E7766

E7766, a Macrocycle-Bridged Stimulator of Interferon Genes (STING) Agonist with Potent Pan-Genotypic Activity

Dae-Shik Kim,*[a] Atsushi Endo,[a] Francis G. Fang,[a] Kuan-Chun Huang,[b] Xingfeng Bao,[b]
Hyeong-wook Choi,[a] Utpal Majumder,[a] Young Y. Shen,[a] Steven Mathieu,[a] Xiaojie Zhu,[a]
Kristen Sanders,[a] Thomas Noland,[a] Ming-Hong Hao,[a] Yu Chen,[a] John Y. Wang,[a]
So Yasui,[c] Karen TenDyke,[a] Jiayi Wu,[b] Christy Ingersoll,[a] Kara A. Loiacono,[a]
Janna E. Hutz,[a] and Nadeem Sarwar[a]

A strategy for creating potent and pan-genotypic stimulator of interferon genes (STING) agonists is described. Locking a bioactive U-shaped conformation of cyclic dinucleotides by introducing a transannular macrocyclic bridge between the nucleic acid bases leads to a topologically novel macrocycle- bridged STING agonist (MBSA). In addition to substantially enhanced potency, the newly designed MBSAs, exemplified by clinical candidate E7766, exhibit broad pan-genotypic activity in all major human STING variants. E7766 is shown to have potent antitumor activity with long lasting immune memory response in a mouse liver metastatic tumor model. Two complementary stereoselective synthetic routes to E7766 are also described.

Cyclic dinucleotides (CDNs) including cyclic 3’,5’-diguanylic acid (c-di-GMP; 1)[1] and 2’,3’-cyclic guanosine monophosphate– adenosine monophosphate (cGAMP; 2),[2] are natural ligands for stimulator of interferon genes (STING), which has been found to be one of key mediators for innate immune response (Figure 1).[3]
c-di-GMP (1) is derived from bacteria pathogens while cGAMP (2) is produced endogenously by cyclic GMP-AMP synthase (cGAS) as a response to cytosolic double strand DNA (dsDNA).[4] Under- standing the STING pathway from molecular level structures to biological functions has increased the viability of STING as a drug

[a] Dr. D.-S. Kim, Dr. A. Endo, Dr. F. G. Fang, Dr. H.-w. Choi, Dr. U. Majumder, Dr. Y. Y. Shen, S. Mathieu, X. Zhu, K. Sanders, Dr. T. Noland, Dr. M.-H. Hao, Dr. Y. Chen, Dr. J. Y. Wang, K. TenDyke, C. Ingersoll, K. A. Loiacono,
Dr. J. E. Hutz, Dr. N. Sarwar Eisai Inc.
35 Cambridgepark Drive Cambridge, MA 02140 (USA)
E-mail: [email protected] [b] Dr. K.-C. Huang, Dr. X. Bao, J. Wu
H3 Biomedicine
300 Technology Square FL5 Cambridge, MA 02139 (USA)
[c] S. Yasui
Analytical Research Laboratories
Pharmaceutical Science & Technology, Eisai Co., Ltd. 5-1-3 Tokodai, Tsukuba-shi
Ibaraki 300-2635 (Japan)
Supporting information for this article is available on the WWW under https://doi.org/10.1002/cmdc.202100068

Figure 1. Representatives of natural cyclic dinucleotides.

