Simple preparation and purification of ethanol-free solutions of 3′-deoxy-3′-[18F]fluorothymidine by means of disposable solid-phase extraction cartridges

Available online at www.sciencedirect.com

Nuclear Medicine and Biology 39 (2012) 540 – 550 www.elsevier.com/locate/nucmedbio

Simple preparation and purification of ethanol-free solutions of 3′-deoxy-3′-[ 18 F]fluorothymidine by means of disposable solid-phase extraction cartridges☆,☆☆ Claudio Pascali a,⁎, Anna Bogni a , Lorenza Fugazza b , Claudio Cucchi a , Ornella Crispu c , Luca Laera a , Ren Iwata d , Greta Maiocchi a , Flavio Crippa a , Emilio Bombardieri a a

Fondazione IRCCS Istituto Nazionale dei Tumori, V. Venezian, 1-20133, Milan, Italy Present address:Advanced Accelerator Applications, V. Ribes, 5-Colleretto Giacosa, Italy c Present address: Dani Instrument Spa, V. Brianza, 87-20093 Cologno Monzese, Milan, Italy d Cyclotron and Radioisotope Center, 6-3 Aoba, Aramaki, Aoba-ku, Sendai 980-8578, Japan Received 12 July 2011; received in revised form 20 September 2011; accepted 2 October 2011 b

Abstract Introduction: 3′-Deoxy-3′-[ 18F]fluorothymidine ([ 18F]FLT) shows great potential as a tracer for proliferative studies with PET. However, its regular application is often limited by low radiochemical yields and the use of a troublesome HPLC separation. Moreover, a high content of ethanol, at least one-fold higher than the European Pharmacopoeia and US Pharmacopoeia's established limit, is always present in the final product. The present study reports an optimization of the reaction conditions and a simple and straightforward purification step which affords a solution of [ 18F]FLT suitable for human use. Methods: Several conditions and materials were tested for both the nucleophilic substitution and purification step. The latter was achieved by means of a series of commercial solid-phase extraction cartridges. Very conveniently, the whole one-pot synthesis was carried out on commercial automated modules using basically the same setup employed for the synthesis of [ 18F]FDG. Results: Under routine conditions, radiochemical yields of 37% [decay-corrected to start of synthesis (SOS)] were achieved in ca. 39 min from SOS, with radiochemical purities N98% (usually N99%). The negligible radiolysis observed could be easily suppressed by adding 0.5% of EtOH. Typical unlabelled chemical impurities detected were thymidine (0.15 ppm), thymine (0.28 ppm) and stavudine (0.05 ppm). Conclusions: A reliable, simple and efficient preparation of [ 18F]FLT has been developed, able to afford an ethanol-free solution of the tracer with no need for any HPLC purification. Because of its similarity to the [ 18F]FDG synthesis, the method can be readily implemented on basically all the commercial modules developed for this common radiotracer. © 2012 Elsevier Inc. All rights reserved. Keywords: [ 18F]FLT; PET tracer; Fluorothymidine; SPE cartridge purification; Automation

1. Introduction Positron emission tomography (PET) is a well-established technique used in oncology for the diagnosis, staging and ☆ Preliminary results of this work were presented at the 15th European Symposium on Radiopharmacy and Radiopharmaceuticals held in Edinburgh, Scotland (UK) on April 8–11, 2010. ☆☆ This research has been partially supported by the Funds “5 per mille” of the Italian Ministry of Health and by the AIRC/Regione Lombardia Project “Molecular and Cellular Imaging of Cancer”, Milan. ⁎ Corresponding author. Tel.: +39 02 23903405; fax: +39 02 2367874 E-mail address: [email protected] (C. Pascali).

0969-8051/$ – see front matter © 2012 Elsevier Inc. All rights reserved. doi:10.1016/j.nucmedbio.2011.10.005

monitoring of tumour response to therapy. Although [ 18F] FDG remains, thanks also to its high sensitivity, the most widely used imaging agent for tumours, false-positive findings can occur in inflammatory lesions [1]. An inflammatory response to treatment is the reason why elevated uptakes of [ 18F]FDG can persist in the early stage during chemotherapy even though tumour themselves are not growing [2]. To circumvent this pitfall, much effort has been devoted to the development of more selective tracers. In this way, agents that measure tumour proliferation may have greater specificity and their uptake may decrease more rapidly in response to successful therapy [3]. Among these,

C. Pascali et al. / Nuclear Medicine and Biology 39 (2012) 540–550 11

[ C]thymidine was introduced to specifically target the increased biosynthesis of DNA in tumours [4,5]. However, its regular clinical use was prevented by the rapid degradation by thymidine phosphorylase present in blood, liver, spleen and tumour tissue, leading to the formation of several 11C-labelled metabolites. 3′-Deoxy-3′-[ 18F]fluorothymidine ([ 18F]FLT) is a fluorine-18-labelled analogue of thymidine, introduced in the late 1990s, able to withstand in vivo degradation by thymidine phosphorylase owing to the replacement of the hydroxyl group in the 3′-position with a fluorine atom [6]. Results show retention of [ 18F]FLT within the cell to provide a measure of cellular thymidine kinase activity and therefore an approximate measure of cellular proliferation [7–12]. This feature makes it attractive not just for the imaging of tumours but even more so for the early evaluation of tumour response to chemo- or radiation therapy [2,7,13]. However, more work, especially in terms of clinical evaluation in a large patient population, is still required in order to attain a better understanding of this radiotracer potential and limits. Unfortunately, the fairly complex production procedure has so far limited the number of sites capable of synthesizing [ 18F]FLT. In fact, even if the first attempt to synthesize [ 18F] FLT [14] has been steadily followed by considerable improvement and changes [15–19], especially in terms of radiochemical yields, the procedures remain time consuming, mainly due to the semipreparative HPLC purification usually applied. In addition, the retention time of FLT has been observed to be pH dependent, with best resolution obtained apparently under neutral pH, a consideration that has forced some authors to perform a neutralization with AcONa just before injection [20–22]. Last, the high ethanol content (5–15%), often present in the mobile phase, could also be a source of poor peak resolution and therefore a font of impurities in the collected fraction [19]. As a result, the limits described above have been recently addressed by several publications trying to overcome the problems by replacing the HPLC purification with one based on solidphase extraction (SPE) cartridges [20,22–29]. Although encouraging, most of these works lacked essential information (typically, the chemical purity which could be entirely missing or ill-determined with unsuitable chromatographic methods) or showed some drawbacks themselves, such as insufficient radiochemical purities or quite complex procedures requiring two reactors or elaborate workups, and therefore difficult to implement on most simple automation modules such as those used for the synthesis of [ 18F]FDG. Most disturbingly, all these methods still yielded a product with an ethanol content of roughly an order of magnitude higher than the European Pharmacopoeia's threshold, a fact that should prevent the human use of the tracer, unless specifically justified. Based on our recent patent [30], a simple and efficient automated procedure for the synthesis and SPE cartridgebased purification of [ 18F]FLT, able to provide a basically

