|Home | About | Journals | Submit | Contact Us | Français|
This paper describes a microfluidic system to screen and optimize organic reaction conditions on a submicrogram scale. Optimization of reaction conditions is required to achieve high efficiency and selectivity in organic reactions. Combinatorial methods1 and high-throughput screening2 are powerful tools for optimization. To perform solution-phase synthesis, typical microtiter plates or reaction blocks for parallel synthesis run reactions on the scale of mL/reaction1 and are less applicable to precious substrates (e.g., products of long synthetic sequences and natural products that can be isolated only in small quantities). To address this problem, one approach used arrayed micro-wells in combination with a robotic liquid sampler on the scale of ~125 nL per reaction.3 To reduce the use of robotics and to minimize evaporation, others used microchannels4–6 to perform reactions, including synthesis of pyrazoles with UV detection (5 μL per reaction)6 and optimization of glycosylation conditions5 (~2 mg reagent per reaction).
Here, we report a screening method that consumes substrates on the scale of less than 1 μg per reaction. The system uses discrete droplets (plugs)7 as microreactors8 separated and transported by a continuous phase of a fluorinated carrier fluid. Such approach is not limited to microfluidics–fluorinated fluids were used previously to segment samples in NMR9 and PCR10 in tubes to prevent dispersion of sample solutions. Previously, we demonstrated the use of a microfabricated PDMS plug-based microfluidic system to perform assays and crystallization experiments in aqueous solutions with optical detection.11 Here, we developed an approach that does not require microfabrication of microfluidic devices,12 is applicable to synthetic reactions in organic solvents, and uses detection by MALDI-MS.
The system consisted of three components: preformed cartridge11,13 of reagent plugs, a PEEK Tee, and a receiving tubing (Figure 1). A cartridge is an array of discrete plugs surrounded by fluorinated carrier fluid; each plug is composed of a solution of a different reagent. The cartridge was prepared by serially aspirating the reagents into a piece of Teflon tubing prefilled with carrier fluid. A commercially available PEEK Tee connected the cartridge and the inlet tubing, containing as little as a submicroliter volume of a solution of the substrate. Fluorinated carrier fluid (FC-70) was used to fill the two syringes (Figure 1) and the connecting tubing, enabling no-loss manipulation of submicroliter volumes of solutions. FC-70 has low miscibility with organic reagents and reasonably low viscosity (Supporting Information). To perform the reactions, the flow was induced with the two syringes, and the reagent plugs were sequentially merged with the substrate solution. After all resulting plugs flowed out of the Tee into the receiving tubing, the flow was stopped, and the receiving tubing was sealed. After incubation, the reaction plugs were deposited onto a MALDI plate, evaporated, and then a solution of MALDI matrix was deposited over, dried, and analyzed by MALDI-MS.2,14,15 We were helped by work on single-neuron analysis,16 where ~150 nL droplets were used. Here, the low surface tension of the fluorinated carrier fluid facilitated deposition of ~30 nL droplets.
Control experiments were performed to characterize merging. Relative volumetric flow rates of substrate and reagent solutions controlled the ratio in which they combined. Reagent plugs were typically 15 nL in volume, and each combined with 11 nL substrate solution during merging (established by high-resolution digital microphotographs). To avoid potential cross-contamination between neighboring plugs during merging, we introduced one long “blank” solvent plug between every two reagents.11 About 22 nL of substrate solution was deposited into each blank plug. Fluorescence measurement showed that the cross-contamination was below 1% (Supporting Information).
