Discovery of a Novel Series of Potent, Selective, Orally Available, and Brain-Penetrable C1s Inhibitors for Modulation of the Complement Pathway

A novel series of non-amidine-based C1s inhibitors have been explored. Starting from high-throughput screening hit 3, isoquinoline was replaced with 1-aminophthalazine to enhance C1s inhibitory activity while exhibiting good selectivity against other serine proteases. We first disclose a crystal structure of a complex of C1s and a small-molecule inhibitor (4e), which guided structure-based optimization around the S2 and S3 sites to further enhance C1s inhibitory activity by over 300-fold. Improvement of membrane permeability by incorporation of fluorine at the 8-position of 1-aminophthalazine led to identification of (R)-8 as a potent, selective, orally available, and brain-penetrable C1s inhibitor. (R)-8 significantly inhibited membrane attack complex formation induced by human serum in a dose-dependent manner in an in vitro assay system, proving that selective C1s inhibition blocked the classical complement pathway effectively. As a result, (R)-8 emerged as a valuable tool compound for both in vitro and in vivo assessment.


■ INTRODUCTION
The complement system is a component of the innate immune system consisting of three pathways, namely, the classical, lectin, and alternative pathways. Activation of the complement system induces phagocytosis via opsonization by C3b deposition, anaphylatoxin (C3a, C4a, and C5a) production, and C5b-9 membrane attack complex (MAC) formation to eliminate pathogens from the body in healthy individuals. 1 On the other hand, the complement system also plays a role in various diseases. For example, impaired complement activity increases susceptibility to severe infections and also contributes to the development of autoimmune diseases, 2 whereas overactivation of the complement pathway contributes to chronic inflammation and tissue injury. 3 In addition, the components of the complement system are expressed in the brain as well as peripheral tissues, and it has been reported that C1q and downstream complement proteins (e.g., C3) are involved in synapse elimination in the brain, suggesting pathological roles of the complement system in central nervous system diseases as well. 4 Because of the involvement of the complement system in various diseases, it has attracted significant attention as a target for drug development. 5 For instance, a C1 esterase inhibitor is used for patients with hereditary angioedema, and eculizumab, a C5 inhibitor, is applied for the treatment of atypical hemolytic uremic syndrome and paroxysmal nocturnal hemoglobinuria. 6−8 Various types of complement inhibitors are under clinical development for many other complement-mediated indications. 5 C1s is a member of the chymotrypsin-like serine protease family, and it activates the downstream cascade of the classical pathway in the complement system. 9 Thus, in theory, inhibition of C1s could decrease or stop the complement cascade, and therefore, C1s could serve as a novel therapeutic agent for various complement-mediated diseases.
C1s cleaves its substrates C4 and C2 at Arg-Ala and Arg-Lys bonds, respectively. 10 Because one of the key recognition motifs is the ionic interaction between Asp626 in the S1 pocket of C1s and Arg of the substrate, amidines or guanidines that interact with Asp626 can act as S1 recognition motifs for C1s inhibitors. In fact, amidine derivatives, such as 1 and 2, have been reported as potent C1s inhibitors ( Figure 1). 11,12 The docking study of a related analogue of 2 suggested that the amidine moiety interacted with Asp626 in the S1 site. 12 Despite showing potent inhibitory activity against C1s, 1 was reported to have poor in vivo pharmacokinetics (PK) properties likely due to the basic amidine moiety. 11,13 While a follow-up study utilizing a PEGylation strategy led to improvement of PK properties for parenteral dosing, 13 to the best our knowledge, there are no reports on potent and orally available C1s inhibitors. On the other hand, as shown in Figure 2, several orally available serine protease inhibitors bearing a non-amidine S1 binder have been reported. 14,15 Therefore, exploration of non-amidine-based compounds was considered a promising approach to discover potent and orally available C1s inhibitors.
