Discovery of GLPG3667, a Selective ATP Competitive Tyrosine Kinase 2 Inhibitor for the Treatment of Autoimmune Diseases

Tyrosine kinase 2 (TYK2) mediates cytokine signaling through type 1 interferon, interleukin (IL)-12/IL-23, and the IL-10 family. There appears to be an association between TYK2 genetic variants and inflammatory conditions, and clinical evidence suggests that selective inhibition of TYK2 could produce a unique therapeutic profile. Here, we describe the discovery of compound 9 (GLPG3667), a reversible and selective TYK2 adenosine triphosphate competitive inhibitor in development for the treatment of inflammatory and autoimmune diseases. The preclinical pharmacokinetic profile was favorable, and TYK2 selectivity was confirmed in peripheral blood mononuclear cells and whole blood assays. Dermal ear inflammation was reduced in an IL-23-induced in vivo mouse model of psoriasis. GLPG3667 also completed a phase 1b study (NCT04594928) in patients with moderate-to-severe psoriasis where clinical effect was shown within the 4 weeks of treatment and it is now in phase 2 trials for the treatment of dermatomyositis (NCT05695950) and systemic lupus erythematosus (NCT05856448).


■ INTRODUCTION
Janus kinases (JAKs) are a class of nonreceptor cytoplasmic tyrosine kinases involved in the signaling of more than 60 cytokines and growth factors.The four members of this class (JAK1, JAK2, JAK3, and tyrosine kinase 2 [TYK2]) 1 function in pairs to phosphorylate signal transducer and activator of transcription (STAT) proteins when activated by a ligand. 2 Activated STATs dimerize and migrate to the nucleus where they regulate gene transcription. 3uring the past decade, a number of JAK inhibitors such as 1 (tofacitinib), 2 (baricitinib), 3 (ruxolitinib), 4 (filgotinib), and 5 (upadacitinib) have been approved to treat patients with inflammatory diseases. 4All these molecules inhibit JAK1 (Figure 1), which is considered the key target owing to its major role in JAK/STAT-dependent signaling. 5Some of the molecules also target other JAKs, leading to a broader spectrum of activity but with potentially some liabilities. 5−10 IL-12 and IL-23 are cytokines that play an important role in various inflammatory diseases. 11−19 The remarkable efficacy of these antibodies 20,21 has generated interest in targeting TYK2 owing to its role in IL-12 and IL-23 signaling. 22ype I IFN is also a key driver in some inflammatory and autoimmune diseases such as systemic lupus erythematosus (SLE) and dermatomyositis.Anifrolumab is a monoclonal antibody that blocks IFNα/β signaling and has recently been approved for the treatment of patients with SLE. 23The TYK2selective inhibitor 6 deucravacitinib (BMS-986165) has displayed good efficacy in a phase 2 trial in patients with SLE, 24 linking the interest in type I IFN and TYK2 in this class of disease.−28 In addition, several genetic studies have identified associations between TYK2 gene single nucleotide polymorphisms and an increased or decreased risk of developing inflammatory or autoimmune diseases. 8,29,30ata from clinical trials of two small-molecule TYK2 inhibitors (Figure 2) have recently been published.Deucravacitinib 6 is an allosteric inhibitor that selectively blocks TYK2 kinase activity by stabilizing its pseudokinase domain. 31fficacy of deucravacitinib has been demonstrated in patients with psoriasis, 32 psoriatic arthritis, 33 and SLE, 34 with limited adverse effects, highlighting the potential benefits of this class for the treatment of inflammatory diseases.Ropsacitinib 7 (PF-06826647) 35 is an adenosine triphosphate (ATP) mimic TYK2-selective inhibitor that displayed efficacy in a phase 2b psoriasis trial. 36More recently, new TYK2 inhibitors have been described or announced, notably, 8 (zasocitinib)�an allosteric TYK2-selective inhibitor�which was described as effective in a small phase 1b study in psoriasis patients 37 and in a phase 2b study, 38 and VTX958, 39 which is also an allosteric inhibitor that just completed a phase 1 study in healthy volunteers. 39n a manuscript currently in development, we described a 1H-imidazo [4,5-c]pyridine series capable of dual JAK1/TYK2 inhibition and selective against JAK2 and JAK3. 40In the present article, we describe our efforts to progress to a TYK2selective inhibitor with reduced potency for JAK1 and suitable for clinical progression.These efforts culminated in the discovery of 9 (GLPG3667), a selective TYK2 inhibitor targeting the catalytic domain, which is currently in clinical development for the treatment of inflammatory diseases (Figure 2).

