Radiofluorination of a NHC–PF₅ adduct: toward new probes for ¹⁸F PET imaging

Abstract

The radiofluorination of N-heterocyclic carbene (NHC) phosphorus(V) fluoride adducts has been investigated. The results show that the IMe-PF₅ derivative (IMe = 1,3-dimethylimidazol-2-ylidene) undergoes a Lewis acid promoted ¹⁸F–¹⁹F isotopic exchange. The resulting radiofluorinated probe is remarkably resistant to hydrolysis both in vitro as well as in vivo.

A growing area of radiochemistry is concerned with the discovery of radiolabeled prosthetic groups which, once appended to tissue- or disease-specific biomolecules, provide a modular access to novel positron emission tomography (PET) imaging agents.¹ To date, most prosthetic groups contain a group 13 element^2–15 or a group 14 element^16–19 which serves as a binding site for the fluoride anion. Undoubtedly, boron-based prosthetic groups pioneered by Perrin are the most developed ones. The most versatile example of such a prosthetic group is the zwitterionic ammonium trifluoroborate (I) which can be incorporated in a wide range of peptide based radiotracers (Chart 1).^20–25 In parallel to these advances, our interinstitutional team introduced zwitterionic phosphoniumtrifluoroborates (II)^26,27 and NHC-BF₃ adducts (III) which, like I, can be conjugated to biomolecules.^27,28 Following up on these results, we were attracted to the fluorophilic properties of phosphorus(V) compounds.^29–31 Indeed, based on computed gas phase fluoride ion affinity data (346 kJ mol⁻¹ for BF₃ and 380 kJ mol⁻¹ for PF₅), which show that P(V) fluorides³¹ may be more Lewis acidic than boron (III) derivatives, it occurred to us that phosphorus analogs of III might be ideally suited for application in PET.

Chart 1.

To explore this idea and expand on the limited chemistry of radiofluorinated phosphorus compounds,^32,33 we decided to investigate the radiofluorination of the N-heterocyclic carbene (NHC) phosphorus(V) fluoride derivatives 1 and 2. Compound 1 was synthesized as described in the literature. To access compound 2, we first synthesized and structurally characterized the potassium salt of the known anion [PF₅Ph]⁻³⁴ via the “one pot” oxidation of PPhCl₂ using bromine in the presence of KF (Scheme 1 and Fig. S7, ESI†). This salt, whose ¹⁹F and ³¹P NMR spectra are consistent with those of other [PF₅Ph]⁻ salts reported previously,³⁴ was successfully converted into the target compound 2 in 68% yield by the addition of n-BuLi at −78 °C to a mixture of imidazolium salt and K[PF₅Ph] (Scheme 1). The ¹⁹F NMR analysis of the crude mixture at room temperature after n-BuLi addition shows a doublet (J_PF = 849 Hz) for 2 at −43.9 ppm and a cis product with an approximate ratio of ~1:1 (Scheme 1). The cis adduct is characterized by three ¹⁹F resonances with a relative integration of 2:1:1, respectively. These resonances include a doublet of virtual triplets at −57.6 ppm (J_PF = 783 Hz, J_FF = 40 Hz) and two doublets of doublets of triplets at −43.2 ppm (J_PF = 699 Hz, J_FF = 49 Hz, J_FF′ = 40 Hz) and −61.0 ppm (J_PF = 838 Hz, J_FF = 49 Hz, J_FF′ = 40 Hz). Heating this mixture at 66 °C for 26 h shows isomerisation of the cis product into the trans product 2 (Fig. S5 and S6, ESI†). Compound 2 is further characterized by a ³¹P NMR resonance at 141.1 ppm split into a quintet (J_PF = 849 Hz). The ¹H NMR spectrum shows a characteristic singlet for the methyl substituents while the ¹³C NMR spectrum shows two doublets of quintets at 150.0 ppm (J_CF = 43 Hz, J_CP = 297 Hz) and 159.8 ppm (J_CF = 71 Hz, J_CP = 334 Hz) corresponding to the phenyl ipso-carbon and carbene carbon, respectively. These assignments align with those reported for other NHC-PF₄Ph derivatives.³⁵ The structure of 2 has been confirmed by X-ray diffraction, which shows that the carbene–phosphorus C(1)–P(1) bond (1.898(2) Å) is only slightly longer than the C(6)–P(1) bond (1.839(2)) involving the phenyl group (Fig. 1).

Scheme 1.

Fig. 1.

ORTEP diagram of 2. Thermal ellipsoids are shown at the 50% probability level. Hydrogen atoms are omitted for clarity. Selected bond distances (Å) and angles (°): P(1)–C(6) = 1.839(2); P(1)–F(1) = 1.634(1); P(1)–F(2) = 1.642(1); P(1)–F(3) = 1.631(1); P(1)–F(4) = 1.645(1); P(1)–C(1) = 1.898(2); C(1)–P(1)–(C6) = 178.96(6); F(2)–P(1)–F(4) = 175.75(4); C(6)–P(1)–F(1) = 91.76(6).