541

ethanol-free product, will be presented and discussed herein. As a proof of its straightforward adaptability, the whole synthesis was easily implemented and carried out on two different commercial [ 18F]FDG synthesizers with no major change required. 2. Materials and methods 2.1. General [ 18F]Fluoride was produced by the 18O(p,n) 18F nuclear reaction on [ 18O]H2O (N97% enriched; Cambridge Isotope Laboratories, Andover, MA, USA) using a Scanditronix MC 17 cyclotron and a 1.1-ml silver target. Borosilicate reaction vials (Pierce) were coated with Aquasil (0.77% solution in water) following the instruction given by the manufacturer (Pierce Chemical). Accell QMA-carbonate Light, Alumina N Light and tC18 Plus (400 mg), and Light (145 mg) Sep-Pak were obtained from Waters (Milford, MA, USA). Chromafix PS-HCO3 (45 mg) cartridges were purchased from Macherey-Nagel (Duren, Germany). MaxiClean IC-H cartridges (0.5 ml) were obtained from Alltech. Millex-GS 0.22-μm (vented) sterile filters were purchased from Millipore. All SPE cartridges were conditioned before use: Alumina N with water for injection (WFI, 2 ml) followed by air; PSHCO3 and QMA with EtOH (2 ml), WFI (4 ml) and air; IC-H with MeOH (1 ml), WFI (2 ml) and air. The entire assembly filter-tC18 Sep-Pak was conditioned according to the following procedure: (1) the tC18 Sep-Pak was rinsed with MeOH (3–4 ml); (2) the filter was filled with WFI and fitted at the top of the tC18; (3) WFI (6–8 ml) was slowly passed through the two now joined components; (4) air was flushed with a syringe to empty the filter. The FLT precursor, 1-(2′-deoxy-3′-O-(4-nitrobenzenesulfonyl)-5′-O-(4,4′-dimethoxytrityl)-β-D-threo-pentafuranosyl)-3-N-(tert-butyloxycarbonyl)thymine (3-N-boc-5′-ODMTr-3′-O-nosyl-lyxothymidine), was purchased from ABX Advanced Biochemical Compounds (Radeberg, Germany) and Huayi Isotopes (China), with declared purities above 95% and 97%, respectively. The cold standard 3′-deoxy-3′-fluorothymidine was purchased from ABX. 3′-Chloro-3′-deoxy-thymidine (chlorothymidine or CLT or zidovudine impurity B) was obtained from LGC Promochem. 2′,3′-Didehydro-3′-deoxythymidine (stavudine or d4T, N98%) was purchased from Sigma. Tetrabutylammonium hydroxide 30-hydrate (TBAOH·30H2O, 99%) was purchased from Fluka, while 4,7,13,16,21,24hexaoxa-1,10-diazabicyclo[8.8.8]hexacosane (Kryptofix K2.2.2, 99%) and dry CH3CN (DNA grade, 99%) were purchased from VWR. Potassium carbonate (99%) and potassium bicarbonate (98%) were obtained from Aldrich. Other chemicals and solvents were of analytical grade and were obtained from conventional chemical suppliers. Products were tested for chemical and radiochemical purity by HPLC on a Rocket Altima C18 column (53×7 mm,

542

C. Pascali et al. / Nuclear Medicine and Biology 39 (2012) 540–550

3 μm) eluted at 1 ml/min with either EtOH/H2O 10:90 or 30:70 (v/v), with the latter conditions used to check for possible long-retained impurities. The HPLC output was routinely monitored for chemical and radiochemical purity by a UV/Vis detector (Series 200, Perkin Elmer) set at 267 nm wavelength in series with a Radiomatic Flo-One A515 (Canberra-Packard). Alternatively, to better understand the nature of some impurities, a UV/Vis photodiode array detector (PDA type SPD-MDAvp, Shimadzu) was also used. UV/Vis spectra (from 190 to 800 nm) were acquired by the PDA during the analysis for each peak and afterward compared via software (Class VP 7.3) to those of pure standards to confirm their identity and exclude the presence of any co-eluting compound. Additionally, each batch of [ 18F]FLT was analyzed by gas chromatography (Hewlett-Packard mod. 7890; Agilent Innowax column 30 m×0.53 mm, 1 μm; helium flow rate 4 ml/min; FID detector) to determine the content of any residual solvent. K2.2.2 in the final product solution was determined semiquantitatively according to two different thin-layer chromatographic methods reported in the literature [31,32]. As for TBA, it was detected by HPLC on C18 Luna column

(100×4.6 mm, 100 A, Phenomenex) eluted at 0.6 ml/min with 4 g/L solution of toluenesulfonic acid/acetonitrile (25:75, v/v) with the UV detector set at 254 nm. 2.2. Preparation of K2.2.2/K2CO3, K2.2.2/KHCO3 and TBAHCO3 solutions For convenience, the title solutions were prepared in large batches from which 1-ml aliquots were then dispensed in single-use septum-capped vials. The 1-ml CH3CN/H2O (90:10, v/v) Kryptofix solutions contained either 20 mg of K2.2.2 and 3.5 mg of K2CO3 or 15 mg of K2.2.2 and 3.5 mg of KHCO3. The 30 mM tetrabutylammonium bicarbonate (TBAHCO3) solution in CH3CN/H2O (90:10, v/v) was prepared by dissolving TBAOH·30H 2O in H2O and bubbling CO2 till pH 8–9 before adding the required amount of CH3CN. 2.3. Radiochemical synthesis of [ 18F]FLT The automated one-pot syntheses and following purification by disposable SPE cartridges were carried out on either a Modular-Lab (Eckert & Ziegler; reactor made of Aquasil-treated borosilicate) (Fig. 1) or a normal TracerLab

Fig. 1. Modular-Lab (Eckert & Ziegler) control panel developed for the synthesis of [ 18F]FLT.