■ RESULTS AND DISCUSSION
Scaffold-Hopping Exercise to Improve JAK1/TYK2 Selectivity.To initiate this project, a 3H-imidazo[4,5- Examples of approved selective and nonselective JAK1 inhibitors. 4 (Note: This is not an exhaustive list of all licensed JAK inhibitors targeting JAK1.) Figure 2. Investigational tyrosine kinase 2 inhibitors. 31,35,37]pyridine series was designed and synthesized as part of a scaffold-hopping exercise.A comparison of this scaffold with the 1H-imidazo [4,5-c]pyridine series, both bearing a carbonylamino substituent at the top, clearly indicated that the 3Himidazo [4,5-b]pyridine scaffold possessed a more favorable TYK2 selectivity profile (Figure 3A).Comparison of matched pairs 10 40 and 11, both possessing a cyclopropane carboxamide at the top (positions C4 and C7, respectively), showed a 21-fold improvement in JAK1/TYK2 selectivity (Figure 3B).Docking of 10 and 11 in JAK1 and TYK2 structures showed very similar binding modes in both proteins (Figure 4A).In the kinase hinge region, the imidazole nitrogen established a hydrogen bond with the backbone NH of Val981 (TYK2 numbering), while the hydrogen borne by the carbon between both nitrogen atoms formed a weak C−H−O interaction with the backbone carbonyl of Glu979.The amide NH was also involved in a hydrogen bond to the carbonyl oxygen of Val981 in the hinge region.The nitrile group pointed toward the Glyrich loop, but without any clear interaction.The ethyl substituent filled a hydrophobic pocket (lined by Leu1030, Gly1040, and the backbone of Asn1028; residues not shown for clarity).Quantum mechanics minimization of 10 and 11  docking poses and overlay of the resulting structures showed that the different position of the nitrogen atom in the bicyclic core induced a shift in the location of the aryl and cyano groups that point toward the Gly-rich loop (Figure 4B).Engagement of the Gly-rich loop has often been cited as a way of modulating the selectivity profile within the JAK family, 35     .All compounds were potent TYK2 inhibitors (half maximal inhibitory concentration [IC 50 ] < 1 nM).Compared with 11, the urea 12 maintained good selectivity for TYK2 versus JAK1 and showed improved lipophilic efficiency (LipE), an important parameter in lead optimization; 41 however, there was a major 37-fold selectivity with the pyrimidin-4-amine 13.All compounds displayed very good selectivity (>30×) for TYK2 against JAK3, but low selectivity against JAK2 (<5×).A lower biochemical selectivity against JAK2, though undesirable, was less concerning because JAK2 inhibitors have been shown to undergo larger biochemical to cellular potency shifts than JAK1 and TYK2 inhibitors. 42Different hypotheses were developed to explain this disconnect, i.e., the low ATP K m of JAK2 compared with other JAK family members 43 or a difference in contribution to signaling between the JAK family members. 44All compounds had a high in vitro unbound intrinsic clearance 45,46 of >100 L/h/kg in mouse liver microsomes.
Docking of 13 in TYK2 and JAK1 structures suggested that the aminopyrimidine group points toward the solvent along the hinge region, with the amino group establishing a hydrogen bond with the backbone oxygen of Pro982 (Figure 5).These interactions are common to both TYK2 and JAK1 and do not account for the observed TYK2 selectivity.However, molecular dynamics simulations of this compound in both proteins showed that in TYK2, the amino group can also interact with the oxygen atom of Tyr980 (Phe958 in JAK1).This hydrogen bond was not stable but appeared frequently during the simulation, suggesting a weak interaction in TYK2 but not in JAK1.
SAR Exploration at the C5 Position.The next focus was the C5 position of the 3H-imidazo [4,5-b]pyridine scaffold, retaining the pyrimidin-4-amine at the C7 position in light of its improved TYK2 selectivity profile (Table 2).A key strategy was to explore the use of alkyl tails in this position to reduce the number of aromatic rings to increase saturation, reported as an approach to improving clinical success. 48Compounds 14 and 15 led to a reduction in potency for TYK2 compared with 13, and insufficient improvements in in vitro clearance despite lower lipophilicity.TYK2 potency was regained with 16 but JAK1/TYK2 selectivity was reduced approximately 3-fold compared with 15 and the metabolic stability in mouse microsomes did not improve, likely owing to the relatively high lipophilicity.The replacement of the methylamino linker with an NH linker in 17 provided a major improvement in terms of in vitro clearance, while retaining potency and selectivity toward TYK2; however, in vitro clearance in mouse microsomes was still above 10 L/h/kg.The O-linker in 18 also provided an improvement in in vitro clearance, though not to the same extent as the NH linker.Efforts were also made to improve the compounds with an aromatic group at the C5 position.The introduction of a pyridine moiety in 19 led to a considerable reduction in in vitro clearance, while conserving potency and selectivity.Replacing the ethyl group in the aromatic tail of 19 with a methyl group in 20 led to approximately a 2-fold reduction in in vitro clearance and an improvement in JAK2/TYK2 selectivity.While the in vitro clearance was still rather high, these improvements clearly demonstrated that modification of the aromatic tail had potential to improve metabolic stability.Utilizing the lessons learned from matched pairs 17 and 18, the linker connecting the aromatic tail to the core was modified in 21 and 22 to further improve in vitro clearance; the O-linked 22 had an in vitro clearance of below 4.1 L/h/kg.Despite some loss of JAK1/TYK2 selectivity, the improved metabolic stability of 22 was considered to provide a balanced profile.The less potent O-linked 23 also showed good metabolic stability, with an in vitro clearance value in mouse liver microsomes <10 L/h/kg.

