Discovery of a photoactivatable dimerized STING agonist based on the benzo[b]selenophene scaffold

Stimulator of interferon genes (STING) agonism presents a powerful weapon for cancer immunotherapy. This study reports a novel dimerized STING agonist diBSP01, which exhibited promising STING binding and activation properties in vitro, based on the benzo[b]selenophene scaffold. Meanwhile, shielding the pharmacophores of diBSP01 with photoremovable protecting groups (PPGs) resulted in the generation of the first photoactivatable STING agonist, caged-diBSP01, that exerted no biological potency in the absence of light stimulation while regaining its STING agonistic activity after 400 nm irradiation. Optically controlled in vivo anticancer activity was also proven with caged-diBSP01 in a zebrafish xenograft model. Our study provides insights into developing novel STING agonists for cancer treatment and a solution for precise STING activation to avoid the on-target systemic inflammatory response responsible for normal cell damage caused by systemic STING agonism.

The cyclic GMP-AMP synthase (cGAS)-stimulator of interferon genes (STING) pathway is one of the most vital components in initiating innate immune responses upon endogenous (dead and tumor cells) or exogenous (bacterial and viral infections) stimulation. 1 Cytosolic aberrant double-stranded DNA (dsDNA) is detected by cGAS, followed by the generation of cyclic GMP-AMP (cGAMP) that serves as a second messenger to bind to and activate STING, aer which the endoplasmic reticulum (ER) membrane-anchored STING translocates to perinuclear compartments such as the Golgi apparatus. 2,3 STING then recruits and promotes the autophosphorylation of TANKbinding kinase 1 (TBK1) at the C-terminal tail and the phosphorylated TBK1 in turn phosphorylates STING tails that further recruit and deliver transcription factor interferon regulatory factor 3 (IRF3) to TBK1 for its phosphorylation. [4][5][6] In addition, TBK1 recruitment is also of great importance, activating transcription factor nuclear factor kB (NF-kB). 7 The activated IRF3 dimer and NF-kB together enter the nucleus and act synergistically to promote the production of type I interferons (IFNs) and other proinammatory factors (TNFa, IL6, and CXCL10), which ultimately initiates innate immune system and subsequent adaptive immune responses. 8,9 In particular, the ER membrane protein STING has been regarded as crucial in regulating such effects and has received increasing attention from scientists not only to shed light on its physiological functions but also to develop modulators targeting the STING adaptor for various diseases. [10][11][12] Indeed, given the potential benets of STING activation in tumor immunotherapy, numerous companies and institutes have dedicated themselves to discovering STING agonists to treat cancer. Most of the agents currently in clinical trials are rst-generation STING agonists, which are cyclic dinucleotide (CDN) derivatives that mimic the natural ligand cGAMP. These agents have their own deciencies, including high molecular weight, poor membrane permeability, and poor stability, which impede their further application. 13 The discovery of next-generation STING agonists with improved druggability and pharmacokinetic/ pharmacodynamic proles has led to several promising non-CDN agonists, represented by diABZI-Compound 3 from GSK, 14 SR717 from Scripps Research, 15 and MSA-2 from Merck, 16 all of which exhibit encouraging efficacy in vitro and in vivo. Moreover, several impressive studies of the structure-activity relationships of these molecules have been reported in succession. [17][18][19][20] Developing activators targeting STING has provided an impetus to cancer immunotherapy, and STING agonists with novel chemical entities are urgently required for propelling their application. However, the systemic administration of STING agonists is likely to induce a robust and inevitable on-target 'cytokine storm' that may cause damage to normal tissues, limiting the further clinical use of these agents. 21,22 Recently, the emerging role of the photoactivation strategy has provided a solution to this challenge. Upon pharmacological inactivation of a small molecule by concealing a functional group with a photoremovable protecting group (PPG), temporal and spatial control can be accomplished aer irradiation with ultraviolet (UV)-visible light. 23,24 Given the on-demand activation properties, multitudes of pharmacological agents disguised with various types of PPGs have been designed for precise treatment. [25][26][27][28] Despite the extensive applications of this strategy in prodrug design, no photoactivatable STING agonists have been reported yet.
