5-(4-((4-[18F]fluorobenzyl)oxy)-3-methoxybenzyl)pyrimidine-2,4- diamine: A selective dual inhibitor for potential PET imaging of Trk/CSF-1R

Abstract

The tropomyosin receptor kinases (TrkA/B/C) and colony-stimulating factor-1 receptor (CSF-1R) are widely recognized as important therapeutic targets in oncology. The 2,4-diaminopyrimidine inhibitor GW2580 (9) has previously been identified as a highly selective, low-nanomolar inhibitor of TrkB, TrkC, and CSF-1R.

In this study, fluorinated derivatives of compound 9 were designed, synthesized, and evaluated through enzymatic assays. Among these, the highly potent inhibitor 10 was identified, which retained the selectivity profile of the non-fluorinated lead compound 9. Additionally, the radiosynthesis of [18F]10 was successfully developed.

The findings from the biological evaluation of compound 10, along with the radiosynthesis of [18F]10, support further investigation of this tracer as a potential PET imaging probe for TrkB, TrkC, and CSF-1R.

In recent years, the expanding knowledge of protein tyrosine kinases and their role in abnormal signal transduction in cancer has driven the development of numerous targeted small-molecule tyrosine kinase inhibitors (TKIs) for cancer treatment. TKIs have emerged as one of the fastest-growing classes of anticancer drugs, with 16 FDA-approved inhibitors introduced over the past twelve years and hundreds more currently in development.

Despite these clinical advancements, the relatively low response rates of many TKIs reaching the market highlight the need for more effective tools to facilitate drug development and identify patients most likely to benefit from treatment. The current reliance on invasive approaches, such as tumor biopsies, provides only partial information regarding specific target expression and mutation status.

In this context, the development of radiolabeled TKIs for positron emission tomography (PET) imaging offers a promising avenue for advancing drug development and personalized medicine. PET imaging has the potential to provide critical insights into target expression, binding kinetics, potential toxicity, and treatment efficacy, making it a valuable tool in optimizing therapeutic strategies.

Trk receptors play a crucial role in the development and maintenance of the nervous system. However, their overexpression in various neural and non-neural malignancies, including breast, pancreatic, lung, and neuroendocrine tumors, has been associated with aggressive tumor phenotypes and poor prognosis. Over the past decade, significant efforts have been dedicated to the development of Trk ligands, particularly ATP-competitive inhibitors, for cancer treatment.

Currently, the inhibition of Trk receptors is being investigated in six clinical trials, along with numerous preclinical studies, highlighting the growing interest in targeting these receptors as a therapeutic strategy in oncology.

A recent comprehensive kinase inhibitor analysis revealed that the orally bioactive diaminopyrimidine colony-stimulating factor-1 receptor (CSF-1R) inhibitor GW2580 (9) strongly inhibits Trk receptors, particularly TrkB. The dissociation constants (Kd) for GW2580 were found to be 2.2 nM for CSF-1R, 630 nM for TrkA, 36 nM for TrkB, and 120 nM for TrkC. Notably, GW2580 exhibits one of the most selective kinase inhibition profiles among known kinase inhibitors, with no significant inhibition of other kinases at Kd values below 3 μM. This high degree of selectivity presents a significant advantage in the context of PET imaging, where target specificity is crucial.

Given its exceptional selectivity, we hypothesized that GW2580 could serve as a promising scaffold for developing PET-TKI probes with potential imaging applications for CSF-1R. CSF-1R plays a critical role in regulating mononuclear phagocyte differentiation and proliferation and is central to several macrophage-mediated pathological conditions.

In particular, tumor-associated macrophages (TAMs), which infiltrate tumor microenvironments and rely on CSF-1R for survival and differentiation, have been strongly linked to poor prognosis in various cancers. Therefore, adapting GW2580 into a dual Trk/CSF-1R PET probe could be highly beneficial in cases where cancer cells overexpress Trk receptors while CSF-1R is abundant in the stromal cells due to extensive TAM infiltration. This approach could offer a powerful imaging tool for assessing tumor progression and the tumor microenvironment.

The structure of GW2580 contains two aromatic methoxy moieties that could potentially be labeled with carbon-11 (t1/2 = 20 min). However, fluorine-18 presents superior nuclear properties (t1/2 = 109 min; 97% β+; Emax (β+) = 0.64 MeV), which allow for more flexible radiosynthesis and result in high-quality PET images. Additionally, the incorporation of fluorine into bioactive molecules is known to enhance physicochemical properties and improve oxidative and hydrolytic metabolic stability.

This study describes the design, synthesis, and biological evaluation of a series of fluorinated analogs of GW2580. The selected derivatives were specifically designed to be accessible as 18F-isotopologues. Among them, a new potent fluorinated Trk(B/C)/CSF-1R inhibitor, compound 10, was identified and subsequently labeled with fluorine-18. Furthermore, an extensive selectivity analysis across a panel of 342 kinases confirmed that compound 10 retains the exceptional selectivity of the non-fluorinated lead compound, GW2580 (9).

