Identification of Selective Imidazopyridine CSF1R Inhibitors

Abstract

Colony stimulating factor-1 receptor (CSF1R or c-FMS), a class III receptor tyrosine kinase expressed on members of the mononuclear phagocyte system (MPS), plays a key role in the proper functioning of macrophages, microglia, and related cells. Aberrant signaling through CSF1R has been associated with a variety of disease states, including cancer, inflammation, and neurodegeneration. In this Letter, we detail our efforts to develop novel CSF1R inhibitors. Drawing on previously described compounds, including GW2580 (4), we have discovered a novel series of compounds based on the imidazo[4,5-b]pyridine scaffold. Initial structure–activity relationship studies culminated in the identification of 36, a lead compound with potent CSF1R biochemical and cellular activity, acceptable in vitro ADME properties, and oral exposure in rat.

Colony stimulating factor-1 receptor (CSF1R or c-FMS) is a member of the class III family of receptor tyrosine kinases that includes PDGFRα, PDGFRβ, c-KIT, and FLT3.¹ This transmembrane protein, expressed on cells of the mononuclear phagocyte system (e.g., monocytes, macrophages, microglia, and osteoclasts), has two known ligands, colony stimulating factor-1 (CSF1) and IL-34. Binding of either ligand results in receptor dimerization, autophosphorylation, and downstream signaling.² In this manner, CSF1R signaling regulates key cellular functions, including proliferation, survival, differentiation, activation, and migration. Dysregulation of the CSF1R signaling pathway, whether by overexpression or inappropriate activation, has been implicated in a variety of disease states, including cancer,³ inflammation,⁴ neurodegeneration,⁵ and bone disorders.⁶ As such, there has been significant interest in small molecules capable of inhibiting CSF1R over the last two decades.^7,8

Among the more than 80 kinase inhibitors that have been approved for clinical use, many inhibit CSF1R, including imatinib (IC₅₀ = 21 nM), sunitinib (IC₅₀ = 5 nM), dasatinib (IC₅₀ = 2 nM), and axitinib (IC₅₀ = 78 nM).⁹⁻¹¹ However, none of these compounds demonstrate selectivity for CSF1R. Pexidartinib (1) is the most CSF1R-selective drug currently on the market (Figure 1), yet it still potently inhibits other members of the class-III RTK family (CSF1R IC₅₀ = 13 nM, c-KIT IC₅₀ = 27 nM, FLT3 IC₅₀ = 160 nM).¹² It was approved in 2019 as a treatment for tenosynovial giant cell tumors (TGCTs), a group of rare benign neoplasms involving the synovium, bursae, or tendon sheaths that are driven by CSF1 overexpression.¹³ While CSF1R-specific kinase inhibitors have yet to receive FDA approval, there are ongoing clinical trials with compounds that are more selective than the previously mentioned drugs, including sotuletinib (2) (CSF1R IC₅₀ = 1 nM; c-KIT IC₅₀ = 3.2 μM, PDGFR-β IC₅₀ = 4.8 μM, FLT3 IC₅₀ = 9.1 μM)¹⁴ and vimseltinib (3) (CSF1R IC₅₀ = 2 nM; c-KIT IC₅₀ = 0.48 μM, PDGFR-α IC₅₀ = 0.43 μM, PDGFR-β IC₅₀ = 2.3 μM).^15,16 Our efforts to identify CSF1R-selective inhibitors are detailed in this communication.

Figure 1.

Selective CSF1R inhibitors.

From the outset, we considered morphing previously reported CSF1R inhibitors to identify the starting points for our discovery program. We examined a variety of potential starting points, including GW2580 (4), a 2,4-diaminopyrimidine analog reported by Conway et al. in 2005.¹⁷ Not only is this compound potent against CSF1R, but it is also very selective. In 2008, Zarrinkar et al. demonstrated that against a panel of 317 kinases, the only significant off-target activity observed for 4 was against the tropomyosin receptor kinase family (Trk A, B, C).¹⁸ The crystal structure of GW2580 bound to CSF1R¹⁹ indicated that 4 binds to the target in a DFG-out manner with the diaminopyrimidine making three hydrogen-bonding interactions to hinge region residues Cys666 and Tyr665 (Figure 2). The catechol ether moiety binds to an adjacent allosteric pocket, with the oxygen atom of the benzyl ether forming a hydrogen-bonding interaction with Asp796.

Figure 2.

