Article pubs.acs.org/ac
Cite This: Anal. Chem. XXXX, XXX, XXX−XXX
Isolating Single Euglena gracilis Cells by Glass Microfluidics for Raman Analysis of Paramylon Biogenesis Nobutoshi Ota,†,∇ Yusuke Yonamine,‡,∇ Takuya Asai,§ Yaxiaer Yalikun,† Takuro Ito,∥,⊥ Yasuyuki Ozeki,§ Yu Hoshino,# and Yo Tanaka*,† †
Center for Biosystems Dynamics Research, RIKEN, Suita, Osaka 565-0871, Japan Research Institute for Electronic Science, Hokkaido University, Sapporo, Hokkaido 001-0021, Japan § Department of Electrical Engineering and Information Systems, The University of Tokyo, Tokyo 113-8656, Japan ∥ Japan Science and Technology Agency, Kawaguchi, Saitama 332-0012, Japan ⊥ Department of Chemistry, School of Science, The University of Tokyo, Tokyo 113-0033, Japan # Department of Chemistry, Kyushu University, Fukuoka 819-0395, Japan
Downloaded via KEAN UNIV on July 17, 2019 at 19:25:55 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
‡
S Supporting Information *
ABSTRACT: Time-course analysis of single cells is important to characterize heterogeneous activities of individual cells such as the metabolic response to their environment. Single-cell isolation is an essential step prior to time-course analysis of individual cells by collecting, culturing, and identifying multiple single-cell targets. Although single-cell isolation has been performed by various methods previously, a glass microfluidic device with semiclosed microchannels dramatically improved this process with its simple operation and easy transfer for time-course analysis of identified single cells. This study demonstrates isolating single cells of the highly motile microalgae, Euglena gracilis, by semiclosed microchannels with liquid flow only. The isolated single cells were identified in isolating channels and continuously cultured to track, by Raman microscopy, for the formation of subcellular granules composed of polysaccharide paramylon, a unique metabolite of E. gracilis, generated through photosynthesis. Through low-temperature glass bonding, a thin glass interface was incorporated to the microfluidic device. Thus, the device could perform the direct measurements of cultured single cells at high magnification by Raman microscopy with low background noise. In this study, the first demonstration of sequential monitoring of paramylon biogenesis in a single identified E. gracilis cell is shown.
S
These methods can perform single-cell manipulation with high precision and/or in a relatively short time. Meanwhile, these methods require additional equipment be attached to the microfluidic device that is too expensive or large to introduce at the laboratory level. To isolate single cells only by liquid flow in microfluidics, sophisticated channel designs have been adopted for cells with low motility.17,18 Among various types of cells, microorganisms are difficult to measure with single-cell isolation and time-course analysis because of their high motility. Conventional cell culturing environments are not suitable for tracking and observing individuals of highly motile microorganisms. When a microorganism is isolated in a small chamber, circulation of culture medium is required to keep the cells in an active condition. Although previous devices17,18 have succeeded in isolating single cells in separate chambers, they need strong flow to keep
ingle-cell analysis reveals the importance of individual cells by characterizing their roles in a diverse group of cells such as organs1 and bacterial communities.2 Besides spatial distribution of organelles and biomolecules, diversity of cells is characterized by time-course activity of individual cells such as enzymatic activities,3 protein levels,4 and responses to drug candidates.5 To track time-course information on single cells, it is important to measure individual cells without undesired influence from the surrounding environment,6 including molecules secreted by the surrounding cells. Thus, isolation of single cells is frequently employed to eliminate an undesired influence from the neighboring cells before single-cell analysis.7−9 Considering operations of single-cell isolation, microfluidic devices are a powerful tool to simplify the process of isolation with diminished labor and reduced chance of damaging cells10−12 compared with conventional pipetting.13 There have been some reports of single-cell manipulation with microfluidics by using techniques such as controlling liquid flow,14 mechanical isolation by actuators,15 and optical tweezers.16 © XXXX American Chemical Society
Received: February 25, 2019 Accepted: June 11, 2019
A
DOI: 10.1021/acs.analchem.9b01007 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry
