Compact, cost-efficient microfluidics-based stopped-flow device
- First Online:
- Cite this article as:
- Bleul, R., Ritzi-Lehnert, M., Höth, J. et al. Anal Bioanal Chem (2011) 399: 1117. doi:10.1007/s00216-010-4446-5
- 313 Views
Stopped-flow technology is frequently used to monitor rapid (bio)chemical reactions with high temporal resolution, e.g., in dynamic investigations of enzyme reactions, protein interactions, or molecular transport mechanisms. However, conventional stopped-flow devices are often overly complex, voluminous, or costly. Moreover, excessive amounts of sample are often wasted owing to inefficient designs. To address these shortcomings, we propose a stopped-flow system based on microfluidic design principles. Our simple and cost-efficient approach offers distinct advantages over existing technology. In particular, the use of injection-molded disposable microfluidic chips minimizes required sample volumes and associated costs, simplifies handling, and prevents adverse cross-contamination effects. The cost of the system developed is reduced by an order of magnitude compared with the cost of commercial systems. The system contains a high-precision valve system for fluid control and features automated data acquisition capability with high temporal resolution. Analyses with two well-established reaction kinetics yielded a dead time of approximately 8-9 ms.
KeywordsStopped flowMicrofluidicsLow sample amountsShort dead time
First introduced in 1940 by Chance  and improved in 1964 by Gibson , stopped-flow technology is a highly versatile and convenient method for monitoring fast reactions. It is often used to study the dynamic interactions of biomolecules or other macromolecules, where it can provide insights into the molecular details of enzyme functions, protein–protein or protein–DNA interactions, as well as molecular transport mechanisms [3–5]. Another application is in the investigation of protein folding kinetics . Understanding these kinds of processes is of growing importance not only for basic research but also to the pharmaceutical industry. For example, stopped-flow technology is key in elucidating catalytic enzyme mechanisms and functions of specific proteins in an organism. Reactant concentrations can be monitored photometrically, and changes in enzyme conformation during substrate binding or product release can be tracked using fluorescent dyes. In some cases, changes in spectral properties of the reactants can be exploited to optically detect the binding process. A prerequisite, however, is that the substances be fully mixed before the onset of binding. Clearly, then, manual mixing is inadequate for observing processes on the timescale of milliseconds; a more efficient mixing method is needed . Also, highly time resolved data acquisition must be available as part of the setup.
As a result of these stringent requirements, commercial stopped-flow devices have tended to be complex in design, incorporating various components (i.e., light source, amplifier electronics, detection unit, mixing station) that in the aggregate make these systems bulky, expensive to purchase and maintain, and prone to mechanical maladjustments. Typical equipment often contains two step-motor-driven syringes which push the reactants (e.g., enzyme and substrate), at very high velocity and in one go, first through a mixing chamber and then into the optical detection cell. The detection unit, in turn, is triggered by a stop syringe. The processes within the detection cell may be monitored via changes in absorption or fluorescence intensity. Such monitoring can, however, only be initiated after the so-called dead time has elapsed, which denotes the time span between the first contact of the reactants and the beginning of measurements in the detection region which are usable for kinetics analysis. It is thus dependent on both the mixing and the transport times. In effect, only processes happening on timescales longer than the dead time can be monitored. The dead time is determined mainly by the fluid flow rate, the mixing efficiency, the volume traversed by the fluid between the mixing chamber and the detection cell, and the volume of the detection cell [8, 9].
The intricate setups of conventional stopped-flow systems necessitate relatively high sample and reagent volumes during the first operating cycle. In serial experiments conducted without changing chemicals, however, sample consumption is very low. Still, the volumes required in the first cycle are often already prohibitive as samples and reagents are expensive or hard to isolate. Volumes of up to 3 ml during setup and, subsequently, of 1 ml per experiment are not uncommon. Extensive test series probing various parameters, as are commonly performed, thus require considerable reactant quantities. Despite this high rate of sample consumption, however, most of the volume is consumed rather wastefully in the flushing and washing steps. Here is where our microfluidics-based stopped-flow system provides substantial improvements in simplicity and cost-efficiency: the use of injection-molded disposable microfluidic chips reduces the required sample volume to 20 μl even in the first experiment as flushing is no longer necessary; sample costs are hence lowered significantly; handling is simplified; and adverse cross-contamination effects are prevented.
The chip layout and material choice allow for the use of narrow-bandwidth light-emitting diodes (LED) as light sources in the UV–vis spectrum instead of conventionally used xenon and mercury arc lamps. These UV and visible LEDs can easily be replaced with ones covering different wavelengths as the case may demand, thus broadening the applicability of our approach. Furthermore, UV–vis-sensitized photodiodes covering a wide range of wavelengths are employed instead of photomultipliers for detection.