GW2580 (4) bound to CSF1R

The plan to design novel chemotypes involved identifying motifs capable of engaging the hinge region of CSF1R while simultaneously maintaining the ability to form a hydrogen-bonding interaction with Asp796. Using this approach, we were able to identify >20 mono- and bicyclic heterocycles theoretically capable of acting in such a manner (data not shown). To narrow our choices, we ranked these heterocycles in terms of novelty and chemical tractability,²⁰ leading us to explore the use of a 2-aminobenzimidazole as our scaffold.

As detailed in Scheme 1, the synthesis of the desired benzimidazole derivative 8 was achieved via cyanogen bromide-mediated cyclization of o-phenylenediamine 7a which in turn was readily available from amine 6a and 2-fluoronitrobenzene (5a). To our delight, 8 proved to be an effective inhibitor of CSF1R, achieving an in vitro activity and selectivity profile consistent with that of 4 (Table 1).

Scheme 1. Synthesis of Initial Benzimidazole Analogs.

Reagents and conditions: (a) DIPEA, CH₃CN, 80 °C, 1–16 h; (b) Fe, NH₄Cl, 1,4-dioxane/aq EtOH, reflux, 4 h, or FeSO₄·7H₂O, Zn, NH₄Cl, aq EtOH/THF, 70 °C, 1.5 h, or Zn, NH₄Cl, aq THF, rt, 1 h; (c) BrCN, CH₃CN/aq MeOH, rt 16 h (7a, 7c–7g) or 1,3-bis(methoxycarbonyl)-2-methyl-2-thiopseudourea, AcOH, MeOH, reflux, 16 h, then 5.0 N NaOH, reflux, 16 h (7b) or CH(OC₂H₅)₃, p-TsOH·H₂O, EtOH, reflux, 30 min (7h); (d) Cs₂CO₃ or K₂CO₃, DMF, 70–75 °C, 16–18 h (6g–6n) or NaH, DMF, rt, 48 h (6o).

Table 1. Comparison of CSF1R Activity and Kinase Selectivity (Ratio of IC₅₀ Values) for GW2580 (4) and Compound 8.

compd	CSF1R IC50(nM)	c-KIT/CSF1R	FLT3/CSF1R	PDGFRα/CSF1R	PDGFRβ/CSF1R	TrkA/CSF1R	TrkB/CSF1R
4	20	951	725	3368	2799	11	2
8	52	441	579	938	>1000	8	2

With an encouraging starting point identified, we began an examination of the catechol portion of the molecule to understand the structure–activity relationship (SAR) associated with this region of 8. Synthesis of most of the follow-on analogs of 8 was accomplished using two approaches. Compounds 17–19 and 22 were prepared as described for 8 using the appropriate S_NAr substrate (5a, 5b) and benzylamine input. Compound 10 was also prepared via an o-phenylenediamine intermediate, 7b, although in this case cyclization was accomplished with 1,3-bis(methoxycarbonyl)-2-methyl-2-thiopseudourea rather than cyanogen bromide. Additional analogs were prepared using an alternative strategy that relied on alkylation of either 2-aminobenzimidazole (5c) or benzimidazole (5d) with an appropriate benzyl bromide. In this fashion, R₁ derivatives 9, 11–16, and 23 were prepared. The synthesis of benzimidazole analogue 24 which lacked the 2-amino functionality was accomplished by the reaction of intermediate 7h with triethyl orthoformate in the presence of an acid catalyst.

Several SAR trends emerged from this effort, and they are detailed in Table 2. The inclusion of a substituent para to the benzyl ether linkage was favored over a meta substituent (e.g., 9) or proton (e.g., 10). The size and lipophilicity of the para substituent also had an impact on the CSF1R activity. Potency improvements tracked with increases in both these properties for the groups in question (p-SCH₃ > p-OCHF₂ > p-Cl > p-F), perhaps as a consequence of increased nonspecific interactions with lipophilic residues in the allosteric pocket. Further, more highly substituted aromatic moieties were tolerated provided that an appropriate 4-substituent (e.g., OCH₃) was included. While selectivity against class III RTK family members c-KIT and FLT3 remained high for compounds 9–16, most of these early analogs were less selective against TrkA/B than 8. Compound 15 stood out as the exception to this observation: this analog exhibited increased selectivity (∼4-fold) versus TrkA/B relative to 8. Based on these observations, we were encouraged to believe that improving kinase selectivity might be achievable. In addition to selectivity issues, compounds 8–16 had poor physiochemical properties, including low aqueous solubility and high (>4.0) calculated LogD values, indicating additional areas for optimization. Despite these limitations, some instances of moderate lipophilic efficiency (2–3) encouraged us to further improve the scaffold.