Journal of Medicinal Chemistry pubs.acs.org/jmc
Drug Annotation The compounds described in Table 2 indicated that aromatic groups at the C5 position conferred superior selectivity and, in some cases, potency compared with their alkyl counterparts.Analysis of a set of compounds bearing the aminopyrimidine group at the C7 position reinforced these conclusions (Figure S1, Supporting Information).The compounds bearing alkyl substituents had lower potency and selectivity compared with the compounds bearing aromatic tails.
The improved in vitro clearance of compounds 17, 18, 21, and 22 in mouse liver microsomes appears to have stemmed from modifications of the linker.These modifications were further analyzed by considering the lipophilicity.The inherent stability of a set of compounds bearing an aminopyrimidine group at the C7 position was assessed, using the lipophilic metabolic efficiency metric (LipMetE) 49 to remove the contribution of lipophilicity (Figure 6).This analysis indicated an NH-or an O-linker was more stable than an NMe-linker.Owing to the additional hydrogen bonding donor of the NH linker, a property which should be carefully considered for compound progression, 50 the O-linked 22 was selected for further optimization.
Advanced SAR Exploration at the C7 Position.Two subseries of compounds with carboxamide or aniline groups connected to the C7 position protruding toward the solvent were designed (Table 3) to improve on the thermodynamic solubility of 22, which was low (<100 μg/mL) in all media.
In the carboxamide subseries, potency and LipE lost with 24 were regained by replacing the benzene ring with pyridine in 25; however, both compounds had low JAK1/TYK2 selectivity (<10×).The morpholine in 26 and the N-methylpiperazine in 27 all increased potency and JAK1/TYK2 selectivity by at least 3-fold compared with 25.Compounds 24−27 all had good in vitro metabolic stability in mouse liver microsomes around 10 mL/h/kg.The chirally pure 28 showed exquisite JAK1/TYK2 selectivity (63×), but metabolic stability was compromised.In the aniline subseries, 29 displayed good potency and JAK1/TYK2 selectivity.The replacement of the pyridine with a pyridazine in 9 improved selectivity 2-fold, with a slight increase in potency and LipE.As with the carboxamides, the introduction of basic groups was generally associated with an increase in selectivity.30 and 31 showed (41×) and (78×) JAK1/TYK2 selectivity, respectively, but metabolic stability was eroded with 31.
Most of the compounds shown in Table 3 had an attractive profile with good potency, LipE, selectivity, and metabolic stability (in mouse liver microsomes).Some compounds were therefore subjected to more advanced absorption, distribution, metabolism, excretion, and toxicity profiling (Table 4).In general, permeability was moderate to good but efflux ratios were high, indicating that the compounds could be P-glycoprotein substrates. 51The carboxamide 27 with basic bearing group had a particularly high efflux ratio.Neutral carboxamide 26 had better solubility than 22, with moderate solubility in acidic media.Basic groups in 27 and 30 further improved solubility in acidic media, as expected.9 showed a considerable 10-fold improvement in solubility in fasted state simulated gastric fluid compared with 22.The compounds had acceptable in vitro metabolic stability in mouse and human liver microsomes.Human ether-a-go-go-related gene (hERG) inhibition was low for all compounds except 30, which bears a basic piperazine (68% inhibition at 10 μM).
In Vivo Pharmacokinetic Profiles of Advanced Compounds.Compounds 26, 27, and 9 were progressed to mouse pharmacokinetic (PK) experiments (Table 5).Compounds 26 and 27 had moderate in vivo clearance.The volume of distribution was moderate for all compounds, and the basic group of 27 did not lead to a high volume of distribution.The combination of these factors led to half-lives of less than 1 h for all compounds.Bioavailability of 26 and 27 was low to moderate.9 displayed lower clearance, with a volume of distribution of ∼2 L/kg, resulting in a good half-life (1.6 h) and improved bioavailability.