analysis of microorganisms,26−30 time-course analysis has not been reported for identified single cells with high motility due to lack of an appropriate platform for single-cell isolation, culturing, and observation for these cells. We used a thin glass layer for the interface of optical measurements and a thick glass layer as a solid support for the microstructure to fabricate a totally glass-made microfluidic device. This device has three important advantages: (1) the device is compatible with Raman spectroscopy for single-cell analysis because glass has high transmittance in visible and near-visible light regions, (2) the device can be made thin to meet the short working distance requirement of high-magnification lenses, and (3) the device resists deformation induced by applied pressure due to the physical rigidity of the glass, unlike conventional microfluidic devices made of soft materials such as poly(dimethylsiloxane) (PDMS) which can be readily deformed, causing damage to or loss of culturing cells especially in semiclosed microchannels. Even with these clear advantages of glass, microfluidic devices have been rarely composed of glass only because of challenges on glass bonding. When a device is composed of two glass layers with different thicknesses, glass bonding has a further restriction. Direct glass−glass bonding can be performed by glass fusion at high temperature, usually exceeding 400 °C.31 However, this method is not suitable for bonding two glass plates with different thicknesses. These glass plates show a difference in thermal expansion due to different rate of temperature change during heating and cooling. Furthermore, glass is softened at the temperature of fusion bonding. It can cause undesired closing of shallow regions in a semiclosed microstructure. Employing an intermediate gluing layer, e.g., adhesive polymer layer, is an alternative method of glass−glass bonding.32 However, the slight amount of polymer remaining on the glass induces background noise in sensitive optical measurements even after multiple washes. Microfluidic devices with a glass interface and a polymer substrate might be another option, although sensitive optical measurements are suffered by background noise from the polymer layer of these devices. In addition, deformation can easily occur with a polymer structure compared with glass. To overcome these technical difficulties, our glass bonding method at low temperature33 has been applied to fabricate the glass microfluidic device with a semiclosed microstructure. Using this glass microfluidic device, we have isolated a motile microalga, Euglena gracilis, to analyze its metabolic property via Raman spectroscopy. E. gracilis has gained attention among researchers due to its production of unique metabolites. In particular, E. gracilis accumulates β-1,3-glucan, known as paramylon, as a product of its photosynthetic fixation of carbon34−36 and converts paramylon into wax esters under anaerobic conditions.37−40 The wax esters can be refined into diesel oil which is suitable as a biofuel for aircraft and rocket engines. To increase the production efficiency of lipids from microalgae such as E. gracilis, it is essential to reveal the detailed metabolic process. In this study, we demonstrate sequential monitoring of paramylon biogenesis in a single identified E. gracilis cell via Raman spectroscopy and Raman imaging combined with a glass microfluidic device for single E. gracilis isolation and culturing.
cells isolated in certain spots when using motile microorganisms, and strong flow potentially stresses or damages cells. Microwell-array devices can isolate more than hundreds of single cells with a simple procedure to observe.19−21 In these devices, a simple procedure is typically used so that quick and high-throughput preparation of single cells can be carried out. For culturing isolated cells, some studies employed microwells connected to microchannels since isolated cells in microwells require fresh medium for culturing. These microstructures are effective for culturing and observing single cells.19,20,22,23 In contrast, highly motile cells, that move a hundred-fold faster than moving cells like leukocytes, can escape from designated microwells through channels so that time-course analysis cannot continue. Other devices have used shallow and wide channels with a dam structure to capture cells24,25 while the medium was circulated with weak flow to culture microorganisms successfully, but identification of individual cells was difficult in time-course analysis because continuous monitoring of highly motile cells is not practical for hours. When subcellular observation is performed under high magnification, it is particularly difficult to track highly motile cells moving randomly in one chamber. To overcome these challenges, we have designed a microfluidic device having parallel semiclosed microchannels (i.e., there is a gap between the microchannel walls and a lid) with dam structures for trapping and easily tracking infused single cells, which continuously supplies culture medium to the isolated cells with a simple feed pump without applying high pressure (Figure 1; see the Design of the Glass Microfluidic Device section for details). For time-course analysis of microorganisms, we have selected Raman spectroscopy as a label-free method to avoid any influence of chemical labeling on cellular metabolism. Although Raman spectroscopy has been applied to single-cell