Three fluorinated derivatives of GW2580 were rationally designed based on the available co-crystal structure of TrkB with GW2580 (PDB code: 4AT5). The objective was to maintain the potency and selectivity of the lead compound while ensuring compatibility with 18F-labeling. The design strategy involved structural modifications on the para-methoxybenzyl (PMB) ring, which occupies the selectivity hydrophobic pocket formed by residues Ile616, Leu611, Leu608, and Leu688.

Meanwhile, the diaminopyrimidine fragment interacting with the hinge region and the 1-(benzyloxy)-2-methoxybenzene central ring, which interacts with Asp710 from the DFG motif, remained unchanged. Analysis of the hydrophobic back pocket suggested that only minor modifications could be accommodated at the ortho- and para-positions of the tail fragment. However, the meta-position of the PMB ring was found to be solvent-exposed, indicating that it could tolerate bulkier modifications.

Based on this rationale, three fluorinated derivatives were synthesized and evaluated: fluoroaryl derivatives 10 and 11 (with fluorine introduced at the ortho- and para-positions for labeling) and the 2-fluoroethoxy derivative 12. These compounds were then subjected to further biological assessment to determine their potential as PET imaging probes.

First, we synthesized inhibitor 9 following the patent procedure reported by Shewchuk et al. to provide its first complete characterization (see Supporting Information). To obtain compounds 10–12, we employed two different synthetic approaches.

The first method involved the alkylation of the common 5-phenol-2,4-diaminopyrimidine intermediate 20 (Scheme 1). We anticipated that this intermediate could also serve as a nonradioactive precursor for the synthesis of [18F]10 and [18F]11 via a simple alkylation process using 18F-fluorobenzyl halides.

Initially, we attempted to synthesize 20 through the direct catalytic hydrogenolysis of 9. However, this approach resulted in very low yields under various tested conditions. As an alternative, we pursued the synthesis of 20 by deprotecting either O-MOM- or O-THP-protected 2,4-diaminopyrimidine intermediates 18 and 19. These intermediates were obtained through condensation/cyclization reactions using the protected vanillin derivatives 14 and 15. The reactions were performed using either the 3-morpholinopropionitrile/aniline exchange strategy or 3-ethoxypropionitrile, followed by treatment with guanidine (Scheme 1).

Among these approaches, the cyclization with the β-morpholinopropionitrile intermediate proved to be significantly more efficient than the synthesis of 19. Deprotection of the MOM group in 18 successfully yielded the phenol intermediate 20 with an overall yield of 28% from vanillin. In contrast, attempts to use the more labile THP protecting group in combination with the 3-morpholinopropionitrile/aniline exchange strategy failed to produce 19 or 20. This failure was likely due to acid-promoted aminolysis of the THP fragment in the presence of aniline hydrochloride.

Alkylation of intermediate 20 with suitable benzyl halides successfully yielded compounds 10–12 in good overall yields. However, the high polarity introduced by the common diaminopyrimidine moiety, present in both 20 and its alkylated derivatives, often resulted in mixtures of residual starting material and final products that were difficult to separate. To address this issue, we adapted a linear synthesis strategy similar to the approach used for obtaining compound 9, as exemplified in the synthesis of 11 and 12.

Once synthesized, the compounds were evaluated for their inhibitory activity against TrkA, TrkB, TrkC, and CSF-1R, and their potency was compared to that of GW2580 (9). Under the tested conditions, lead inhibitor 9 exhibited moderate intra-Trk isoform selectivity compared to previously reported Kd values from binding assays.

The fluorine-for-methoxy substitution had minimal impact on TrkB inhibition, as compound 10 demonstrated an IC50 of 119 ± 38.7 nM, compared to 132 ± 12.0 nM for 9. This potent fluorinated TrkB inhibitor (10) also showed similar potency against TrkC (135 ± 5.66 nM) and CSF-1R (169 ± 27.6 nM), along with slightly improved selectivity for TrkA. These results suggest that compound 10 possesses sufficient affinity for PET imaging of tumors.

Additionally, fluorinated compound 10 displayed a reduced total surface polar area (TSPA) and an increased cLogD/cLogP compared to 9, falling within a range favorable for PET radiotracers. Conversely, inhibitor 12 exhibited 2–4.5-fold lower potency against Trk receptors and a 28-fold reduction in CSF-1R inhibition relative to 9.

Interestingly, substituting a single hydrogen atom with fluorine at the ortho-position in compound 11 had a drastic negative impact on potency across all four kinase targets, resulting in more than a 100-fold decrease in activity. As expected, derivative 20, which lacks the benzyloxy tail fragment, did not exhibit any kinase inhibition.

To understand the unexpected potency difference between compounds 9/10 and 11, a molecular modeling study was conducted using the X-ray co-crystal structures of the TrkB-GW2580 complex (PDB ID: 4AT5) and the CSF-1R complex (PDB ID: 3LCO) with the FITTED (FORECASTER platform).