In our case, a uniform illumination of the flow channel over a length of 10 mm is achieved with integrated on-chip mirrors that reflect light toward the detection cell and the detector. Coupling the photodiode with simple but effective amplifier electronics offers excellent sensitivity and resolution over the spectral range relevant in most biological applications (i.e., 250-1,000 nm).
The overall cost of the stopped-flow device at current component prices, including costs for housing, electronics, and optical elements, is in the low four-digit euro range, which is roughly an order of magnitude lower than that of commercially available solutions. Moreover, economies-of-scale effects are expected to further lower the per-unit cost at higher production volumes.
The functionality of our system is evaluated by reaction III: labeling histone deacetylase like amidohydrolase (HDAH) from Bordetella with fluorescent dyes . This bacterial amidohydrolase is structurally highly homologous to, and thus serves as an easily produced proxy for, the human histone deacetylases currently under investigation as possible targets for cancer therapy. Because of this high homology, the transferability of drug screening results is expected in many cases.
Materials and methods
Poly(methyl methacrylate) (PMMA; Plexiglas® 7N) was chosen as the polymer chip material owing to its high degree of transparency in the UV range and low autofluorescence, which allows dynamic protein investigations at a wavelength of 280 nm (the intrinsic fluorescence wavelength of tryptophan).1 Structures for fluid transport and mirrors for light reflection were machined into injection-molded chip blanks by precision milling and laser ablation technologies. The outer chip dimensions are 43 mm × 64 mm. The channels for reactant inlets are 1 mm × 1 mm, having a semicircular shape at the bottom (radius 0.5 mm). The T junction has a depth of 0.4 mm and a width of 0.5 mm. The length of the tapered channel is 9.5 mm and the diameter at the connection of the detection cell is 1 mm. The detection cell is 10 mm long and 1 mm in diameter. The volume of the complete structure is 64.5 μl, where the volume from the T junction to the end of the detection cell is 16.5 μl (Figs. 2, 3). The structured chip is bonded with microencapsulated foil (40 μm, polyolefin, HJ Bioanalytik, Mönchengladbach). To enhance the wettability of the channel surfaces and suppress unspecific adsorption processes, the surfaces were processed with ethylenediamine or ethanolamine as described by Stein et al. .
For the electronics, controlled high-precision magnetic valves purpose-made for miniaturized instruments (The Lee Company, Westbrook, CT, USA) were used. They combine the advantages of a very fast response time (less than 1 ms), a small dead volume (a few microliters), and a low operating voltage (12 V). Valves resistant to pressures up to 120 psi (approximately 8 bar) are available. Since these valves are prone to contamination with small particles, it is recommended that the liquids be filtered before induction (preferably using pores with diameters of 10 μm). Syringe pumps are used to load the samples and to pump them without pulsation (cetoni, Korbußen/Gera, Germany).
The detection unit consists of a LED for excitation (Roithner Lasertechnik, Vienna, Austria) and photodiodes for absorption and fluorescence detection (Silicon Sensor, Berlin, Germany). The photodiodes are equipped with a cutoff or band-pass interference filter for wavelength selection. Alternatively, absorbing filters made of a bulk polymer such as polycarbonate (Makrolon® UV 2099, 2 mm, Röhm, Darmstadt, Germany) or polystyrene (1 mm, Greiner Bio-One, Solingen, Germany) could be used to separate the excitation light from the fluorescence signal if their peak wavelengths are sufficiently distant from each other (e.g., deep-UV excitation at 280 nm with UV fluorescence at 335 nm, or UV excitation at 385 nm and visible fluorescence at 480 nm).
The beam propagation inside the chip prevents direct exposure of the fluorescence detector to the excitation beam. Since the photoelectric current of the diodes is in the picoampere to microampere range, a transimpedance amplifier (sglux SolGel Technologies, Berlin, Germany) is used for signal enhancement and to transform current signals to voltage signals, the latter with amplification factors of 107 V/A (off-line) or 1010 V/A (in-line) and output signals not exceeding ±4 V. The high amplification factors used require the system to be shielded electrically with an aluminum cover, which also serves as an optical shield blocking ambient light.
As mentioned already, fast mixing is a major requirement in highly time resolved measurements. Various mixers with mixing times down to the millisecond range are discussed in the literature [21–25]. Commercially available stopped-flow instruments most often contain ball mixers  utilizing the turbulent wake which occurs downstream from a spherical surface. Since mixing is ultimately a diffusion process on microscopic length scales, mixers generally aim to increase the contact area of the reactants to achieve faster mixing [21–25]. This is often achieved by subjecting the flow to turbulent flow conditions.