Table 2. Impact of Modification of Benzyl Ether Substituent on CSF1R Activity/Selectivity (Ratio of IC₅₀ Values)^a.

compd	R1	R2	R3	CSF1R IC50 (nM)	c-KIT/CSF1R	FLT3/CSF1R	TrkA/CSF1R	TrkB/CSF1R	Sol pH 7.4 (μg/mL)	cLogDb	LipE
8	OCH3	H	H	52	>500	>500	8.5	2.0	n/d	4.0	3.3
9	H	OCH3	H	1113	>45	>45	3.0	0.5	n/d	4.1	2.0
10	H	H	H	2541	>20	>20	1.1	0.2	n/d	4.1	1.5
11	SCH3	H	H	32	>500	>500	3.4	1.1	<0.2	4.6	2.9
12	OCHF2	H	H	67	>500	>500	1.8	0.5	n/d	4.2	3.0
13	Cl	H	H	347	>150	>150	1.0	0.2	n/d	4.7	1.8
14	F	H	H	1035	>50	>50	1.4	0.3	n/d	4.1	1.9
15	OCH3	F	H	60	>800	>800	32	8.4	n/d	4.1	3.2
16	OCH3	H	F	80	>600	>600	9.3	2.4	n/d	4.2	3.0

^an/d = not determined.

^bAll measured LogD values were >3.50.

From these early compounds, it was clear that an appropriately substituted terminal aromatic group was a key factor in driving potency against CSF1R. We next examined whether heterocycles might also be tolerated in this position (Table 3). Pyridine analogs of 8 demonstrated the potential utility of this modification. Compounds 17 and 18 were nearly an order of magnitude more potent against CSF1R than analog 19, indicating that while polar N atoms are tolerated in this region, proper placement is required to maintain potent activity. In all three cases, however, the pyridyl analogs were less active than the corresponding phenyl analogs, and TrkA/B selectivity remained poor. While the loss of CSF1R potency was disappointing, the potential benefit of increased polarity offered by the pyridine moiety suggested that this modification might still be worthy of additional investigation. In contrast to the benzyl ether analogs in Table 2, 17–19 demonstrated some modest level of aqueous solubility. Still, the improvements were small, and the overall drug-likeness of these analogs as indicated by their LipE values remained suboptimal. Derivatives with more substantial modifications to this terminal group (e.g., 20, 21) suffered even greater losses in activity compared to the phenyl/pyridyl-containing compounds.

Table 3. Impact of Heterocycles/Saturated Carbocycles on CSF1R Activity/Selectivity (Ratio of IC₅₀ Values)^a.

^an/d = not determined.

Our next SAR investigation of 8 focused on the hinge binding 2-aminobenzimidazole motif (Table 4). As we observed for the catechol moiety, small structural changes to the hinge binder resulted in significant changes in CSF1R activity. Transforming the 2-aminobenzimidazole of 8 to a 2-aminoimidazo[4,5-b]pyridine led to 22, an analogue with a 4-fold improvement in potency. Surprisingly, removal of the 2-amino functionality from 8 to generate 23 was also beneficial for CSF1R activity, an indication that the amino functionality, while tolerated, is unnecessary for potent inhibition of the target. The related imidazo[4,5-b]pyridine analog 24 was also an effective inhibitor, albeit with slightly less activity than 22 and 23. We also noted that the imidazo[4,5-b]pyridine analogs 22 and 24 demonstrated reduced levels of selectivity versus TrkA/B compared to the benzimidazole 23. In addition to lacking the 2-amino functionality, 24 also incorporated the previously highlighted pyridyl-for-phenyl swap on the catechol moiety of the molecule. This combination resulted in several important firsts for this series: significant improvement to aqueous solubility, increased LipE values (>4), and submicromolar cellular activity, as evidenced by the compound’s ability to prevent phosphorylation of the target in CSF1-stimulated HEK293 cells engineered to overexpress hCSF1R.

Table 4. Impact of Alterations to the Hinge Binding Motif on CSF1R Activity/Selectivity (Ratio of IC₅₀ Values)^a.

compd	R1	X	Y	CSF1R IC50 (nM)	pCSF1R IC50 (nM)	c-KIT/CSF1R	FLT3/CSF1R	TrkA/CSF1R	TrkB/CSF1R	Sol pH 7.4 (μg/mL)	LogD	Caco-2b	LipE
22	NH2	N	CH	13	n/d	>1000	>1000	1.7	0.9	n/d	3.3	n/d	4.6
23	H	CH	CH	12	n/d	n/d	>800	58	16	<0.2	>3.5	n/d	n/d
24	H	N	N	56	91	>800	>800	6.0	1.5	70	2.9	320	4.3

^an/d = not determined.

^bUnits: 10^–7 cm/s.