Figure 1. Schematic illustration of the device. Arrows indicate direction of flow. (a) Top view of the device. (1) E. gracilis sample was introduced through the inlet port closest to the isolating channels. (2) The injected E. gracilis was directed into the isolating channels by medium flow. As shown in the cross section, E. gracilis was trapped at the dam of the isolating channels for spectroscopic observation while liquid passed over the dam. (3) Excess E. gracilis was flushed away by medium flow while opening the exit for excess cells. (b) Cross section of the device along the x-axis indicated by the red dashed line in panel a. (c) Cross section of the device along the y-axis indicated by the orange dashed line in panel a. B
DOI: 10.1021/acs.analchem.9b01007 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry
■
DESIGN AND THEORY Design of the Glass Microfluidic Device. We designed a glass microfluidic device to perform isolation, cultivation, and time-course observation of single E. gracilis cells (Figure 1). Parallel microchannels were in two (inlet and outlet) rows to isolate multiple E. gracilis cells simultaneously. At the end of the parallel microchannels, a dam structure was placed to keep isolated cells while culturing medium flows through from the inlet to the outlet. This structure provides a fresh environment for the isolated E. gracilis and allows easy observation of identified cells at the end of each microchannel. If the device is made of soft material such as PDMS, deformation of microchannels can occur, which leads easily to loss of isolated cells when pressure is applied by liquid delivery. For the parallel isolating channels, a semiclosed microstructure was used because channels of a closed structure are clogged by trapped E. gracilis that will disturb circulation of medium. By using the semiclosed channel with a gap between microchannels’ walls and a lid, medium liquid can freely go through the device without applying a high pressure even after capturing cells in each channel. For Raman spectroscopy, excitation light passes through the glass device and the emitted Raman signal is received at the other side of the device (Figure 1). Also, the microfluidic device composed of thin glass plates meets conditions for observation under a high-magnification objective lens while resisting deformation of the microchannels. In addition, the presented design of the device requires relatively low pressure for liquid delivery, while other designs, such as microwells with narrow side channels, require high pressure. To observe biogenesis of paramylon granules at the subcellular level, it is important to use a thin glass component with a semiclosed channel design that allows low pressure to circulate medium. Raman Spectroscopy. Raman spectroscopy is an effective tool for noninvasive analysis of cellular metabolism with direct detection of inherent molecular vibrations.41,42 Consequently, it can monitor time-course changes in biological activity of a specific single cell in the isolating channels through a series of measurements. Furthermore, Raman spectroscopy is compatible with a glass microfluidic device, ensuring low background noise. Raman signals move to lower wavenumber position when an atom in the molecule is replaced by a heavier isotope. The red shift is attributable to the fact that the Raman shift of a polyatomic molecule is inversely proportional to the square root of the atomic mass.43 We incorporated stable isotopes (2H, 13C) into E. gracilis cells through photosynthetic metabolism by replacing H2O and CO2, which are substrates of photosynthesis, with stableisotope-labeled 2H2O and 13CO2.44 Consequently, the inorganic substrates were converted to stable-isotope-labeled organic products including carotenoids and paramylon. By detecting the red shift of the Raman signals, the metabolic activity of the cells can be determined. First, we labeled carotenoids of E. gracilis with 2H by incubating in 2H2O culture medium because the cell consistently retains carotenoid under a normal culture condition, and then probed the red shift of the 2H-carotenoids by resonance Raman spectroscopy. Previously, we investigated the C−2H bond formation attributed to 2H-glucose and its metabolites by spontaneous Raman spectroscopy (Figure S1).44 However, it was limited to an invasive method because