Compounds 26, 27, and 9 were tested in vitro in cellular and whole blood assays to assess potency and selectivity against the JAK family members in more relevant systems (Table 6).In the cellular assays performed on fresh human peripheral blood mononuclear cells (PBMCs), the three compounds showed dose-dependent inhibition of IFNα-induced STAT1 phosphorylation (JAK1/TYK2-dependent), IL-2-induced STAT5 phosphorylation (JAK1/JAK3-dependent) and granulocyte macrophage colony-stimulating factor (GM-CSF)-induced STAT5 phosphorylation (JAK2-dependent) and displayed more than 10-fold selectivity when comparing the assays on TYK2-dependent pathways (e.g., IFNα-driven assays) with the assays on JAK1/JAK3-and JAK2-dependent assays.In human whole blood assays, the potency of the three compounds was highest in the TYK2-dependent IFNα-induced STAT1 phosphorylation assay compared with the IL-6-induced  STAT1 phosphorylation (JAK1-dependent), GM-CSF-induced STAT5 phosphorylation, and IL-2-induced STAT5 phosphorylation assays.Comparison with data obtained with a JAK1-selective inhibitor (filgotinib) and the TYK2-selective inhibitor deucravacitinib further validate the TYK2 selectivity of compound 9. Overall, these results indicate that compounds 26, 27, and 9 are selective TYK2 inhibitors in human PBMC and whole blood assays, further reinforcing the data obtained with the biochemical assays.Of note, potency of all compounds decreased in whole blood assays compared with PBMC.This is likely due to the absence of serum in the PBMC assays that delivers free fraction activity, while in whole blood, part of the molecules are trapped by plasma protein as frequently documented. 52,53hese three compounds were subsequently tested in an in vitro micronucleus test in the human lymphoblastoid TK6 cell line in the presence or absence of S9 fraction.Compounds 26 and 27 induced structural DNA damage, in contrast to compound 9 which did not and was selected for further profiling.
In Vitro Profiling of 9. Due to its good global profile, 9 was further tested for general kinase selectivity against a panel of 365 kinases (performed at Eurofins Discovery, Cerep, France), at a concentration of 1 μM (Figure S2, Supporting Information).The compound only hit 21 non-JAK kinases with at least 50% inhibition.Biochemical IC 50 was <100 nM for three kinases (AURKB, FLT3, and FLT4), and cellular IC 50 values (generated at Reaction Biology, Freiburg, Germany) were in the micromolar range for AURKB and FLT3 (FLT4 was not tested) (Table S1, Supporting Information).No offtarget activity was reported in the diversity panel (performed at Eurofins Discovery, Cerep, France) at 10 μM for any of the 97 binding, enzyme, and uptake assays.
Additional Profiling of 9. Further in vitro profiling is listed in Table 7. Compound 9 was highly stable in vitro and in mouse, rat, dog, and human liver microsomes and hepatocytes.It was highly permeable in Caco-2 cells, with a low efflux ratio.The IC 50 for cytochrome P450 (CYP) inhibition was >33 μM for each of the CYP isoenzymes tested and there was no evidence of time-dependent inhibition of CYP3A4.Additionally, there was no meaningful mRNA induction of CYP3A4 in cryopreserved human hepatocytes (2.2% vs rifampicin at 10 μM).The mutagenic potential of 9 was also investigated with a bacterial reverse mutation test (Ames test) using three different bacterial strains, which was negative.As reported above, compound 9 did not induce structural DNA damage in an in vitro micronucleus test in the human lymphoblastoid TK6 cell line in the presence or absence of S9 fraction.9 was highly stable in plasma of mouse, rat, dog, and human, and chemically stable in solutions of pH 1.2, 5.0, 7.4, and 9.0 for up  to 24 h, with >80% of compound remaining at 2 and 24 h (Table S2, Supporting Information).The profiling of 9 did not reveal liabilities and the compound passed the preclinical toxicity studies to advance to clinical studies.Additionally, compound 9 was docked and its binding mode was compared to that of the earlier compound 13 in the SAR campaign (Figure 7).The resulting model suggests that in the kinase hinge region, the imidazole nitrogen establishes a hydrogen bond with the backbone NH of Val981 (human TYK2 numbering), as for compound 13.The nitrile group points toward the Gly-rich loop, but without any clear interaction.The methyl substituent (ethyl in compound 13) fills a hydrophobic pocket (lined by Leu1030, Gly1040, and the backbone of Asn1028; residues not shown for clarity).The NH linker could also be involved in a hydrogen bond to the carbonyl oxygen of Val981 in the hinge region (as is the amide NH in compound 13).The morpholine moiety points toward the solvent-exposed region.Two water molecules might bridge the oxygen of the morpholine moiety to the backbone of Asp988, and its side chain to the ether linker of the ligand.These water-mediated interactions could not be present in compound 13, due to the absence of acceptor atoms in the substituents.
