Continuous purification of reaction products by micro free-flow electrophoresis enabled by large area deep-UV fluorescence imaging
- 490 Downloads
Microreactors have gained increasing attention in their application toward continuous micro flow synthesis. An unsolved problem of continuous flow synthesis is the lack of techniques for continuous product purification. Herein, we present a micro free-flow electrophoresis device and accompanying setup that enables the continuous separation and purification of unlabeled organic synthesis products. The system is applied to the separation and purification of triarylmethanes. For imaging of the unlabeled analytes on-chip a novel setup for large area (3.6 cm2) deep ultra violet excitation fluorescence detection was developed. Suitable separation conditions based on low conductivity electrophoresis buffers were devised to purify the product. With the optimized conditions, starting materials and product of the synthesis were well separated (R > 1.2). The separation was found to be very stable with relative standard deviations of the peak positions smaller than 3.5% over 15 min. The stable conditions enabled collection of the separated compounds, and purity of the product fraction was confirmed using capillary electrophoresis and mass spectrometry. This result demonstrates the great potential of free-flow electrophoresis as a technique for product purification or continuous clean-up in flow synthesis.
KeywordsContinuous flow Free-flow separation Flow microreactor synthesis Ultraviolet fluorescence
Micro flow synthesis has gained a lot of attention recently because of its many advantages in reaction optimization and synthesis automation [1, 2, 3, 4]. The main strength of microreactors are that many commonly performed reaction steps like dosing of reagents, heating, cooling, and mixing can be conveniently implemented in a continuous fashion. However, other crucial operations in multi-step synthesis are thus far not easily transferred to continuous flow. For example, intermediate solvent exchanges, purification of intermediates, and most importantly the purification of the product are often performed in a discontinuous off-chip manner.
Research into continuous purification systems has only taken off in recent years. Among others, the Jensen group has been developing multistep microreactor networks with intermediate phase separations  and distillation  devices. Other continuous work-up strategies are micro evaporation , spray drying , and liquid–liquid extraction .
Micro free-flow electrophoresis (μFFE) has been identified as a promising technique for purification of reaction products emerging from microreactors [10, 11]. Yet, only few publications have demonstrated work in this direction. Most research on μFFE is concerned with separating biologically relevant molecules like peptides or proteins [11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21]. We have previously shown the feasibility of integrating a microreactor with subsequent separation by free zone electrophoresis  and isoelectric focusing . In those publications, fluorescent labeling of amino acids and peptides was performed and the reaction mixtures were continuously separated. However, recovery of the analytes after separation was not a focus in those reports.
Separation and purification of more conventional organic synthesis products has not been shown yet. This is a result of various technical challenges arising when the effluent of an organic synthesis is to be separated. Among others, the solubility of synthesis products, the abundance of neutral species, and compatibility issues of synthesis solvent and separation buffer can be challenging .
Furthermore, the observation of the separation, which is necessary for optimizing separations on the device, is challenging. The prevailing detection technique in μFFE is fluorescence imaging in the visible light (VIS) range [23, 24]. Obviously, this requires intrinsically fluorescent molecules or necessitates a labeling reaction. Unfortunately, the majority of synthesis products are not fluorescent in the VIS range nor is labeling of the products usually desired. Great progress has been made in recent years with regards to novel detection techniques for μFFE. Among others, detection systems were devised based on mass spectrometry [25, 26, 27, 28], deep ultra violet (deep-UV) fluorescence excitation [29, 30], surface enhanced Raman spectroscopy , cell-based signaling , and detection via saccharide specific fluorescent probes .
Despite this progress with regards to detection techniques, to date a continuous separation and purification of organic synthesis products using μFFE has not been shown. Herein, we present our most recent work with μFFE, including the first continuous separation and fractionation of unlabeled organic synthesis products. For on-line observation of the separation, a novel detection setup for large area deep-UV excited fluorescence detection was developed. The detection system was instrumental for optimization of the chip system and the separation conditions.