long measurement time was needed to obtain a relatively weak nonresonance Raman signal and quenching of a strong autofluorescence background was imperative. On the other hand, Raman signals of carotenoids are significantly enhanced when the wavelength of the excitation light corresponds to that of the electron transition of carotenoids (resonance Raman scattering).45 We selectively highlighted and noninvasively analyzed Raman signals of the carotenoids over other signals of cellular constituents with appropriate excitation light. Second, we labeled paramylon of E. gracilis with 13C by incubating with culture medium that contains NaH13CO3 as the 13CO2 source. Under a paramylon induction condition, the cell significantly accumulates paramylon granules, which should be monitored with high-resolution Raman imaging. Paramylon is a suitable indicator for photosynthetic activity reflecting prompt incorporation of 13C since it is a direct product of photosynthetic carbon fixation. On the basis of the red shift, we detected and visualized 13C-paramylon with a stimulated Raman scattering (SRS) microscope, which has high spatiotemporal resolution.46
■
EXPERIMENTAL SECTION Chemicals and Materials. Deionized water was generated with the Milli-Q Integral 15 (Merck Millipore, Billerica, MA, U.S.A.). The following reagent grade chemicals were purchased from Wako Pure Chemical (Osaka, Japan): acetone, isopropyl alcohol, ethanol, and HCl (35 wt %). Photoresist (OFPR-800) was purchased from Tokyo Ohka Kogyo (Tokyo, Japan). Hydrofluoric acid (HF, 49 wt %) was purchased from Morita Chemical Industries (Osaka, Japan). For glass etching, HF and HCl were diluted by deionized water to obtain a mixture of 10 wt % HF and 0.7 wt % HCl. Borosilicate glass plates of 70 mm × 30 mm with 0.2 or 0.7 mm thickness were purchased from Matsunami Glass (Osaka, Japan). Fabrication of Glass Microfluidic Devices. The glass microfluidic devices were fabricated by the previously described method33 with modifications. In short, the glass plates were cleaned by H2SO4/H2O2 solution for 15 min, and subsequently, Cr and Au layers were deposited on the cleaned glass surface by sputtering (EIS-220, Elionix, Tokyo, Japan). Photoresist was spin-coated on the metal layers and exposed to UV light for 14.5 s to develop the microchannel pattern. A coated glass plate had the pattern of a deep microchannel and another plate had the pattern of a shallow microchannel (Figure 1 and Figure S2). Through Cr, Au, and glass etching, the microchannels were patterned onto the glass surface. Then, inlet and outlet ports were made on the glass plate with the shallow microchannels by drilling. The etched glass plates were then treated with H2SO4/H2O2 solution to activate the glass surface. These activated glass plates were aligned manually for prebonding. For permanent bonding of the glass, a weight of 1.2 kg was placed on the two aligned 0.7 mm glass plates which were heated in a furnace (KDF-900GL, Denken, Oita, Japan) at 620 °C, or a force of 450 N was applied to the aligned 0.7 mm plate and 0.2 mm plate at 250 °C. After glass bonding, PDMS ports were attached on the device via surface activation by oxygen plasma, followed by 120 °C baking for 1 h. Simulation of Device Deformation. Single-liquid-phase fluidic simulation with fluidic−solid interaction was conducted by using the VOF (volume of fluid) laminar flow model to compare the distribution of layer deformation. Simulation with the finite volume method was implemented by computational fluid dynamics software (COMSOL 5.4a, COMSOL Inc., Los C
DOI: 10.1021/acs.analchem.9b01007 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry Angeles, CA, U.S.A.). Furthermore, for convenience of calculations, we used simple conditions of simulation in which there was a fixed bottom layer and a wide inlet channel (0.01 mm height). The conditions of inlet and outlet were set as a mass flow inlet and outlet. The flow condition was set to 5 μL/min. Preparation of Isolating Channels in Glass Microfluidics. The fabricated glass microfluidic device was connected to the liquid delivery system and placed under a microscope (Figure 2a). To use all isolating channels, the glass device was flushed with water at a flow rate of 50 μL/min to remove air bubbles. Appropriate conditions of water flushing were investigated by using microfluidic devices with the gap of 0.5−1.5 μm. The microfluidic devices were flushed by water for 10 min at 50 μL/min controlled by a syringe pump (Harvard Apparatus, Massachusetts, U.S.A.). After flushing, the number of isolating channels filled with water was counted. Also, the change in number of isolating channels filled with water was tested with a device with the gap of 0.5 μm. The device was flushed by water at 50 μL/min for 60 min. Every 10 min, the number of isolating channels filled with water was counted. Once air bubbles in the device were removed completely, water was replaced by culture medium prior to isolation of E. gracilis. To theoretically investigate the flow in the isolating channels, single-liquid-phase fluidic simulation was conducted by using the VOF laminar flow model. A pair of inlet and outlet isolating channels with the dam structure of the glass microfluidic device were modeled to figure out the flow