Regarding selectivity against other serine proteases, 2 showed excellent selectivity against C1r, a serine protease located upstream of C1s. However, there is room for improvement on the selectivity against other serine proteases, including the related complement cascade protease (e.g., factor D) and other general serine proteases (e.g., trypsin). 12 We considered that it might be difficult to secure excellent selectivity among serine proteases while an amidine was utilized as an S1 binder against C1s. This motif is used to acquire high affinity via ionic interactions and thus relies on strong basicity. Thus, it was hypothesized that discovery of a non-amidine S1 binder would be crucial to eventually achieve extremely high selectivity.
Recently, non-amidine-based small-molecule C1s inhibitors were disclosed, 16−18 although the detailed properties, such as PK properties and selectivity, are unknown. We have separately performed medicinal chemistry research without such information, and herein, we describe the discovery of a novel series of potent, selective, orally available, and brain-penetrable C1s inhibitors.

■ RESULTS AND DISCUSSION
With the aim of acquiring a promising starting point for pursuing non-amidine-based C1s inhibitors with high lead-likeness, various approaches have been implemented, such as an enzymatic assay-based high-throughput screening (HTS), an affinity selection with a DNA-encoded library, and an NMRbased fragment screening. As a result, there were few tractable and viable hits. Among them, although its C1s inhibitory activity was weak, 3 was identified as a non-amidine-based hit compound (Figure 3), which encouraged us to initiate medicinal chemistry research to identify a novel series of potent, selective, orally available, and brain-penetrable C1s inhibitors.
We constructed a docking model of 3 according to the reported X-ray crystal structure of the apo form of human C1s (PDB ID: 1ELV; 19 Figure 4A). The model indicated that the isoquinoline of 3 should act as an S1 binder to make an ionic interaction with Asp626. To enhance affinity in the S1 site, we developed two strategies: (1) acquisition of an additional interaction and (2) enhancement of the interaction with Asp626. We first considered if acquisition of other interactions at the S1 site could be possible. Aside from an ionic interaction between a basic moiety and an aspartic acid residue in the S1 site, acquisition of a Cl−π interaction between a chlorine atom and a tyrosine residue has been reported as an effective approach to discover orally available serine protease inhibitors, as exemplified by orally active factor Xa inhibitors ( Figure 2). As shown in Figure 4B, Tyr228 is located deeper in the S1 pocket than Asp189, and the factor Xa inhibitor rivaroxaban interacts with Tyr228 via the Cl−π interaction. 14 Because C1s also has the corresponding tyrosine residue (Tyr665), the Cl−π interaction acquisition is an attractive approach to design potent and orally available C1s inhibitors. However, a comparison between factor Xa and C1s revealed a key structural difference in the S1 site. As shown in Figure 4A (left), in the case of C1s, ligands appear unable to access Tyr665 due to steric repulsion with the Ser627 residue (Ala190 residue in the corresponding site of factor Xa), which indicates that it should be difficult to utilize the Cl−π interaction for design of C1s inhibitors. Subsequently, we focused on enhancement of affinity with Asp626. Regarding the interaction with Asp626, 3 shows a monodentate binding mode. Acquisition of an additional interaction through a bidentate binding mode could be expected to enhance affinity to C1s, and the docking model of 3 suggested that there is spatial room for a substituent to acquire an additional interaction with Asp626 ( Figure 4A (right)).    Therefore, we initiated modification of the isoquinoline of 3 to achieve a bidentate interaction with Asp626.