This study describes the discovery of a novel dimerized STING agonist diBSP01, which exhibited promising binding properties and agonistic potency in vitro, based on the benzo[b] selenophene scaffold. In addition, a photo-sensitive STING agonist is presented for the rst time by using a coumarinbased PPG on carboxylic acid moieties representing the critical pharmacophores for STING binding. As expected, caged-diBSP01 was incapable of activating the STING pathway, while light irradiation resulted in the efficient cleavage of the PPG moiety and liberation of the parent drug responsible for the restored agonistic activity. In addition, the light-controllable in vivo anticancer activity of caged-diBSP01 was well characterized.
Our molecular design was initiated with a recently disclosed STING agonist, MSA-2 (Fig. 1A), identied by Merck. 16 MSA-2 is the rst orally available non-nucleotide STING agonist having tremendous potency in vivo in various mouse syngeneic tumor models. It has a relatively simple yet ingenious chemical structure, which indicates the presence of abundant space for chemical modication. Selenium (Se) is a chalcogen participating in a wide range of physiological events. As a favorable bioisostere related to oxygen and sulfur (S), Se-containing compounds are appreciated not only for their improved physicochemical proles but also for their remarkable anti-inammatory, antitumoral, antibacterial, and/or antiviral properties. [29][30][31] In addition, Se is capable of activating immune cells, resulting in the reversal of immunosuppression in the tumor microenvironment, and promoting the secretion of proinammatory cytokines such as IFN-g. 32 Considering the promising prospects of incorporating Se into bioactive molecules and the potential antitumor immune responses of Se compounds, a bioisostere strategy was implemented to replace the S in MSA-2 with Se so as to exert the potential synergistic Concentration-dependent SPR binding of diBSP01 to immobilized hSTING H232 (above) and corresponding Langmuir-binding isotherm (below). (C) Validation of the interaction between diBSP01 and hSTING H232 by ITC at 25°C. Experiments were performed three times and values are presented as mean ± standard deviation (SD).
antitumor effect of Se and STING activation, which resulted in the generation of a novel compound with the benzo[b]selenophene scaffold, namely BSP01 (Fig. 1A, see the ESI † for the synthesis). However, this simple modication of BSP01 resulted in no binding affinity towards hSTING H232 protein in a differential scanning uorimetry (DSF) assay (Fig. S1A †) and a dramatically decreased activity in an interferon-stimulated response element (ISRE) luciferase reporter assay in both human-derived cells (ISG-THP1 cells) and murine-derived cells (ISG-RAW264.7 cells) (Fig. S1B †), suggesting that BSP01 is a weak STING agonist. The underlying mysteries emerging from these results remain unclear; however, we speculated that the incorporation of selenium might increase the electron cloud radius of the benzoselenophene ring, leading to a change in the dihedral angle of C2/C3 from the butyric acid side chain (Fig.  S2 †) and thus an improper orientation of the terminal carboxyl in the binding pocket and weak interactions with Arg238, which is a crucial residue for STING binding and activation. 16 MSA-2 can preorganize in solution to form a bioactive noncovalent dimer, and then enter the ligand-binding pocket of STING and exert its pharmacological effects. Hence, Merck's researchers expended considerable efforts in searching for more potent covalent dimer STING agonists. 16 Dimer 3 that tethers benzothiophene between 5-positions with an all-carbon propyl linker was nally selected as a surrogate of MSA-2 using a computational screening method based on the crystal structure of the MSA-2-STING complex combined with phenotypic experiments. In addition, these dimers had a broad accommodation for different linker compositions and positions, provided that the oxobutanoic acid side chains stretched properly in the binding pocket and that the carbonyl and carboxyl groups bonded correctly to the key amino acids.