All substances were used as received without further purification. Polyethyleneglycol diacrylate (PEG-DA, MWavg 250, and 575 g*mol–1), 3-(trimethoxysilyl)propyl methacrylate (TPM), 2,2-dimethoxy-2-phenylacetophenone (DMPA), 1-napthol, and 3-(cyclohexylamino)-1-propanesulfonic acid (CAPS) were purchased from Sigma-Aldrich (Darmstadt, Germany). Chloroform, n-heptane, and acetonitrile (ACN) were purchased from VWR (Dresden, Germany). Na2CO3, NaHCO3, NaCl, H2SO4, and H2O2 were from Carl Roth (Karlsruhe, Germany). HPMC was from Ferak Laborat (Berlin, Germany), and Triton X-100 from Riedel de Haen (Seelze, Germany). The other synthesis materials as shown later were synthesized in the work group of Professor Schneider (University of Leipzig).
μFFE chips were produced largely following our previously published procedures using liquid-phase lithography between two glass slides . For the lid of the device, access holes for fluidic and electrical contacting were powder blasted (Barth, Königstein, Germany) in quartz microscopy glass slides (Science Services, Munich, Germany). The bottom quartz glass slide was sputtered with platinum to form electrodes for electrical contacting. Platinum deposition was performed in two steps: First, the slides were thoroughly cleaned using acetone, isopropanol, ultrapure water, and piranha solution (H2SO4, H2O2, 2:1). As a positive photo resist AR-P 3510 (ALLRESIST, Strausberg, Germany) was applied to the glass slide using a spin coater following the manufacturer’s recommendations. After a prebake of the substrate, a photomask was applied and exposed (18 s) using a mask aligner (MA-6; SÜSS MicroTec, Munich, Germany). The photoresist was developed using AR 300-26 developer agent (ALLRESIST). In the second step, the glass slides were transferred to a sputter coater (BAE 250, Bal-Tec). A layer of chromium (60 nm) and platinum (350 nm) were deposited in sequence. After stripping of the photoresist using acetone, the platinum electrodes remained on the glass slide.
μFFE device operation
The μFFE chip was connected to Nemesys syringe pumps (Cetoni, Korbussen, Germany) equipped with glass syringes from ILS (Stützerbach, Germany). All but the electrode channels were connected by capillary tubing (i.d. 150 μm, o.d. 360 μm; Machery-Nagel, Düren, Germany). The electrode channels were connected using Teflon tubing (i.d. 500 μm, o.d. 1.58 mm; ESKA, Hamburg, Germany). The analyte inlet was further connected to a HPLC injection valve (Knauer, Berlin, Germany) equipped with a 20 μL sample injection loop.
Figure 2 shows a frontal view of the separation bed and highlights the different inlets; namely, analyte, buffer, and electrode stream. The separation buffer consisted of 20 mM CAPS (pH 10.0) containing 0.1% HPMC. The electrode channels were flushed with the same CAPS buffer but supplemented with 1 M NaCl and 1 mM Triton X-100.
The deep-UV fluorescence detection setup was based on a confocal fluorescence lifetime microscope MicroTime200 (PicoQuant, Berlin, Germany). The system was additionally equipped with a deep-UV extension consisting of a deep-UV capable photo multiplier tube (PMA 165-N-M; PicoQuant) and a 40× quartz objective (Partec, Münster, Germany). As the excitation source, the fourth harmonic of a Nd:YVO4 laser (Cougar, Time-bandwidth) was coupled to the auxiliary light path. The beam was reflected by dichroic mirrors (266 RazorEdge, Semrock) and guided into the objective of the modified epifluorescence microscope (iX 71; Olympus). The fluorescence emission was collected with the same objective and guided through a quartz tube lens, a 100 μm confocal pin-hole, and an emission filter (short pass 532 nm, SP532-RS; Semrock) onto the photo multiplier. The laser and detection unit were synchronized using the time correlated single photon counting electronics provided by the MicroTime200. Data acquisition was performed using the accompanying SymphoTime32 software (PicoQuant). To enable large area scanning, the microscope was further fitted with a motorized specimen stage (TANGO 2 Desktop; Märzhäuser Wetzlar, Wetzlar, Germany), which was controlled by the accompanying computer software (SwitchBoard; Märzhäuser Wetzlar). A photograph of the optical setup and its components is provided in the Electronic Supplementary Material (Fig. S1).
The software solution translating fluorescence intensity over time to pseudo colored intensity images as described in the Results section was implemented in Python (https://python.org). Libraries used for implementation of the necessary transformations were numpy (https://numpy.org), scipy (https://scipy.org), photon-tools (https://github.com/bgamari), PyQT4 (Riverbank Computing, Wimborne, UK), and plotly (https://plot.ly). For more general data analysis and plotting, OriginPro 8.5 (OriginLab, Northampton, MA, USA) was used. pKa values of the different synthetic compounds were calculated using ChemDraw 15.0 Professional (Perkin Elmer, Hamburg, Germany).