target for treatment of various diseases including pathogenic infections, cancers and autoimmune diseases.[5]
Given the demonstrated capability of STING agonists to induce a tumor antigen-specific immune response in multiple preclinical tumor models, tremendous efforts have been on- going to develop STING agonists for cancer immunotherapy as a single agent, as well as a combination agent with immune checkpoint blockades to potentially expand their clinical benefits.[6] Despite the great enthusiasm, early phase 1 clinical studies for STING agonists ADU-S100[7] and MK1454[8] haven’t shown impressive anticancer activity although both compounds were found to be tolerable. In addition, chronic and prolonged STING pathway activation showed senescence-associated secre- tory phenotype (SASP) and is correlated with evasion of senescence and immune-suppression.[9] All these recent clinical and preclinical findings suggest that a better understanding about the relationship of STING pathway and tumor immunity, and better STING agonists are necessary to modulate STING pathway as a cancer immunotherapy approach.
Modification of natural ligands (CDNs) such as shown in Figure 2 has been the major approach to create drug candidates.[5,6] Despite those intense efforts, the essential macro- cyclic structure of the natural ligands has largely been main- tained. In addition, discovery of a pan-genotypic STING agonist is much needed given that there are four major known STING variants[10] whose responses to CDN agonists are variable. The REF variant in particular, present at a frequency of 10–15% worldwide, has been found to show poor response to the bacteria-derived 3’,3’-CDNs.[11] CDNs adopt a U-shaped conforma- tion with two nucleobases π-parallel upon binding to STING dimer proteins during the biological event of innate immunity activation.[12] With a view toward rigidifying the CDN in the binding conformation, we considered a new additional macro-

ChemMedChem 2021, 16, 1–5 1 © 2021 Wiley-VCH GmbH

Figure 2. Structures of conventional CDN STING agonists and macrocycle- bridged STING agonist (MBSA).

cycle construct linking the two purine bases as shown in Figure 2. We theorized that these new macrocycle-bridged STING agonists (MBSAs) would enhance the binding properties due to favorable entropic contribution. In order to test this hypothesis, a synthetic route to these topologically novel bridged-bicyclic macrocycles would need to be established. In this article are described: 1) a ring-closing metathesis (RCM) route to macro- cycle-bridged STING agonist E7766, 2) a complementary stereo- selective phosphitylative macrocyclization approach, and 3) dem- monstration of unique biological profile (including potent activity against the REF variant) of E7766.[13]
Synthesis of E7766 commenced with Mitsunobu reaction of commercially available 3[14] with allyl alcohol, leading to alkene 4 in 73% yield (Scheme 1). Hydrolysis of phosphoramidite 4 followed by sequential removal of 2-cyanoethyl (CE) and 4,4’- dimethoxytrityl (DMT) groups afforded H-phosphonate mono- ester 5.[15] Coupling of alcohol 5 and phosphoramidite 4 followed by sulfurization[16] with 3-[(dimethylaminomethylene)amino]-3H- 1,2,4-dithiazole-5-thione (DDTT) and detritylation gave com- pound 6 as a 1:1 mixture of phosphorothioate diasteromers.[17]
Treatment of hydroxyl H-phosphonate 6 with 2-chloro-5,5- dimethyl-1,3,2-dioxaphosphorinane 2-oxide (DMOCP) followed

Scheme 1. First -generation synthetic route for E7766.

by 3H-benzo[c][1,2]dithio-3-one and 2-nitrobenzylbromide[18] pro- vided a mixture of three phosphorous diastereomers with 4:1 selectivity at the newly generated phosphorothioate linkage. After prep-HPLC separation, SpRp isomer 7 was obtained along with the RpRp isomer (8; 5.6 and 4.7% yield respectively from 4). The phosphorous stereochemistry was assigned by NMR data based on literature precedents,[19] and later confirmed by X-ray structures of E7766 and its RpRp isomer (vide infra).[20] Having established an appropriately functionalized CDN substrate (albeit with modest stereoselectivity), the transannular RCM was attempted. Gratifyingly, RCM of the SpRp isomer (7) with Hoveyda-Grubbs secnd generation catalyst (HG-II)[21] afforded the desired macrocycle-bridged construct as trans-alkene 9 in 39% yield. After sequential removal of two 2-nitrobenzyl groups with PhSH/TEA[18] and two benzoyl groups with ammonium hydroxide, E7766 was obtained in 70% yield. The structure and stereo- chemistry of E7766 were unambiguously determined by X-ray analysis.
While the viability of the macrocycle-bridged CDN synthesis had been demonstrated, low selectivity installing the phosphor- othioate stereogenic centers and modest yield in the trans- annular RCM limited the overall efficiency of the route. A complementary synthetic route that connected the two purine bases prior to forming the phosphorothioate linkages was considered. Pre-installation of the covalent carbon linker would obviate the need for RCM while increasing the potential for stereochemical communication through intramolecular phos- phorothioate formation.
The second-generation synthesis of E7766 began with Mitsunobu reaction of known compound 10[22] with diol 13 to give product 14 in 68% yield (Scheme 2). Mitsunobu reaction of alcohol 14 with mono-TBS ether 12, prepared from diol 11 via selective silylation, afforded alcohol 15 in 67% yield. Treatment of alcohol 15 with diphenyl phosphite[23] followed by detritylation provided H-phosphonate monoester 16. Intramolecular phosphi- tylation with DMOCP followed by sulfurization led to a 3:1 diastereomeric mixture of the cyclized products. The major product Sp phosphorothioate 17[24] was isolated by silica gel column chromatography in 41% yield from 15. The 2nd macro- cyclization, transannular macrocyclization, was realized by the

