By testing nearly every possible single amino acid substitution in human STING, Zhang and colleagues establish a sequence–function atlas that maps spatial control points and enables the interpretation of human variants. Structural and trafficking analyses reveal how oligomeric geometry and subcellular location differentially regulate STING-driven type I interferon production, NF-κB signaling, and LC3B lipidation, offering a powerful framework for better understanding STING biology in host defense and disease.The cyclic GMP-AMP synthase (cGAS)-STING pathway detects mislocalized DNA and drives antiviral, antitumor, and pathological inflammatory responses.1,2 STING resides in the endoplasmic reticulum (ER), where 2′3′-cGAMP binding triggers conformational changes, oligomerization, and trafficking through the Golgi and post-Golgi system, leading to type I interferon (IFN) production, NF-κB-dependent inflammation, and noncanonical autophagy. How STING combines rapid responsiveness with restraint while routing distinct outputs remains a central question.Structural analyses and characterization of pathogenic gain-of-function mutations causing STING-associated vasculopathy with onset in infancy (SAVI) have identified key regulatory elements,3,4,5,6 but known variants cover only a small portion of STING’s sequence space. Zhang et al. address this gap by combining deep mutational scanning with measurements of ligand-independent and ligand-induced IFN signaling and LC3B lipidation.7 Their near-complete atlas links sequence variation to activity and output selection.Autoactivating substitutions cluster in spatially separated regions across the transmembrane domain and its ER-luminal face, the ligand-binding domain, and higher-order assembly interfaces, revealing a distributed autoinhibitory architecture. This organization may keep STING near an assembly threshold while making it vulnerable to substitutions that stabilize signaling-competent states.Cryo-electron microscopy shows that these states are structurally distinct (Fig. 1a–d). For instance, W119K/M120K remodels the luminal transmembrane bundle and forms highly curved, active-like filaments, supporting strong IFN signaling largely from the ER. By contrast, D319K/D320K forms straighter oligomers — with an active-like ligand-binding domain but an apo-like transmembrane bundle — and requires forward trafficking. I165H creates a localized preactivated state near the cGAMP-binding pocket. Thus, STING can reach signaling competence through several conformational routes rather than through a binary inactive-to-active switch.Fig. 1: Oligomeric architecture and subcellular trafficking govern STING signaling outputs.Full size imagea cGAMP converts resting wild-type STING dimers into moderately curved oligomers that traffic from the ER to Golgi/post-Golgi compartments, enabling type I IFN production, NF-κB, and LC3B lipidation. b W119K/M120K stabilizes highly curved, constitutively active, largely ER-associated oligomers or filaments. These drive strong ER-derived type I IFN production despite limited trafficking, but weak NF-κB and reduced LC3B lipidation. c D319K/D320K forms planar, active, trafficking-competent oligomers with the potential to engage all three STING signaling outputs. d I165H stabilizes a preactivated dimer narrowed near the ligand-binding domain. Subsequent oligomerization remains unresolved; productive trafficking enables signaling. e Before ligand binding, L100E is shown as a resting dimer with an open ligand-binding domain. cGAMP induces ER-confined higher-order oligomers that preserve type I IFN induction but induce no detectable NF-κB activation or LC3B lipidation. Its depiction as a long, curved assembly is schematic because its architecture remains unresolved. All oligomer lengths and membrane curvatures are schematic and not to scale.Output-separating mutants reveal the functional consequences of these differences. L100E forms ligand-induced oligomers but remains at the ER, retaining strong IFN induction while nearly abolishing LC3B lipidation and NF-κB activation. Together with ER-active W119K/M120K, it uncouples oligomer formation from trafficking (Fig. 1b, e). In these engineered settings, appropriate higher-order assembly supports IFN signaling at the ER, whereas LC3B lipidation requires post-ER membrane remodeling and NF-κB activation requires downstream trafficking.7 These mutants do not show that wild-type STING normally bypasses trafficking; they reveal the contribution of subcellular location when assembly and movement are uncoupled.The findings suggest that organelles are active signaling environments, not