C. Pascali et al. / Nuclear Medicine and Biology 39 (2012) 540–550

FX-FDG (single, General Electric; reactor made of glassy carbon) module using basically the same setup of an [ 18F] FDG preparation. After synthesis, an automated cleaning procedure was performed in between the runs. [ 18F]Fluoride (1.2–18.5 GBq) was captured on an anion-exchange cartridge (Chromafix PS-HCO3 or Accell QMA-carbonate) and eluted with a 1.0-ml solution of TBAHCO3. As an alternative, analogous volumes of K2.2.2/K2CO3 or K2.2.2/KHCO3 were also tested and later abandoned. After a vacuum-assisted evaporation till dryness at 75°C under a gentle stream of helium, the FLT precursor (10–25 mg in 1 ml dry CH3CN) was added and the reaction mixture was allowed to react in the closed reactor at 100°C for 5 min. After a partial evaporation down to ca. 350–500 μl, a hydrolysis step with HCl 1 M (2 ml) at 90°C for 3.5 min took place. The reaction mixture was then reduced in volume (ca. 0.5–0.7 ml) by vacuum-assisted evaporation to remove the CH3CN still present, cooled down to 40°C, passed through a small strong cation-exchange cartridge (MaxiClean IC-H) and conveyed to a vented sterile filter (Millex-GS) joined to a reversed-phase cartridge (two tC18 Sep-Pak Light cartridges or a single tC18 Sep-Pak Plus) and from here to the waste. An additional 0.8 ml of WFI (“H2O #1” in Fig. 1) was used to rinse the reaction vessel and the cation-exchange cartridge. WFI (16–20 ml, “H2O #2”) was then used to remove most of the impurities from the tC18 Sep-Pak and send them to the waste. Finally, the product was eluted with more WFI (10–15 ml, “H2O #3”) throughout an Alumina N Light Sep-Pak and sterile filter and collected in the vented sterile and pyrogen-free product vial. Typical elution flow rates were 2–3 ml/min. 2.4. Stability test of [ 18F]FLT solutions The radiochemical purity of a solution of [ 18F]FLT (10 ml, 2.4 GBq), obtained starting from 15 mg of precursor, was determined at the end of synthesis (EOS) by radio-HPLC. The solution was then split in two 5-ml fractions stored in closed vials. To one of these, the OH radical scavenger EtOH (25 μl) was added so as to achieve the European Pharmacopoeia's limit content of 0.5%. Radiochemical purity for both fractions was checked after 4.5–5 h. 2.5. Cold elution test of a stavudine–FLT mixture through a tC18 Sep-Pak with and without the use of a vented filter In order to assess possible discrepancies in the purification caused by a disruption of the reversed-phase bed, a solution made of stavudine (2.5 mg) and FLT (0.4 mg) in 0.5 ml of WFI was loaded on a tC18 Sep-Pak Plus previously conditioned with MeOH (4 ml) and WFI (8 ml). To simulate the worst loading condition of the reaction mixture on the tC18 Sep-Pak cartridge during the transfer from the reactor, a 10-ml syringe full of air was then emptied through the tC18

543

Sep-Pak, after which the elution was carried out with WFI at ca. 2–3 ml/min. Fractions of 1.5 ml were collected and analyzed by HPLC to determine the content of both compounds. These data were then used to build their elution profile, reported as percentage of compound eluted within a given volume (Fig. 3, top, solid lines). The test was repeated on a second tC18 (belonging to the same lot) placing this time a vented filter on top of it to prevent any air from disrupting the sorbent bed (Fig. 3, top, dotted lines). 2.6. Cold elution test of a stavudine–FLT–K2.2.2 mixture through a vented filter — tC18 Sep-Pak In order to assess the influence of K2.2.2 on the separation of stavudine from FLT, the same procedure above described for the vented filter- tC18 Sep-Pak Plus setup, was applied this time to a mixture of stavudine (2.5 mg), FLT (0.4 mg) and K2.2.2 (3 mg) in 0.5 ml of WFI. The resulting elution profile is shown in Fig. 3 (bottom). 3. Results A summary of the overall [ 18F]FLT produced under different reaction conditions is shown in Table 1, while the radioactivity distribution in all components and solutions is reported in Table 2. Radiochemical yield and overall process times varied mostly according to the phase-transfer agent, base, precursor amount and tC18 Sep-Pak used, with the last one affecting the volumes of H2O #2 and H2O #3. Thus, using TBAHCO3 and 20 mg of precursor, 10-ml sterile and pyrogen-free solutions of [ 18F]FLT were obtained within 37 min and 30% radiochemical yields, decay-corrected to start of synthesis, when two tC18 Light were used, while these values increased to 39 min, 15 ml and 37%, respectively, using one tC18 Plus. Basically, two tC18 Light (2×145 mg) allowed for smaller volumes — and thus shorter elution times — of H2O #2 and H2O #3 when compared to a single tC18 Plus (400 mg). On the other hand, the extra amount of adsorbent in the latter type guaranteed a better separation (see Discussion). In both cases, yields could be raised by a few extra percentile points by simply increasing the collected volume, i.e., H2O #3, although to the detriment of the product concentration. Table 1 Overall [ 18F]FLT produced as a function of the amount of precursor and the type of phase-transfer agent used Precursor (mg)

10 15 20 25

Total [ 18F]FLT eluted, decay-corrected to SOS⁎ (%) Using TBAHCO3

Using K2.2.2/K2CO3

Using K2.2.2/KHCO3

27.3 36.0 42.9 58.4

– 24.1 28.5 –

– 41.8 47.8 –

⁎ Determined as sum of [ 18F]FLT present in the “waste” and “product” fractions plus the residual [ 18F]FLT washed out from the cartridges at EOS using 5 ml of 50% EtOH.

544

C. Pascali et al. / Nuclear Medicine and Biology 39 (2012) 540–550

Table 2 Activity distribution at synthesis completed, when 15 mg of precursor was used a % of activity Product Waste QMA or PS-HCO3 IC-H Filter on tC18 Plus tC18 Plus Al2O3 N Cleaning Balance

29.6±8.1 24.1±4.3b 3.0±1.6 11.2±4.2c 0.4±0.2 15.6±1.7 4.7±1.1 4.8±0.6 94.4±3.1

a Data reported are decay corrected at SOS and represent the mean and standard deviations of 20 experiments. b 18 [ F]FLT content: 6.2±3.2. c On two experiments where the rinsing with H2O #1 was not applied, this value was ca. 19%.