The inhibitory activity of each synthesized compound against human C1s was evaluated. As shown in Table 1, incorporation of an amino moiety at the 1-position of the isoquinoline to furnish 1-aminoisoquinoline was not effective (4b). However, replacement of the 1-aminoisoquinoline with 1-aminophthalazine dramatically enhanced C1s inhibitory activity by around 40fold (4c). Interestingly, removal of the 1-amino moiety of 4c to afford a phthalazine resulted in a remarkable drop in potency (4d) with much weaker potency than 4a, which demonstrated that the combination of the phthalazine scaffold and the amine moiety is crucial to exhibit potent C1s inhibitory activity. On the basis of the calculated pK a value of each compound, it was considered that a certain level of basicity was essential to inhibit C1s (4a vs 4d) and that 4c likely got a bidentate interaction with Asp626 to enhance C1s inhibitory activity. On the other hand, it is difficult to explain the difference in IC 50 values between 4b and 4c only by basicity. To better understand the structural requirement of the 1-aminophthalazine moiety, we obtained crystal structural information of a 1-aminophthalazine 4e in complex with C1s, which, to the best of our knowledge, was the first report of a cocrystal structure between human C1s and a small-molecule ligand. As shown in Figure 5A, the structure revealed that the amino moiety at the 1-position of the phthalazine ring and the nitrogen atom at the 2-position made a bidentate interaction with Asp626 in the S1 pocket of C1s as expected, thereby boosting potency. Furthermore, the nitrogen atom at the 3-position of the phthalazine ring does not undergo hydrogen bonding with the C1s protein directly and possibly plays a role in facilitating improved penetration into the S1 pocket compared to the 1-aminoisoquinoline (4b). In addition to potent C1s inhibitory activity, 4e showed good selectivity against not only C1r, MASP-2, and factor D, serine proteases involved in the complement pathway, but also other serine proteases such as thrombin and trypsin (IC 50 > 30 μM, respectively) ( Figure 5B). As benzamidine shows high basicity (pK a = 11.6), 20 it was expected that these weakly basic nonamidine compounds could be a good starting point to pursue selective and orally available C1s inhibitors. Thus, we identified 4c and 4e as novel attractive lead compounds of selective C1s inhibitors possessing the 1-aminophthalazine moiety as a unique non-amidine-based S1 binder.
On the basis of the binding pose of 4e cocrystallized with C1s, a structure-based approach was performed to further enhance potency. Figure 6A shows the superposition of gigastasin, 21 a peptidic C1s/MASP-2 inhibitor, with the 4e-C1s complex. It has been reported that gigastasin occupies the S2 and S3 site of C1s through interaction with Cys64 (P2) and Lys63 (P3), respectively. In the case of 4e, the amide moiety was located  around the S2 site. Because the S2 site is a lipophilic environment surrounded by His475 and Phe526 ( Figure 6B), 4e is not likely to occupy the S2 site effectively. Furthermore, there are no substituents of 4e located in the S3 site. Therefore, enhancement of affinity in the S2 site and acquisition of a novel interaction in the S3 site were attractive targets for enhancement of C1s inhibitory activity. As shown in Figure 6C, it was hypothesized that the migration of the lipophilic benzene ring of 4e closer to the piperazine motif would enhance the affinity via effective utilization of the S2 site. Regarding the S3 site, an in silico approach was performed to identify novel interactions effectively.
As shown in Table 2, removal of the carbonyl moiety of 4f resulted in the enhancement of C1s inhibitory activity (5a). Replacement of the nitrogen atom attached to the terminal benzene ring with a carbon atom decreased potency (5b). Furthermore, the regioisomer of the piperidine analogue 5c showed dramatically reduced potency, which indicated that the nitrogen atom attached to the 1-aminophthalazine ring plays an important role for C1s inhibition via an electron-donating effect to make the 1-aminophthalazine more basic and/or a conformational effect to occupy the S1 and S2 site exquisitely. As a result, among 5a−5c, phenylpiperazine 5a was considered as the optimum moiety for effective C1s inhibition. Next, we optimized around the terminal benzene ring. A methyl scan indicated that the meta-position was well-tolerated for substituent installation (5d−5f). Regarding this meta-position, incorporation of a fluoro or chloro moiety (5g and 5h) also resulted in comparable potency to 5a, whereas introduction of polar moieties such as methoxy or cyano led to a decrease in potency (5i and 5j), suggesting that lipophilic substituents should be favorable in this area, surrounded by His475 and Phe526. To investigate if the terminal benzene effectively occupies the S2 site, a docking study was performed with 5g. As shown in Figure 7, the terminal benzene ring was accommodated in the S2 site and predicted to make a CH−π interaction with Phe526, which likely contributed to enhance-ment of C1s inhibitory activity. The piperazine moiety of 5g was thought to be a suitable linker to allow the key pharmacophores, the 1-aminophthalazine moiety and the terminal benzene ring, to adopt the optimal spatial configuration. Thus, we found the advanced lead series with submicromolar IC 50 values, and 5g was selected for further chemical modification.
