Multistep Synthesis of a Radiolabeled Imaging Probe Using Integrated Microfluidics

See allHide authors and affiliations

Science  16 Dec 2005:
Vol. 310, Issue 5755, pp. 1793-1796
DOI: 10.1126/science.1118919


Microreactor technology has shown potential for optimizing synthetic efficiency, particularly in preparing sensitive compounds. We achieved the synthesis of an [18F]fluoride-radiolabeled molecular imaging probe, 2-deoxy-2-[18F]fluoro-d-glucose ([18F]FDG), in an integrated microfluidic device. Five sequential processes—[18F]fluoride concentration, water evaporation, radiofluorination, solvent exchange, and hydrolytic deprotection—proceeded with high radio-chemical yield and purity and with shorter synthesis time relative to conventional automated synthesis. Multiple doses of [18F]FDG for positron emission tomography imaging studies in mice were prepared. These results, which constitute a proof of principle for automated multistep syntheses at the nanogram to microgram scale, could be generalized to a range of radiolabeled substrates.

Continuous-flow microreactors (13) have recently been used to manipulate individual chemical processes on nanoliter to microliter scales. The advantages of such chemical reaction circuits include enhanced heat transfer performance, faster diffusion times and reaction kinetics, and improved reaction product selectivity (46). For example, in microfluidic environments, triphasic hydrogenation (7) can be achieved with higher reaction efficiency, the inorganic synthesis of high-quality CdSe nanocrystals has been demonstrated (8), and chemical processes involving highly reactive intermediates can be executed with superior reaction selectivity (9). However, challenges remain in applying the technology to sequential syntheses of fine chemicals and pharmaceuticals.

In multistep procedures, flow-through systems are plagued by cross-contamination of reagents from different steps; side reactions and poor overall yield result from the inability to confine each individual step. Microfluidic batch devices with integrated microvalves show promise for the automation of multiple, parallel, and/or sequential chemical processes on a single chip under digital control. By analogy, this technology has already been successfully applied to biological problems (10).

A compelling application is in the preparation of organic compounds bearing short-lived isotopes whose emission permits detailed mapping of biological processes in living organs. In conjunction with positron emission tomography (PET) (11), the development of sensitive radiolabeled molecular probes is crucial for expanding the capability of target-specific in vivo imaging for biological research, drug discovery, and molecular diagnostics. The United States already has a vast network of PET cyclotron production sites in place as convenient sources for radiolabeled precursors [e.g., [18F]fluoride, [11C]CO2, and [11C]methyl iodide (MeI)] and a few labeled biomarkers. The capacity for diversifying radiolabeled probe structure is therefore limited only by the cost, speed, and efficiency of synthetic methods. A central challenge in this regard is the half-life of the radiolabels.

The synthesis (12) of the [18F]-labeled molecular imaging probe 2-deoxy-2-[18F]fluoro-d-glucose ([18F]FDG) in an integrated microfluidic chip was chosen as a proof-of-principle study. This compound is the most widely used radiolabeled molecular probe, with more than 1 million doses for patient diagnosis produced in the United States in 2004 and a similar number in the rest of the world (13). The brief half-life of [18F]fluorine (t1/2 = 110 min) makes rapid synthesis of doses essential. Today, [18F]FDG is routinely produced in about 50 min with the use of commercial synthesizers (14), which are expensive (∼$140,000) and produce ∼10 to 100 doses in a single run. Obtaining high yields with short synthesis times is even more critical for molecular imaging biomarkers bearing positron-emitting radioisotopes with shorter half-lives, such as 11C (t1/2 = 20 min) and 13N (t1/2 = 10 min). A unique aspect of PET molecular imaging probes is that only nanogram masses per dose of the radiopharmaceuticals are administered to subjects.

The radiopharmaceutical requirements of expedited chemical kinetics and low-mass quantities of product, together with the emerging need to expand and diversify the catalog of molecular imaging probes, provide a unique opportunity for the use of integrated microfluidics. In addition, the preparation of [18F]FDG provides a conceptual model for the preparation of other molecules (including pharmaceuticals) because it includes common steps required in many chemical syntheses.

We developed a microfluidic chemical reaction circuit (Fig. 1) capable of executing the five chemical processes of the syntheses of both [18F]FDG and [19F]FDG within a nanoliter-scale reaction vessel. Conceptually, however, the chip was designed to demonstrate the digital control of sequential chemical steps, variable chemical environments, and variable physical conditions, all on a single chip. It was also designed to produce sufficient quantities of [18F]FDG (100 to 200 μCi) for mouse imaging. The chip thus has the capability of synthesizing the equivalent of a single mouse dose of [18F]FDG on demand. The device accelerates the synthetic process and reduces the quantity of reagents and solvents required. This integrated microfluidic chip platform can be extended to other radiolabeled molecular imaging probes.

Fig. 1.