The hydrolytic stability study of 1 and 2 was evaluated using a previously published method.³⁶ The compounds were dissolved in D₂O–CD₃CN (8/2 vol) at pH 7.5 ([phosphate buffer] = 500 mM) and the hydrolysis reaction was monitored by ¹⁹F NMR spectroscopy. While the salt K[PF₅Ph] shows a complete hydrolysis in less than 5 min, both carbene adducts 1 and 2 are highly water stable. Compound 2 undergoes a slow hydrolysis releasing free fluoride with a pseudo-first order rate constant (k_obs) of 2.3 × 10⁻⁵ min⁻¹ (Fig. S8 and Table S3, ESI†). Surprisingly, we did not observe any free fluoride signal for 1 after five days, indicating that 1 can be considered as “eternal” (Fig. 2). It is more stable than the NHC-BF₃ analogue which shows a hydrolytic rate constant (k_obs) of 1.2 × 10⁻⁶ min⁻¹ under the same conditions.²⁸

Fig. 2.

¹⁹F NMR spectrum of 1 in D₂O–CD₃CN (8/2 vol) phosphate buffer solution at t = 0 and t = 5 days.

Next, we investigated the radiofluorination of these compounds (Scheme 2). Compound 1 could be successfully radiolabeled via ¹⁸F–¹⁹F isotopic exchange using SnCl₄ as a promoter, a method that we pioneered in the preparation of [¹⁸F]-BODIPY dyes.³⁷ In this experiment, compound 1 was mixed with 5 equiv. of SnCl₄ in MeCN and combined with a solution of [¹⁸F]-TBAF in the same solvent (Table 1). The reaction mixture was then shaken for 10 min at different temperatures. After being quenched by the addition of water, the radiolabeled compound ([¹⁸F]-1) was loaded on a Sep-Pak cartridge (Sep-Pak Plus tC18). Then, the excess tin reagent and by-products were removed using water. [¹⁸F]-1 was eluted off the cartridge with MeCN. A portion of the resulting MeCN solution was subjected to HPLC analysis. The identity of [¹⁸F]-1 was confirmed by the comparison of its elution time with that of its non-radioactive analog 1 (Fig. 3).

Scheme 2.

Table 1.

Radiosynthetic results for [¹⁸F]-1

Entry	[1] (µmol)	SnCl4 (equiv.)	Temp. (°C)	Time (min)	SAa (mCi µmol−1) (n = 3)c	RCYb (%) (n = 3)c
1	0.9	5	25	10	No [18F]-1 observed
2	0.9	5	60	10	22.6 ± 0.6	4.3 ± 0.3
3	0.9	5	80	10	33.9 ± 1.5	6.6 ± 0.4

^aSpecific activity is determined by dividing the product activity by the amount of the product (based on the integration of UV-HPLC and comparing with the UV chromatogram of the standard).

^bRCY = activity of the isolated product/starting ¹⁸F activity. All yields are decay corrected.

^cEach experiment was repeated 3 times.

Fig. 3.

Left: UV trace of 1 as the standard reference. Right: Radio-HPLC profile of ¹⁸F-[1].

As illustrated in Table 1, the radiochemical yields (RCY) of [¹⁸F]-1, calculated based on the radio-activity of the isolated product and the starting radio-activity, are quite low (4–6% decay corrected RCY). These low yields originate from the stability of the P–F bonds which impedes the ¹⁸F–¹⁹F isotopic exchange process. We found that increasing the reaction temperature leads to higher radiochemical yields (entries 1–3). However, when a high reaction temperature (100 °C) was employed, [¹⁸F]-1 was not detected by either of the two HPLC modalities (radio and UV), indicating precursor decomposition (Fig. S9). Similar issues were encountered in the radiofluorination of 2, for which all efforts proved unsuccessful including those involving different types of activators such as SnCl₂, SnCl₄, TMSOTf, HCl, and KHF₂.

The stability of [¹⁸F]-1 was first investigated in phosphate buffer solution (0.01 M, pH 7). [¹⁸F]-1 displayed >98% radiochemical purity even after an incubation time of 3 hours. This result suggested that [¹⁸F]-1 might be extremely stable under physiological conditions. The stability of [¹⁸F]-1 was further evaluated in a murine model. The probe [¹⁸F]-1 (0.1 mCi) was injected into female nude mice and static microPET scans were obtained at 3 hours after the injection. As shown in Fig. 4, the microPET/CT images showed an obvious localization in the bladder indicating that [¹⁸F]-1 was cleared through the urinary track. More importantly, no bone uptake was observed suggesting that the [¹⁸F]-fluoride release was insignificant even 3 hours post-injection.

Fig. 4.

Decay-corrected whole-body microPET-CT images of nude mice from a static scan at 3 h after injection of [¹⁸F]-1. (A) Coronal image and (B) sagittal image.

In conclusion, we report an organophosphorus [¹⁸F]-radiotracer based on a N-heterocyclic carbene. Owing to Coulombic effects between the imidazolium and phosphate moieties, this probe is remarkably resistant to hydrolysis. It can nevertheless be radiolabeled by isotopic exchange when SnCl₄ is used as an acidic promoter and can be imaged using PET for as long as three hours post-injection. We are now exploring ways to functionalize this adduct such that it can be used as a prosthetic group for targeted tissue and disease imaging.