Inspired by these observations, a dimerization strategy was implemented on BSP01 to restrict the conformation and reduce the entropy loss during the spontaneous dimerization process, which might ultimately increase the agonistic activity. Merck's alternative dimer 3 showed loss of interactions of the 5-methoxy groups with Ser162 because of its all-carbon linker between 5positions, 16 which was not desired because we wanted to preserve the original interactions as much as possible. Therefore, we revisited the co-crystal structure of the MSA-2-STING complex and extracted the information that the 5-O of one monomer and the 5 ′ -O of the other monomer were adjacent to each other in the noncovalently dimerized MSA-2, with a distance of 3.1 Å (Fig. S3 †). This implied that a three-carbon atom linker with a C-C bond length of 1.54 Å is a good choice for covalent connection of BSP01 between 5-O and 5 ′ -O, which resulted in the formation of diBSP01 (Fig. 1A).
Following the synthesis of diBSP01 (see the ESI † for details), the thermostability of STING protein with diBSP01 was rst evaluated. diBSP01 could increase the thermal denaturation temperature of hSTING H232 in a concentration-dependent manner (Fig. S4A †). In addition, diBSP01 also showed strong affinity towards hSTING R232 and hSTING HAQ protein with DT m = 16.2°C and 27.0°C, respectively ( Fig. S4B and C †). These results indicated that diBSP01 restored its affinity to STING protein through dimerization. Surface plasmon resonance (SPR) and isothermal titration calorimetry (ITC) assays were then performed to conrm the binding ability of diBSP01 to STING. diBSP01 could induce the upregulation of the uorescence signal in a concentration-dependent manner with a dissociation constant (K D ) of 49 nM in SPR assay (Fig. 1B). Besides, in ITC assay, the binding of diBSP01 with STING effectively triggered a heat change and the binding ratio (N) was calculated to be 1.3 (Fig. 1C), suggesting that the stoichiometric ratio of the diBSP01-STING complex was 1 : 1. Moreover, the calculated K D was 218 nM. Also, DH = 10.10 kJ mol −1 , while −TDS = −48.1 kJ mol −1 , which implied a binding mode driven by entropy (Fig. 1C).
The 2.7 Å co-crystal structure of diBSP01 bound to hSTING H232 was resolved to fully understand the interactions between diBSP01 and STING. As illustrated in Fig. 2A, diBSP01 was buried in the ligand-binding pocket as a monomeric molecule. The ligand adopted a U-like shape by folding the connecting linker leading to the parallel orientation of the two benzoselenophene rings with a distance of 3.6 Å between the two surfaces ( Fig. 2A). Furthermore, the p-p stacking between the two rings was believed to be responsible for stabilizing this molecular conformation.
By merging the diBSP01-STING complex with the MSA-2-STING complex, it is found that diBSP01 coincided with MSA-2 in terms of the spatial location in the pocket and mimicked the interactions in the same way as MSA-2 ( Fig. 2B and C). Methoxyl and etheroxyl groups acted as hydrogen bond receptors to form hydrogen bonds with Ser162 located at the bottom of the pocket. In addition, the benzoselenophene rings were parallel to Tyr167, leading to the p-p stacking effect. The folded butyric acid side chains extended into the inside of the protein, where carboxyl and ketone moieties form intricate hydrogen bond networks with Arg238 and Thr263, inducing the formation of a b-sheet lid sealing the pocket and resulting in a 'closed' conformation responsible for STING activation and translocation. 4 In detail, the carbonyl and hydroxyl groups of carboxylic acids interacted with Arg238 of the opposite STING (e.g., chain B) and Thr263 of the proximal STING monomer (e.g., chain A), respectively, while ketones formed the cross-linking hydrogen bonds with Arg238 from the STING homodimer.
