Applications of cryo-EM in drug development for STING

Abstract

STING is a critical adaptor protein in the cGAS-mediated DNA-sensing innate immune pathway. Binding of the second messenger cGAMP generated by cGAS to STING induces the high-order oligomerization and activation of the STING dimer. STING is a promising target for diseases associated with the cGAS/STING pathway such as cancer and autoimmune diseases. Recent applications of cryo-EM to STING have led to exciting progress in the understanding of its regulatory mechanism. Cryo-EM structures of STING bound to either cGAMP mimetics or novel small molecule ligands not only revealed the action mechanisms of these ligands but also suggested new ways to modulate the activity of STING for therapeutic purposes. Some of these recent studies are highlighted here.

Introduction

Innate immunity serves as an essential defense against viral and bacterial infections. One of the innate immune pathways is triggered by the DNA sensor cyclic GMP-AMP synthase (cGAS), which recognizes viral or bacterial DNA within the cytosol [1]. DNA-activated cGAS generates the second messenger 2’,3’-cyclic-di-GMP-AMP (cGAMP), which activates stimulator of interferon genes (STING), a membrane protein localized at the endoplasmic reticulum (ER) (Figure 1) [2,3]. Activated STING is translocated to the Golgi apparatus, where it forms high-order oligomers that recruit and activate TANK-binding kinase 1 (TBK1), which phosphorylates both STING and the transcription factor IRF3 [4,5]. Phosphorylated IRF3 triggers the expression of type I interferons, leading to antimicrobial immunity and inflammatory responses. The cGAS-STING pathway can also be activated by genomic or mitochondrial DNA aberrantly entering the cytosol under pathological conditions. This pathway therefore has become an important target for treating infections, autoimmune diseases, neurodegenerative diseases and cancer [6–8]. As a key player in this pathway, STING has been extensively studied for elucidating its regulatory mechanisms and exploiting it as a drug target. In the past few years, cryo-EM structural analyses have made fundamental contributions to the mechanistic understanding of STING. In several cases, the high-resolution structures of STING bound to various small molecule agonists provided clear insights into the binding modes and the action mechanisms. This article will highlight these recent progresses.

Figure 1.

Model of STING activation by cGAMP and the subsequent activation of TBK1 and IRF3.

Regulatory mechanisms of STING revealed by cryo-EM structures.

STING consists of four transmembrane (TM) helices, a cytosolic ligand-binding domain (LBD) that binds cGAMP, and a C-terminal tail (CTT) responsible for TBK1 recruitment (Figure 1) [4]. Previous crystal structures of the LBD have shown that the dimeric LBD adopts a butterfly-shape, with the crevice at the junction of the two wings constituting the cGAMP binding site [9,10](reviewed in [4]). cGAMP binding induces conformational changes that involve both an inward rotation of the two wings and the formation of a four-stranded β-sheet lid that covers cGAMP. However, it remained unclear how these conformational changes lead to the activation of TBK1. This critical mechanistic question has been addressed by the first set of cryo-EM structures of STING and its complex with TBK [11,12]. Full-length STING exhibits a domain swapped dimeric architecture in the TM domain (TMD), which is linked to the LBD through a short connector helix [12]. Strikingly, cGAMP induces a 180°-rotation of the LBD relative to the TMD, which in turn triggers local structural rearrangement of a loop on the side surface of the LBD [12]. These conformational changes together allow the side-by-side packing between neighboring STING dimers, ultimately leading to the formation of high-order oligomers [12,13]. STING uses the conserved TBK1-binding motif (TBM) (residues 369–377) in the CTT to bind TBK1 [11,14]. This binding mode however sterically hinders the phosphorylation of the critical Ser366 residue in STING by TBK1 [11]. The problem is solved by the high-order oligomers of STING, in which a subset of STING serves as recruiters of TBK1, while the CTT of STING not bound to TBK1 can be phosphorylated [11]. This elegant mechanism provides an explanation for the essential role of the high-order oligomerization in STING signaling.

