Discovery of Potent and Selective Tricyclic Inhibitors of Bruton's Tyrosine Kinase with Improved Druglike Properties.

In our continued effort to discover and develop best-in-class Bruton's tyrosine kinase (Btk) inhibitors for the treatment of B-cell lymphomas, rheumatoid arthritis, and systemic lupus erythematosus, we devised a series of novel tricyclic compounds that improved upon the druglike properties of our previous chemical matter. Compounds exemplified by G-744 are highly potent, selective for Btk, metabolically stable, well tolerated, and efficacious in an animal model of arthritis.

Bruton’s tyrosine kinase (Btk) plays a critical role in the function of B cells and myeloid cells, making it an attractive target for the treatment of immunological disorders such as rheumatoid arthritis (RA), systemic lupus erythematosus (SLE), multiple sclerosis (MS)[1−4] as well as B-cell lymphomas.[5] Accordingly, there have been significant efforts from the pharmaceutical community toward identifying Btk inhibitors for clinical evaluation.[6−14] Of these, the most advanced compound to date is ibrutinib, recently approved for treatment of Mantle Cell Lymphoma (MCL), chronic lymphocytic leukemia (CLL), and Waldenstrom’s macroglobulinemia and under evaluation in additional indications.[15]

We have previously reported on the discovery of several series of novel and potent Btk inhibitors that possess exquisite selectivity for Btk over other kinases.[9−11,16] Their high potency and selectivity stems from the ability of compounds such as 1 to create an induced binding fit in the protein via a rearrangement of the activation loop.[3] Other kinases, including members of the tyrosine kinase family, cannot rearrange in the exact same way due to differences in the amino acid sequence within this loop.[11] The tetrahydrobenzothiophene moiety within 1, held rigidly in place by other contacts the inhibitor makes with the protein, presents a hydrophobic surface that “attracts and sequesters” Y551, ultimately creating a lipophilic specificity pocket (“H3”). A common feature in this series is the incorporation of a secondary amide that tethers the distal bicyclic ring to the central benzene. We set out to address metabolism and permeability liabilities associated with the amide moiety.

In particular, we hypothesized that molecules with an amide tethered back onto the tetrahydrobenzothiophene, such as 2a, would offer a molecule with one less exposed N–H donor and reduce the number of rotatable bonds, potentially improving potency and permeability. Crystal structures of related compounds bound to Btk indicated the amide N–H was not interacting with the protein.[3] Additionally, we incorporated a hydroxyl group in the central benzene ring that was well positioned to interact with nearby K430 and D539 residues, with the hope of improving binding affinity. Table shows the potencies of these compounds for Btk in a biochemical assay, as well as their potencies in an in vitro mouse splenocyte CD86 cell based assay that was used as a downstream pharmacodynamic marker for Btk activity (assay protocols in SI). Pleasingly, the newly generated tricyclic compound 2a was roughly 4-fold more potent against Btk than the uncyclized precursor 1. Compound 2b, which includes a hydroxymethyl group on the central benzene ring, had a ∼10-fold increase in Btk binding potency than 2a and was roughly 40-fold more potent than compound 1. With the improved potency and/or decreasing lipophilicity, the lipophilic ligand efficiency (LLE) increased sequentially from 1 (3.6) to 2a (4.2) to 2b (5.8), indicating enhanced druglike properties. The cellular potency of compound 2b (IC50 = 0.087 μM), as measured by inhibition of CD86 surface expression on B cells, is 8-fold higher than that of 1 (IC50 = 0.74 μM). Additionally, 2b is 2-fold more orally bioavailable (90%) than 1 (45%) and has similarly low clearance in rat.

Table 1

Improved Properties of Tricyclic H3 Btk Inhibitors

	IC50 (μM)a				Rat
ID	Btk	CD86	LLEb	hPPBc	CLd	F%e
1	0.042	0.74	3.6	99%	2	45%
2a	0.011		4.2	99%
2b	0.001	0.087	5.8	97%	2	90%

Assay protocols in Supporting Information (SI), n ≥ 2.