To occupy the S3 site effectively, an in silico-based hot spot prediction was performed. As shown in Figure 8, presumed hot spots were found in the deep region of the S3 site, occupied by the main chain of Lys63 (P3) of gigastasin. To access that region from 5g, we considered that an axial substituent at the 3-position of piperazine should be optimal. As shown in Table 3, among the two enantiomerically pure 3-Me analogues, (S)-6 was found to enhance potency over 5g, which confirmed the validity of the approach for the S3 site. Furthermore, the main chain of Gly656 of C1s was found in close proximity to the predicted hot spots, suggesting that C1s inhibitory activity could be enhanced by acquisition of hydrogen bonding in the S3 site while keeping the active conformation of 5g. To obtain an additional interaction, incorporation of a heteroaromatic ring possessing nitrogen and/ or oxygen atoms should be a suitable approach, and thus, we designed the isoxazole derivative (R)-7 ( Figure 9A). The predicted binding mode of (R)-7 ( Figure 9B) supported the hypothesis, demonstrating that the oxygen and/or nitrogen atom of the isoxazole ring could interact with the NH of Gly656. In addition, the methyl moiety at the 5-position of the isoxazole ring exists in close proximity to the C1s wall, which can be expected to acquire an additional lipophilic interaction. As a synthetic strategy, we applied photoredox chemistry with silicon amine protocol (SLAP) reagents to furnish the key substitutedpiperazine moiety efficiently ( Figure 9C). 22 As shown in Table 4, introduction of the 5-methyl-isoxazole ring dramatically enhanced C1s inhibitory activity ((R)-7), indicating that the 5-methyl-isoxazole moiety effectively occupied the S3 site. On the other hand, it was revealed that membrane permeability of (R)-7 was suboptimal ( Table 4). As the 1-aminophtalazine moiety, possessing two hydrogen bond    donors, was thought to contribute to poor permeability, a fluorine atom was incorporated at the 8-position of the aminophthalazine ring to cap one of the hydrogen bond donors via intramolecular hydrogen bonding. As a result, (R)-8 showed dramatically improved permeability (Table 4) while showing potent C1s inhibitory activity with an IC 50 value of 36 nM.
The properties of (R)-8 are shown in Figure 10A. Compound (R)-8 showed potent inhibitory activity against both human and mouse C1s. As for selectivity against other serine proteases, (R)-8 showed excellent selectivity over C1r, MASP2, factor D, thrombin, trypsin, kallikrein, and plasmin, demonstrating that there was no compromise on selectivity in the course of optimization from the lead 4e. As shown in Figure 10B, (R)-8 significantly inhibited MAC formation induced by human serum in a dose-dependent manner, which proved that selective inhibition of C1s could effectively block the classical complement pathway. As indicated in Figure 10C, (R)-8 showed plasma exposure after oral administration in mice (10,30, and 100 mg/kg) in a dose-dependent manner. Furthermore, (R)-8 was also distributed into the brain ( Figure 10D) after oral administration. Collectively, (R)-8 could emerge as a valuable tool compound for in vivo assessment by modulating the classical complement pathway.
In summary, we have explored non-amidine C1s inhibitors and discovered 1-aminophtalazine as a unique S1 binder to C1s. The following structure-based approach based on the crystal structure of 4e−C1s led to the identification of a novel series of potent and selective C1s inhibitors. Compound (R)-8, showing potent C1s inhibitory activity with excellent selectivity against other serine proteases, was shown to effectively inhibit MAC formation in the in vitro assay system. Furthermore, (R)-8 was found to be orally available and brain-penetrable in mice. Taken together, (R)-8 was identified as a valuable in vitro and in vivo tool compound of C1s inhibitors. We believe that this approach, starting from the investigation of a novel S1 binder, can be utilized for identification of other orally available selective serine protease inhibitors.