Given the promising behaviour of diBSP01 shown in STING's binding properties, the biological evaluation of diBSP01 for activating the STING pathway was subsequently conducted. In the ISRE reporter assay, diBSP01 regained its STING activation efficacy and exhibited similar activity related to MSA-2 with a normalized fold-change value of 34 and 32, respectively (Fig.  S5A †). Meanwhile, diBSP01 was effective in cGAS knock-out (KO) ISG-THP1 cells whereas it was incompetent in STING KO ISG-THP1 cells (Fig. S5A †), indicating that diBSP01 reliably targeted the STING adaptor in the cGAS-STING pathway. In THP1 cells, diBSP01 could signicantly induce the expression of downstream genes including IL6, CXCL10, and IFNb and showed a more potent STING activation ability than MSA-2 (Fig.  S5B †). Intriguingly, diBSP01 exhibited much more effective gene expression levels at a lower concentration compared with MSA-2 ( Fig. S5B †), implying that diBSP01 could still exert its pharmacological potency in a lower dose regimen. Since diBSP01 is a covalent dimer molecule, the equivalent dimer of MSA-2, namely diMSA-2, was prepared (following the same synthetic protocol as for diBSP01) to effectively assess the STING agonism capacity of our proposed compound. As shown in Fig. S5C and D, † diBSP01 exhibited comparable activities to those of diMSA-2 for the activation of the STING pathway in ISG-THP1 cells and the stimulation of downstream gene expression, indicating that diBSP01 was a potent dimer agonist targeting STING.
STING activation by small molecules is a promising strategy for cancer treatment. Nevertheless, systemically administered STING agonists potentially trigger non-negligible on-target immuno-inammatory responses throughout the whole body as a result of the widespread expression of STING in tumor and normal cells. Indeed, BALB/c mice intraperitoneally administered with 30 mg kg −1 diBSP01 exhibited complete intolerance, with 90% mortality on the rst day and all mice died within 4 days and a reduced dose of 10 mg kg −1 also resulted in 90% deaths within 7 days (Fig. S6A †), indicating diBSP01 had a formidable toxicity prole. Consequently, STING activation is urgently needed only in the tumor microenvironment. In view of this, a light-controlled STING agonist modied on the diBSP01 structure was developed as a proof-of-concept.
Given the pivotal role of carboxyl groups in recognizing key amino acid residues (Arg238 and Thr263) in the ligand-binding pocket veried by the co-crystal structure, blocking the terminal carboxyl moieties might be conducive to the temporary abolition of STING agonism. Coumarins are one of the most widely investigated PPGs due to their fast photodeprotection prole under biocompatible conditions and safety considerations upon photorelease. 33 Therefore, a commonly used diethylaminocoumarinyl-4-methyl (DEACM, Fig. 3) group was selected as the PPG of choice for covalent connection with diBSP01 via an ester bond. The linking of the two carboxyl groups in diBSP01 to two molecules of DEACM resulted in the rst photoactivatable STING agonist caged-diBSP01 (Fig. 3, see the ESI † for the synthesis) that has potential for spatiotemporal control of STING pathway activation by light.
We rst investigated the optical properties and photodeprotecting behaviors of caged-diBSP01 (Fig. S7 †). In terms of the UV absorption proles, caged-diBSP01 has two absorption maxima corresponding to the wavelengths of around 340 and 375 nm, which were almost the maximum UV absorption wavelengths of diBSP01 and PPG, respectively (Fig. S7A †). Furthermore, an extremely quick and complete conversion of caged-diBSP01 was observed under 400 nm irradiation using HPLC and UV spectra. The absorption at 375 nm gradually decreased with the extension of irradiation time, suggesting the occurrence of a photouncaging process (Fig. S7B †). The HPLC analysis indicated that caged-diBSP01 could liberate its active form diBSP01 and the corresponding PPG moiety with full conversion within 120 s (Fig. S7C †). It is noteworthy that an intermediate that only deprotected one side of PPGs was observed during this process and it was veried by LC/MS analysis (Fig. S8 †), which makes sense mechanistically. Unexpectedly, another by-product with a retention time between those of PPG and diBSP01 was detected, and its molecular weight was determined to be 277 using LC/MS (Fig. S8 †). Although we had no idea about the structure of this by-product, we speculated that it was a molecule related to PPG. Meanwhile, caged-diBSP01 was remarkably stable in PBS at 37°C in the dark over 24 h of monitoring (Fig. S9 †). These results together suggested that caged-diBSP01 was capable of causing subsequent STING activation assay under physiological conditions with quite short biocompatible irradiation (400 nm).