LLE = pIC50(Btk) – cgLogD.

Human plasma protein binding.

Total clearance (mL/min/kg) at 1 mg/kg i.v. dose formulated using a mixture of EtOH/Cremophor/water for 1 (solution) or PEG400/EtOH/water for 2b (solutions).

F% = oral bioavailability after a 5 mg/kg oral dose (n = 3) formulated using a mixture of EtOH and Cremophor for 1 (solution) or PEG400/EtOH/Tween80/water for 2b (suspension).

Modeling of 1 and 2b in the Btk catalytic domain is shown in Figure . The tricyclic system did not alter the overall predicted ligand binding mode, and preserved the interactions at the kinase selectivity H3 pocket with the neighboring hydrophobic residues. Specific interactions with Y551 and K430 were maintained. The hydroxymethyl on the central benzene ring projected into a water-filled cavity, and formed H-bonds with K430 and D539.

Figure 1

Superposition of modeled structures of 1 (magenta) and 2b (cyan) in the Btk catalytic domain. H-bonds to the protein are shown as black cylinders with dashed lines. The active site surface is shown color coded by lipophilicity (green, nonpolar; purple, polar).

The success of our initial attempts at creating improved Btk inhibitors by the employment of a tricyclic moiety encouraged us to explore additional tricycles at this position as well as carefully designed bicycles. Table shows a selected subset of such compounds. Compounds 2b–6 are examples of inhibitors that contain distal 6–5–6 tricyclic ring systems. Since the H3 site is lipophilic, with nonpolar residues flanking much of this pocket, it is not surprising that the most potent compounds, 2b and 3 (Btk IC50 = 0.001 and 0.006 μM respectively), contain all-carbon fused cyclohexane rings.

Table 2

SAR for H3 Groups

Assay protocol in SI, n ≥ 2.

Kinetic solubility was measured at pH 7.4.[17]

Compound 4, with a heteroatom in the right-hand saturated ring, displayed reduced potency (Btk IC50 = 0.081 μM), while 5 and 6 were much less potent, presumably due to the reduced lipophilicity of the tricycle. Modeling indicated that there was an opportunity to further flesh out the H3 pocket and that gem dimethyl substituted 6–5–5 systems might offer some advantages by occupying more of the lipophilic H3 pocket. The resulting 6–5–5 compounds 7 and 8 of 2b and 3, respectively, were equipotent to slightly more potent against Btk (IC50 = 0.002 μM), and both had improved cell potencies.

We also examined bicyclic 6–6 fused (9), 5–5 fused (10-11), and 5–6 fused (12) ring systems containing appropriately placed tert-butyl groups to extend into the H3 pocket. Among this group, the 5–5, 5–6 fused 10 and 12 had the highest enzyme binding potency (Btk IC50’s = 0.003 μM) but were relatively less potent in the CD86 cellular assay.

Finally, we surveyed 7–6 ring systems (13 and 14). These compounds were less appealing due to their decreased biochemical and cell potencies. The cell potencies of all compounds mostly tracked with biochemical potencies, with the Btk/CD86 IC50 ratio ranging from 6- (4) to 64- (10) fold. The only major outlier was the bicyclic compound 12, which had a cell shift of 400-fold. Presumably, this shift was due to differences in plasma protein binding and cell permeability of this compound. Other researchers have reported on a variety of bicyclic H3 moieties within alternate series;[18−20] however, we are the first group to describe highly potent and Btk-selective tricyclic 6–5–6 and 6–5–5 compounds.