In Vivo Profiling of 9. PK profiles were further explored in mouse, rat, and dog (Table 8).Compound 9 was characterized by low clearance, moderate volume of distribution, moderate half-life, rapid absorption following oral administration, and good oral bioavailability.In dogs, a modified formulation for oral administration was used due to low solubility and allowed to reach a bioavailability of 35%, which was considered suitable for progression.
Efficacy of 9 in a Murine Model of IL-23-Induced Psoriasis.As shown by Gerstenberger et al. (2020), 35 TYK2 inhibitors display limited selectivity for TYK2 in mouse compared with human due to decreased inhibition of TYK2 and increased inhibition of JAK1.For that reason, we only assayed our compound in a unique mouse model to verify that it displayed pharmacological effects when given orally to the animals.Compound 9 was evaluated at three doses (once daily [q.d.] oral administration at 3, 10, and 30 mg/kg) in the mouse model of IL-23-induced psoriasis. 54,55The exposure increased dose-proportionally between the dose range (Figure 8) allowing a limited coverage of the TYK2-dependent IFNα pathway and an absence of coverage of the JAK1-dependent IL-6 pathway for the two higher doses used.
No dose-dependency was observed, and the effect of 9 was similar to the effect of a TYK2 inhibitor which was used as a positive control 56 for preventing ear pinna thickening induced   by intradermal injections of IL-23 (Figure 9A).While phosphorylated STAT3 (pSTAT3) is rarely expressed in healthy ear skin, IL-23 triggering led to a marked induction of STAT3 phosphorylation in keratinocytes and in some immune cells infiltrated in the dermis which is in line with the fact that STAT3 is the main target of IL-23.In the same experiment, compound 9 and the positive control TYK2 inhibitor prevented STAT3 phosphorylation in the epidermis and dermis at all tested doses, confirming target engagement and effect of the compound at the three doses used (Figure 9B, 9C).The absence of a dose−response relationship is likely due to the involvement of other pathways not impacted by TYK2 inhibition, and to the fact that mass balance studies in mice showed that compound 9 was highly concentrated in the skin (data not shown), suggesting that despite unfavorable PK, it displayed a strong effect in this psoriasis-like model.In addition, it is noteworthy that Leit et al. ( 2023) 52 observed the same partial effects with the TYK2 inhibitor TAK-279, while this compound is more potent than compound 9 and displayed more favorable PK.Other compounds from the same series also gave a partial effect, strongly suggesting that in this model TYK2 inhibitors may not be able to fully counteract all the effects of IL-23.
We also looked at the presence of neutrophils in the ears of mice. Figure 9D shows that IL-23 injection induces a massive infiltration of neutrophils essentially in the dermis.Compound 9 was able to prevent this infiltration with no apparent a dose− response relationship (Figure 9E).Taken together, these data define a minimal effective dose of 3 mg/kg q.d. and define 9 as a selective TYK2 inhibitor that blocks the IL-23 pathway in vivo. 58uman Dose Prediction and Clinical Efficacy of Compound 9. Based on pharmacokinetics in mouse, rat, and dog, body weight allometry (rule of exponents with correction for maximum lifespan potency) predicted human blood clearance of 0.17 L/h/kg.Human volume of distribution at steady state (V ss ) was estimated at 1.85 L/kg using the geometric mean of V ss in mouse, rat, and dog.Human bioavailability was assumed to be 50% minimum.The human target concentration was based on coverage of human whole blood assay IFNα/pSTAT1 IC 50 (623 nM) for 6 h in blood.Based on these assumptions, the human efficacious dose of compound 9 was predicted to be 100 mg, which is in line with efficacy signals seen with GLPG3667 at 150 mg q.d. in a phase 1b study of GLPG3667 in patients with moderate-tosevere psoriasis, 59 in which compound 9 given orally at a dose of 150 mg q.d.displayed efficacy after only 4 weeks of treatment with PASI 75 comparable with what was observed with deucravacitinib without any safety alert. 57