■ CHEMISTRY
As shown in Scheme 1, 4a and 4d were synthesized by the Pdcatalyzed coupling reaction of 9a and 9d with 10, respectively. Scheme 2 describes the synthesis of 4b and 4c. 9b and 9c were treated with 2,4-dimethoxybenzylamine to afford 11b and 11c followed by a coupling reaction with 10 and deprotection to give the target compounds 4b and 4c, respectively. Scheme 3 illustrates the synthesis of 4e, 4f, and 5a−j. 11c was subjected to a coupling reaction with tert-butyl piperazine-1carboxylate to afford 12, which was deprotected to give the piperazine dihydrochloride salt 13. Subsequent acylation and deprotection afforded the target compounds 4e and 4f. A coupling reaction of 13 with 1-chloro-3-iodobenzene followed by deprotection gave the target compound 5h. 11c was coupled with the corresponding substituted phenyl piperazines or 4phenyl piperidine followed by deprotection to give the target compounds 5a, 5b, 5d−g, 5i, and 5j. A coupling reaction of 11c with boronic ester gave 14, which was deprotected to afford tetrahydropyridine 15. The subsequent coupling reaction, deprotection, and hydrogenation gave the target compound 5c.   were coupled with 11c followed by deprotection to afford the target compounds (R)-6 and (S)-6, respectively.
The key intermediate, isoxazole-substituted piperazine 20, was prepared in a single step from the SLAP reagent 19. 22 The reaction worked well in gram-scale synthesis. A coupling reaction of 20   with 1-bromo-3-fluorobenzene afforded 21, which was converted to Boc-protected piperazine 22. Subsequent deprotection gave the precursor 23, which was optically resolved to afford (R)-23. A subsequent coupling reaction with 11c and deprotection gave the target compound (R)-7.

■ CONCLUSIONS
To discover potent, selective, orally available, and brainpenetrable C1s inhibitors, we explored a series of non-amidine compounds based on HTS hit 3. On the basis of the predicted binding pose of 3, isoquinoline, the presumed S1 binder, was replaced with 1-aminophthalazine to enhance C1s inhibitory activity by around 10-fold while exhibiting good selectivity against not only C1r, MASP-2, and factor D, serine proteases involved in the complement pathway, but also other serine proteases such as trypsin and thrombin. The crystal structure of a complex of C1s and a small-molecule inhibitor was solved, which demonstrated that the 1-aminophthalazine moiety of 4e fits tightly in the S1 site and suggested further opportunities to enhance C1s inhibitory activity by structure-based chemical modification around the S2 and S3 site. We successfully enhanced C1s inhibitory activity by acquisition of an additional interaction in the S2 pocket via plausible CH−π interactions with Phe526. Furthermore, in silico-based hot spot prediction aided our investigation into suitable substituents to introduce in the S3 pocket to obtain additional interactions. As a result, the 5methyl-isoxazole analogue (R)-7, designed to interact with Gly656 and the nearby wall in the S3 site, further boosted C1s inhibitory activity showing a single-digit nanomolar IC 50 value. Although membrane permeability of (R)-7 was suboptimal, introduction of a fluorine atom at the 8-position of the 1aminophthalazine significantly improved the permeability, which led to identification of (R)-8 as a potent, selective, orally available, and brain-penetrable C1s inhibitor. Compound (R)-8 significantly inhibited MAC formation induced by human serum in a dose-dependent manner in the in vitro assay system, proving that selective inhibition of C1s blocked the classical complement pathway effectively. Oral administration of (R)-8 in mice showed reasonable exposure both in plasma and the brain, demonstrating that (R)-8 could work as a valuable tool compound for both in vitro and in vivo assessment by modulating the classical complement pathway. ■ EXPERIMENTAL SECTION Chemistry. 