No major difference in radiochemical yields and synthesis duration was observed between the two automated modules. The assumed higher inertness of glassy carbon toward fluoride was compensated by coating the borosilicate vial with a siliconizing agent (Aquasil), a technique successfully tested in a previous study on [ 18F]FDG [33]. Product and impurities were identified by HPLC by comparison of their retention time with those of standards. Further confirmation was given by the software-mediated comparison of the UV absorbance spectrum obtained with the PDA detector on sample and standards. Starting from 20 mg of precursor and using a tC18 Plus, typical unlabelled chemical impurities present in a 15-ml final product were thymidine (0.15 ppm), thymine (0.28 ppm) and stavudine (0.05 ppm). At the early stage of this study, negligible amounts of furfuryl alcohol (b0.05 ppm) and the chlorine analogue chlorothymidine (b0.05 ppm) were also very occasionally found, to be later totally removed once the method was optimized. All impurities were well below their limits of toxicity, thus authorizing the human use of the [ 18F]FLT solution. Typical radio-HPLC chromatograms for “product” and “waste” solutions are shown in Fig. 4A and B, respectively. Sample injections were also repeated, increasing the amount of EtOH in the mobile phase to 30% over a 60-min long run, so as to observe any possible long-retained compound. No additional impurities were detected apart from those mentioned above. In view of the very low height of the impurities' peaks, the use of two isocratic conditions was preferred to a gradient because — at least on our own system — it offered a definitely better baseline and, consequently, an increased signal-to-noise ratio. Radiochemical purities were always N98% (generally N99%). At the low–medium activity levels used for our animal studies, only a very slight decrease in radiochemical purity from 99.6% to 97.4% was observed over a 4.5- to 5-h range for the EtOH-free solution of [ 18F]FLT. No change at all was observed instead when a 0.5% EtOH was added to the product solution. Specific activity at EOS for the 2.4-

GBq, 10-ml [ 18F]FLT solution used for the stability test was 290 GBq/μmol. Organic solvents, when present at all, were found only in traces, well below the limits set by the European Pharmacopoeia [34]. No TBA was detected in either the final product solution (detection limit: 10 ppm) or the waste. When K2.2.2 was used as phase-transfer agent, its content in the product solution was below the detection limit for the method (20 ppm), although considerable amounts — up to 10% of the original quantity used — were present in the waste.

4. Discussion The main aims of this work were (a) to find a way to produce [ 18 F]FLT which could be easily and readily implemented in most of the ubiquitous automated modules used for the radiosynthesis of [ 18F]FDG and (b) to achieve reasonable and consistent radiochemical yield and purities without HPLC purification and using only reagents and materials already thoroughly tested and of common use. Thus, innovative routes such as that suggested by Kim et al. [35] and Lee et al. [18,36] making use of protic solvents and apparently offering higher yields were not taken into consideration in view of contrasting reports from other authors [20]. Likewise, the use of ionic liquids as reaction medium to decrease the amount of precursor and eliminate the azeotropic drying process was ruled out for the reason that it did not work well with aqueous volumes of [ 18F] fluoride above 100 μl [37]. Moreover, the above procedures still required a time-consuming and delicate HPLC purification. Also, a new interesting precursor, which promised to simplify the purification step through a 2,5′-anhydro moiety able to provide an intramolecular protection strategy for both the thymidine amide and the 5′-hydroxyl group, did not offer at the moment satisfactory yields [38]. Hence, it was decided to work only on well-established procedures and reagents, for which much data were already available in the literature. According to several reports and comparison studies, 3N-boc-5′-O-DMTr-3′-O-nosyl-lyxothymidine (Fig. 2) seems to be the precursor affording the highest incorporation yields of [ 18F]fluoride and therefore most widely used [21]. For this reason, it was also selected as a substrate in this study, a choice further supported by the sometimes contrasting wide range of reaction conditions and yields reported in the literature, which gave rise to the notion that there was still a margin of improvement to the reaction itself. At first, a temperature of 160°C was set for the SN2 reaction. This value was changed after a few experiments as judged too inconvenient for most module apparatus. This is because heating, and even worse, cooling, is usually a timeconsuming operation. Besides, the overpressure generated inside the reactor (N300 kPa) by the combination of high temperature and low-boiling solvent could be a source of leaks, an event often observed on the TracerLab's valves

C. Pascali et al. / Nuclear Medicine and Biology 39 (2012) 540–550

545

Fig. 2. Synthesis of [ 18F]FLT and observed by-products.

directly connected to the reactor. Such setback was also reported by Oh et al. [19] who resorted to split the heating into two separate phases — a short one at 150°C, followed by an 8-min-long cooling down to 85°C — in order to bypass the problem. Although both the above pitfalls could be overcome on the Modular-Lab, since its modular approach allowed it to be fitted with high-pressure–resistant valves while the Peltier unit could speed up the cooling, we discarded these drastic reaction conditions because they are too limiting for an implementation of the method on a wider range of commercial systems. According to several authors, high temperatures as well as the amount of precursor in the range of 30–45 mg are necessary in order to get good incorporation values of [ 18F] fluoride [19,20,22,39]. However, more recently, some works have brought into question these points, directing instead the attention to the competitive elimination reaction leading to stavudine and the need to suppress it by drastically decreasing the amount of base (K2.2.2/K2CO3) employed, since, in the presence of an excess of base, the precursor is consumed quickly by the E2 elimination mechanism before the SN2 reaction is completed [40,41]. 1 Likewise, reaction conditions which could most likely accelerate the E2

1 On the other end, a considerable reduction in the amount of base must be weighed against the negative impact it might have on the recovery of [ 18F]fluoride from the anion-exchange cartridge.

reaction and contribute to rapid precursor decomposition, such as excessive heating, should be avoided. Accordingly, and in agreement with other authors' reports, a lower temperature of 100°C was set for the SN2 reaction [40]. In addition, three different conditions were tested regarding base and phase-transfer agent: (a) K2.2.2/K2CO3; (b) K2.2.2/KHCO3; and (c) TBAHCO3. The use of the last one in substitution for the more thermally stable K2.2.2 was allowed by the lower reaction temperature adopted in this study [42]. As shown in Table 1, the use of the more basic K2CO3 yielded lower incorporations of [ 18F]fluoride in comparison to KHCO3. Reports comparing K2.2.2 and TBA in their use as phasetransfer agents during [ 18F]fluoride nucleophilic substitution are contrasting, although K2.2.2 is generally regarded as better able to cope with metals in the water and thus provide more reliable results and better yields [43]. This remark was also confirmed in our own study, with slightly higher [ 18F] fluoride incorporations provided by K2.2.2/KHCO3 compared to those by TBAHCO3 (Table 1). However, under the specific conditions employed in this synthesis, K2.2.2 was definitely less retained than TBA by the strong cationexchange resin during the purification phase. In fact, while no TBA could be detected in either waste or product solutions, relevant quantities of K2.2.2 were actually found in the waste. Even though its quantity in the product was below the threshold value set by the European Pharmacopoeia, the coeluting K2.2.2 was deemed to negatively interfere with the