ChemMedChem 2021, 16, 1–5 www.chemmedchem.org 2 © 2021 Wiley-VCH GmbH

Scheme 2. Second-generation synthetic route for E7766.

phosphoramidite method.[25] Removal of bis-TBS groups in 17 with TEA-3HF, S-alkylation with 2-nitrobenzylbromide, and selective phosphitylation of the resulting diol provided phos- phoramidite 18. Without isolation, the resulting product was then treated with pyridinium trifluoroacetate followed by DDTT to give a 5:1 mixture of SpRp and SpSp isomers.[26] The desired SpRp isomer (19) was isolated by silica gel column chromatog- raphy in overall 44% yield over three steps. Treatment of 19 with PhSH/TEA followed by ammonium hydroxide afforded E7766 in 85% yield.
To elucidate the binding mode of E7766 to STING, X-ray co- crystal structures of E7766 with human WT (PDB ID: 6XF3) and REF (PDB ID: 6XF4) STING proteins were obtained as depicted in Figure 3. In both cases, E7766 presented at the dimer interface

Figure 3. X-ray co-crystal structures of E7766 with human WT and REF STING proteins.
of STING proteins. As expected, E7766 in the co-crystal structures has essentially the same conformation as E7766 alone shown in Scheme 1, consistent with the design concept of MBSA: locking the STING bioactive conformation. The olefin linker in E7766 creates distinct hydrophobic interaction with the aliphatic portion of the side chain of Arg238 in both cases.
Biochemical binding and reporter cell assay results of E7766 along with natural ligand 2’,3’-cGAMP (2) are summarized in Table 1. Based on Kd measured by isothermal titration calorimetry (ITC), E7766 has higher binding affinity than 2’,3’-cGAMP (2; Kd = 0.04 ti 0.01 vs. 0.07 ti 0.01 μM) due to favorable entropic contri- bution to the binding even though E7766 has the 3’,3’-linkage system which is known to be entropically less favored than the 2’,3’-linkage upon binding to STING.[27] In the THP-1 cell-based IRF reporter assay, four cell lines expressing each of four main human STING variants were used. As shown in Table 1, E7766 exhibited single digit μM EC50 activities for all these four human STING variants, which are 8–55 fold more active than 2’,3’-cGAMP (2).
For in vivo evaluation of anti-cancer activity, E7766 was tested in a tumor model mimicking liver metastasis using murine colon cancer CT26 cells, where CT26 tumors grew both under skin and in the liver of the same animal, with a single intratumoral injection of 10 mg/kg E7766 in the subcutaneous tumor. E7766 showed statistical significant survival benefit compared to the vehicle group (p < 0.0001, log-rank Mantel-Cox test, Figure 4). Treatment with E7766 eradicated the injected subcutaneous tumor and the non-injected liver tumors as well in ChemMedChem 2021, 16, 1–5 www.chemmedchem.org 3 © 2021 Wiley-VCH GmbH Figure 4. Animal survival curves of E7766 and vehicle-treated CT26 liver metastatic tumor model (left) and individual tumor volume plot of follow-up CT26 tumor rechallenging study (right). Arrows denote compound admin- istration (left) and tumor rechallenge (right). nine out of ten treated animals. All the treatment did not induce severe toxicity (< 20% body weight loss) in the tested animals. On day 90, the nine tumor-free mice from E7766 treatment were