the presence of N-PhIMT, and subsequent oxidation gave c-di-4′-thioadenosine dimer 11 in 93% yield (2 steps), much the same yield as in the corresponding reaction to give 5. After detritylation of 11, the resulting dimer 12 was subjected to the cyclocondensation reaction using a diamidite reagent to give a fully protected c-di-4′-thioAMP 14 in 66% yield (3 steps). These results indicated that our phosphoramidite strategy for constructing the CDN skeleton was useful, even for preparing chemically modified CDN analogues. The Bz and CE group of 14 were then deprotected with saturated ammonia in methanol at room temperature. After concentration to dryness, the residue was suspended in H₂O, and the resulting white solid was treated with Et₃N·3HF at 65 °C in methanol. Finally, purification of the reaction mixture using solid phase extraction cartridges afforded the pure 1.

To evaluate how the replacement of oxygen with sulfur on the furanose ring affected the activation of STING, we also prepared asymmetrical analogue 2, in which an oxygen atom on one of two furanose rings was substituted with a sulfur, in a similar manner to that described for 1 while using 3 and 10 as starting materials. All CDN analogues were able to be prepared without HPLC purification in this study.

Next, we evaluated the immunostimulatory effect of 4′-thiomodified CDN analogues 1 and 2 using HEK-293T-derived interferon regulatory factor (IRF) and IFN-β reporter cells (293T-Dual™ Null cells). Since HEK-293T cells are known to have a non-functional STING pathway, we stably transfected R232 hSTING, which is the most prevalent isoform (occurrence of approximately 60% in the human population),⁴⁶ into 293T-Dual™ Null cells. As shown in Fig. 2a, significant IFN-β induction was observed when compound 1, which has two sulfur atoms on each furanose ring, was simply added to the cell culture medium of 293 T-Dual R232 hSTING cells, and its induction degree was much more potent than that of the parent compound c-di-AMP. The asymmetrical 4′-thiomodified CDN analogue 2 also showed a stronger immunostimulatory effect than c-di-AMP. In addition, both the 4′-thiomodified CDN analogues 1 and 2 exhibited much more potent IFN-β inductions than c-(RpRp)-di-Aps, which has two sulfur atoms instead of the two nonbridging oxygens of the phosphodiester linkages. Compounds 1 and 2 did not respond to the HEK-293T cells lacking a functional STING pathway (293T-Dual™ Null cells), indicating that they induced IFN-β responses as STING agonists.

Fig. 2. The evaluation of the INF-β induction by STING activation. The degree of INF-β induction by CDNs was measured in a) 293T-Dual™ Null and 293T-Dual R232 hSTING cells and b) 293T-Dual A230 hSTING cells. CDNs (20 nmol mL⁻¹) were added to the cells. INF-β induction was measured by a luciferase reporter assay. Luminescence data are the mean relative light units ± standard error of the mean (SEM) of at least 3 experiments (n = 3). *P < 0.05 and ****P < 0.0001 by Student's t-test.

To further compare the immunostimulatory effects, we also evaluated the agonist activities of 1 and 2 to A230 hSTING, a non-synonymous variant of STING (Fig. 2b). A230 hSTING contains a single amino acid substitution G230A. G230 is located in the flexible loop that forms a lid above the CDN binding pocket, and the G230A variant, A230 hSTING, is able to respond to lower concentrations of CDNs.⁴⁶ As a result of simple addition to the cell culture medium of 293T-Dual A230 hSTING, 4′-thiomodified CDN analogues 1 and 2 showed a superior induction of IFN-β to c-di-AMP and c-(RpRp)-di-Aps. In addition, 1 and 2 both showed potent IRF3 induction as a result of STING activation (Fig. S1 in ESI†).

There are three possible factors responsible for the improvement in the immunostimulatory effects of compounds 1 and 2: 1) enhancement of the stability against PDE-mediated degradation, 2) improvement in the membrane permeability, and 3) changes in the binding mode to the STING arising from the replacement of oxygen with sulfur on the furanose ring. Regarding the potential involvement of factor 1), we previously reported that compound 1 displayed resistance against PDE specific for CDNs, YybT.³³ Since the immunostimulatory effects of compounds 1 and 2 actually increased over time from 24 to 72 h post-compound addition (Fig. 2 and S1†), the enhanced stability of 1 and 2 arising from the replacement of O4′ with sulfur might have led to an improvement in STING activation. To determine the potential involvement of the other two factors, evaluations of other 4′-thiomodified CDNs, such as c-di-4′-thioGMP, are currently underway.

In conclusion, we developed practical synthetic procedures for CDN analogues using phosphoramidite chemistry. Using this procedure, 4′-thiomodified c-di-AMP analogues (1 and 2) were successfully prepared without the need to perform tedious purification using HPLC. The resulting 4′-thiomodified CDN analogues 1 and 2 showed potent and long-term immunostimulatory effects by activating STING. Our results indicate that the replacement of oxygen with sulfur on the furanose ring is a useful strategy for designing bioisosteres of c-di-AMP.

Methods

Synthesis of CDNs

The detailed synthetic procedure and full characterization data are presented in the ESI.†

Immunostimulatory effects: cell culture and transfections

293T-Dual™ Null cells (Invivogen, San Diego, CA, USA), derived from the HEK293T cell line by the stable integration of a SEAP reporter construct (to monitor the activation of the transcription factor IFN-stimulated response elements) and a luciferase reporter construct (to monitor the expression of IFN-β), were propagated in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 100 μg mL⁻¹ streptomycin, 100 units per mL penicillin, normocin (100 μg mL⁻¹), hygromycin (100 μg mL⁻¹), and zeocin (100 μg mL⁻¹) in a humidified atmosphere with 5% CO₂. Stable R232 hSTING- or A 230 hSTING-expressing cell lines (293T-Dual R232 hSTING or 293T-Dual A230 hSTING) were obtained by transfection of the expression vector pUNO1-hSTING-WT or pUNO1-hSTING-A230 (Invivogen) into 293T-Dual™ Null cells using Lipofectamine 2000 (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer's instructions. Stable clones were selected for resistance to blasticidin (10 mg mL⁻¹) and then further screened using Western blotting.

Cell stimulation and the SEAP/luciferase assay

293T-Dual R232 hSTING or 293T-Dual A230 hSTING cells were seeded in a 96-well plate (3 × 10⁴ cells/100 μL per well) in DMEM with 10% FBS, 100 μg mL⁻¹ streptomycin, 100 units per mL penicillin, normocin (100 μg mL⁻¹), hygromycin (100 μg mL⁻¹), zeocin (100 μg mL⁻¹), and blasticidin (10 μg mL⁻¹). After incubation for 24 h, the cells were treated with each CDN. After further incubation for the intended duration, the supernatant was collected, and the IRF-3/INF-β induction was analyzed based on luminescence using QUANTI-Luc (InvivoGen) and a SEAP Reporter Assay kit (InvivoGen) according to the manufacturer's instructions.

Author contributions

N. S. T. designed and performed the experiments, analyzed data, and wrote the paper. M. K., Y. I., and Y. K. performed experiments. N. M. supervised the research and proofread the paper. All authors discussed the results and contributed to the final manuscript.

Conflicts of interest

The authors declare no competing financial interests in association with the present study.