With a variety of novel tricyclic compounds in hand that had excellent potency for Btk, we focused our efforts on designing compounds with improved physiochemical properties. As demonstrated in Table , although compounds 2b, 7, 8, and 9 had the best cellular potency (CD86 IC50 < 100 nM), they unfortunately all had low kinetic solubility (<2 μM) at physiological pH. Kinetic solubility as measured in an internal high throughput assay[17] was used as an initial gauge of intrinsic solubility. Low solubility is a primary cause of reduced oral bioavailability of the crystalline materials preferred for clinical development, and often presents formulation challenges in dose escalations for safety studies. This was indeed the case for many compounds in this chemical series. To find compounds with improved solubility, modifications of the left-hand portion of the inhibitor that extends into a partially solvent exposed area of the protein region (“H2”, Figure ) were explored and found to be well tolerated. Of the many compounds generated,[21] we determined that compounds with a substituted pyridinopiperazine group (Table ), as opposed to the pyrimidine of the previous compounds (Table ), offered an exciting new subseries with much improved kinetic solubility.

Table 3

Combination of the Best H2 and H3 Moieties for Optimal Potency and Druglike Properties

Assay protocol in SI, n ≥ 2.

Projected hepatic clearance using human or rat hepatocytes.

Kinetic solubility was measured at pH 7.4.[17]

Measured permeability using Madin-Darby Canine Kidney (MDCK) Epithelial cell lines, A to B; MOD: 1–10 (10–6 cm/s) and HIGH: >10 (10–6 cm/s).

Total clearance (mL/min/kg) at 1 mg/kg i.v. dose formulated using a mixture of PEG400/ethanol/water for 7, 15, and 17–20 (solution).

F% = oral bioavailability after 5 mg/kg oral dose (n = 3) formulated using a mixture of PEG400/ethanol/Tween80/water for 7 (suspension), PEG400/ethanol/water for 17 (suspension), 15, 18–20 (solution).

A small subset of compounds containing combinations of the best H2 and H3 groups are highlighted in Table . Installing a tertiary basic amine containing left-hand H2, which is primarily protonated at neutral pH, indeed helped improve the solubility from 1.4 μM (7) to 18 μM (15). Reducing the basicity from a calculated pKa of 7.8 (15) to 6.3 (16, substituted with an oxetane)[22] abrogated the improved solubility. This prompted us to focus on H3 groups with higher polarity than that of 7. Gratifyingly, all four compounds containing a more polar tetrahydroindole H3 group (17-20) maintained high kinetic solubility regardless of the amine basicity. However, the moderate CD86 cellular potencies of 17 and 18 (IC50‘s of 0.107 and 0.137 μM, respectively) mirrored that of the 0.2 μM potency seen for 3, which bears the same polar H3 group. A potency breakthrough was realized when a single fluorine atom was added to the middle linker benzene ring (19–20). This fluorine is well positioned to engage the backbone carbonyl carbon of G409 in a dipole–dipole interaction (Figure s1 in the SI), further stabilizing the binding orientation of these inhibitors in the Btk catalytic domain. The resultant CD86 cellular potency improved 5-fold and 13-fold for the two pairs 17 → 19 and 18 → 20, respectively.

All compounds in Table , except 17, were low to moderately cleared and had reasonable bioavailability in in vivo rat PK studies. In vitro hepatic clearance represented by RHEP data underestimated the in vivo clearance; however, the relative stability trend was consistent from in vitro to in vivo studies. It was also pleasing to see that human hepatic clearance ranged from low to moderate. Given the totality of data including safety assessment (not discussed here), compounds 7 (G-744) and 20 rose to the top of the list with the most balanced overall profile.

In addition to potency and druglike properties, a prominent goal of developing a clinically viable kinase inhibitor lies in controlling selectivity. Poor selectivity may have a profound effect on drug safety, especially for nononcology indications where chronic dosing requires an exquisitely clean safety profile. Therefore, both G-744 and 20 were profiled against a panel of 285 active recombinant human kinases. In particular, G-744 demonstrated >1000-fold Btk biochemical selectivity against all kinases tested except for EphA7 and Fgr, against which it still showed robust selectivity of 428-fold and 868-fold, respectively (Figure s3 in the SI). Due to G-744’s superb kinase selectivity (superior to 20), we realized it could be an excellent tool molecule to probe the biology of Btk, as the results would not be confounded by off-target activity and the need for interpretation. Thus, we performed a full characterization of this molecule, with key data summarized in Table . In addition to preventing cellular functions in murine B-cells such as B-cell receptor (BCR)-mediated CD86 induction with an EC50 of 64 nM, G-744 also inhibited BCR-stimulated B-cell proliferation in human B-cells (EC50 = 22 nM). In human monocytes, production of the inflammatory cytokine TNFα following activation with immune complexes was abrogated by G-744 (EC50 = 33 nM). In human whole blood, G-744 demonstrated potent inhibition of BCR-stimulated CD69 expression on B-cells with an EC50 of 87 nM.