Capillary electrophoresis experiments were conducted on a commercial instrument (P/ACE MDQ; Beckman Coulter, Krefeld, Germany). The instrument was equipped with a fused silica capillary (i.d. 50 μm) with a total length of 60 cm (50 cm effective) and a diode array absorbance detector. Injections were performed by applying pressure (35 mbar) for 15 s to the sample vial. For separation, a potential of 20 kV was applied to the thermostatted capillary (25 °C). Samples from the reaction mixture were diluted in separation buffer prior to separation. Samples from the fractions collected at the outlets of the μFFE chip were injected carefully. The starting materials and reaction products of the triarylmethane synthesis were spectroscopically characterized using absorbance and fluorescence spectrometers (V-650 and FP-6200; Jasco, Groß-Umstadt, Germany).
Results and Discussion
For purification of these products and separation from catalysts and starting materials, Saha et al. employed column chromatography . For a successful transfer of the purification step to continuous free-flow (zone) electrophoresis the molecules should carry a charge or be ionizable in a suitable buffer. Therefore, the pKa values of the hydroxyl groups of the reaction partners (1-3) were calculated. The pKa values were in the range of 9–10. This suggested that deprotonation should occur at a buffer pH of about 10 or higher. To confirm the pKa values and to prototype possible buffers for μFFE separations, capillary electrophoresis (CE) experiments were conducted. A carbonate buffer (20 mM, pH 10) was selected as a starting point. Figure 3b shows the electropherogram of a reaction mixture containing 100 mM each of the starting materials and 5 mM of diphenyl phosphate catalyst after 45 min reaction time in a vial. All substances were found to migrate after the electroosmotic flow (EOF) and therefore carried negative charge as expected. The peak identity was confirmed using standards and the obtained migration time matched observed migration times of standards. Full baseline separation of the three compounds (1-3) was achieved in less than 7 min with this protocol (Fig. 3b). Only the acid catalyst (4) was not fully separated from 1-naphthol (2).
The analytes were then screened with regard to their spectroscopic behavior. Macroscopically, solutions of the compounds in acetonitrile were optically clear, suggesting no appreciable absorbance in the VIS range, which was confirmed by absorbance measurements (data not shown). The analytes only showed absorption below 350 nm with strong absorption maxima at or below 250 nm. The substances were further characterized with regard to their fluorescent properties and specifically with respect to deep-UV (<300 nm) excitation. Figure 3c shows the fluorescence emission spectra of the starting materials and one representative triarylmethane product. All species involved, with the exception of the diphenyl phosphate, displayed fluorescence emission in the region of 300–450 nm upon excitation with 266 nm.
Large area deep-UV fluorescence imaging
To visualize the separation on the μFFE device, we opted to employ deep UV fluorescence detection. It has been previously shown that deep-UV excitation fluorescence detection is a valuable tool to detect unlabeled analytes in microfluidic systems in general [36, 37] and to visualize μFFE separations in particular [29, 30]. However, these instrumental developments were not applicable for the current approach because they only enabled visualization of a small portion of the separation bed. Herein, we employed μFFE devices with a separation bed area of up to 3.6 cm2.
The resolution of the resulting images depends on the scanning speed of the electronic specimen stage, the number of lines scans along the bed, and the detection volume of the detector system. In a typical experiment, as shown herein, 20 scans across the separation bed were performed equally distributed from the inlets to the outlets of the chip (~8 lines per cm). Every line scan took about 2.5 s to complete; thus a complete image was acquired after 50 s. While this might be slow in comparison to directly imaging using, e.g., a CCD-camera , it is more than sufficient for μFFE separations, which ideally show a steady positioning of analyte bands. For comparison, the residence time in the separation shown below was 55 s. In essence, as long as the sample is continuously introduced and the residence time of the sample exceeds the acquisition time, the resulting images accurately reflect the band position in the separation bed. The distinct advantage of the developed setup is the large area that can be scanned and imaged. Previous deep UV imaging setups for μFFE were restricted to imaging of 0.12 cm2  and 0.57 cm2 . The newly devised approach is able to image the whole separation bed of devices herein, which covered an area of 3.6 cm2 (1.4 cm × 2.6 cm, W × L). Furthermore, the maximum translation of the electronic stage is even bigger and might allow imaging of large areas or multiple sites on larger integrated devices in the future. Additionally, because the underlying detector system of the microscope was a commercially available time correlated single photon counting system, fluorescence lifetimes of the analytes bands could be obtained as well.