More recently, cryo-EM analyses of full-length chicken STING revealed double-layered high-order oligomers, with the two layers held together by head-to-head interactions of the LBD [15]. This double-layered assembly has been proposed to represent an autoinhibited state of STING, by hindering the ER-to-Golgi translocation of STING and/or the recruitment of TBK1 [15,16]. Cryo-EM has also contributed to the understanding of the termination of STING signaling by clathrin-associated adaptor protein complex 1 (AP-1), which delivers STING to endolysosomes for degradation [17]. A cryo-EM structure of AP-1 bound to the phosphorylated STING-CTT shows how the phosphorylation enhances the recognition by AP-1, establishing a negative-feedback mechanism to prevent harmful persistent immune signaling of STING [17].

Activation of STING by C53 through targeting a cryptic pocket in the TMD.

Compound 53 (C53) was recently identified by Curadev as a human STING specific agonist with an EC₅₀ of ~185 nM [18]. Distinct from negatively charged cGAMP, C53 does not contain a charged group and is predominantly hydrophobic. cGAMP or C53 individually shows weak induction of oligomerization of purified full-length STING in detergent solution [19]. In contrast, co-treatment with both cGAMP and C53 yields robust formation of high-order oligomers of STING. These findings indicate that C53 binds to a distinct site in STING and can potentially work synergistically with cGAMP to facilitate STING activation. The high-order STING oligomers induced by cGAMP and C53 together possess a positive curvature, which may promote their enrichment at the ridge of the ER membrane and vesicles and thereby facilitate the ER-to-Golgi translocation [19]. A similar curvature has been observed in the cryo-EM structure of STING oligomers bound to cGAMP alone [15], suggesting that C53 facilitates the intrinsic oligomerization mechanism of STING, rather than inducing an artificial oligomeric state. It should be noted that, given its critical role in signaling, the high-order oligomerization of STING in cells is subjected to additional regulations such as the lipid environment, localization to different subcellular compartments and post-translational modifications [5,20,21].

The cryo-EM structure of STING/cGAMP/C53 complex reveals that cGAMP and C53 binds the LBD and TMD of STING, respectively (Figure 2) [19]. In particular, the “C”-shaped C53 is located at a hydrophobic pocket at the luminal side of the TMD of each STING dimer, around the 2-fold symmetry axis. The binding pocket is formed between two TM2 and two TM4. On one side of the binding pocket, the outer edge of the oxindole core of C53 interacts with TM2 from one protomer; meanwhile, the two methyl groups of the oxindole core of C53 contact TM4 from the other protomer. On the other side of the binding pocket, the 2-Cl-6-F-phenyl and trifluoro-phenyl rings of C53 make interactions with TM2 of one protomer and TM4 of the other protomer, respectively. TM3 as well as the loop connecting the TM3 and TM4 also make direct contacts with C53, further stabilizing the interaction between C53 and STING.

Figure 2.

Cryo-EM structure of human STING bound to both cGAMP and C53. The left and right panels show the side and top views of the structure, respectively (PDB ID: 7SII). The inset shows an expanded view of C53 with the cryo-EM density displayed in semi-transparent mode.

A structural comparison of STING with or without the C53 bound unveils significant conformational changes in TMs around C53. Specifically, TM2 and TM4 undergo substantial sideways movements upon the C53 binding. These movements result in substantial dilation of the TMD, which promotes the TMD-TMD interactions between STING dimers that stabilize the high-order oligomerization. This allosteric activating mechanism of C53 is orthogonal to that of cGAMP, providing an explanation for their synergistic or additive effects on STING activation.