Falk and Commenge  recently demonstrated that the main factor determining mixing times is the pressure drop across the mixer. They further showed that the mixing time is governed by a common empirical scaling behavior with respect to the pressure drop across a wide array of mixer geometries and flow rates. The task of finding an appropriate mixer thus reduces to selecting a geometry that will produce the required pressure drop at the pertinent flow rate. Using this result, which holds in both the laminar and the turbulent flow regimes, we conclude that a basic T mixer geometry  suffices to obtain the required mixing performance.
Adjusting the pressure drop via variations in the spatial dimensions of the mixing section allows this mixer to easily fit into the overall layout of our system. At the junction of the T structure, the two solutions collide and merge at high velocities. The reactants are mixed by the induced vortices, and the reaction starts within the short mixing channel . Upon leaving the mixing channel, the mixture is propelled directly into the optical detection cell.
Dead times of commercial systems
Dead time (ms)
Applied Photophysics SX 18MV
Applied Photophysics SX MV
2 (Fluo), 8 (CD)
Applied Photophysics SPF-17
AVIV Instruments stopped-flow tower interface with AVIV 202 SF spectrometer
TgK Scientific HPSF-56
Unisoku RSP-1000 (Japan)
Unisoku type USP-539
The flow may be accelerated by increasing the applied pressure. As a result, turbulences will tend to be amplified, leading to faster mixing and shorter dead times (as defined above).
Dead time determination by absorption and fluorescence measurements—reactions I and II
The dead time of the stopped-flow device is defined as the time interval from the onset of the reaction of interest to the time when it can first be monitored. It depends mainly on the mixing time, the flow rate, and the dead volume and can be determined from signals (e.g., absorption or fluorescence) measured in simple first-order reactions [8, 9]. To examine the dead time of our device, we measured reaction I (by absorption) and reaction II (by fluorescence) as described in “Introduction.” In both cases, a signal (absorption or fluorescence) is obtained which varies as the chemical reaction progresses. Although the reaction starts immediately upon first contact of the reactants, this initially happens outside the detection zone.
For each set of experiments at different reactant concentrations, we repeat the procedure just described. The rate constant, k, of the reaction is obtained for each concentration directly as part of the respective fit. To determine the dead time, we proceed as follows. First, the median-normalized and averaged signal plots for the different reactant concentrations (as shown, e.g., in Fig. 5b) are shifted relative to each other such that the endpoints of their fits coincide. The objective here is to gauge all graphs to a common asymptote representing the saturation stage of the reaction. It is pertinent that the drift parameter, b, varied only very slightly (typically less than 2%) across fits for different concentrations, justifying our normalization procedure.
The point at t < 0 where the graphs intersect marks the lower end of the dead time interval. The upper end is given by t = 0. Following Peterman , we plot the logarithm of the graphs after subtracting their long-time limits (thereby “linearizing” them; see, e.g., Fig. 5c) to better discern their common crossing point. The dead time is then given by the distance on the time axis between this point and t = 0.
HDAH expression, isolation, and labeling with fluorescamine–reaction III
HDAH was expressed in, and purified from, bacterial strain XL1-Blue-FB188-HDAH. The strain and protocols were kindly provided by C. Hildmann (TU Ilmenau, Ilmenau, Germany). The HDAH concentration was determined using a bicinchoninic acid protein determination kit (Interchim) as prescribed by the manufacturer. When excited at 390 nm, fluorescamine (4-phenylspiro-[furan-2(3 H),1-phthalane]-3,3′-dione) exhibits strong fluorescence at 475-490 nm when bound to proteins. For the reaction of 3 μM HDAH in 0.2 M borate, a buffer with a pH of 8.6 was mixed 1:1 with 1-4 mM fluorescamine in dimethyl sulfoxide. For optical detection, a 385-nm LED was used. The amplifier gain was set to 5·107 V/A.
The data from reaction III were fitted to the expression a + bt + c exp(-kt) in the same fashion as for reactions I and II to obtain reaction rate constants, k. Since the dead time is not a critical parameter for this analysis, it was not determined again here. The signal plots used (Fig. 7) show clearly undisturbed data, justifying our analysis procedure.
Results and discussion
Dead time determination by absorption measurements—reaction I
The first reaction used to determine the dead time of the system was the reduction of DCIP by ascorbic acid (Fig. 5). Data were acquired at a sample rate of 10,000 s-1 (individual measurements, no integration was performed) starting 10 ms after the closing of stop valve V-6. Analyses of seven independent measurements revealed an average rate constant, k, of (94 ± 13) s-1. After averaging and data fitting, we obtained a rate constant, k, of 91.1 s-1 (Fig. 5a).