μFFE separation and purification
The starting point for optimizing the μFFE separation conditions was the carbonate buffer (20 mM, pH 10) employed in CE measurements as shown in Fig. 3b. While this buffer enabled electrophoretic baseline separations in CE using 50 μm i.d. capillaries, it proved to be less suitable in μFFE. At higher electric fields, the μFFE separations were poorly reproducible due to excessive joule heating. Therefore, lower conductivity buffers were investigated as separation media. Specifically, CAPS (10 mM, pH 10) was found to be a viable alternative. The prepared CAPS buffered had a conductivity of 0.6 mS*cm–1 (Carbonate buffer 2.7 mS*cm–1) and performed otherwise similarly in CE separations. It was found that supplementing the separation buffer with HPMC (0.1% w/w) to suppress residual EOF helped to stabilize the separation conditions. The separation electrolyte with the high pH did not damage the PEG walls of the microfluidic system.
The developed electrolyte system was then employed to separate a mixture of analytes resembling a reaction mixture. The sample contained 1 mM each of the analytes (1-3) in a buffer acetonitrile mixture (10 mM CAPS pH 10, 10% ACN). A 20 μL sample was injected and the effluents of the different outlets of the separation bed were collected. Over a period of 20 min, fractions of about 70 μL were obtained from each individual outlet. The separation was monitored simultaneously to ensure stable band positioning in front of the outlets.
It could be shown that μFFE is capable of separating and resolving native unlabeled synthesis products. This was demonstrated for a synthesis of triarylmethanes as a target reaction. To enable observation of the separation, a novel deep-UV excitation fluorescence detection system was developed. In contrast to previous deep-UV imaging approaches, the developed system enabled large area imaging of the separation bed. The chip system and separation conditions were optimized to separate the components of a synthesis mixture. It was not only possible to achieve separation but also to collect the separated bands from an artificial sample. The purity of the collected fractions was confirmed using capillary electrophoresis and mass spectrometry.
To the best of our knowledge, this is the first demonstration of the separation of small synthetic molecules separated and individually collected using μFFE. The next step could be the direct connection of a micro flow reactor and μFFE to perform synthesis and separation on a single chip. The presented approach toward product purification is currently limited to synthesis products that show appreciable fluorescence under deep-UV excitation, which is however, more broadly applicable then, e.g., detection in the VIS region. While we used a sophisticated research laser microscope in the present set-up, the dramatic developments in photonic technology such as deep-UV LEDs and image sensors should allow realizing more economic detection set-ups in the future, thereby facilitating more widespread use of the approach. Especially for further automation in micro flow synthesis where continuous separation techniques are needed, we believe that the application of μFFE has a great potential to supplement the chemist’s toolbox.
The authors gratefully acknowledge funding by the Deutsche Forschungsgemeinschaft (DFG) through grant FOR 2177.
Compliance with ethical standards
Conflict of interest
The authors declare no conflict of interest.
- 4.Plutschack MB, Pieber B, Gilmore K, Seeberger PH. The Hitchhiker’s Guide to Flow Chemistry. Chem Rev. 2017; https://doi.org/10.1021/acs.chemrev.7b00183.
- 22.Herzog C, Poehler E, Peretzki AJ, Borisov SM, Aigner D, Mayr T, Nagl S. Continuous on-chip fluorescence labelling, free-flow isoelectric focusing and marker-free isoelectric point determination of proteins and peptides. Lab Chip. 2016;16:1565–72.Google Scholar
- 24.Novo P, Janasek D. Current advances and challenges in microfluidic free-flow electrophoresis – a critical review. Anal Chim Acta. 2017; https://doi.org/10.1016/j.aca.2017.08.017.
- 30.Poehler E, Herzog C, Lotter C, Pfeiffer SA, Aigner D, Mayr T, Nagl S. Label-free microfluidic free-flow isoelectric focusing, pH gradient sensing and near real-time isoelectric point determination of biomolecules and blood plasma fractions. Analyst. 2015;140:7496–502.Google Scholar