The cryo-EM study of human STING bound to both cGAMP and C53 elucidates the role of the TMD in STING oligomerization and activation, and more importantly identifies a new agonist-binding site in the TMD, opening avenues for developing agonists targeting this site. Compounds like C53 targeting the TMD site might surpass cGAMP or cGAMP analogs as STING agonists for therapeutic applications, due to their greater potential to cross the plasma membrane and reach STING on ER or Golgi. This work also raises an intriguing speculation that an endogenous ligand, such as a metabolite or lipid, binds to STING in a similar fashion as C53 and thereby facilitates STING activation in cells. Indeed, several previous studies have suggested roles of lipids and cholesterol in modulating STING signaling [21–25].

Recently, human STING was proposed to act as a proton channel through the center pore in the TMD [26,27]. Proton efflux mediated by STING increases the pH within the Golgi apparatus and post-Golgi vesicles, which is required for the activation of the inflammasome and autophagy downstream of STING. Interestingly, C53 functions as a channel blocker, by binding to the TMD and obstructing the pore. Therefore, C53 can selectively activate the TBK1/IRF3/Interferon branch of STING signaling while inhibit the inflammasome and autophagy branches, making it a great tool for dissecting the functions of the cGAS-STING pathway. It has been shown that sulfated glycosaminoglycans (GAGs) in Golgi bind the luminal end of the STING-TMD, near the C53-binding site and thereby promote the high-order oligomerization of STING [20]. An interesting question is whether GAGs block the channel activity of STING in a manner similar to C53. It should be noted that the proton conducting rates of STING in liposomes is low [26,27]. It remains to be seen whether human STING is indeed a proton-specific channel or conducts ions non-specifically through the TMD pore, and whether the channel activity is a broadly conserved function of STING from all species.

NVS-STG2 induces STING oligomerization by acting as molecular glue.

Through a functional screen of a chemical library containing 250,000 compounds and subsequent optimization, Novartis discovered NVS-STG2, a novel STING agonist that is distinct from both cGAMP analogues and C53 [28]. NVS-STG2 was able to significantly curtail tumor growth in vivo, despite the modest potency in activating STING in cell-based assays (EC₅₀ in the single digit μM range).

NVS-STG2 alone induced larger oligomers of STING than those induced by cGAMP or C53 [28]. Remarkably, combining NVS-STG2 with cGAMP, C53, or both led to the formation of much larger oligomers. This indicated that NVS-STG2 induces STING activation by binding to a third site that does not overlay with either the cGAMP- or C53-binding site. To understand the binding and activation mechanism of STING by NVS-STG2, the cryo-EM structures of STING in complex with cGAMP/NVS-STG2 or cGAMP/C53/NVS-STG2 in the high-order oligomeric form were determined at 4 Å and 2.9 Å resolution, respectively (Figure 3). The two structures are nearly identical to each other and to the previous cGAMP/C53-bound structure. This indicates that NVS-STG2 induces the same conformational changes in the TMD of STING as C53, supporting the idea that these conformational changes are an intrinsic part of the oligomerization and activation mechanism of STING.

Figure 3.

Cryo-EM structure of human STING bound to both cGAMP and NVS-STG2. The left and right panels show the side and top views of the structure, respectively (PDB ID: 8FLM). The inset shows an expanded view of NVS-STG2 with the cryo-EM density displayed in semi-transparent mode.

The cryo-EM structures show that two NVS-STG2 molecules pack side-by-side mainly through the central phenyl group and occupy a cavity formed between the TMDs of two neighboring STING dimers [28]. Specifically, NVS-STG2 interacts with TM2, TM3, and TM4 of one STING dimer and TM2 and TM4 of the adjacent dimer. The carboxylic acid group of NVS-STG2 makes electrostatic interactions with R94 and R95, further contributing to interaction between STING and STG2. Through this binding mode, NVS-STG2 acts as a “molecular glue” between neighboring STING dimers and thereby promotes the formation of high-order oligomers. Several key residues in the NVS-STG2-binding site are not conserved in mouse STING, underlying the fact that NVS-STG2 activates human STING but not mouse STING.