■ CHEMISTRY
The preparation of the core 37, required for synthesis of the compounds listed in  methylamine yielded a mixture of regioisomers 35a/35b, which were reduced using sodium dithionite.The crude mixture of regioisomers 36a/36b was cyclized by condensation with triethyl orthoformate.The scaffold 37 with the desired regioselectivity was obtained after silica flash chromatography (Scheme 1).
The preparation of the core 45, required for synthesis of the compounds listed in trifluoroacetic acid (TFA).Halogenated pyridine 43 was subjected to aromatic nucleophilic substitution with 2 M methyl amine in THF to yield 44, which was cyclized using trimethyl orthoformate and formic acid to yield core 45 (Scheme 3).
Finally, compounds 27 and 28 were obtained by late-stage amide coupling with 54, which was obtained via Buchwald− Hartwig cross-coupling of intermediate 51b with methyl 6aminonicotinate 52, followed by LiI-promoted saponification (Scheme 7). 56CONCLUSIONS TYK2-selective compounds belonging to an 3H-imidazo [4,5b]pyridine series were obtained by scaffold hopping from 3Himidazo[4,5-c]pyridine-based compounds possessing dual activity against JAK1/TYK2.During SAR optimization, potency and selectivity toward TYK2 was increased.Modification of the linker between the core and an aromatic subunit improved metabolic stability, and further refinements, produced through derivatization of a vector protruding toward the solvent front, led to the discovery of 9 (GLPG3667).
TYK2 selectivity for GLPG3667 was confirmed in PBMC and whole blood assays.GLPG3667 also reduced dermal ear inflammation in a mouse IL-23-induced model of psoriasis.The PK profile in preclinical species was favorable.A phase 1b study of GLPG3667 in patients with moderate-to-severe psoriasis is completed and reported clinical efficacy, 59 and GLPG3667 is now in phase 2 trials for the treatment of dermatomyositis (NCT05695950) and SLE (NCT05856448).
High-resolution mass spectrometry (HRMS) samples were prepared at 0.1 mg/mL concentration in MeCN either by dilution of DMSO stock solution (10 mM) or by dissolving solid compound.HRMS data were recorded by eluting samples using an Agilent 1260 Infinity II ultra-HPLC system on C18 column (Waters Acquity BEH C18, 2.1 × 150 mm, 1.7 μm particle size) linked to the Agilent 6540 UHD Q-Tof mass spectrometer.LC method was generic using 0.1% formic acid in water and 0.1% formic acid in MeCN from 3% to 97% of organic modifier over 5 min (flow rate: 0.5 mL/min).Ionization mode used was ESI positive with mass range from 100 to 3200 Da.The melting points (mp) were taken in open capillaries on a Mettler Toledo MP50 apparatus and are uncorrected.
Molecular Dynamics Simulations.Molecular dynamics simulations were run using the Desmond package included in the Schrodinger suite.Ligand geometries were obtained from the docking studies and parametrized using OPLS3 force field.The complexes generated were initially solvated in a cubic box with SPC water molecules, leaving at least 10 Å between the solute atoms and the border of the box.The systems were then neutralized with seven Na + counterions, and NaCl salt concentration of 0.15 M was added.The systems generated were equilibrated and gradually heated from 0 to 300 K using the Desmond default equilibration protocol.After the equilibration, the systems were subjected to 50 ns MD simulations in the NPT ensemble at 1.01325 bar (by Martyna−Tobias−Klein barostat) and 300 K (by Nose−Hoover chain thermostat), setting a cutoff of 9 Å for the short-range nonbonded interactions.Trajectories were saved every 50 ps.Analysis of the simulation was based on protein and ligand RMSD values, hydrogen bond occupancy along 50 ns simulation, and visual inspection of the binding geometries.