(A) Schematic representation of a chemical reaction circuit used in the production of 2-deoxy-2-fluoro-d-glucose (FDG). Five sequential processes are shown: (i) concentration of dilute fluoride ion with the use of a miniaturized anion exchange column located in a rectangle-shaped fluoride concentration loop, (ii) solvent exchange from water to dry MeCN, (iii) fluorination of the d-mannose triflate precursor 1, (iv) solvent exchange back to water, and (v) acidic hydrolysis of the fluorinated intermediate 2a (or 2b) in a ring-shaped reaction loop. Nanogram amounts of FDG (3a, 3b) are the final product. The operation of the circuit is controlled by pressure-driven valves, with their delegated responsibilities illustrated by their colors: red for regular valves (for isolation), yellow for pump valves (for fluidic metering circulation), and blue for sieve valves (for trapping anion exchange beads in the column module). (B) Optical micrograph of the central area of the circuit. The various channels have been loaded with food dyes to help visualize the different components of the microfluidic chip; colors are as in (A), plus green for fluidic channels. Inset: Actual view of the device; a penny (diameter 18.9 mm) is shown for comparison.

Some of the components required for conducting sequential chemical processes within microfluidics are similar to those previously demonstrated for biological analysis: isolation of distinct regions on the chip with micro-mechanical valves for nanoliter chemical reactions (15), acceleration of diffusion-dominated mixing within a confined volume with a rotary pump (16), and creation of in situ affinity columns (10). However, two additional technical advances were required to perform effective chemical synthesis. First, an in situ ion exchange column was combined with a rotary pump to concentrate radioisotopes by nearly three orders of magnitude, thereby optimizing the kinetics of the desired reactions. Second, the gas-permeable poly(dimethylsiloxane) (PDMS) matrix allows solvent exchange to occur within the microfluidic channel through direct evaporation, thereby allowing for the sequential execution of chemical reactions in PDMS-compatible solvents (17). A solution inside a PDMS-based microfluidic reactor can be heated above its normal (atmospheric) boiling point to provide further kinetic enhancement. Pressure is mediated not only by the heat supplied to the chip, but also by the porosity of the PDMS matrix. Thus, PDMS plays a role akin to the safety valve of a pressure cooker that regulates the “cooking pressure” within a critical range. Our device permits computer-controlled mixing of spatially isolated reagents under individually regulated solvent and temperature conditions.

The production (12) of [18F]FDG is based on five sequential chemical processes (Fig. 1A): (i) concentration of the dilute [18F]fluoride mixture (18) solution (<1 ppm, specific activity ∼5000 to 10,000 Ci/mmol), obtained from the proton bombardment of [18O]water at a cyclotron facility (19); (ii) solvent exchange from water to acetonitrile (MeCN); (iii) [18F]fluoride substitution of the triflate group in the d-mannose triflate precursor 1 in dry MeCN; (iv) solvent exchange from MeCN to water; and (v) acidic hydrolysis of the fluorinated intermediate 2a to obtain [18F]FDG (3a).

The concentration of [18F]fluoride mixture (18) obtained from a proton bombardment of [18O]water is usually below 1 ppm. We created a miniaturized anion exchange column (Fig. 2) in the microfluidic device to concentrate the [18F]fluoride mixture solution to ∼100 ppm. Sieve valves (Fig. 2B) were created using a square-profiled fluidic channel and a control membrane. Actuation of this membrane prohibits the passage of large particles while still permitting the solution to pass along the edges of the channel. Using these sieve valves to trap anion exchange beads, we obtained the anion exchange column (Fig. 2, C and D) for the concentration of the [18F]fluoride mixture.

Fig. 2.

Schematic representations of the operational mechanisms of (A) a regular valve having a round-profiled fluidic channel, and (B) a sieve valve having a square-profiled fluidic channel. When pressure is introduced into the control channels, the elastic membranes expand into the fluidic channels. In a regular valve, the fluidic channel is completely sealed because of the perfect fit between the expanded membranes and the round profile of the fluidic channel. In a sieve valve, the square-profiled fluidic channel is only partially closed, which allows fluid to flow along the two edges. Sieve valves can be used to confine solid objects within the fluidic channel but allow liquid to flow through. (C) Schematic illustration of the loading of anion exchange beads into a column module incorporating one fluidic channel and five sieve and five regular valves (×, closed valve). A suspended solution of anion exchange beads is introduced into the column modules where five sieve valves and five regular valves operate cooperatively to trap anion exchange beads inside the fluidic channel (total volume 10 nl). A miniaturized anion exchange column for fluoride concentration is achieved when the fluidic channel is fully loaded. (D) A snapshot of the bead-loading process in action.

We performed a proof-of-concept trial (Fig. 3) with the use of nonradioactive [19F]fluoride. The acquired experimental parameters could be used directly for the production of radioactive [18F]FDG. For the concentration of dilute fluoride (process i, Fig. 1A), a NaF solution (5 ppm) was loaded into the anion exchange column (Fig. 3A). The loading rate (5.0 nl/s) was controlled with a metering pump. After the fluoride solution was loaded completely, a K2CO3 solution (0.25 M, 18 nl) was introduced to fill the rectangular loop. The circulating pump module was then turned on so that the K2CO3 solution (0.25 M, 18 nl) could loop through the column continuously to produce a concentrated KF solution.

Fig. 3.