rechallenged with injection of CT26 cells subcutaneously. Seven out of the nine animals (78%) completely rejected the tumor rechallenge, indicating establishment of an effective antitumor immunity against CT26 tumor in the large majority of those E7766-treated animals. In contrast, all naïve animals (5 out of 5) that received the same tumor cell challenge grew tumors. In conclusion, E7766, the first macrocycle-bridged STING agonist (MBSA), was designed to lock the STING active CDN conformation, and synthesized according to two complementary synthetic approaches utilizing RCM in the first route and stereo- selective intramolecular phosphorothioate formation in the second route. E7766 showed superior and pan-genotypic activity against four major human STING variants than cGAMP. Intra- tumoral administration with E7766 was highly efficacious in CT26 live metastasis tumor model. A single-crystal structure of E7766 and a co-crystal structure of E7766 bound to STING provide a structural explanation for its potent biological activity. Based on the promising preclinical results, E7766 is currently undergoing phase I clinical assessment (NCT04144140 and NCT041109092). Further applications of the MBSA design and synthesis approaches to other members of this class and towards additional therapeutic indications are being pursued. Acknowledgements We thank our colleagues J. Cutter and Dr. Hui Fang (Eisai Inc.) for HPLC purification, and Dr. Marsha Eno (Eisai Inc.) for HRMS data. Conflict of Interest The authors declare no conflict of interest. Keywords: cyclic dinucleotides · drug design · immuno- oncology · macrocycles · STING agonists [1]D. L. Burdette, K. M. Monroe, K. Sotelo-Troha, J. S. Iwig, B. Eckert, M. Hyodo, Y. Hayakawa, R. E. Vance, Nature 2011, 478, 515–518. [2]J. Wu, L. Sun, X. Chen, F. Du, H. Shi, C. Chen, Z. J. Chen, Science 2013, 339, 826–830. [3]a) H. Ishikawa, G. N. Barber, Nature 2008, 455, 674–678; b) H. Ishikawa, Z. Ma, G. N. Barber, Nature 2009, 461, 788–792. [4]A. Ablasser, M. Goldeck, T. Cavlar, T. Deimling, G. Witte, I. Rohl, K. P. Hopfner, J. Ludwig, V. Hornung, Nature 2013, 498, 380–384. [5]H. Zhang, Q.-D. You, X.-L. Xu, J. Med. Chem. 2020, 63, 3785–3816, and references cited therein. [6]G. Berger, M. Marloye, S. E. Lawler, Trends Mol. Med. 2019, 25, 412–427, and references cited therein. [7]F. Meric-Bernstam, S. K. Sandhu, O. Hamid, A. Spreafico, S. Kasper, R. Dummer, T. Shimizu, N. Steeghs, N. Lewis, C. C. Talluto, S. Dolan, A. Bean, R. Brown, D. Trujillo, N. Nair, J. J. Luke, J. Clin. Oncol. 2019, 37, 2507–2507. [8]K. J. Harrington, J. Brody, M. Ingham, J. Strauss, S. Cemerski, M. Wang, A. Tse, A. Khilnani, A. Marabelle, T. Golan, Ann. Oncol. 2018, 29, viii712. [9]J. Kwon, S. F. Bakhoum, Cancer Discovery 2020, 10, 26–39. [10]Four major known STING variants with allele frequency:[11] WT (R232) 57.9%, HAQ (R71H, G230 A, R293Q) 20.4%, REF (R232H) 13.7%, AQ (G230A, R293Q) 5.2%. [11]G. Yi, V. P. Brendel, C. Shu, P. Li, S. Palanathan, C. Cheng Kao, PLoS One 2013, 8, e77846. [12]P. V. Krasteva, H. Sondermann, Nat. Chem. Biol. 2017, 13, 350–359, and references cited therein. [13]D.-S. Kim, F. G. Fang, A. Endo, H.-W. Choi, M.-H. Hao, X. Bao, K.