1 H NMR and 13 C NMR spectra were recorded on a Bruker AVANCE III (300 MHz) or a Bruker Advance III plus (400 MHz) spectrometer. Chemical shifts are given in parts per million (ppm) downfield from tetramethylsilane (δ) as the internal standard in a deuterated solvent, and coupling constants (J) are in Hertz (Hz). Data are reported as follows: chemical shift, integration, multiplicity (s = singlet, d = doublet, t = triplet, q = quartet, m = multiplet, dd = doublet of doublets, td = triplet of doublets, and br s = broad singlet), and coupling constants. Unless otherwise noted, reagents and solvents were obtained from commercial sources and used without further purification. Thin-layer chromatography (TLC) was performed on silica gel 60F254 plates (Merck) or NH TLC plates (Fuji Silysia Chemical Ltd.). Chromatographic purification was performed on a Purif-Pack (SI or NH, Fuji Silysia Chemical Ltd.) or on a UNIVERSAL column (silica or amino, YAMAZEN Corporation). LC−MS analysis was performed on a Shimadzu liquid chromatography−mass spectrometer system, operating in the APCI (+ or −) or ESI (+ or −) ionization mode. Analytes were eluted using a linear gradient of 0.05% TFA-containing water/acetonitrile or a 5 mM ammonium acetate-containing water/acetonitrile mobile phase and detected at 220 nm. The purities of compounds submitted for biological evaluation were >95% as determined by analytical HPLC unless otherwise noted. Analytical HPLC was performed with a corona charged aerosol detector (CAD), a nanoquantity analyte detector (NQAD), or a photodiode array detector. The column was a Capcell Pak C18AQ (50 mm × 3.0 mm I.D., Shiseido, Japan) or L-column 2 ODS (30 mm × 2.0 mm I.D., CERI, Japan) with a temperature of 50°C and a flow rate of 0.5 mL/min. Mobile phases A and B under a neutral condition were a mixture of 50 mmol/L ammonium acetate, water, and acetonitrile (1:8:1, v/v/v) and a mixture of 50 mmol/L ammonium acetate and acetonitrile (1:9, v/ v), respectively. The ratio of the mobile phase B was increased linearly from 5 to 95% over 3 min and 95% over the next 1 min. Mobile phases A and B under an acidic condition were a mixture of 0.2% formic acid in 10 mmol/L ammonium formate and 0.2% formic acid in acetonitrile, respectively. The ratio of the mobile phase B was increased linearly from 14% to 86% over 3 min and 86% over the next 1 min. Elemental analyses were performed by Sumika Chemical Analysis Service, and all results were within ±0.4% of the calculated values. Yields are not optimized.
PAMPA. A PAMPA Evolution system (pION) was applied for the assay. After incubating the test compound (10 μmol/L) for 3 h at 25°C, the permeation coefficient value was calculated with determining the test compound by LC/MS/MS.
Evaluation of Membrane Permeability with Human Multidrug Resistance 1 (MDR-1) Expressing Cells. The transcellular transport study was performed as reported previously. 31 In brief, the cells were grown in an HTS Transwell 96-well permeable support (pore size of 0.4 μm, 0.143 cm 2 surface area) with a polyethylene terephthalate membrane (Corning Life Sciences, Lowell, MA, USA) at a density of 1.125 × 10 5 cells/well. The cells were preincubated with M199 at 37°C for 30 min. Subsequently, transcellular transport was initiated by the addition of M199 to apical compartments containing 1 μmol/L test compounds and terminated by the removal of each assay plate after 2 h. The aliquots in the opposite compartments were subjected to measurement for compound concentration by LC−MS/ MS. Permeability was calculated using the permeated compound concentration.
Animal Experiments. All animal experiments were performed in compliance with the Guidelines for the Care and Use of Laboratory Animals of Takeda Pharmaceutical Company Ltd.
Pharmacokinetics Analysis in Mice. Male C57BL/6J mice were purchased from The Jackson Laboratory Japan, Inc. (Kanagawa, Japan). All animals were with free access to food and water prior to the treatment and were healthy throughout the experimental period. The compound was dissolved in 10% DMSO/10% Cremophor EL/20% PEG400/60% 0.1 mol/L citric acid solution and administered orally at doses of 10, 30, or 100 mg/kg. Blood samples were collected into heparinized tubes at the designated time points (0. 5, 1, 2, 4, 8, and 24 h postdosing), and brains were also collected at 2 and 24 h postdosing. Plasma was separated from the blood samples by centrifugation. The plasma and brain concentrations were quantified with high-performance liquid chromatography−tandem mass spectrometry. The lower limit of quantitation was 1 ng/mL for plasma and 5 ng/mL for brains.

* sı Supporting Information
The Supporting Information is available free of charge at https://pubs.acs.org/doi/ 10.1021