Quantum Mechanics Geometry Optimization.Geometry optimization of the docking poses of 10 and 11 in TYK2 was carried out with the Jaguar package 73 included in the Schrodinger suite.−76 All other parameters were kept to their default values.
JAK1, JAK2, JAK3, and TYK2 Assays.The inhibitory effect of the described compounds against human TYK2 and its selectivity over JAK1, JAK2, and JAK3 was evaluated in the ADP-glo, radioactive assay kinase biochemical assays (Supporting Information; protocols).
Cellular Assay, Mouse and Human Whole Blood Assays.These protocols can be found in the Supporting Information.
The study was performed according to the Animal Institutional Care and Use Committee of Galapagos.The animal care unit is authorized by the French "Ministere de l'Alimentation et de l'Agriculture" (Agreement No C93−063−06).Animal experiments were performed according to ethical guidelines of animal experimentation with respect to European Directive 2010.After a 7-day acclimatization period, the mice were randomly assigned to a treatment group (n = 10 per group), ensuring a homogeneous body weight distribution.On day 1, the left ears of the mice were shaved under anesthesia with a cocktail of ketamine and xylazine.Intradermal injections in the left ears of either 1 μg rmIL-23 (R&D system) in 20 μL of PBS/0.1% BSA or 20 μL of PBS/0.1% BSA alone were performed daily from day 1 through day 4. Compound 9 or TYK2 inhibitor used as positive reference 54 were dosed orally once a day from day 1 through day 5. On day 4, blood sampling was performed for compound 9 treated mice at the retro-orbital sinus (under isoflurane anesthesia) at different time points and plasma was prepared to assess the concentration of circulating compound.The thickness of the left ear was measured daily before an intradermal injection of rmIL-23, using an electronic gage.On day 5, the mice were weighed and anesthetized by isoflurane inhalation.After measurement of ear thickness, mice were sacrificed by cervical elongation at T max +1 h postdosing (120 min for TYK2 inhibitor and 75 min for 9).The left ear was collected excluding cartilage: half ear was fixed in 4% formalin for pSTAT3 evaluation.The half ears were embedded in paraffin blocks and two different 4 μm thick sections were stained with anti-pSTAT3 antibody (Cell Signaling, #9145L) or with an antibody recognizing the neutrophil marker, 6A608, (Santa Cruz Biotechnology, reference sc-71674) by immunohistochemistry with the BOND-RX automate (Leica).The brown staining was quantified by image analysis (Calopix, Tribvn) as the immunopositive cell number per area of epidermis and dermis.Only STAT3 was analyzed from the STAT family as it is the main target of IL-23.
Additional figures and tables, description of in vitro assay systems, assay statistics, experimental procedures for compounds, analytical data for compounds ( 1 H and 13