Given its balance of potency, selectivity, and ADME properties, we focused our effort around developing the SAR of 24. Further substitution of the hinge-binding heterocycle offered the opportunity to grow the molecule in previously unexplored vectors. In doing so, we hypothesized that these extended compounds might realize additional interactions with the protein and represented an opportunity for improvements to the ADME properties. The synthesis of the appropriately functionalized 6-carboxy intermediate 29 or 6-iodo intermediate 30 was accomplished as shown in Scheme 2. Briefly, the regiospecific synthesis was accomplished by S_NAr substitution of appropriate chloropyridine 25 or 26 with intermediate 6c. Subsequent zinc-mediated reduction of nitroarene 27 followed by acid-catalyzed ring closure with triethyl orthofomate and ester hydrolysis provided 29. Similarly, 28 was converted to 30 via the two-step reduction/cyclization protocol.

Scheme 2. Synthesis of Key Intermediates 29 and 30.

Reagents and conditions: (a) DIPEA, CH₃CN, 100 °C, 1 h; (b) Zn, NH₄Cl, aq THF, rt, 30 min; (c) CH(OC₂H₅)₃, p-TsOH·H₂O, EtOH, reflux, 30 min, then 1.0 N NaOH, MeOH, rt, 30 min (for 29).

Both 29 and 30 proved to be valuable for the synthesis of additional analogs as detailed in Table 5. In general, the incorporation of a 6-substituent proved beneficial. Amides, amines, ethers, heterocycles, and alkyne derivatives—all readily accessible from either 29 or 30—provided increased potency in both the biochemical and cellular CSF1R assays versus the 6-unsubstituted analog 24. The observed increases in activity were greater in the enzymatic assay (3–14× more potent) than in the cellular phosphorylation assay (up to 6.5× more potent). Rarely did a 6-substituent fail to offer a potency boost (e.g., 6-cyano derivative 37). As observed in earlier compounds, the selectivity against other class III RTKs remained high, but potent TrkA/B activity persisted. As the most potent analog in the series, 36 was also profiled in a murine bone marrow-derived macrophage assay (mBMDM) to determine the compound’s ability to thwart CSF1-mediated proliferation of cells that natively express CSF1R. The results indicated that 36 was a potent inhibitor of macrophage proliferation. In addition to improvements in CSF1R inhibitory activity, many of analogs 31–37 retained the improved ADME properties initially observed with 24. Table 6 provides additional information about the in vitro ADME performance for these more advanced analogs, including dramatic improvement in LipE values compared to earlier compounds (cf.8–34). At this stage of the program, we also confirmed that the binding mode of our new series matched our expectations (Figure 3). Cocrystallization of 32 with CSF1R demonstrated that the compound binds to the protein in a DFG-out conformation, the key hydrogen-bonding interactions with residues Cys666 and Asp796 are observed, and the amide functionality at the 6-position of the hinge-binding motif extends toward solvent-exposed space.

Table 5. Impact of 6-Substituent on CSF1R Activity/Selectivity (Ratio of IC₅₀ Values)^a.

^an/a = not applicable; n/d = not determined.

Table 6. In Vitro ADME Properties for Analogs 31–37^a.

compd	LogD	LipE	Sol pH 7.4b	Caco-2c	CLint hd	CLint rd	Cyp inhe
31	1.9	5.9	>400	190	9	52	>10
32	1.6	6.3	>450	63	25	26	>10
33	2.3	5.5	>450	290	33	130	>10
34	0.1	7.9	>400	5	13	10	>10
35	1.6	6.5	13	20	21	<7.2	>10
36	1.5	6.9	57	51	16	19	6.6 (3A4)M
37	2.8	4.1	<0.2	n/d	n/d	n/d	n/d
4	2.6	5.1	1.2	110	n/d	n/d	>10

^an/d = not determined.

^bUnits: μg/mL.

^cUnits: 10^–7 cm/s.

^dUnits: μL min^–1 mg^–1.

^eUnits: μM.

Figure 3.

Compound 32 bound to CSF1R (PDB entry 8W1L).

To assess the pharmacokinetic (PK) profile of this new series of CSF1R inhibitors, we selected 32 and 36 as representatives of the class. Single-dose PK experiments in male Sprague-Dawley rats (1 mg/kg iv, 3 mg/kg po) with either 32 or 36 demonstrated that the two analogs had similar exposure profiles (moderate F, moderate Cl, short T_1/2). The most significant difference between the two analogs was the volume of distribution. The value for 36 was more than 6 times that of 32, likely due to the presence of a basic amine in 36 (Table 7).

