Automated synthesis of radiopharmaceuticals for PET: an apparatus for [1-11C]labelled aldoses

This paper describes an instrumentation system for positron emission tomography (PET). A variety of [1-11C]labelled aldoses, such as [1-11C]-D-glucose, and galactose by a modification of the Kiliani-Fischer method have been produced. The instrumentation is fully automatic and consists of a synthesis system and control system. The synthesis system has the following functions: supplying reagents; performing reactions; purifying 11C labelled aldose; and preparing an injectable solution of 11C labelled aldose. These operations are performed by the control system in a remote control room. In a preliminary, hot experiment an injectable solution of [1-11C]-D-glucose was obtained. In addition, the operator is exposed to minimal radiation. The radioactivity of [1-11C]-Dglucose was 47 MBq, and the preparation time was 49 min.


Introduction
Positron emission tomography (PET) [1] is a non-invasive imaging technique which can obtain biofunctional information from humans and animals using radiopharmaceuticals containing a positron emitter (for example 11C, 150 and 18F). There is currently great interest in the production of radiotracers for PET. [-1-11C]labelled aldoses are very useful radiotracers for regional cerebral glucose metabolism [2] and tumor markers [3]. However, there are some major synthetic problems in their preparation. 11C has a radioactive half-life of only 20"4 min and decays with the evolution of X-rays (energy 511 keV). Moreover, the synthetic scale has to be very small because only pico-mol order of 11C can be obtained by 14N(p, z)11C reaction using a cyclotron. To overcome these difficulties, the preparation process has to be rapid, reproducible and on a micro scale. So the development of a rapid, stereoselective reaction and automation of the process are very important and the authors have been working on this.
The first synthesis of [ 1-11C]-D-glucose using the classical Kiliani-Fischer method was reported by Shiue el al. [4], and recently Schoeps et al. I-5] reported the preparation of [1-11C]-D-glucose from [11C]-nitromethane using a Nef reaction. In these approaches, the final product is obtained as a mixture of [ 1-11C]-D-glucose and mannose, and the ratio of D-glucose to mannose was reported to be from 0"25 to 0"5. More recently, Carmen et al. [6-1 have improved the ratio ofD-glucose to mannose by the reaction of D-arabinose with NH411CN in pH 8" borate buffer-the ratio was improved to 1"80 _ _ _ 0"57 in favour of D-glucose. Using these synthetic methods, several groups [-7, 8] have developed remote or automated instruments for preparing [1-11C]_D_glucose. However, the majority of these instruments were not fully automatic and were not very flexible. This paper describes a new method for preparing optical isomers by changing the reaction conditions and a new instrument set up which is fully automatic and can synthesize other [1-11C]labelled aldoses.

Development of a rapid synthetic method
Micro-scale synthetic study ofaldoses was performed using a mock-up apparatus and cold experiments. The focus was on a modification of the Kiliani-Fischer method.
In a similar manner, rapid synthetic methods for other aldoses were also investigated.

Construction of automated instrumentation
The automated instrument was built for the synthetic method. To minimize the operator's exposure to radiation, the hardware was designed to produce an injectable solution through a remote control. The apparatus was designed for laboratory use and for ease of improvement of the hardware and software. The software was programmed with Hyakuninriki (Asahi Electronics Co. Ltd, Japan) which operates under MS-DOS.

Hot experiment
After examining the synthesis of the aldoses in the cold test, an attempt was made to produce an injectable solution of [-1-11C]-D-glucose from HllCN [11] gas, which is prepared with a cyclotron and a H llCN gas generator--see figure 1o The production of 11C was accomplished by the nuclear reaction of accelerated protons with high pressured 14N 2 gas. This reaction was performed with a cyclotron and a target chamber. The pico mole quantities of 11C undergo rapid oxidation to 11CO2 in the target chamber. The 11CO2 gas was then transferred to the H 11CN gas generator and converted to These operations were performed with four personal computers, and could be monitored with CCD video cameras in the remote controlling room. The product was analysed with a radioisotope dose calibrator and a HPLC system by remote monitoring and controlling.