546

C. Pascali et al. / Nuclear Medicine and Biology 39 (2012) 540–550

separation on the tC18. In fact, much more [ 18F]FLT had to be given up in the waste fraction in order to keep the chemical purity of the product at comparable levels, so that, in the end, the final radiochemical yield was actually lower than that obtained with TBAHCO3. Accordingly, TBA was selected as a phase-transfer agent in the routine preparations. As shown in Table 1, even operating at relatively low temperature (100°C), slightly better incorporations of [ 18F] fluoride were achieved with increased amount of precursor, a result somewhat in contrast with Teng et al. [22] who observed this trend only at much higher temperatures. The hydrolysis was initially accomplished by the straightforward addition of HCl at the end of the 5-min nucleophilic substitution reaction. This procedure, although simple and efficient, presented a major disadvantage in the following step, i.e., the removal of CH3CN prior to the transfer to the SPE cartridges, a condition essential for achieving a good, reproducible purification [17]. In fact, the large volume involved (1 ml CH3CN+2 ml HCl) required an excessively long vacuum-assisted evaporation to safely reduce the CH3CN content to acceptable levels (a few microliters). The opposite alternative, i.e., the almost complete removal of CH3CN before HCl, was also tested but was soon abandoned because, apart from affording lower radiochemical yields, minimum variations in helium flow, vacuum and starting volume could easily lead to overdrying and thus to decomposition or, at least, a difficult resolubilization in HCl. Not surprisingly, an unidentified sloweluting unlabelled impurity was observed in the HPLC chromatogram when the latter approach was followed. Therefore, since some CH3CN is necessary to assist the hydrolysis, it was judged more convenient to perform a very brief evaporation to reduce the volume down to 350–500 μl before adding HCl. By so doing, the final removal of the organic solvent before the purification step was accomplished in roughly one third of the time previously needed. Similar conclusions on how the extent of CH3CN evaporation after [ 18 F]fluorination affects the overall radiochemical yield were reported by Tang et al. [20] and Oh et al. [19] who observed with a complete evaporation a reduction of 5% and 10%, respectively, compared to the case which left 0.2–0.5 ml of residual CH3CN. With a complete evaporation, formation of a black oil residue difficult to dissolve in HCl, as well as a related increase in the residual activity in the reactor after synthesis, was also reported [19]. Starting point for the isolation of the product from the reaction mixture was the existing HPLC literature employing a C18 column eluted with water containing 5–15% EtOH. The logical step was to transfer the same conditions to a tC18 Sep-Pak. However, in order to fully comply with European Pharmacopoeia and US Pharmacopoeia's limits on EtOH content for an intravenous injectable solution, 100% WFI was selected as eluent. Literature data regarding the HPLC purification of [ 18F] FLT reaction mixtures often describe the process as far from ideal, especially when large amounts of precursor (35–45

mg) are used, forcing some authors to adopt a pre-purification step. A possible explanation for that might be that large amounts of K2.2.2 present in the crude mixture might spoil the separation by either interacting with the other co-products or “overloading” the adsorbent capacity. By running some cold tests consisting in loading a solution made up of FLT, K2.2.2 and stavudine on a tC18 Sep-Pak, eluting with water and observing the collected fractions with UV, it was possible to observe how amounts of K2.2.2 close to those found in the “waste” were able to anticipate the elution and spoil the separation (Fig. 3, bottom). Although a better flow control during the passage through the IC-H cartridge helped to decrease the amount of K2.2.2 passing through and, accordingly, to reduce the problem, better results could be achieved by replacing K2.2.2 with TBA, since its trapping by the cation-exchange resin was more efficient and less susceptible to high flow rates. A source of poor and/or irreproducible separation was the potential presence of gas bubbles on the tC18 SepPak. The importance for silica-based sorbents to form a homogeneously wet, compact bed and the need to prevent them from running dry so as to avoid the formation of preferential channels is a well-known problem made, in our case, more critical by the need to run the whole process in an automated mode. A simple and reliable solution to this issue was achieved by placing an ordinary wetted 0.2-μm sterile vented filter on top of the tC18 SepPak to prevent air and any possible insoluble side-product from passing through. The vented filter placed on top of the Sep-Pak was essential in maintaining the integrity of the reversed-phase bed and thus ensuring an efficient and reliable separation. This method was quite easy to implement into the automated process and gave good results. Evidence of that can be seen in Fig. 3 (top), representing the elution profile of a mixture of stavudine and FLT through a tC18 Sep-Pak, when air was allowed or not allowed (i.e., with the vented filter) to pass through after the loading phase. As expected, when air was let in, the compounds eluted earlier and their separation spoiled to some extent. One more aspect likely to affect the separation could be the flow rate. Since the loading of the reaction mixture onto the reverse-phase Sep-Pak, i.e., the transferring step from the reaction vessel, was reckoned to be the most critical for the separation, it was done at 1–2 ml/min, whereas the following rinsing and elution were done as a precautionary measure at 2–3 ml/min (see discussion below), even though higher values (4–5 ml/min) were occasionally reached without any major change on the separation. 2

2

A certain measure of flow control was easily achieved on the TracerLab by placing a 3-way valve on the external line supplying the inert gas. By so doing it was possible to select, when a lower flow was required, a gas line set at a lower pressure.

C. Pascali et al. / Nuclear Medicine and Biology 39 (2012) 540–550

547

Fig. 3. Top: Elution profile of a solution (0.5 ml) of stavudine (2.5 mg)+FLT (0.4 mg) eluted with WFI from a tC18 Sep-Pak Plus, with and without the support of a vented filter. Bottom: Elution profile of a solution (0.5 ml) of stavudine (2.5 mg)+FLT (0.4 mg)+K2.2.2 (3 mg) eluted with WFI from a tC18 Sep-Pak Plus, when a vented filter was placed on top of it. For comparison, the elution profile of the same mixture but without K2.2.2 is also reported.