pattern and the velocity distribution. Simulation with the finite volume method was implemented by computational fluid dynamics software (Fluent, Ansys 19.0, Ansys Inc., Canonsburg, PA, U.S.A.). Furthermore, for convenience of calculations, we used simple simulation conditions in which a fixed channel and two fixed inlets were employed, and other influences were ignored. The inlet and outlet were set as a mass flow inlet and pressure outlet, respectively. The flow condition was set to 0.0079 μL/ min, corresponding to (1 μL/min)/(126 channels), to calculate the streamlines from an inlet isolating channel to an outlet isolating channel. Algal Material and Culture Conditions. E. gracilis NIES48 was provided by the Microbial Culture Collection at the National Institute for Environmental Studies (NIES, Tsukuba, Ibaraki, Japan). The autotrophic medium AF-6 was used for culture (working volume: 20 mL). Cells were grown in culture flasks (polystyrene suspension culture flask with filler cap, 50 mL volume, Greiner Bio-One) in static conditions under 14 h light to 10 h dark cycle illumination (approximately 100 μmol/ m2 s−1) at 28 °C. Before culturing with stable isotope medium, E. gracilis cells were grown in normal AF-6 medium at pH 6.6 for at least 3 days as a preculture. During the exponential growth phase, cells in the preculture were transferred to 2H2O AF-6 medium (30 vol %; prepared with deuterium oxide, 99.9%, Wako Pure Chemical, Osaka, Japan) for production of 2 H-carotenoids or nitrogen-deficient AF-6 medium (without NH4NO3 and NaNO3) including 20 mM of 13C-sodium bicarbonate (13C, 99%; Cambridge Isotope Laboratories, Inc., Tewksbury, MA, U.S.A.) for induction of 13C-paramylon. E. gracilis cells in medium were manually introduced into the glass microfluidic device from the injection port with a 1 mL syringe (Terumo, Tokyo, Japan), and then the cells flowed into the isolating channels with the corresponding medium (5 μL/ min). The number of occupied channels was controlled to
Figure 2. Setup of the instrument. (a) The whole instrument including glass microfluidic device, liquid delivery system, and Raman spectroscope. Yellow box: the glass microfluidic device disconnected from the liquid delivery system. Arrows indicate flow direction. (b) The ratio of isolating channels filled with water after water flushing for 10 min at 50 μL/min. Channels filled with water were readily used to trap E. gracilis. Microfluidic devices composed of two 0.7 mm plates were tested (n = 3), excluding the gap of 1.1 μm, and microfluidic devices composed of a 0.7 mm plate and a 0.2 mm plate were tested (n = 4) for the gap of 1.1 μm. (c) Change in the ratio of isolating channels filled with water by continuous water flushing at 50 μL/min. The data was obtained from a microfluidic device composed of two 0.7 mm plates with the gap of 0.5 μm (n = 3). D
DOI: 10.1021/acs.analchem.9b01007 Anal. Chem. XXXX, XXX, XXX−XXX
Analytical Chemistry
■
approximately 25% of the total number of channels (approximately 30 out of 126 channels). The excess cells were removed by flushing with the culture medium (10 μL/ min) with the exit port opened. Isolated E. gracilis cells were incubated in the channels with culture medium flow (1 μL/ min) and light illumination (approximately 100 μmol/m2 s−1) at room temperature. As a control sample, E. gracilis cells were fixed with 0.25% glutaraldehyde solution and introduced into the glass microfluidic device in the same manner. Extraction of 13C- and 12C-Labeled Paramylons from E. gracilis Cells. 13C- and 12C-paramylon were extracted from E. gracilis according to a previously reported procedure.47 Cells were cultured in AF-6 media until stationary phase. The media was then replaced with nitrogen-deficient AF-6 medium (without NH4NO3 and NaNO3) including 20 mM of 13Csodium bicarbonate (13C, 99%; Cambridge Isotope Laboratories, Tewksbury, MA, U.S.A.) or 12C-sodium bicarbonate (Nacalai Tesque, Kyoto, Japan) for induction of paramylons. The cells were cultured for 5 days. An aliquot of 1.5 × 107 cells was collected and resuspended in 1.8 mL of deionized water, and then mixed with 0.2 mL of an ice-cold 30% perchloric acid. The mixture was vortexed for 1 min before centrifugation at 1100g for 2 min. The supernatant was discarded, 2 mL of 1% sodium dodecyl sulfate was added to the pellet, and the mixture was vortexed for 1 min before being heated in boiling water for 15 min. The sample was then centrifuged at 1100g for 5 min, the supernatant was discarded, and these steps were repeated twice. The resulting pellet was freeze-dried and analyzed by SRS microscopy. Resonance Raman Spectroscopy of Deuterated E. gracilis Cells. Resonance Raman spectra were acquired using a confocal Raman microscope (inVia, Renishaw, Wottonunder-Edge, U.K.) equipped with an integrated Leica microscope. A 50× dry objective was used to acquire Raman signals from single E. gracilis cells. Laser light was targeted on