Table 7. Rat PK Results for 32 and 36.

	AUC0–∞ (ng h mL–1)	Cmax (ng/mL)	T1/2 (h)	F (%)	Cl (mL min–1 kg–1)	Vdss (mL/kg)
1 mg/kg iv 32	874	–	0.8	–	19	748
3 mg/kg po 32	740	175	1.7	26	–	–
1 mg/kg iv 36	542	–	2.0	–	31	4890
3 mg/kg po 36	625	108	1.7	36	–	–

We next examined the ability of 36 to impact TNF-α production in a murine model of LPS-induced acute inflammation.¹⁷ In the event, 36 was dosed orally at 25, 50, and 100 mg/kg to C3H/HeN mice 45 min prior to priming of the MPS system with an intraperitoneal dose of CSF1.²¹ After 3.5 h, the mice were challenged with an intraperitoneal dose of LPS. Plasma TNF-α levels were determined by an ELISA assay 1.5 h after LPS administration. As shown in Figure 4, 36 demonstrated the ability to reduce TNF-α levels at all doses, and the data suggested a correlation between dose and effect. However, a closer examination of the results revealed that the experiment did not reach statistical significance as determined by one-way ANOVA (p = 0.055). Excessive variation in the TNF-α levels in the control-group animals likely contributed to these findings.

Figure 4.

Impact of 36 on TNF-α production in the LPS-induced acute inflammation model.

As this Letter establishes, we have successfully identified a novel series of CSF1R inhibitors. Early SAR studies resulted in the identification of potent compounds (i.e., 36) with functional activity in relevant CSF1R-expressing cells. Representative examples have demonstrated acceptable in vitro AMDE properties, moderate oral bioavailability in rats, and in vivo activity in a murine model of acute inflammation. While these early analogs represent a promising lead series, additional work to optimize kinase selectivity (particularly against the Trk kinases), in vivo exposure, and efficacy in relevant disease models remains. The results of these continued efforts will be reported in due course.

Glossary

Abbreviations

ANOVA

analysis of variance

CSF1

colony stimulating factor-1

CSF1R

colony stimulating factor-1 receptor

FLT-3

feline McDonough sarcoma-like tyrosine kinase-3

HEK293

human embryonic kidney 293 cells

IL-34

interleukin-34

LipE

lipophilic efficacy

LPS

lipopolysaccharide

mBMDM

murine bone marrow-derived macrophage

MPS

mononuclear phagocyte system

RTK

receptor tyrosine kinase

SAR

structure–activity relationship

TGCT

tenosynovial giant cell tumor

Trk

tropomyosin receptor kinase

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.4c00110.

• Experimental procedures for the synthesis of compounds 8–37 and characterization data (¹H/¹³C NMR, HRMS); NMR spectra and UPLC chromatograms for compounds 32 and 36; biochemical and cellular assay conditions; experimental procedure for in vivo murine LPS acute inflammation model; description of cocrystallization method and refinement statistics for 36 (PDF)

Accession Codes

Atomic coordinates for the X-ray structures of compound 36 (PDB entry 8W1L) bound to CSF1R are available from the RCSB Protein Data Bank (www.rscb.org).

Author Present Address

^∥ DEM Biopharma, Cambridge, MA 02139, United States

Author Present Address

^⊥ Vertex Pharmaceuticals, Boston, MA 02210, United States

Author Present Address

^# Ipsen Bioscience Inc., Cambridge, MA 02142, United States

Author Present Address

^∇ Dragonfly Therapeutics, Inc., Waltham, MA 02451, United States

Author Present Address

^○ Revolution Medicines, Redwood City, CA 94063, United States

Author Present Address

^◆ Paraza Pharma Inc., Montreal, QC, Canada H4S 2E1

Author Present Address

^¶ Omega Therapeutics, Cambridge, MA 02141, United States

Author Present Address

^■ Novartis BioMedical Research, Cambridge, MA 02139, United States

Author Contributions

The manuscript was written by J.L.K. Compound design and synthesis were conducted by J.L.K., B.G., S.L., L.M., M.M. (design), A.S.S., P.S., and L.L.W. Structural biology/crystallography was conducted by J.B., M.K., and J.L. Kinase enzymatic and cellular assays were conducted by G.A., T.G., S.H., and B.L. L.W. conducted the mBMDM assay. M.C., M.Fe., and K.K. conducted the murine LPS model. M.Fi. directed all in vitro/in vivo ADME. The overall project was directed by J.L.K. and A.E. All of the authors reviewed and approved the final version for submission.

The authors declare the following competing financial interest(s): All authors are or were employees of Sanofi and may have stock and/or stock options.