Discussion

As the diversity of nucleotide second messengers continues to expand, particularly in bacteria,41,50 it has become increasingly important to determine whether these molecules can elicit signaling in human cells either through alternative pathways or via STING. Given that human STING responds to both endogenous metazoan 2′3′-cGAMP and selected bacterial CDNs,1,2,3,42,43,44,45 we systematically evaluated whether other nucleotide second messengers can activate STING in human cells. Previous nucleotide screens investigating STING signaling have used diverse human and mouse cell types expressing different STING variants and have monitored distinct cellular responses, making it difficult to analyze and compare STING activation across studies.29,98,99,100,101 Moreover, purine-containing CDNs have traditionally been the primary focus when analyzing STING agonists. Using our comprehensive nucleotide library, we evaluated the ability of diverse CDNs from various organisms, including those containing pyrimidine bases as well as other cyclic nucleotides such as cyclic mononucleotides and oligonucleotides, to activate endogenous STING in THP-1 monocyte cells and pancreatic cancer cells. We observed a striking specificity of nucleotide responses, identifying a distinct set of nucleotides that induced signaling (27 CDNs in total) and others that failed to do so, underscoring the remarkable selectivity that cyclic nucleotide receptors exhibit toward their respective ligands. Overall, our findings support the idea that additional nucleotide receptors beyond STING, with similarly high specificity, may be uncovered by monitoring nucleotide-mediated signaling in other cell types and experimental contexts.

In line with our data, the purine-containing CDNs were reported to be the most potent STING agonists in other nucleotide screens with different cellular models and readouts: (1) in suppression of virus infection with SARS-CoV-2, chikungunya virus (CHIKV), West Nile virus (WNV), and Zika virus (ZIKV) in human fibroblasts (HFF-1) and Calu-3 cells98,99; (2) in induction of IRF3 reporter activation in the mouse macrophage cell line analyzing 3′3′ CDNs composed of A, G, C, and U nucleobases100 and (3) in interferon reporter activation in HEK293T cells expressing the major STING variants and analyzing inosine-containing CDNs in combination with A, G, C, and U nucleobases, across 2′3′, 2′2′, and 3′3′ linkages.101 Analysis of pyrimidine-containing nucleotides has focused primarily on CDNs with 3′3′ linkages, which we show are not STING’s preferred linkage type. Indeed, in the context of 2′3′ and 2′2′ linkages, robust STING activation was observed for the mixed purine-pyrimidine CDNs 2′2′-cUIMP and 2′3′-cUIMP.101 This study, together with our observation that human STING is activated by the pyrimidine-containing CDN 2′3′-cUAMP, challenges the previously proposed strict preference for purine-containing CDNs. It indicates that combinations of different linkage types broaden the repertoire of CDNs capable of activating STING to include pyrimidine-containing nucleotides. In addition, our results extend this picture by demonstrating that even within the preferred 2′3′ linkage context, nucleobase positioning is a critical and previously underappreciated variable: pyrimidine substitution at specific positions reduces STING activation or abolishes it entirely. Signaling by pyrimidine-containing CDNs introduces an exciting dimension to the STING research field, particularly regarding the development of non-hydrolysable analogues for therapeutic use and the characterization of their cellular transport and degradation.22,23,24,25,26,27

Many of the CDNs and STING activators in our library belong to a rapidly expanding list of bacterial CDNs that play important roles in bacterial homeostasis, virulence and anti-phage defense.102,103 STING activation by bacterial CDNs is critical for mounting protective innate immune responses during infection of Gram-positive bacteria Listeria and Gram-negative Chlamydia trachomatis.45,104,105 These findings suggest that sensing bacterial CDNs or their degradation products represents a valuable strategy for detecting invading bacteria, either through STING or additional nucleotide-sensing receptors.106,107 For instance, STING activation may be particularly relevant in the gut, for sensing multiple microbiota-derived CDNs, in addition to detecting pathogenic disruption of the epithelial barrier.108,109,110 Our investigation is limited to the THP-1 monocyte and PDAC cell lines, and we cannot exclude the possibility that these nucleotides may signal in other human cell types, particularly if alternative receptors, other than STING, are expressed differentially across tissues and not present in THP-1s or PDACs.