the cell using an integrated color camera and a motorized XYZ stage. Raman scattering was excited with a 532 nm laser (JUNO 532, 150 mW, Showa Optronics Co., Tokyo, Japan) with a 5% laser filter unless otherwise stated. Each resonance Raman spectrum was acquired in the range between 440 and 2150 cm−1 with an exposure time of 1 s. The system was run with a diffraction grating having a groove density of 1800 lines/mm (vis). The spectral resolution and the stability with the setup above are 5.6 and ±0.1 cm−1, respectively. Stimulated Raman Scattering Imaging of 13C-Labeled E. gracilis Cells. SRS imaging of isolated E. gracilis cells was conducted in a similar manner as reported previously.46,48,49 We modified some procedures as follows. To obtain Raman spectral bases of three constituents (12C-paramylon, 13Cparamylon, and chlorophyll) in E. gracilis cells, we used our SRS microscope with seven spectral points (2860, 2880, 2896, 2910, 2925, 2937, and 3050 cm−1) (Figure S3 and Figure 5c). We used the three spectral bases for pseudoinverse matrix calculation and obtained the spatial distributions of the three constituents from each SRS image of E. gracilis. The number of pixels of the Raman image is 500 × 500 pixels for the field of view of 80 μm × 80 μm at a mapping speed of 30 frames/s (0.033 s per image; 0.23 s in total to obtain a data set of seven images at the seven spectral points with one accumulation, respectively). For enlarged views, the obtained images were cropped to show in 160 × 160 pixels (26 μm × 26 μm).
Article
RESULTS AND DISCUSSION
Flow in Glass Microfluidic Devices. To simplify preparation of single cells with an easy instrumentation, it is desired that microfluidic devices have a minimal number of accompanying instruments. Thus, the present microfluidic devices were designed to use only a liquid delivery system to obtain isolated single cells. To obtain the maximum number of isolated single cells, the appropriate condition for water flushing was investigated to make all isolating channels available. Since circulation of liquid in the microfluidic devices was influenced by the gap at the dam structure, the number of isolating channels filled with water was counted after water flushing (50 μL/min) was applied for 10 min to the microfluidic devices with various gaps. The results of the gap versus number of water-filled isolating channels are shown in Figure 2b. Although a smaller gap required a longer time to remove air bubbles, the device with the 0.5 μm gap needed only 1 h to eliminate all bubbles in the isolating channels (Figure 2c). Thus, 2 h of water flushing were employed for all experiments to remove air bubbles certainly. When water was introduced to the device in a dry condition, water entered the middle of the inlet and outlet rows of the isolating channels first, and then the edges of the rows (Figure S4). Therefore, the middle of the inlet row was mainly monitored during isolation of cells. Conditions for Isolating Single E. gracilis Cells. A fabricated glass microfluidic device was connected to the liquid delivery system to check the capability for single E. gracilis cell isolation. E. gracilis cells are typically 30−100 μm long and 5− 10 μm wide. To trap E. gracilis at the end of the isolating channels, it was important to set an appropriate gap above the dam structure. When the gap was too small, air bubbles persistently remained in isolating channels to reduce the number of available channels for isolation. On the other hand, when the gap was too large, E. gracilis could escape from the end by swimming across the gap spontaneously or being flushed away by flow. To use all isolating channels without losing E. gracilis, we employed the dam structure that had a 1 μm depth gap above it that was 50 μm long. Each isolating channel was 10 μm deep, 19 μm wide, and 3 mm long (Figure S2). Flow rates of E. gracilis introduction and circulation of culture medium were also important to minimize stress on E. gracilis and avoid escape of E. gracilis from the isolating channels. When E. gracilis cells were introduced, flushing by culture medium delivered the cells to the end of the isolating channels. In our setup, the flow rate of 5 μL/min was the maximum flow rate that forced E. gracilis cells to stay at the end of the isolating channels (Figure 3a) although this flow rate was too fast to culture E. gracilis for hours. When flow rate was 5 μL/min or above, the cells responded by rounding their body, most likely to reduce the stress from the pressure of the fast flow. When excessive flow was applied, cells were forced to slide into a small gap or were even damaged (Figure 3a). In addition, using the fast flow rate made it difficult to control introduction of E. gracilis; it could easily result in introduction of multiple E. gracilis cells per isolation channel. Besides the appropriate flow rate for E. gracilis introduction, an appropriate flow rate of circulating culture medium was also investigated. On the basis of the behavior of E. gracilis, a flow rate of 1 μL/min was the minimum to avoid the escape of cells from isolating channels (Figure 3b). At this flow rate, cells E