To examine whether caged-diBSP01 could activate the STING pathway in a light-controlled manner, the initial study was focused on an ISRE luciferase reporter assay. ISG-THP1 cells were treated with N,N-dimethylformamide (DMF), PPG, caged-diBSP01, or diBSP01 followed by the absence of or exposure to 400 nm irradiation for 60 s. Aer 24 h of incubation, a uorescence signal was detected and normalized to the vehicle control. As shown in Fig. 4A, treatment of caged-diBSP01 with 400 nm light signicantly increased the uorescence intensity compared with that of caged-diBSP01 without irradiation whereas PPG showed no STING activation potency regardless of light exposure. Moreover, caged-diBSP01 treated with 400 nm light for 60 s exhibited the same level of STING pathway activation as diBSP01 did, indicating that caged-diBSP01 was fully converted to its active form diBSP01 in a relatively short duration. In addition, caged-diBSP01 showed time-dependent STING activation properties, which reached its peak within a minute (Fig. 4B).
Encouraged by the aforementioned results, the lightcontrolled expression of downstream genes was determined to further conrm the potency of this photoactivatable agonist. In THP1 cells, the dose-dependent expression of IFNb was observed aer treatment with caged-diBSP01 followed by light stimulation, and the IFNb mRNA level in the caged-diBSP01 group increased to an identical level compared with that in the diBSP01 group at the same concentration of 12.5 mM (Fig. 4C). This result again demonstrated that caged-diBSP01 could effectively and efficiently liberate its bioactive molecule diBSP01 responsible for the regained STING activation efficacy. Subsequently, the induction of IL6, CXCL10, and IFNb genes was determined. As expected, the treatment of caged-diBSP01 with 400 nm light signicantly induced the expression of all of these three genes while the absence of irradiation showed no activity (Fig. 4D). Similarly, PPG could not activate the STING pathway irrespective of the presence or absence of light (Fig. 4D).
Western blotting experiments were then conducted to visualize the changes of STING downstream proteins so as to verify the protein levels in the cGAS-STING pathway of this photoactivatable system. The results showed that caged-diBSP01 aer irradiation remarkably induced the phosphorylation of STING, TBK1, and IRF3. In contrast, caged-diBSP01 without light stimulation could not liberate the active form diBSP01, hence possessing similar levels of protein expression compared with the vehicle control (Fig. 5A). Meanwhile, PPG could still not activate the STING-TBK1-IRF3 axis in the presence or absence of light (Fig. 5A). Furthermore, the regulation of STING and its downstream proteins by caged-diBSP01 treated with light also exhibited dose-dependent properties (Fig. 5B). Together all these assessments indicated that caged-diBSP01 is a promising photo-sensitive STING agonist prodrug which can undergo a rapid and complete release of its parent drug diBSP01 aer 400 nm irradiation and was proven to be effective in activating the cGAS-STING pathway in a light-controlled manner.