The identification of an additional agonist-binding site ushers in new prospects for targeting STING with small molecule agonists for cancer therapy. Combined treatment with different drugs that target each of the three ligand-binding sites in STING may prove more effective than individual compounds. Finally, the cryo-EM structures provide a basis for designing more potent agonists. On the other hand, modifications of C53 and NVS-STG2 based on the structures may generate antagonists of STING that bind the same sites but prevent the oligomerization.

HB3089 activates STING without inducing the 180°-rotation.

Dimeric amidobenzimidazoles (diABZIs) are unusual cGAMP-mimetic agonists of STING in that they bind the cGAMP-binding pocket in the LBD of STING, but do not induce the closure of the lid as cGAMP does [29]. The precise mechanism by which diABZIs induces STING activation remained unclear. Through rational design based on diABZI, a recent study successfully developed a diABZI analog with higher asymmetricity, namely HB3089 [30]. HB3089 is able to bind to various STING isoforms and activate STING signaling potently. Impressively, HB3089 exhibits improved pharmacokinetics and anti-tumor efficacy as compared to diABZI.

The cryo-EM structure of the full-length STING dimer bound to HB3089 shows that HB3089 sits in the cGAMP-binding pocket as designed [30] (Figure 4). HB3089 does induce an inward rotation of the two wings of the STING-LBD, but the lid at the top of the ligand-binding site remains open, consistent with that seen in the diABZI-bound structure of the dimeric LBD of STING. An unexpected feature of this structure is that HB3089 does not induce the 180°-rotation of the LBD relative to the TMD, the conserved mechanism in the activation of STING by cGAMP. This stark difference indicates that HB3089 triggers STING activation through a different mechanism, which appears to be also used by some activating mutants found in patients with STING-associated vasculopathy with onset in infancy [30]. Consistently, the authors did not observe high-order oligomerization of STING in the presence of HB3089. Nevertheless, the TMD of STING undergoes subtle conformational changes upon the binding to HB3089. Particularly, TM2 and TM4 of STING move inward, meanwhile TM1 rotates around its own helical axis. The authors propose that these HB3089-induced conformation changes of STING-TMD might facilitate the dissociation of STING with the ER-resident protein STIM1 and thereby promote the ER-to-Golgi trafficking. The enhanced translocation, rather than higher-order oligomerization, may underlie HB3089-mediated activation of STING. Further studies are required to test whether this mechanism is specific to HB3089 or also employed by the endogenous ligand cGAMP. Regardless, HB3089 exemplifies yet another means for modulating the activity of STING for potential therapeutic applications.

Figure 4.

Cryo-EM structure of human STING bound to HB3089. The left panel shows the HB3089-bound structure (PDB ID: 8GT6), with the expanded view of HB3089 and the cryo-EM density shown in the inset. The cryo-EM structure of chicken STING bound to cGAMP (PDB ID: 6NT7) is shown in the right panel for comparison.

Future perspectives

The multiple ways of activating STING discussed here underscore the remarkable mechanistic complexity built into STING, a relatively small (the size of the dimer is ~70 kDa) and seemly simple protein. The insights provided by the cryo-EM structures showcase its great potential in facilitating the understanding and design of drugs for STING specifically and other target proteins in general[31]. However, it remains challenging to solve cryo-EM structures to true atomic resolution, especially for small proteins such as the dimeric state of STING. This is particularly relevant to STING bound to antagonists, which often act by blocking the high-order oligomerization of STING [32,33]. Structures at atomic resolution with individual atoms of the bound small molecules clearly resolved are required for precise structure-based drug design. Tremendous efforts are currently being directed at improving sample stability, cryo-EM grid surface engineering and image processing algorithms [34], which together will hopefully push the resolution and protein size limit of cryo-EM in the near future, turning cryo-EM into a routine tool for accelerating drug discovery.