The volumes of H2O #2 and H2O #3 were empirically determined to optimize recovery and purification of [ 18F] FLT, while keeping the final volume within reasonable values. Higher radiochemical yields might easily be achieved by cutting the volume of H2O #2 and, consequently, increasing the level of unlabelled impurities carried over. Conversely, an increase of H2O #2 would afford an almost impurities-free product, although in lower radiochemical yields. As expected, the less the precursor used, the better the separation, the less the percentage of [ 18F]FLT lost in the waste. Under our typical conditions (15 mg precursor; TBAHCO3; tC18 Plus; stavudine content in the final product b0.05 ppm), ca. 20–30% of the starting activity was recovered in the waste, 29–35% of which was [ 18F]FLT (Fig. 4B), and the radiochemical yield decay corrected was

20–25%. As an indication, when 25 mg of precursor was used, these values increased to 25–35% and 60–70%, respectively, while the radiochemical yield was 34–39%. An even worse trend was observed with K2.2.2, since its presence, as clearly shown in Fig. 3 (bottom), negatively affected the separation. Thus, the slightly higher 18Fincorporations achievable with K2.2.2/KHCO3 were practically annulled by the spoiled separation. Obviously, this could be compensated by either employing bigger solidphase extraction cartridges or improving the blocking of K2.2.2 on the strong cation-exchange resin. However, both measures would imply other negative side effects, like a larger volume required for the elution from both cartridges. As far as the precursor is concerned, complaints regarding the quality of the commercial starting material

548

C. Pascali et al. / Nuclear Medicine and Biology 39 (2012) 540–550

Fig. 4. Radio- (top) and UV- (bottom) HPLC profiles of (A) a typical SPE cartridge-purified [ 18F]FLT solution; (B) a “waste” fraction. Activities are decay corrected at the time of injection. Conditions: Rocket Altima C18 column (53×7 mm, 3µm) eluted at 1 ml/min with EtOH/H2O 10:90. UV wavelength: 267 nm.

were quite common among users until a few years ago. In our hands, a decrease of yield — sometimes accompanied also by unusual amounts of or unidentified unlabelled impurities — was observed whenever an “old” precursor, i.e., taken out from a 1-g bottle repeatedly open, was used or when, because of a delivery problem, a new lot of precursor was left at room temperature for about 2 weeks. The reason probably lay in some decomposition, possibly the removal of a protecting group, taking place with time and accelerated by the air exposure and the room temperature. A total failure was even recorded on one occasion when, because of a delay with the cyclotron bombardment, the reagent solution was left standing in the module for 3 h. Obviously, such a problem should be put in perspective by the use of commercial single-dose vials under argon as well as of a starting material of higher purity grade. Apart from the precursor, a big role on reproducibility was played by the tC18 Sep-Pak. Several different lots were used in the course of this study. Differences among them in terms of packing, particle size and carbon contents had annoying repercussions on back pressure and separation, requiring some correction (in the ±20% range) for H2O #2 and H2O #3 in terms of both volume and flow rate. Anyway, when the Sep-Pak belonged to the same lot, the elution profile was always reproducible. Though these variations seem to have declined with more recent batches, the problem must be kept under control by checking the characteristics on the leaflet each time a new lot is introduced in production. Before realizing the existence of this problem, for instance, the flow rate almost doubled up

and, worst of all, the separation of [ 18F]FLT from the hot and cold impurities proved inadequate. A check afterward of the material leaflet showed then an abnormally large size for the particulate. Because of these problems, reversedphase silica-based materials from other brands were investigated. While some of these proved to be worse, others, like for instance the High Capacity C18 Maxi-Clean (Alltech) and Chromafix C18ec (Macherey-Nagel), afforded even better separations but were nonetheless abandoned because of the considerably larger volumes — and therefore longer time — needed for H2O #2 and H2O #3. For the same reason, it is advised against using two joined tC18 Sep-Pak Plus. Conversely, switching from one tC18 Sep-Pak Plus (400 mg) to two Light Sep-Pak (290 mg) helped to cut down the volumes. However, considering that sorbent quantity affects the maximum amount of sample loadable onto a column (as a rule of thumb, about 5% of the sorbent quantity), the extra leverage offered by the Plus type was judged more convenient, especially taking into account the above observations on the poor quality of these tC18 Sep-Pak. Polymer-based reversed-phase sorbents such as OASIS (Waters), Chromafix HR-P (Macherey-Nagel) or STRATA (Phenomenex) were tested, too. These kinds of materials, which have the appealing property of better withstanding the disruption caused to the packing from any air bubble passing through, would have made the vented filter on top unnecessary, thus simplifying both setup and automation. Unfortunately, experiments showed [ 18 F]FLT to be irreversibly trapped unless some organic solvent was used for the elution.

C. Pascali et al. / Nuclear Medicine and Biology 39 (2012) 540–550

Despite the variation of the factors involved in the purification process (crude reaction mixture composition, tC18 Sep-Pak's consistency, flow rates), the elution profile was quite consistent, so that a cautious choice for the volume of H2O #2 could always make certain of a good outcome. An expected impurity of the present synthesis was stavudine, arising from the competitive base-catalysed βelimination reaction on the precursor [44] and indicated as the favorite reaction to take place in the course of the nucleophilic substitution step [28]. Considering that stavudine, used also as an antiviral drug, is permanently incorporated into the DNA resulting in cytotoxicity [45], it is important to minimize the contamination with the compound. Since the chromatographic separation of this impurity from FLT, perfectly achievable on a HPLC system, could not be reasonably expected to be similarly efficient on a tC18 Sep-Pak setup and in the absence of real-time UV monitoring, it was central to keep the amount of stavudine in the reaction mixture as low as possible. As mentioned earlier, recent studies have correlated the generation of stavudine to the concentration and also to the time of exposure to the base [40,41]. Accordingly, the use of TBAHCO3 allowed to reduce this impurity compared to the amount generated with K2.2.2/K2CO3. A further reduction was next achieved by slightly stressing the hydrolysis step, since stavudine is decomposed under acid conditions to thymine and furfuryl alcohol [46]. By increasing this time from 2 to 3.5 min, our indicative cold tests — not reported in this study — showed the hydrolysis of stavudine to increase from 97% to 99%. As a result, the residual amount of stavudine still present in the final product solution was almost negligible. Moreover, furfuryl alcohol also readily undergoes decomposition under these conditions [47,48], leading to a hydrophilic product (1.25 min retention time on the analytical HPLC) more easily washed out into the waste by H2O #2 during the purification step.