Our analysis reveals a broad set of STING agonists that induce activation to varying degrees, which likely corresponds to differences in ligand binding strength and their ability to drive STING dimer closure. High-affinity ligands such as 2′3′-cGAMP and 3′3′-c-di-AMP engage deeply within the binding pocket and induce a closed dimer conformation. The 2′3′ linkage imposes unique structural constraints that promote this closed conformation leading to STING polymerization and downstream activation.4,40,67,68 In contrast, the earliest described bacterial STING agonist, 3′3-c-di-GMP, functions as a weaker activator both in cells and in vitro, driving a more open, apo-like STING conformation and an alternative cooperative mode of activation.4,29,67,98,99 Downstream signaling, in particular, is a critical aspect to consider in the development of effective STING agonists. Despite promising preclinical studies demonstrating the induction of antitumor immunity, most developed STING agonists have failed to translate these effects in clinical trials.19,21,111 For instance, the non-hydrolysable 2′3′-c-di-AMPSS (Rp,Rp) is under clinical development for the treatment of metastatic, solid tumors or lymphomas, but has so far shown limited efficacy in clinical trials.33,34 Additional strategies are being explored18,112 and may include gene therapies to deliver enzymes in cells for synthesis of STING-activating ligands111 or mRNA encoding constitutively active STING.113 Other approaches involve modifications of STING agonists including altering the position of thiophosphate substitution,114,115 modifying phosphodiester linkages,116 incorporating locked nucleic acids117 or dideoxy derivatives,118 designing sugar-modified analogues,119 inosine-containing CDNs,101,120 non-nucleotide based agonists,121,122,123 intermetallic nanoparticles encapsulating CDNs.124 Collectively, these findings demonstrate that the STING ligand-binding pocket accommodates a broad range of hydrolysable, non-hydrolysable, and chemically modified CDNs, providing multiple avenues for the development of more effective STING agonists. While our screen did not reveal nucleotide signaling pathways beyond STING under the tested conditions, this likely reflects the selectivity of CDN receptor recognition rather than an absence of such pathways in humans. The diversity of uncharacterized nucleotidyltransferases and enzymes that synthesize nucleotide second messengers represents a rich area for future exploration, and defining their products and cognate receptors will be an important next step toward a comprehensive understanding of nucleotide-based innate immune signaling.

Limitations of the study

We provide a systematic functional comparison of a structurally diverse nucleotide library that we anticipate will serve as a resource for the STING agonist and innate immunity community. A limitation of our study is that no STING-independent signaling was detected for any compound in our library. Several factors may account for this, including the absence of relevant receptors in the selected cell types or a structural mismatch between the tested compounds and their true endogenous counterparts. Future studies employing additional cell lines and endogenous nucleotide variants will be important to fully define the scope of CDN-based signaling beyond STING. Furthermore, electroporation was chosen to efficiently deliver nucleotides inside the cells for analysis of intracellular signaling as many nucleotides in our library may lack dedicated cellular importers and to avoid their degradation by extracellular phosphodiesterases. Electroporation does not allow to study intercellular transport, uptake and signaling of nucleotides and may have induced some cellular events, which affect signaling by different nucleotides. In addition, the primary cell lines used, THP-1 monocytes expressing the HAQ allelic variant and DANG pancreatic cancer cells expressing the WT allelic variant, may not capture the full diversity of human STING responses across allelic backgrounds and tissue contexts. Functional screening in additional cell lines carrying other STING alleles will be important to determine whether the observed linkage and nucleobase preferences are general properties in all human STING alleles.