The reaction speed can be influenced by varying the concentration, yielding half-lives shorter than 1 ms. For the reaction, 5-20 mM ascorbic acid in 0.2 M NaCl at pH 4 was mixed 1:1 with 140 μM DCIP. For the optical detection, a 525-nm LED and a detector with an interference filter (536 nm, band width 40 nm) were used. The amplifier gain was set to 109 V/A.
Repeating the same experiments with different ascorbic acid concentrations (2.5-10 mM) reveals a nearly linear dependence of the rate constant on the ascorbic acid concentration (inset in Fig. 5b). The dead time is obtained as described above and amounts to approximately 9 ms, which is competitive with conventional stopped-flow systems, e.g., from TgK Scientific (Bradford-on-Avon, UK), whose dead times are specified as 8-10 ms.
Rate constants for the DCIP reduction by ascorbic acid, however, are difficult to compare since for pH 4 Tonomura et al.  only reported data for a higher ascorbic acid concentration (20 mM). Moreover, between pH 4 and pH 5, the rate constants are rather sensitive to the pH of the reaction mixture (pH 4.07, k = 838 s-1; pH 4.91, k = 224 s-1 ). Yet, the rate constant of 149 s-1 (Fig. 5) determined with our stopped-flow device for 10 mM ascorbic acid lies in the anticipated range, since a linear extrapolation of pH-dependent data given in  to lower concentrations results in k values of 419 and 112 s-1 for pH 4.07 and pH 4.91, respectively.
Dead time determination by fluorescence measurements–reaction II
Another standard reaction for the determination of the dead time is the pseudo-first-order quenching reaction of NATA by NBS, which exhibits rate constants, k, ranging from 70 to 155 s-1 . Experiments were conducted using 10 μM NATA (Sigma-Aldrich) mixed 1:1 with 200-400 μM NBS (Sigma-Aldrich), leading to total NBS concentrations of 100-200 μM. The excitation wavelength was 280 nm, and the emission was measured above 300 nm using a polystyrene filter. All solutions were freshly made in degassed 50 mM phosphate buffer, with a pH of 7 in a nitrogen atmosphere. For the optical detection, the LED was combined with a polystyrene absorption filter. Being able to conduct measurements in the UV range is of particular interest in investigations involving the intrinsic fluorescence present in proteins.
Functionality evaluation using HDAH–reaction III
Conclusion and outlook
A microchip-based automated stopped-flow device has been presented whose small size and significantly reduced reagent consumption represent an important advance for the measurement of highly time resolved reaction kinetics. The present system now puts applications previously not considered feasible, owing either to prohibitive costs or insufficient amounts of available reagent volumes, within the realm of possibility. Based on a microfluidic polymer chip and a compact custom-made instrument, the system is flexible in application, easy to handle, and cost-efficient in both acquisition and operation. The overall cost of the system presented here (low four-digit euro range) is roughly an order of magnitude lower than that of commercially available systems. Economies-of-scale effects are expected to further reduce per-unit costs and thus open up a path to providing low-cost solutions to end users.
Allowing linear absorption measurements up to about A490 = 1.5 and a lower detection limit of fluorescein (excitation wavelength, 490 nm; emission wavelength 530 nm; amplification 107 V/A) at 50 nM, it is as efficient as most conventional photometers (the most sensitive fluorescence spectrometers have detection limits down to 5 nM fluorescein ). The integration of a stop valve helps make the fluid control more robust. The possibility to wash the system after each experiment allows for consecutive series of measurements on the same chip.
To validate the performance and to determine the dead time of the microchip-based stopped-flow device, three different reactions were investigated. Also, measurements in the UV range can be conducted, e.g., for investigations of the intrinsic fluorescence of proteins. Our investigations showed agreement with data published in the literature for all three test reactions. In particular, two reactions showed dead times of approximately 8-9 ms, which is competitive with conventional stopped-flow systems. In the third reaction, kinetic parameters were reproduced consistent with the published reaction scheme.
Although our system is already competitive with commercially available stopped-flow systems, a further reduction in dead times to approximately 0.25 ms will be required to approach the performance level of today’s best-performing systems. Another goal is the further reduction of sample volumes. In this pursuit, reducing the channel length before the mixing structure and behind the detection cell may prove advantageous. Therefore, we envisage the integration of fast switchable valves on the chip. Despite the turbulent flow conditions prevalent in our system, the mixing efficiency can and must be enhanced. Introducing structures to that effect should shorten mixing times, which would be desirable especially for cases of asymmetric reagent volumes. Application of higher pressures will reduce the dead time further but may come at the price of reduced reproducibility due to higher turbulence. By careful design of the mixing structure, it may be possible to reduce dead times below the 1-ms mark.