Table 4

Additional G-744 Potency Data (mean ± SEM)

Assay	Ki,app or EC50 (nM)
Btk biochemical, (Ki,app)	1.28 ± 0.13 [n = 3]
Mouse splenocyte B cell CD86	55 and 75 [n = 2]
Human Whole Blood B cell CD69	87 ± 30 [n = 11]
Human B cell proliferation	22 ± 3 [n = 17]
Human monocyte TNFα production	33 ± 6 [n = 12]

In pharmacokinetic experiments, G-744 exhibited low to moderate clearance in four preclinical species (Table ). Sufficient oral exposures were achievable using a crystalline formulation in both mouse and rat despite low kinetic solubility.

Table 5

Preclinical DMPK Profiling of G-744

	In vitro		In vivo
Species	LM CLhepa	Hep CLhepb	CLc	F%d
mouse	35	2	22	45e – 77f
rat	18	3	16	23e – 37f
dog	15	1	7	27f
cyno	18	31	17
human	8	11

Projected hepatic clearance using liver microsomes (mL/min/kg).

Projected hepatic clearance using hepatocytes (mL/min/kg).

Total clearance in vivo (mL/min/kg).

F% = oral bioavailability after 5 (amorphous) or 100 mg/kg (crystalline) dose in mouse; 5 (amorphous) or 100 mg/kg (crystalline) dose in rat; 5 mg/kg dose in dog (n = 3).

Methylcellulose/Tween80/water suspension of crystalline free base material.

Hydroxypropylmethethylcellulose/Na citrate/water pH 3 suspension of amorphous material.

Given the favorable DMPK profile of G-744 in rat, we examined its efficacy in the developing collagen-induced arthritis (CIA) model in Lewis rat.[3,23] Oral dosing with G-744 at 6.25, 12.5, and 25 mg/kg b.i.d. maintained plasma concentrations above the IC50, IC70, and IC90, respectively, for inhibition of Btk Y223 phosphorylation in whole blood (Figure s2 in the SI). As shown in Figure , all three doses resulted in a significant dose-dependent inhibition of ankle thickness between day 10 and day 17 (onset of increase in ankle diameter on day 9). The 25 mg/kg dose (97% inhibition of the area under the ankle thickness vehicle curve) showed comparable efficacy to dexamethasone. In naïve rats treated with vehicle, the ankle diameters did not change over the course of the study and ankles from normal rats were significantly different (P < 0.05) compared with the CIA rats treated with vehicle. In addition, G-744 was also highly efficacious in preventing Interferon α-driven lupus nephritis.[24]

Figure 2

Results from a CIA study in rats. Treatment with G-744 protects Lewis rats from collagen-induced arthritis. Female Lewis rats (n = 10 per group) with developing CIA were dosed orally with G-744 as indicated (b.i.d.) or with 0.05 mg/kg dexamethasone daily starting on day 0. Daily ankle diameter measurements are shown as mean ± SEM and were significantly (by ANOVA) reduced toward normal for all drug-treated rats (significant days 10–17) as compared to the vehicle control.

In summary, orally bioavailable Btk inhibitors with novel tricyclic head groups were discovered through structure and property based drug design. Improved molecules were discovered, and an outstanding tool molecule, G-744, was identified with excellent potency, favorable DMPK properties, and superb kinase selectivity. G-744 demonstrated efficacy equivalent to Dexamethasone in a rat CIA model at 25 mg/kg b.i.d. dosing. Such findings further solidified our commitment to Btk as a therapeutic target. The chemistry culminating in a clinical candidate will be described in a subsequent manuscript.