Chemistry
In order to investigate the possibility of selectively synthesizing either isomer by changing reaction conditions, reaction rate and the steroselectivity of cyanohydrin formation was examined. As a typical example, the reaction of compound 1 with one equivalent of sodium cyanide in a mixture of organic solvent and alkali buffer The yields of the cyanohydrins, gluconotrile 2_ and mannononitrile _3, were measured using HPLC. The total yield curve producing _2 and 3_ versus reaction time is shown in figure 2. The initial reaction rate calculated from the summation yield of _2 and 3_, depended upon the organic solvent. However, the total yield curves appeared to level offwithin 5 min. Interestingly, the formation ratio of_2 to _3 was found to be greatly dependent on the organic solvent and the pH of the buffer--figure 3.
These results can be divided into three groups. First is the glucononitrile _2 selective group, in which a mixture of toluene and alkali buffer was used as the reaction solvent. In these cases, the formation of_2/3_ increased with the reaction time and the pH value of the buffer, and levelled off after 5 min. Second is the non-selective group, in which a mixture of ethyl acetate or diisopropyl ether and alkali buffer was used. Last is the mannononitrile 3 selective group, in which a mixture of ethanol and alkali buffer was used. The formation ratio reached the steady state after 2 min. It seems that the increasing ratio with time is caused by equilibrium reaction between _2 and _3, and that the formation ratio of _2/_3 relates to the polarity of the organic solvent.
In addition, the toluene-buffer condition for preparing _2, which is the precursor of D-glucose, was examined. Figure  3 shows that the stereoselectivity of the cyanohydrin formation depends on the pH. Figure 4 shows the yield curves of _2 when the pH varied from 8"3 to 11"5, but the reaction rate tended to decrease at high pH such as pH 11"5. Thus, the optimum pH was found to be 10"8 in figure 4. To obtain a better understanding of the reaction mechanism, the dissociation rates of 2 and 3 were measured. After cyanohydrins 2 and _3 uTere isolated by silica gel chromatography, each sample (10 gmol) was added to a mixture of toluene (50 l.tl) and aqueous pH 10"8 buffer (1 M Na2CO3-1 M HC1, 50 t.tl). Each mixture was stirred at room temperature and analysed by HPLC buyer.
as shown in figure 5. In both cases, the ratios of_2/_3 varied with reaction time and after 5 min the ratios stabilized to 2"1"1. From these results, it is considered that the stereoselectivity of the cyanohydrin formation is not a result of cyanide attack on aldehyde _1, but, rather, it is due to an equilibrium reaction between the products _2 and 3. For the carbon-11 labelling, it was favourable that the e-quilibrium reaction proceeds rapidly. This method is practical and not moisture sensitive, so it is possible to apply to a microsynthesis of aldoses by combining it with a reductive hydrolysis step. Thus, a one-pot synthesis of D-glucose and mannose was as shown below: to give them in 23"0 and 13"5, respectively. The one-pot reaction was performed within 15 min. Furthermore, we synthesized )-galactose (28"1o) and D-talose (11"9) using a similar method from 2,3:4,5-0-isopropylidene-D-lyxose (_4) [ The automated apparatus consists of the synthesis system and the controlling system. The synthesis system, an auto-manual switch box, and an interface are placed in a radiation shielded room. As these are removable, it is convenient for the cold experiment to be performed elsewhere and to facilitate the maintenance for the apparatus. The computer and its accessories are placed in the remote control room. The general appearance is shown in figures 6 and 7. performed by the controlling system and can be performed manually through the auto-manual switch box, which is useful in the case of the investigation with the cold experiment and the maintenance of the apparatus. As the solenoid valves and other devices of the reagents' supply unit and reaction unit were installed on the punched metal board, it is easy to modify the hardware.

Synthesis system
Reagent and wash solvent supply unit The reagent supply unit [13] has eight reservoirs (12)(13)(14)(15)(16)(17)(18)(19) in figure 8) for liquid reagents and solvents. Each reagent and solvent in the reservoirs is under nitrogen atmosphere, and can be transferred to the reaction flasks in two steps. First, the liquid is allowed to flow from the reservoir into a volumetric tube (0"5 ml) by nitrogen gas pressure. When the tube is full and the photosensor (45-52 in figure 8) is activated, the contents of the volumetric tube are emptied into the reaction flask by nitrogen gas pressure. The same volume of liquid may be repeatedly measured and added to the reaction flask. These operations are performed with three-way solenoid valves, photo-sensors and nitrogen gas pressure. In this system, even moisture or air sensitive liquids can be stored in the reservoirs and transferred to the reaction flasks. After a synthetic run, all of the flow lines can be washed and dried by passing wash solvents which are stored in tanks (10 and 11 in figure 8), and then nitrogen gas through them.