-C. Huang, US Patent 10,246,480, 2019. [14]Improvement of biological activities by 2’ fluorine modification on CDNs was observed in our SAR studies and also reported in the literature; see T. Lioux, M.-A. Mauny, A. Lamoureux, N. Bascoul, M. Hays, F. Vernejoul, A.-S. Baudru, C. Boularan, J. Lopes-Vicente, G. Qushair, G. Tiraby, J. Med. Chem. 2016, 59, 10253–10267. [15]B. Das, G. Mahender, V. Sunil Kumar, N. Chowdhury, Tetrahedron Lett. 2004, 45, 6709–6711. [16]Phosphorothioate modification is known to increase stability and cellular potency of CDNs: L. Li, Q. Yin, P. Kuss, Z. Maliga, J. L. Millán, H. Wu, T. J. Mitchison, Nat. Chem. Biol. 2014, 10, 1043–1048. [17]B. L. Gaffney, E. Veliath, J. Zhao, R. A. Jones, Org. Lett. 2010, 12, 3269– 3271. [18]Z. J. Lesnikowski, M. M. Jaworska, Tetrahedron Lett. 1989, 30, 3821–3824. [19]B. Das, G. Mahender, V. Sunil Kumar, N. Chowdhury, Tetrahedron Lett. 2004, 45, 6709–6711. [20]Deposition numbers 2009202 (for E7766) and 2009203 (for the RpRp isomer (S2) of E7766) contain the supplementary crystallographic data for this paper. These data are provided free of charge by the joint Cambridge Crystallographic Data Centre and Fachinformationszentrum Karlsruhe Access Structures service www.ccdc.cam.ac.uk/structures. S2 exhibits weaker biological activities than E7766: EC50 (μM) in THP-1 cell- based IRF reporter assay (n = 2) = 3.6 (WT), 5.8 (HAQ), 3.5 (AQ) and 11.5 (REF). [21]S. B. Garber, J. S. Kingsbury, B. L. Gray, A. H. Hoveyda, J. Am. Chem. Soc. 2000, 122, 8168–8179. [22]W. Chang, J. Du, S. Rachakonda, B. S. Ross, S. Convers-Reignier, W. T. Yau, J. F. Pons, E. Murakami, H. Bao, H. M. Steuer, P. A. Furman, M. J. Otto, M. J. Sofia, Bioorg. Med. Chem. Lett. 2010, 20, 4539–4543. [23]J. Jankowska, M. Sobkowski, J. Stawinski, A. Kraszewski, Tetrahedron Lett. 1994, 35, 3355–3358. [24]The phosphorous stereochemistry of compound 17 was determined based on X-ray structures of E7766 and the RpRp isomer (S2) of E7766. [25]S. L. Beaucage, M. H. Caruthers, Tetrahedron Lett. 1981, 22, 1859–1862. [26]The SpSp isomer of E7766 exhibits weaker biological activities than E7766: EC50 (μM) in THP-1 cell-based IRF reporter assay (n = 2) = 10.4 (WT), 16.1 (HAQ), 14.4 (AQ), and 71.4 (REF). [27]H. Shi, J. Wu, Z. J. Chen, C. Chen, Proc. Natl. Acad. Sci. USA 2015, 112, 8947–8952. Manuscript received: January 26, 2021 Accepted manuscript online: January 31, 2021 Version of record online: February 25, 2021 ChemMedChem 2021, 16, 1–5 www.chemmedchem.org 4 © 2021 Wiley-VCH GmbH 1 2 COMMUNICATIONS Locking the bioactive conformation of cyclic dinucleotides led to a new class of stimulator of interferon gene (STING) agonists, macrocycle bridged STING agonists (MBSA), exemplified by E7766. Having entropic advantage upon binding STING proteins, E7766 exhibits enhanced potency and pan- genotypic activity.
Dr. D.-S. Kim*, Dr. A. Endo, Dr. F. G. Fang, Dr. K.-C. Huang, Dr. X. Bao, Dr. H.-w. Choi, Dr. U. Majumder,
Dr. Y. Y. Shen, S. Mathieu, X. Zhu, K. Sanders, Dr. T. Noland, Dr. M.-H. Hao, Dr. Y. Chen, Dr. J. Y. Wang, S.
Yasui, K. TenDyke, J. Wu, C. Ingersoll, K. A. Loiacono, Dr. J. E. Hutz, Dr. N. Sarwar
1 – 5
E7766, a Macrocycle-Bridged Stimulator of Interferon Genes (STING) Agonist with Potent Pan- Genotypic Activity