DOI: 10.1021/acs.analchem.9b01007 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry
caused by trapped cells. A previous study indicated that cells were alive only in a narrow window of flow rates and the reported flow rate was too slow (0.2 μL/min) to hold E. gracilis in isolating channels.50 In contrast, semiclosed channels are not clogged by single cells. They provide uniform flow during medium circulation because flow can occur above the dam and the walls of the isolating channels (Figure S5). For time-course analysis of identified cells, multiple cells should not occupy an isolating channel if changes in single cells are to be clearly discussed. When E. gracilis introduction was performed at the flow rate of 5 μL/min, we saw that 30 ± 5 out of 126 isolating channels, corresponding to 19.8−27.8%, were typically occupied by single E. gracilis cells and the rest of the channels remained unoccupied. This percentage of successful single-cell isolations was higher than a previous value (10%) found for Escherichia coli, a microorganism with slower motility.51 Other studies for isolation of highly motile single cell52−54 could perform trapping one cell per operation. Thus, for preparation of highly motile single cells, our device achieved higher throughput than previous studies. Further introduction of the cells reduced the number of unoccupied channels while it increased the number of channels occupied by multiple cells simultaneously. Thus, we aimed at introducing single E. gracilis cells into 30 ± 5 channels before switching the flow rate from the sample introduction rate (5 μL/min) to the culturing rate (1 μL/min). Excess cells that remained in the microfluidic device were removed by liquid flushing through the detour (Figure S7) before culturing started. During the removal of excess cells, all of the outlet ports were closed so that excess cells were directed to flow the detour. This removal process was adequately short; hence, isolated cells remained in the isolating channels. Resonance Raman Spectroscopy of Isolated Single E. gracilis Cells. We evaluated the performance of our microfluidic device for long-term culturing of E. gracilis cells without interfering with the fundamental metabolism. After isolation of E. gracilis cells in the channels, they were cultured for 22 h with a continuous supply of culture medium containing 30% 2H2O (1 μL/min) with light irradiation (approximately 100 μmol/m2 s−1). By conducting resonance Raman spectroscopy of the E. gracilis cells with an excitation wavelength of 532 nm, we observed typical resonance Raman spectra of carotenoids (mainly β-carotene) instead of nonresonant Raman spectra (Figure S8).45 Red shift of the resonance Raman signals was monitored as an indicator that the normal metabolism including photosynthesis was maintained. After 22 h of incubation, the ν1 peak of carotenoids (1522 cm−1) shifted to lower wavenumber position (Figure 4a, red) compared to the spectra after 1 h of incubation (Figure 4a, black) and 6 h of incubation (Figure 4a, green). This result showed that the substituted protons of the CC double bond in a carotenoid molecule were replaced with deuterium atoms. The spectrum after 22 h of incubation shows broadening, which can be attributed to distribution of deuteration degrees in carotenoid molecules even in a single cell. As a control experiment, E. gracilis cells whose biological activity was stopped by glutaraldehyde fixation were introduced into the channel and incubated under the same condition. Since the fixed cells could not swim, they were pushed into the gap more than the active cells by long-term continuous flow of medium (Figure 4b). After 22 h of incubation, intensities of the resonance Raman signals were significantly decreased. This result was attributed to deficiency in recovery of carotenoids
Figure 3. Trapping E. gracilis in isolating channels by medium flow. (a) Introduction of E. gracilis. At a flow rate of 5 μL/min, all E. gracilis cells settled at the end of the isolating channels and tended to round their bodies. At a flow rate of 50 μL/min, some E. gracilis cells were forced to slide into a small gap by the flow. Scale bar: 20 μm. (b) Difference in E. gracilis culturing. Left: isolation of E. gracilis cells either at the end or on the wall of isolating channels at flow rate of 1 μL/min. Right: the same E. gracilis flowed to the end of the isolating channels at a flow rate of 5 μL/min. Scale bar: 100 μm. (c) Simulated velocity distribution of flow in the cross section of an isolating channel. Flow rate was set at 0.079 μL/min, corresponding to (1 μL/ min)/126 channels. Scale bar: 10 μm. (d) Isolated single E. gracilis cells at a flow rate of 1 μL/min indicated by red circles.