Zebrash have been widely used as a model organism for cancer research due to their high fecundity, low cost, accessibility for the modeling of various cancers, and transparent embryos for in vivo imaging. [34][35][36] They were applied for studying biology related to the STING pathway. 37,38 Therefore, in this study, we chose zebrash for developing a tumor model to investigate our photoactivatable STING agonist in vivo. Fluorescently labeled MC38 cells were implanted into the yolk sac of zebrash embryos and zebrash were incubated with different compounds followed by different light treatments. The uorescence intensity was detected 4 days aer implantation and  represented the proliferation of tumor cells. As illustrated in Fig. 6, the caged-diBSP01 group without irradiation showed no signicant difference in uorescence intensity compared with the vehicle group exposed to 400 nm light. The uorescence intensity signicantly decreased aer treatment with caged-diBSP01 with 400 nm irradiation for 60 s, suggesting that a light-induced antitumor activity was exerted. Furthermore, MC38-bearing zebrash were microinjected with compounds into the tumor site and subsequently given the indicated light treatments to better verify our idea of controllable local STING activation. It was found that, compared with the vehicle control, caged-diBSP01 aer irradiation could still signicantly inhibit tumor growth, but no signicant effect was found with caged-diBSP01 in the absence of light (Fig. S10 †). Moreover, diBSP01 exhibited almost no antitumor activity with no signicant difference compared with the vehicle control in the transiently STING knocked out zebrash while diBSP01 still showed a signicant tumor inhibitory effect in normal wild-type zebrash (data not shown), which implied that diBSP01 exerted antitumor efficacy in zebrash in a STING-dependent manner and suggested that diBSP01 could activate the zebra-sh STING protein.
In conclusion, these results demonstrated that caged-diBSP01 is capable of exerting a light-mediated antitumor potency in vivo, offering a new insight into precise STING-dependent innate immune activation and cancer immunotherapy.
Meanwhile, BALB/c mice intraperitoneally administered with up to 200 mg kg −1 caged-diBSP01 exhibited good tolerance, with no death and the body weight changes consistent with those for the vehicle control (Fig. S6 †), indicating that caged-diBSP01 had an improved safety prole aer blocking the terminal carboxylic acids of diBSP01. In addition, cell viability studies of dimer diBSP01 and its photoactivatable prodrug caged-diBSP01 were conducted in both MC38 and THP1 cells (Fig. S11 †). They showed that both compounds had no cytotoxicity to MC38 and THP1 cells under the tested concentrations within 72 and 24 h of incubation, respectively (Fig. S11A-E †). However, diBSP01 showed signicant cytotoxicity to THP1 cells aer long-term incubation (72 h). In contrast, caged-diBSP01 caused a lower decrease in cell viability compared with its parent drug diBSP01 (Fig. S11F †), which implied an improvement in the safety of this photo-sensitive prodrug.
In summary, we discovered a novel STING agonist diBSP01 by the covalent dimerization of two monomers of BSP01, a Sereplaced analogue of MSA-2, which resulted in a dramatic change in the STING agonistic potency of these two molecules. BSP01 showed nearly no STING activation ability, while diBSP01 displayed an excellent performance. The affinity assays demonstrated that diBSP01 was a favorable STING binder, which was further illustrated by its co-crystal complex with STING protein. Subsequent structure-activity relationship studies and in vivo assessment of this sort of dimerized STING agonist containing the benzo[b]selenophene scaffold are ongoing in our laboratory. On the other hand, a photoactivatable STING agonist, caged-diBSP01, was reported for the rst time by disguising carboxylic acids with a coumarin-based PPG. Caged-diBSP01 lost its STING activation properties and could efficiently liberate the parent drug diBSP01 which accounted for the restored biological efficacy aer light treatment. Furthermore, caged-diBSP01 could exhibit an optically controlled anticancer potency in vivo in a zebrash xenogra model. This concept allows the spatiotemporal activation of Zebrafish were incubated with the indicated compounds (all 12.5 mM) followed by different treatments (±400 nm, 60 s). Fluorescence intensity that correlated with the tumor size was detected on 4 days post-implantation and normalized to the vehicle control. Results are shown as mean ± SD (n = 10 for each group). Statistical significance was determined by a one-way ANOVA test. ns, not significant; **P < 0.01; ***P < 0.001; ****P < 0.0001. STING, which could minimize the adverse effects caused by systemic STING agonism, and offers an option for precise treatment of STING-related diseases such as cancer.

Data availability
The datasets supporting this article have been uploaded as part of the ESI. †

Conflicts of interest
There are no conicts to declare.