The binding modes of compounds 10 and 11 within the ATP-binding cavity of TrkB (DFG-out) were analyzed, revealing significant overlap with the resolved crystal structure of 9. While the key hydrogen bonds and hydrophobic interactions at the hinge region and the DFG motif remained consistent across inhibitors, notable differences emerged in the spatial orientation of the tail benzyl moiety within the hydrophobic back pocket.

In compound 10, the fluorobenzyl ring maintains the same perpendicular orientation relative to the central methoxybenzyl ring as seen in 9. This conformation is favored as it represents the lowest energy state within this motif, likely optimizing interactions with hydrophobic pocket residues. In contrast, the ortho-fluoro PMB fragment in 11 adopts a distorted conformation, deviating from the optimal orientation.

The difference in spatial arrangement can be attributed to steric and electrostatic factors. In compound 9, the ortho-hydrogen atoms at positions 2 and 6 are positioned approximately 2.8 Å from the oxygen atoms of the carbonyl groups in residues Asp710 and Val617. This distance is smaller than the sum of the van der Waals radii of fluorine and oxygen (2.99 Å), indicating potential steric compatibility. However, in compound 11, the presence of a fluorine atom at the ortho-position extends the C–F bond length, exacerbating electrostatic repulsion and leading to an unfavorable orientation of the tail group.

Moreover, even in its distorted conformation, the fluorine substituent in 11 is likely restricted to an unfavorable position near Asp710 due to the steric hindrance imposed by Val617 if it were to reorient toward the back of the hydrophobic cavity. Additional conformational effects, such as intermolecular hydrogen bonding involving the solvated ligand, may further contribute to the reduced potency of 11. Similar stabilizing intramolecular interactions have been observed in ortho-fluorobenzylic alcohol and ortho-fluoroarylamide structures, which could further reinforce suboptimal binding conformations.

Apart from its high affinity, the decision to select GW2580 as a lead compound in our radiotracer development program was driven by its exceptional selectivity. The significant impact of minor structural modifications on the four selected kinases prompted us to conduct an extensive kinase selectivity profiling of our fluorinated inhibitors.

Compound 10, along with the less potent derivative 12, was evaluated in enzymatic assays across a panel of 342 kinases. In the presence of 1.0 µM of compound 10, strong inhibition was observed only for TrkB, TrkC, and CSF-1R, with remaining activity falling below 25% compared to the DMSO control. Compound 10 also exhibited moderate inhibition of TrkA, with 55.7 ± 1.4% remaining activity, while showing negligible inhibition of only a few other kinases, including CDK6, HGK/MAP4K4, and TXK.

A similar kinase inhibition profile was observed for compound 12, although it displayed slightly more off-target minor inhibitions. Inhibition in the 10–25% range was noted for CDK6, CK1γ3, DAPK2, ERK5/MAPK7, HGK/MAP4K4, SNARK/NUAK2, and TXK. These findings confirmed that compound 10 retained the high selectivity of its non-fluorinated lead compound toward TrkB, TrkC, and CSF-1R, supporting its potential for further development as a radiolabeled probe.

In preliminary radiolabeling experiments, [18F]10 was synthesized through the alkylation of precursor 20 using [18F]fluorobenzyl bromide. The alkylation reaction proceeded with a radiochemical yield (RCY) of 13%, as determined by high-performance liquid chromatography (HPLC) incorporation yield (non-decay corrected). The reaction was carried out in the presence of Cs2CO3 and tetrabutylammonium iodide (TBAI) at 100 °C for 10 minutes.

[18F]S8 was synthesized following an on-cartridge procedure adapted from Lemaire et al., starting with 4-formyl-N,N,N-trimethylammonium triflate (S5). The radiosynthesis of [18F]S8 was typically achieved with a radiochemical yield (RCY) of 25–30% (non-decay corrected) as a crude mixture, which contained over 85% of [18F]S8 along with residual 4-[18F]fluorobenzaldehyde ([18F]S6).

This mixture was used directly in the alkylation step. While the multi-step approach used, which was conducted manually, proved sufficient for evaluation purposes, it will be challenging to scale up for automated synthesis in routine production. To overcome this limitation, a simpler and more efficient route to [18F]10, such as a diaryliodonium salt strategy, will be developed.

In conclusion, starting from the known inhibitor 9, a potent and highly selective fluorinated TrkB/TrkC/CSF-1R inhibitor, compound 10, was designed and successfully labeled with fluorine-18. The exceptional selectivity profile of 10 was confirmed through comprehensive kinase profiling.

These promising results, combined with preliminary radiosynthesis studies, suggest that [18F]10 could serve as a uniquely selective tool for assessing the in vivo levels of TrkB, TrkC, and CSF-1R using PET imaging. Future studies will involve imaging of Trk-positive tumor-bearing nude mice, particularly TrkB-overexpressing neuroblastoma xenograft models, with results to be reported in due course.