Reaction unit
The reaction unit has two reaction flasks (20 and 21 in figure 8). Reaction flask is used for the hydrocyanation of a precursor with NaXlCN and flask 2 is used for the reductive hydrolysis reaction. Both flasks are about 2 ml in volume and have jackets through which heating/cooling fluid is circulated with circulator and 2 (24 and 25 in figure 8). The reaction ,temperature is maintained at the desired setting by the circulators. The mixing of the reaction mixture in the flasks are accomplished by nitrogen gas bubbling. The bubbling rate can be controlled with two mass flow controllers and 2 (34 and 35 in figure 8). A reaction mixture in flask can be transferred to flask 2 by using nitrogen gas pressure. The reaction mixture in flask 2 can be filtered with a glass filter, which is at the bottom of flask 2, and the filtrate can be transferred to the purification unit by using reduced pressure.

Purification unit
The purification unit consists of three devices: an ion exchange resin column (29 in figure 8) flask 2 is desalted with the resin column and transferred to flask 3 for evaporation. Flask 3 also has a jacket through which heating fluid is circulated with circulator 3 (26 in figure 8), and is about 10ml in volume. The mixing process in flask 3 is performed with the magnetic stirrer (36 in figure 8). The evaporating process is carried out with a vacuum device (27 and 28 in figure 8). The concentrated mixture is then transferred to the autoinjection device through the bubble trap (42 in figure 8) with the roller pump (40 in figure 8). An objective compound is isolated tiom the HPLC column (5 in figure  8) and detected with the refractive index and radiation detector (7 and 8 in figure 8). The eluate containing the objective compound is injected into the flask 4 (23 in figure 8).

Pharmaceutical preparation unit
The pharmaceutical preparation unit has three functions: adjusting the pH of the radiopharmaceutical solution; diluting with saline; and filtration with the filter 3 (56 in figure 8). The aqueous solution of the radiopharmaceutical in flask 4 is neutralized with a dilute acid or alkali solution from the reagent's supply unit. Flask 4 is equipped with the pH sensor (43 in figure 8), the level sensor (44 in figure  8), and the magnetic stirrer 2 (37 in figure 8). The radiopharmaceutical solution in flask 4 is filtered with the membrane filter and roller pump 2 (41 in figure 8). In the way, the lC-labelled compound is available in ready-to-use form for the PET study.

Control Computer and software
The instrumentation is controlled with a personal computer (PC-9821Ap, NEC), which is linked with the other computers and LAN (Local Area Network) as shown in figure 1. An OPTMUX (Opto 22, USA) interface unit is used. The computer software was developed by using Hyakuninriki. The program consists of four processes as follows; hydrocyanation process; reductive hydrolysis process; purification process; and pharmaceutical preparation process. A flowchart of these processes is shown in figure 9.
The hydrocyanation process contains subroutines from 'Add NaCN soln.' to 'Hydrocyanation in FI'. The reductive hydrolysis process contains subroutines of'Add HC1 HCOOH' and 'Reductive hydrolysis in F2'. The purification process contains subroutine from 'Desalt' to 'HPLC', and the pharmaceutical preparation process contains subroutines from 'pH adjustment in F4' to 'Volume adjustment in F4'. The reaction processes are controlled by a time sequential method, and the injection process in HPLC is performed by sequential control using the signal of the photosensor. The processes of pH and (tl-12 min 14 min), all the peaks of the HPLC are ignored. Collection of the eluate is started when the ratio of peak variation becomes > (R)1. Collection is stopped when the ratio of peak variation is < (R)2, and the peak value is less than the threshold value. This systematic procedure is highly efficient for isolating the objective compound even if the HPLC column becomes degraded. A flowchart of the operation for the 'HPLC' subroutine is given in figure   11.