could stick to the wall of the channels and could slowly crawl against the flow, but they could not swim against the flow and were eventually forced to move back to the wall. This could happen because flow near the wall was much slower than at the middle of a channel as was simulated (Figure 3c; flow velocities in different sections of an isolating channel are shown in Figure S5). Flow rate smaller than 1 μL/min allowed the escape of E. gracilis from isolating channels by crawling on the channel wall or swimming against the medium flow. Thus, we selected the flow rate of 5 μL/min for E. gracilis introduction and 1 μL/min for E. gracilis culturing for further experiments to keep the isolated single E. gracilis cells with the minimum stress (Figure 3d). Simulation showed that the deformation of the glass device caused by a flow rate of 5 μL/ min was approximately 2.5 μm in which E. gracilis cells cannot escape from isolating channels (Figure S6). Meanwhile, if the top layer was made of PDMS instead of glass, deformation of the device was approximately 25 μm, and E. gracilis could escape easily from isolating channels. In addition, a semiclosed microstructure, i.e., walls of isolating channels had 1 μm gap above it, played an important role for circulation of culture medium. If isolating channels are a closed microstructure, i.e., walls of isolating channels are fused to the lid, trapped cells clog isolating channels. Although the closed microstructure has an advantage in single-cell isolation because the flow is directed to empty chambers rather than chambers occupied by trapped cells,17,18 it prevents circulation of culturing medium due to nonuniform flow F
DOI: 10.1021/acs.analchem.9b01007 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry
tion. The peak shifts of the active cell samples distributed widely, probably caused by individual difference in cellular metabolic activity of photosynthesis and subsequent carotenogenesis. Since long-term incubation was difficult to perform in a microwell-array device (Figure S9) or other previously reported devices,17,18,22−25 these results demonstrated the ability of our microfluidic device to culture E. gracilis cells under the isolated condition while maintaining the cell metabolic activity. SRS Imaging for Paramylon Metabolism of E. gracilis. Next, we aimed at monitoring the paramylon accumulation process of E. gracilis in our microfluidic device under the condition of paramylon induction. First, we confirmed the significant spectral shift of 13C-paramylon by SRS microscopy using authentic 13C- and 12C-paramylon samples (Figure 5,
Figure 5. SRS images of an identified single E. gracilis cell in an isolating channel. The images were acquired after incubation with paramylon-induction medium containing 20 mM NaH13CO3 for (a) 0.5 h and (b) 22 h. Flow rate of the medium was 1 μL/min. (c) Raman spectral bases of three constituents (12C-paramylon, 13Cparamylon, and chlorophyll) for analyzing SRS images of E. gracilis. The spectra were measured at seven spectral points (2860, 2880, 2896, 2910, 2925, 2937, and 3050 cm−1). (d) An SRS image of extracted paramylons (mixture of 13C- and 12C-paramylons). Pseudocolors of SRS images are assigned as follows: red, 13C-paramylon; green, 12C-paramylon; blue, chlorophyll. Scale bars: 10 μm.
Figure 4. Resonance Raman spectra obtained from individually isolated single E. gracilis cells. (a) Active and (b) inactive E. gracilis cells cultured in isolating channels provided with culture medium containing 30% 2H2O. (c) Time-course changes of the ν1 peak tops of active (red) and inactive (blue) cells (active cells, n = 7; inactive cells, n = 3). Each ν1 peak was fitted by a Gaussian function.
that had been disrupted by photobleaching. This was in contrast with the result of the living cells which still showed strong resonance Raman signals after 22 h of incubation. The resonance Raman spectra of the control cells were measured with increasing laser intensity up to 20 times. The obtained ν1 peak did not show any red shift indicating that 2H-labeling of carotenoids did not occur (Figure 4b). To compare the red shifts of active and inactive cell samples quantitatively, Raman shifts of the ν1 peak tops were obtained by fitting with Gaussian functions and the values were plotted as a function of the incubation time, respectively (Figure 4c). As a result, the active cell samples showed lager red shift in 3 cm−1 after 22 h of incubation than that of the inactive cell sample (