5. Conclusion With the use of data and experiences reported in the literature as starting point, a reliable synthesis of [ 18F]FLT has been developed, able to afford this radiotracer in relatively short time and good radiochemical yields. This result has been achieved by a careful selection of reaction parameters and reagents and, above all, by introducing a novel and simple purification method able to bypass the use of the troublesome HPLC so far required. By applying this patented procedure, remarkably high chemical and radiochemical purities were obtained. Most important, the product's final solution was EtOH-free and therefore fully compliant with the European Pharmacopoeia and US Pharmacopoeia's requirements, an aspect so far neglected in all the other suggested preparations. Because of our initial approach, the implementation of the whole procedure in the two selected commercial modules of

549

synthesis was quite straightforward and successful. As a result, the whole synthesis of [ 18F]FLT can now be carried out by means of the same modules widely used for [ 18F]FDG without any substantial modifications to the related hardware and programme. Furthermore, the similarities with [ 18F] FDG in the preparation steps preceding the synthesis, the synthesis itself and the final cleaning ensure obvious advantages in terms of learning curve and daily use. Although the above-described results originated by a careful choice of conditions, as well as a long-acquired firsthand experience, it is quite reasonable to expect a similar outcome also for newcomers as long as special attention is given to the most relevant points, notably the amount of CH3CN left over to assist the hydrolysis; its almost complete removal prior the purification step; the volumes of H2O #2 and #3 according to the type and lot of tC18 used; and the elution flow rate. The suitability of the [ 18F]FLT thus prepared is further demonstrated by the animal investigations with microPET presently being carried out, while an already approved clinical trial is planned to start soon. Last, as long as strong, high-boiling solvents such as DMSO or DMF are not used, it is reasonable to expect the above-described procedure to work also on other synthetic approaches, which make use of similar but less complex precursors, such as 5′-O-(4,4′-dimethoxytrityl)2,3′-anhydrothymidine. Acknowledgments We acknowledge the financial support of Eckert & Ziegler Eurotope, Berlin, and of the Bundesministerium für Wirtschaft und Technologie of Germany, project IW081029. References [1] Shreve PD, Anzai Y, Wahl RI. Pitfalls in oncologic diagnosis with FDG PET imaging: physiologic and benign variants. Radiographics 1999;19:61–77. [2] Pio BS, Park CK, Pietras R, Hsueh WA, Satyamurthy N, Pegram MD, et al. Usefulness of 3′-[F-18]fluoro-3′-deoxythymidine with positron emission tomography in predicting breast cancer response to therapy. Mol Imaging Biol 2006;8:36–42. [3] Kubota K, Ishiwata K, Kubota R, Yamada S, Tada M, Sato T, et al. Tracer feasibility for monitoring tumour radiotherapy: a quadruple tracer study with fluorine-18-fluorodeoxyglucose or fluorine-18fluorodeoxyuridine,l-[methyl- 14C]methionine, [6- 3H]- thymidine, and gallium-67. J Nucl Med 1991;32:2118–23. [4] Vander Borght T, Pauwels S, Lambotte L, Labar D, De Maeght S, Stroobandt G, et al. Brain tumour imaging with PET and 2-[carbon-11] thymidine. J Nucl Med 1994;35:974–82. [5] Shields AF, Mankoff DA, Link JM, Graham MM, Eary JF, Kozawa SM, et al. [ 11C]Thymidine and FDG to measure therapy response. J Nucl Med 1998;39:1757–62. [6] Shields AF, Grierson JR, Dohmen BM, Machulla HJ, Stayanoff JC, Lawhorn Crews JM, et al. Imaging proliferation in vivo with [F-18] FLT and positron emission tomography. Nat Med 1998;4:1334–6. [7] Been LB, Suurmeijer AJH, Cobben DCP, Jager PL, Hoekstra HJ, Elsinga PH. [ 18F]FLT-PET in oncology: current status and opportunities. Eur J Nucl Med Mol Imaging 2004;31:1659–72.

550

C. Pascali et al. / Nuclear Medicine and Biology 39 (2012) 540–550

[8] Rasey JS, Grierson JR, Wiens LW, kolb PD, Schwartz JL. Validation of FLT uptake as a measure of thymidine kinase-1 activity in A549 carcinoma cells. J Nucl Med 2002;43:1210–7. [9] Sherley JL, Kelly TJ. Regulation of thymidine kinase during the cell cycle. J Biol Chem 1988;263:8350–8. [10] Mier W, Haberkorn U, Eisenhut M. [ 18F]FLT: portrait of a proliferation marker. Eur J Nucl Med Mol Imaging 2002;29:165–9. [11] Kong XB, Zhu QY, Vidal PM, Watanabe KA, Polsky B, Armstrong D, et al. Comparisons of anti-human immunodeficiency virus activities, cellular transport, and plasma and intracellular pharmacokinetics of 3′fluoro-3′-deoxythymidine and 3′-azido-3′-deoxythymidine. Antimicrob Agents Chemother 1992;36:808–18. [12] Lu L, Samuelsson L, Bergstrom M, Sato K, Fasth KJ, Långstöm BJ. Rat studies comparing 11C-FMAU, 18F-FLT, and 76Br-BFU as proliferation markers. J Nucl Med 2002;43:1688–98. [13] Buck AK, Halter G, Schirrmeister H, Kotzerke J, Wurziger I, Glatting G, et al. Imaging proliferation in lung tumours with PET: 18F-FLT versus 18F-FDG. J Nucl Med 2003;44:1426–31. [14] Wilson I, Chatterjee S, Wolf W. Synthesis of 3′-fluoro-3′-deoxythymidine and studies of its 18F- radiolabeling, as a tracer for the noninvasive monitoring of the biodistribution of drugs against AIDS. J Fluor Chem 1991;55:283–9. [15] Grierson JR, Shields AF. Radiosynthesis of 3′-deoxy-3′-[ 18F]fluorothymidine: [ 18F]FLT for imaging of cellular proliferation in vivo. Nucl Med Biol 2000;27:143–56. [16] Machulla H, Blocher A, Kuntzsch M, Grierson J. Simplified labeling approach for synthesizing 3′-deoxy-3′-[ 18F]fluorothymidine ([ 18F] FLT). J Radioanal Nucl Chem 2000;243:843–6. [17] Martin SJ, Eisenbarth JA, Wagner-Utermann U, Mier W, Henze M, Pritzkow H, et al. A new precursor for the radiosynthesis of [ 18F]FLT. Nucl Med Biol 2002;29:263–73. [18] Lee SJ, Oh SJ, Chi DY, Kil HS, Kim EN, Ryu JS, et al. Simple and highly efficient synthesis of 3′-deoxy-3′-[ 18F]fluorothymidine using nucleophilic fluorination catalyzed by protic solvent. Eur J Nucl Med Mol Imaging 2007;34:1406–9. [19] Oh SJ, Modszianoswski C, Chi DY, Kim JY, Kang SH, Ryu JS, et al. Fully automated synthesis system of 3′-deoxy-3′-[ 18F]fluorothymidine. Nucl Med Biol 2004;31:803–9. [20] Tang G, Tang X, Wen F, Wang M, Li B. A facile and rapid automated synthesis of 3′-deoxy-3′-[ 18F]fluorothymidine. Appl Radiat Isot 2010;68:1734–9. [21] Yun M, Oh SJ, Ha HJ, Ryu JS, Moon DH. High radiochemical synthesis of 3′-deoxy-3′-[ 18F]fluorothymidine using (5′-O-dimethoxytrityl)-2′-deoxy-3′-O-nosyl-β-D-threo-pentofuranosyl)thymine and its 3-N-BOC-protected analogue as a labeling precursor. Nucl Med Biol 2003;30:151–7. [22] Teng B, Wang S, Fu Z, Dang Y, Wu Z, Liu L. Semiautomatic synthesis of 3′-deoxy-3′-[ 18F]fluorothymidine using three precursors. Appl Radiat Isot 2006;64:187–93. [23] Nandy SK, Rajan MGR, Korde K, Chaudhari PR, Soni PS. Rapid synthesis of [ 18F]fluoro-L-thym idine with simplified purification using a combination-column. J Labelled Compd Radiopharm 2007;50:S121. [24] Nandy SK, Krishnamurthy NV, Rajan MGR. Evaluation of the radiochemical impurities arising during the competitive fluorination of nosyl group during the synthesis of 3′-deoxy-3′-fluorothymidine, [ 18F] FLT. J Radional Nucl Chem 2010;283:245–51. [25] Nandy SK, Rajan MGR. Fully automated and simplified radiosynthesis of [ 18F]-3′-deoxy-3′-fluorothymidine using anhydro precursor and single neutral alumina column purification. J Radioanal Nucl Chem 2010;283:741–8. [26] Walsh JC, Padgett HC, Ysaguirre T. Method for preparing radiolabeled thymidine having low chromophoric byproducts. 2005, PCT WO 2005/058247 A2. [27] Trump DP. Radiolabeled thymidine solid-phase extraction purification method. 2005, PCT WO 2005/025519 A2.