Figure 1 .
Figure1.Examples of approved selective and nonselective JAK1 inhibitors.4  (Note: This is not an exhaustive list of all licensed JAK inhibitors targeting JAK1.)

Figure 3 .
Figure 3. Selectivity comparison between the (A) 1H-imidazo[4,5-c]pyridine and the 3H-imidazo[4,5-b]pyridine scaffolds and (B) matched pairs 10 and 11.IC 50 values obtained from fluorescence-based biochemical assays using the catalytic domains of the four JAK members.Selectivity determined as the ratio of the IC 50 values for JAK1/TYK2.

Figure 4 .
Figure 4. (A) Compound 11 docked in TYK2 structure (Protein Data Bank code 3LXN).Compound 11 is displayed in green ball-and-sticks.Only key amino acids of TYK2 are shown (gray sticks).Hydrogen bonds are highlighted with yellow dotted lines, and the aromatic H-bond is materialized with a dotted blue line.(B) Quantum mechanics minimized structures of 10 (pink) and 11 (green), overlaid on the imidazole ring.A distance of 1.5 Å could be measured between the nitrogen atoms of the cyano group.

Table 1 .
SAR Generated by Variations at the C7 Position of the 3H-imidazo[4,5-b]pyridine Series a Calculated by Simulation Plus; 47 b geometric mean of at least two experiments; c LipE = pIC 50 − cLog D; d intrinsic unbound clearance in mouse liver microsomes (LM).
so this difference may explain the increased selectivity of 3Himidazo[4,5-b]pyridine compounds.Initial Structure−Activity Relationship Exploration at the C7 Position.The initial structure−activity relationship (SAR) exploration focused on the C7 position of the 3Himidazo[4,5-b]pyridine scaffold (

Figure 5 .
Figure 5. Compound 13 docked in TYK2 structure (Protein Data Bank code 3LXN).Compound 13 is displayed in green ball-and-sticks.Only key amino acids of TYK2 are shown (gray sticks).Hydrogen bonds are highlighted with yellow dotted lines.

Figure 7 .
Figure 7. Representative frame of a molecular dynamics simulation of compound 9 docked in TYK2 structure (Protein Data Bank code 3LXN).Compound 9 is displayed in green sticks.Only key amino acids of TYK2 are shown (gray sticks).Bridging water molecules that were stable during the simulation are displayed.Hydrogen bonds are highlighted with black dotted lines.

Figure 8 .
Figure 8. Pharmacokinetics after 4 days of treatment with 9, dosed at 3, 10, and 30 mg/kg (q.d.).IC 50 values of 9 for the two mouse pathways are deduced from mouse whole blood assays (see Supporting Information).

Figure 9 .
Figure 9.Effect of 9 dosed at 3, 10, and 30 mg/kg (q.d.) and TYK2 inhibitor used as positive control 56 dosed at 30 mg/kg (q.d.) on IL-23-induced psoriasis-like inflammation in the ear skin of mice on (A) quantification of ear thickening and (B) pictures of inflammation and STAT3 phosphorylation in the ear tissue, (C) quantification of phosphorylated STAT3, (D) pictures of neutrophil (NIMP-R14-positive cells) infiltration in the ear tissue, and (E) quantification of neutrophil percentage area.Data shown are average ± standard error of the mean of n = 9−10 mice per group.Groups were compared using a one-way analysis of variance followed by Dunnett's multiple comparisons test and, for part B, t test for two groups comparison: *p < 0.05, **p < 0.01, ***p < 0.001 versus vehicle group.# p < 0.05 for TYK2 inh. group versus Cpd 9 30 mg/kg group.£ p < 0.05 for Cpd 9 10 mg/kg group versus Cpd 9 30 mg/kg group.BSA, bovine serum albumin; C, cartilage; D, dermis; E, epidermis.

Table 2 .
SAR Generated by Variations at the C5 Position a Calculated by Simulation Plus; 47 b geometric mean of at least two experiments; c LipE = pIC 50 − cLog D; d intrinsic unbound clearance in mouse liver microsomes (LM).

Table 3 .
SAR Generated by Variations at the C7 Position

Table 4 .
Selected Absorption, Distribution, Metabolism, Excretion, and Toxicity (ADMET) Properties of Advanced Compounds Solid state form not determined; b intrinsic unbound clearance in mouse/human liver microsomes (LM); c percentage of inhibition at 10 μM compound, single dose; assay conditions: automated patch clamp.NT, not tested. a

Table 6 .
Potency of 9, 26, 27, Deucravacitinib, and Filgotinib in Cellular and Whole Blood Assays human peripheral blood mononuclear cell assays a human whole blood assays b Experiments performed on peripheral blood mononuclear cells triggered with IFNα, IL-2, or GM-CSF.b Experiments performed in human whole blood triggered with IFNα, IL-6, IL-2, and GM-CSF.In all cases, n refers to the number of experiments performed to determine IC 50 values.The results are the geometrical mean of the different tests.NT, not tested. a

Table 7 .
In Vitro Profile of 9

Table 8 .
Mouse, Rat, and Dog Pharmacokinetic Properties of 9 a n = 6 per group, all values were obtained from plasma; b n = 3 per group, all values were obtained from plasma; c vehicle was polyethylene glycol 200/water for injection (60/40; v/v); d vehicle was Solutol HS15/methylcellulose 0.5% (2/98; v/v); e single gavage of a solid dispersion in corn oil.CL u , unbound clearance calculated using plasma protein binding data.
C NMR spectra, and HPLC and LCMS conditions), table of compound SMILE data (PDF)Molecular formula strings with associated data (CSV)