Schematic diagrams showing the four most critical steps of FDG production in the reaction circuit. (A) Concentration of dilute fluoride ion. With the cooperation of regular valves, a dilute fluoride solution (blue) is introduced into the ion exchange column by a metering pump. (B) Evaporation of water from the concentrated KF solution. After transferring the concentrated KF solution from the fluoride concentration loop to the ring-shaped reaction loop, the circuit is heated on a hot plate to evaporate water from the reaction loop. Meanwhile, all of the surrounding regular valves are completely closed and the circulating pump is turned on. (C) Fluorination reaction. After introduction of a MeCN solution (green) of Kryptofix and the d-mannose triflate 1 into the reaction loop, the inhomogeneous reaction mixture is isolated in the reaction loop, mixed using the circulating pump, and heated under a computer-controlled gradient to generate the intermediate 2a (or 2b). (D) Hydrolysis reaction. After evaporation of the MeCN, an HCl solution (blue) is introduced into the reaction loop to hydrolyze the intermediate 2a (or 2b) to give the final product, FDG (3a, 3b).

Because the fluorination (process iii) of the d-mannose triflate precursor requires anhydrous conditions, a digitally controlled hot plate was used to heat the reaction circuit for removing water (process ii) from the concentrated KF solution (Fig. 3B). Dry MeCN was loaded into the reaction loop and the reaction circuit was heated again to completely extrude any remaining moisture. Moisture and MeCN vapor can penetrate and escape the gas-permeable PDMS matrix. Once the circuit had cooled to room temperature, an anhydrous MeCN solution (40 nl) containing the d-mannose triflate 1 (92 ng, limiting reagent) and Kryptofix 222 (364 ng) was introduced into the ring-shaped reaction loop containing the dried KF. This heterogeneous reaction mixture was mixed inside the loop using the circulating pump. During this step (process iii), the circuit was heated (100°C for 30 s and then 120°C for 50 s) to yield the fluorinated intermediate 2b (Fig. 3C), as analyzed by gas chromatography–mass spectrometry (GC-MS). This analysis indicated that the conversion yield for the fluorination process was 98%. After removal of MeCN by direct evaporation, 3 N HCl solution (40 nl) was injected into the circuit, and the hydrolysis (Fig. 3D, processes iv and v) of the intermediate 2b was conducted at 60°C to obtain [19F]FDG in >90% purity, according to GC-MS analysis. The entire synthesis was reproduced on multiple chips.

Radioactive [18F]FDG was also produced in the reaction circuit by starting from the radioactive [18F]fluoride mixture. For this demonstration, only 720 μCi of [18F]fluoride (limiting reagent) in ∼1 μl of [18O]water was used. Because of the relatively high loading rate (∼65 nl/s) applied, only 500 μCi of [18F]fluoride was trapped in the column. Then, 324 ng of d-mannose triflate was introduced into the circuit to obtain 190 μCi of [18F]FDG with a radiochemical yield of 38% and a radiochemical purity of 97.6% (20), according to radio–thin layer chromatography (TLC) analysis. This sequential production of [18F]FDG was completed in automated fashion within 14 min (21), and similar results were observed across multiple runs.

We also designed a second-generation chemical reaction circuit with the capacity to synthesize larger [18F]FDG doses (22). This chip has a coin-shaped reactor (volume 5 μl) equipped with a vacuum vent. It was used to synthesize 1.74 mCi of [18F]FDG, an amount sufficient for several mouse experiments. From the purified and sterilized product (Fig. 4A), two doses (375 μCi and 272 μCi) were used for microPET- and microCT-based molecular imaging of two mouse models of cancer (23). One of the mouse images is shown in Fig. 4B as a two-dimensional projection. This circuit design should ultimately yield large enough quantities (i.e., >100 mCi) of [18F]FDG for multiple human PET scans, which typically use 10 mCi per patient.

Fig. 4.

(A) Analytical TLC profile of the unpurified mixture (blue curve) obtained upon the production of [18F]FDG in the second-generation reaction circuit, indicating that the radiochemical purity of the FDG production is 96.2%. The two peaks have Rf values of 0.0 and 0.36, corresponding to [18F]fluoride and [18F]FDG, respectively. After purification and sterilization, the [18F]FDG (black curve) with 99.3% radiochemical purity was used for mouse microPET/microCT imaging. (B) Projection view of microPET/microCT image of a tumor-bearing mouse injected with [18F]FDG produced in a microfluidic chip.

A major limitation (17) of the current chips involves the PDMS elastomer. This material is not chemically resistant to most organic solvents, so it seriously limits the variety of chemical reactions that can be executed within integrated microfluidic environments. New solvent-resistant elastomeric materials (24) have been introduced for integrated microfluidic chips. Therefore, the above technology shows promise for a broad range of chemical syntheses. In addition to the flexibility of rapidly arranging unit operations on an integrated microfluidic chip for specific reactions, new circuit designs take less than 2 days to proceed from a computer-aided design (CAD)–based proposal to a working chip. Taken together with the low production costs for the chips, these chemical reaction circuits offer an appealing versatility for molecular biomarker and pharmaceutical chemistry, among other applications.

Supporting Online Material

Materials and Methods

Figs. S1 to S7

Movie S1


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article