[28] Husbyn M. Separation process. 2010, US 2010/0105888 A1. [29] Tadino V, Muller M. Process for synthesizing labeled compounds. 2006, PCT WO 2006/133732 A1. [30] Pascali C, Bogni A. Purification procedure for the preparation of the radioactive tracer 3′deoxy-3′-[ 18F]fluorothymidine ([ 18F]FLT). 2008, WO 2008/146316 A1. [31] Mock BH, Winkle W, Vavrek MT. A color spot test for the detection of Kryptofix 2.2.2 in [ 18 F]FDG preparations. Nucl Med Biol 1997;24:193–5. [32] Chaly T, Dahl JR. Thin layer chromatographic detection of Kryptofix 2.2.2 in the routine synthesis of [ 18F]2-fluoro-2-deoxy-D-glucose. Nucl Med Biol 1989;16:385–7. [33] Bogni A, Pascali C, Iwata R, Cavalleri A, de Sanctis V, Decise D, et al. [ 18F]FDG synthesis by Anatech RB-86 robotic system: improvements and general considerations. J Radioanal Nucl Chem 1998;230:45–50. [34] European Pharmacopoeia 6th edition, Residual Solvents. Strasbourg, France. 2008, 603–610. [35] Kim DW, Ahn DS, Oh YH, Lee S, Kil HS, Oh SJ, et al. A new class of SN2 reactions catalyzed by protic solvents: facile fluorination for isotopic labeling of diagnostic molecules. J Am Chem Soc 2006;128:16394–7. [36] Lee SJ, Oh SJ, Chi DY, Lee BS, Ryu JS, Moon DH. Comparison of synthesis yields of 3′-deoxy-3′-[ 18F]fluorothymidine by nucleophilic fluorination in various alcohol solvents. J Labelled Compd Radiopharm 2008;51:80–2. [37] Moon BS, Lee KC, An GI, Chi DY, Yang SD, Choi CW, et al. Preparation of 3′-deoxy-3′-[ 18F]fluorothymidine ([ 18F]FLT) in ionic liquid, [bmim][OTf]. J Labelled Compd Radiopharm 2006;49: 287–93. [38] Windhorst AD, Klein PJ, Eisenbarth J, Oeser T, Kruijer PS, Eisenhut M. 3′-Sulfonylesters of 2,5′-anhydro-1-(2-deoxy-β-D-threo-pentofuranosyl)thymine as precursors for the synthesis of [ 18F]FLT: syntheses and radiofluorination trials. Nucl Med Biol 2008;35:413–23. [39] Barthel H, Cleij MC, Collingridge DR, Hutchinson OC, Osman S, He QM, et al. 3′-Deoxy-3′-[ 18F]fluorothymidine as a new marker for monitoring tumour response to antiproliferative therapy in vivo with positron emission tomography. Cancer Res 2003;63:3791–8. [40] Suehiro M, Vllabhajosula S, Goldsmith SJ, Ballon DJ. Investigation of the role of the base in the synthesis of [ 18F]FLT. Appl Radiat Isot 2007;65:1350–8. [41] Bourgeois M, Mougin-Degraef M, Leost F, Cherel M, Gerstein JF, Le Bars D, et al. Purification of [ 18F]-fluoro-L-thymidine ([ 18F]-FLT) for positron emission tomography imaging. J Pharm Biomed Anal 2007;45:154–7. [42] Culbert PA, Adam MJ, Hurtado ET, Huser JMA, Jivan S, Lu J, et al. Automated synthesis of FDG using TBAHCO3. Appl Radiat Isot 1995;46:887–91. [43] Tewson T. Procedures, pitfalls and solutions in the production of [ 18F] 2-deoxy-2-fluoro-D-glucose: a paradigm in the routine synthesis of fluorine-18 radiopharmaceuticals. Nucl Med Biol 1989;16:533–51. [44] Cleij MC, Steel CJ, Brady F, Ell PJ, Pike VW, Luthra SK. An improved synthesis of 3′-deoxy-3′-[ 18F]fluothymidine ([ 18F]FLT) and the fate of the precursor 2,3′-anhydro-5′-O-(4,4′-dimethoxytrityl)thymidine. J Labelled Compd Radiopharm 2001;44:S871–3. [45] Kaul S, Dandekar KA, Schilling BE, Barbhaiya R. Toxicokinetics of 2′,3′-didehydro-3′-deoxythymidine, stavudine (D4T). Drug Metab Dispos 1999;27:1–12. [46] Shi J, Ray AS, Mathew JS, Anderson KS, Chu CK, Schinazi RF. 2′,3′Didehydro-2′,3′-dideoxynucleosides are degraded to furfuryl alcohol under acidic conditions. Biorg Med Chem Lett 2004;14:2159–62. [47] Krishnan TA, Chanda M. Kinetics of polymerization of furfuryl alcohol in aqueous solution. Angew Makromol Chem 1975;43: 145–56. [48] Batista PS, Ferreira de Souza M. Furfuryl polymerization inside capillaries. Synth Met 1999;101:635–6.