Synthesis of H11CN
The production ofH 11CN was accomplished by an on-line synthesis according to Iwata's method [11]. Production of 11 CO2 was accomplished through 14 N (p, ) 11C reaction by proton bombardment 18 MeV, 15 gA) of 14"7 kg/cm} N 2 gas target using a cyclotron-target system (CYPRIS HM-18, Sumitomo Heavy Industries Co. Ltd). The 11 CO2 gas in the target chamber was transferred to 11CO2 gas concentration equipment (AMCT 01, NKK Corp.) and then the concentrated 11CO gas was transferred to an HllCN gas generator (AMHC 01, NKK Corp.) using He gas (flow rate: 100 ml/min) as a carrier gas, and hydrogenated (flow rate of H 2 gas: 10 ml/min) to give 11CH 4 gas at 200C in the presence of silica-gel supported Ru catalyst. The reaction of 11CH4 gas with NH gas (flow rate: 5 ml/min) at 850C in the presence of Pt catalyst gave off H 11CN gas, which was passed through a P205 (5"0 g) column to remove excess NHa gas, and then 16"9 GBq (at the end of bombardment) of HllCN gas was transferred to the automated apparatus for labelled aldoses.

Hydrocyanation of compound 1_
Nitrogen gas flow rates of the mass flow controller and 2 were set to zero. The outlet of circulator was opened by switching the valve 49. The outlet of reservoir was opened by switching the valves 17 and 33. A mixture solution of NaCN (as a carrier) and pH 10"8 buffer in reservoir was allowed to flow from reservoir into a volumetric tube (0"5 ml) by nitrogen gas pressure (nitrogen gas flow line: 30-31-V13-V14-V34-V33, in figure 8).
When the tube was full and photosensor activated, the contents of the volumetric tube were emptied into reaction flask by switching the valve 17, 33, and 34 (nitrogen gas flow line" 30-2 l-V13-V14-V34-V17). When the volumetric tube was empty and the photosensor was off, the valve 34 was switched. Trapping ofH 11CN gas in flask was carried out by switching valves 1, 3, and 4 (HllCN gas flow line: 32-V3-V1-V4-bottom of the flask 1). The waste gas was exhausted through valve 2 to the waste gas line. When the introduction of HllCN gas to flask was finished, valve 3 was switched. To the mixture in flask was added 0"5 ml of the solution of compound 1 in toluene from reservoir 2 in a similar manner to that of reservoir using photosensor 2 and the valves 18, 35, and 36. The mixture in flask was mixed for 8 min by nitrogen gas bubbling using the mass  stirring was stopped, the aqueous solution was injected to the HPLC system as follows: valves 10 and 11 were switched, the aqueous solution containing air bubbles was delivered to the bubble trap via valve 58 with the roller pump 1, the debubbled solution in the bubble trap was sent to the sample loop via photosensor 9 and the rotary six-way valve, when the end of the solution line was detected by the photosensor 9, the rotary six-way valve was switched and the solution in the sample loop was loaded to the HPLC column.

2O2
The HPLC separation conditions were as follows: column: Bio-Rad Aminex HPX-87P (7"8 mm x 30 cm, 9 gm), mobile phase: water, flow rate: 0"6 ml/min, temperature: 85C, retention time of D-glucose: 13 As the pH value was in the range from 6"5 to 7"5, the addition of the acid in reservoir 6 or the alkali in reservoir 7 was not performed. Thus the solution was diluted with the saline in reservoir 8. The saline was added to flask 4 repeatedly until the level sensor was activated which was set to a volume of 10 ml. Finally, the injectable solution of [1-1C]-D-glucose was obtained by filtration with roller pump 2 and the membrane filter 3. The total synthesis time was 49 min. The product was analysed by remote monitoring and operatingand the analysis data were found to be as follows: chemical yield 2"0 from NaCN, radiochemical yield 1"3 from HICN, radio- (1) Analysis of aldononitrile _2 and 3_" Column: Waters radialpak C-18 (8 mm x 10 cm, 5 lam).
Mobile phase: water.
The analysis of [1-aXC]-D-glucose was performed by remote control. The HPLC analysis was accomplished using a Shimadzu LC-9A pump, a Shodex refractive index detector RISE-61, an Aloka positron detector TCS-R81-3454, a handmade auto sampler, and a Shimadzu C-R4AX two-channel data processor. The radioactivity of the product was measured with a CAPINTEC CRC-712 dose calibrator.