To test this screening system, we used as an example the deacetylation (Scheme 1) of ouabain hexaacetate (Ac6-OUA). We are interested in derivatives of ouabain (OUA) because it is a poisonous cardiac glycoside and is extensively used in studies of neurons17 to block the sodium pump (Na–K–ATPase). Selective deacetylation of Ac6-OUA has been reported only to the ouabain triacetate (Ac3-OUA) in 22% yield.18
We envisioned that screening for optimal conditions would consist of two steps–screening a large number of reagents to evaluate reactivity rapidly, and then repeating the screening with a narrower set of reagents while varying reaction conditions (time, solvents, and concentrations). To determine the best reagents to prepare Ac3-OUA, we merged six cartridges containing 44 reagents with 0.01 M Ac6-OUA solution in 1:1 dioxane:ethanol, incubated the reaction plugs for 1 h at 18 °C, deposited the plugs to the MALDI-plate, and quenched the reactions by adding acidic matrix solution. To evaluate the reaction efficiency by MALDI-MS, we compared the fraction of peak area (A) of Ac3-OUA (defined as AAc3-OUA/∑Aall OUA derivatives) for each reagent (Figure 2). MALDI-MS can be used to characterize reactions quantitatively or semi-quantitatively.2,14,15 We confirmed that there was a strong correlation between the fraction of peak area in MALDI-MS of the compound and its actual fraction in the sample. Measurements among samples with different salt additives were reproducible (Supporting Information).
Consuming ~20 μg of the substrate (~2 μL of solution), we established that inorganic bases showed the highest reactivity for hydrolysis, followed by several organic amines, and all the Lewis acids and lipases tested showed very low reactivity. On the basis of this screening, we chose reagent 39 (0.06 M Na2CO3 in 1:2 H2O: methanol) for 0.1 mmol scale-up reaction. Using 83.7 mg of Ac6-OUA, we obtained 60.3 mg of Ac3-OUA (85% isolated yield) (Scheme 1).
To optimize reaction conditions on a submicrogram scale, we followed the reaction progress with amines or inorganic bases as the reagents (Figure 3). Regioisomers of ouabain tetraacetate (Ac4-OUA) and ouabain pentaacetate (Ac5-OUA) are unstable intermediates in hydrolysis of Ac6-OUA, and their syntheses have not been reported. By quenching reactions at different times, we were able to follow the reaction course in solvents, including DMF, dioxane, acetonitrile, methanol, and water. Using the optimized conditions, we performed 0.1 mmol (83.7 mg) scale reactions and obtained Ac5-OUAs and Ac4-OUAs, with isolated yields of 39% (25% of major regioisomer) and 47% (19% of major regioisomer), respectively (Scheme 1).
A potential problem of microfluidic approaches is large overhead consumption of substrate during interfacing of the chip with the macroscopic world–while an experiment itself may consume only a few microliters of solution, a much larger volume of solution may be lost to fill the connecting lines and delivery channels. To demonstrate the no-loss handling of substrate with this system from start to finish, we dissolved 7 μg of Ac6-OUA in 0.8 μL of DMF, merged the solution with 14 reagent plugs, and obtained MALDI-MS data consistent with other screens.
This solution-phase screening method can be extended to multistep synthesis19 and generates little waste. It will be complementary to the methodologies using substrates attached to solid surfaces.1,15 To optimize this system, we are developing a merging method that will eliminate cross-contamination without introducing blank solvent plugs. Performing reactions under anhydrous and/or anaerobic conditions was not demonstrated here, but it should be possible to use this compact system in a glovebox. Screening reactions at elevated temperatures requires further study to evaluate the dissolution of reagents in carrier fluid. As a preliminary step, we performed several deacetylation reactions at 40 °C and obtained data consistent with the expected reactivity of reagents. In this study, we used MALDI-MS to follow the reaction progress. It is semiquantitative but cannot easily distinguish between isomers. Integrating micro separation and characterization methods (e.g., capillary-LC or microcoil-NMR9) into this system will improve it further. This simple system for performing submicrogram, nanoliter scale reactions should become useful for optimizing reactions of precious substrates over a large chemical space.
This work was supported by the NIH (R01 EB001903), NSF CRC CHE-0526693, and DuPont Young Professor Award. R.F.I. is an Alfred P. Sloan Research fellow. T.H. is a research fellow of the Japan Society for the Promotion of Science.
Supporting Information Available: Materials and methods, control experiments, and additional MALDI-MS data. This material is available free of charge via the Internet at http://pubs.acs.org.