passive stations. The ER, Golgi, and endolysosomal system differ in lipid composition, membrane curvature, ionic conditions, and remodeling capacity. ER cholesterol sets a threshold for STING export,8 phosphoinositides and cholesterol stabilize higher-order assemblies,9,10 and STING-dependent proton flux links the ionic environment of Golgi-derived compartments to LC3B lipidation.11 Sequence may therefore define accessible states while trafficking exposes STING to environments that select the output (Fig. 1). Although specific lipids have not been assigned to individual signaling branches, the output-biased mutants offer tools to test how organelle identity edits the traveling signal.The atlas also grounds the interpretation of human STING variants. It distinguishes established benign and pathogenic alleles, identifies activating variants beyond classical SAVI hotspots, and supports the classification of D319G as a pathogenic variant in a patient with SAVI-like interferonopathy. Cancer-associated variants are enriched for impaired cGAMP responsiveness, consistent with — but not proving — selection against DNA-triggered immune surveillance. Disagreement with computational predictions at flexible surface loops underscores the value of experimental maps for functionally important but structurally inconspicuous regions.Limitations remain. Most variants were screened by ectopic expression in HEK293T cells, with some validated in myeloid cells. Endogenous STING abundance, cell-specific partners, membrane composition, and post-translational modifications could alter the landscape, while mutations may stabilize rarely sampled or nonphysiological states. Physiological studies must define which conformations occur, how organelles control their lifetimes and outputs, and whether they can be selectively targeted.This study is more than a mutation catalog: it provides a functional reference for human STING variants and recasts this fascinating immune-related protein as a spatial signaling machine. The sequence encodes structural possibilities, but the eventual signaling output is also shaped by oligomer geometry, trafficking, and the microenvironmental membrane context. This framework raises the prospect of therapies that redirect where and how STING signals, preserving host defense while limiting inflammation or modulating autophagy. Similar mutational landscapes could reveal organizing principles in other signaling proteins and offer a blueprint for mapping context-dependent cellular signaling.ReferencesIshikawa, H., Ma, Z. & Barber, G. N. Nature 461, 788–792 (2009).Article CAS PubMed PubMed Central Google Scholar Sun, L., Wu, J., Du, F., Chen, X. & Chen, Z. J. Science 339, 786–791 (2013).Article CAS PubMed Google Scholar Ergun, S. L., Fernandez, D., Weiss, T. M. & Li, L. Cell 178, 290–301.e10 (2019).Article CAS PubMed Google Scholar Liu, Y. et al. N. Engl. J. Med. 371, 507–518 (2014).Article CAS PubMed PubMed Central Google Scholar Shang, G., Zhang, C., Chen, Z. J., Bai, X. C. & Zhang, X. Nature 567, 389–393 (2019).Article CAS PubMed PubMed Central Google Scholar Zhang, B. C. et al. Nat. Immunol. 26, 1916–1930 (2025).Article CAS PubMed Google Scholar Zhang, B. et al. Nature 655, 1271–1281 (2026).Article CAS PubMed PubMed Central Google Scholar Zhang, B. C. et al. Nat. Commun. 15, 2760 (2024).Article CAS PubMed PubMed Central Google Scholar Li, J., Tan, J. X., Chen, Z. J., Zhang, X. & Bai, X. C. Nature 652, 499–507 (2026).Article CAS PubMed PubMed Central Google Scholar Fang, R., Jiang, Q., Jia, X. & Jiang, Z. Immunity 56, 500–515.e6 (2023).Article CAS PubMed Google Scholar Liu, B. et al. Science 381, 508–514 (2023).Article CAS PubMed PubMed Central Google Scholar Download referencesAcknowledgementsThe work in the Paludan laboratory is supported by grants from the Lundbeck Foundation (R359-2020-2287), the Novo Nordisk Foundation (NNF23OC0084931), the Danish National Research Foundation (Center for Immunology of Viral infections, DNRF164), and the Swedish Research Council (2024-02549).Author informationAuthors and AffiliationsDepartment of Biomedicine, Aarhus University, Aarhus, DenmarkBao-cun Zhang & Søren R. PaludanCenter for Immunology of Viral Infections, Aarhus University, Aarhus, DenmarkBao-cun Zhang & Søren R. PaludanAuthorsBao-cun ZhangView author publicationsSearch author on:PubMed Google ScholarSøren R. PaludanView author publicationsSearch author on:PubMed Google ScholarCorresponding authorsCorrespondence to Bao-cun Zhang or Søren R. Paludan.Additional informationPublisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.Rights and permissionsReprints and permissionsAbout this article