A range of readily available commercial boron trifluoride precursors has been tested in different reaction media employed on two different synthesizers (Table 1). There was no notable difference in terms of isolated radiochemical yields when different boron trifluoride source reagents were deployed. We chose the boron trifluoride-methanol complex as a starting material for the optimised procedure mainly because it is readily available as a 10% boron trifluoride methanol solution in sealed 10 mL dark glass bottles, which facilitates the use of the reagent.
The radiochemical yields were generally better and more reproducible when more concentrated precursor solutions were used, but the more concentrated solutions provide lower molar activity of the final product. We have found that diluting the commercial source of boron trifluoride to 1/600 by volume still provided reproducible yields with good molar activity (maximum obtained was 216.7 GBq/mmol [18F]TFB at EOS starting from 102.3 GBq [18F]fluoride at EOB).
To compare the molar activity obtained in this study with the results reported in the literature we have normalised the molar activity at EOS by the starting activity at SOS, the resulting normalised MAn expressed in mmol− 1 allows direct comparison of the results without taking into consideration the starting activity. The best average MAn in this report was 1.75 mmol− 1 with the boron trifluoride-methanol complex dissolved in acetonitrile with 1/500 dilution (V/V). This is considerably better compared to most of the published procedures (see Table 1). The best published MAn of 3.8 mmol− 1 was reported by Khoshnevisan et al. [1]. Nevertheless, we were not able to find follow-on publications from this group on large scale high molar activity [18F]TFB studies, probably due to the fact that the impressively high values were achieved using a manual synthesis method employing laborious and moisture-sensitive serial dilutions techniques of the boron trifluoride precursor (boron trifluoride-etherate diluted in anhydrous acetonitrile to 0.8mM). For routine production of the tracer for clinical and pre-clinical studies robustness of the process is of paramount importance.
Another means to improve the molar activity (other than diluting the precursor) could be minimising the mass of precursor in the radiolabelling by reducing the volume of the solution used. Decreasing the volume or increasing the dilution factor leads to less stable radiochemical yields, as evidenced by the standard deviations (STD) of the radiochemical yields (RCY) from the optimisation experiments in Table 2.
The source of fluorine-18 for radiolabelling of [18F]tetrafluoroborate has significant importance. Usually [18F]fluoride comes from the cyclotron target contaminated with various long-lived metal radionuclides and tritium [22]. To prepare [18F]fluoride for the next radiolabelling steps it is usually extracted from the irradiated target material by anion exchange on a solid support (QMA-carbonate cartridge). Consequently [18F]fluoride is displaced from an anion exchanger by eluting with another anion-containing solution (usually carbonate or bicarbonate) with the addition of counter-ion (normally potassium or tetrabutylammonium) and phase-transfer agents (usually Kryptofix 222 or tetrabutylammonium) thus facilitating dissolution of inorganic fluoride in organic solvents for further nucleophilic substitution reactions [23].
The Blower group has reported that no radiolabelling was observed when the customary Kryptofix 222/potassium carbonate mixture was used for fluoride elution from a QMA cartridge [1]. Instead, they proposed the use of sodium chloride – 15C5 crown ether mixture. A possible explanation for this observation could be basicity of carbonate, making unfavourable formation of tetrafluoroborate anion [18].
Ideally for the nucleophilic substitution reaction employed in [18F]TFB radiolabelling, fluoride in the reaction mixture should be without interfering anions. These conditions were achieved by trapping fluoride on a QMA cartridge and reacting it with boron trifluoride etherate directly on the solid support, as reported by deGrado group [14].
In our manual procedure [15] we trapped fluoride-18 from the irradiated water on the week anion exchange cartridge (WAX, Waters corp.) followed by elution with neat ethanol (i.e., no use of any ionic substrates in the eluate). This procedure performed well with nearly quantitative fluoride extraction into the ethanol solution and high incorporation yields of fluoride-18 into the [18F]TFB in the manual setup. Unfortunately, it proved to be non-reproducible in the automated setting: often 40–50% loss of activity was observed either due to fluoride break-through in the trapping step, or its retention on the cartridge in the elution step. We attribute this irreproducible behaviour to the flow-rate factors governing anion exchange on the weak anion exchange resin – with the uncontrollable higher flow rates during target unloading and ethanol elution the process becomes less reliable, than in manual setup. Therefore, we turned to an alternative fluoride preparation procedure.
In this work we have successfully used 2.5% ammonium hydroxide aqueous methanol solution for eluting [18F]fluoride from the QMA cartridge to obtain a reactive source of fluoride that is soluble in polar solvents and free from interfering ions. Evaporation of the eluate gave pure ammonium fluoride NH418F, that reacted quantitatively with boron trifluoride substrates in both methanol and acetonitrile complexation forms.
It shall be noted that excessively high temperatures, both in the fluoride drying step and in the radiolabelling reaction can dramatically reduce the radiochemical yields (Table 3), probably due to thermal decomposition of ammonium fluoride under vacuum at elevated temperatures.
In the published procedures, purification of the final product is achieved by solvent evaporation followed by formulation in physiological buffer solution and [18F]fluoride removal using an alumina solid phase extraction cartridge. It was noted by several researchers that a series of two or more alumina cartridges is needed to achieve the desired 95% radiochemical purity. We had the same experience in the beginning of this work when alumina was used for fluoride removal after product formulation.
Khoshnevisan, with colleagues [1], has supposed that alumina could catalyse hydrolysis of [18F]TFB. This is corroborated by studies of tetrafluoroborate decomposition in the presence of aluminium ions by Katagari et al [24]. Yoshioka et al. have shown also, that TFB is effectively adsorbed on alumina doped with magnesium oxide [25]. In this study we also observed losses of the final product (20–30%) on the alumina cartridge and poor radiochemical purity when only one alumina cartridge was employed (data not presented).
A simple manual experiment, consisting of passing non-radioactive TFB of known concentration through alumina and measuring fluoride and tetrafluoroborate by a conductivity detector in anion-exchange HPLC (Fig. 5) confirms TFB decomposition (rise of fluoride concentration) and its adsorption on alumina (decrease of tetrafluoroborate signal). This observation explains reduction of the radiochemical yields noticed when alumina is employed for fluoride extraction.
In order to overcome this problem, we developed an alternative purification procedure. The crude reaction mixture is dissolved in anhydrous ethanol. The mixture is passed through the Florisil and QMA solid phase extraction cartridges. Florisil is a magnesium oxide loaded silica-based adsorbent. Magnesium fluoride has the lowest solubility in water among all metal fluorides. Florisil effectively traps all the unreacted [18F]fluoride and does not retain [18F]TFB. Thus, the reaction mixture dissolved in ethanol after passage through florisil contains [18F]tetrafluoroborate, acetonitrile and unreacted boron trifluoride-methanol complex. [18F]TFB is efficiently trapped on QMA anion exchange resin and, unreacted neutral boron trifluoride passes through into the waste, thus reducing amount of carrier tetrafluoroborate. This is confirmed by the fact that while 0.39 mmol boron trifluoride equivalent is added into reaction vessel only 0.17 ± 0.07 mmol tetrafluoroborate is measured in the final formulation. After rinsing the QMA cartridge with water (10 mL) the final product is eluted with a physiologic saline solution.
Preclinical imaging
To determine if high molar activity [18F]TFB had effective imaging properties we produced NIS-expressing lung lesions by injecting A549-LN cells into the tail vein of mice. We identified multiple small lung tumours as early as day 48 with high [18F]TFB uptake and high tumour-to-normal lung contrast (Fig. 6). Further, we were able to image lung tumours longitudinally by performing repeated imaging on day 81.