On-Chip Cell Sorting System Using Laser-Induced Heating of a

A high-speed camera (Motion Pro model 1000 Mono, Redlake MASD Inc., San Diego, CA) ... the software alters the laser position to target the GFP in the...
0 downloads 0 Views 519KB Size
Anal. Chem. 2006, 78, 695-701

On-Chip Cell Sorting System Using Laser-Induced Heating of a Thermoreversible Gelation Polymer to Control Flow Yoshitaka Shirasaki,† Jyunichi Tanaka,‡ Hiroshi Makazu,† Koichiro Tashiro,‡ Shuichi Shoji,‡ Shoichiro Tsukita,§ and Takashi Funatsu*,†,|,⊥ Department of Physics and Department of Electrical Engineering and Bioscience, School of Science and Engineering, Waseda University, 3-4-1 Ohkubo, Shinjuku-ku, Tokyo 169-8555, Japan, Department of Cell Biology, Faculty of Medicine, Kyoto University, Sakyo-ku, 606-8501, Japan, Laboratory of Bio-Analytical Chemistry, Graduate School of Pharmaceutical Sciences, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-0033, Japan, and Core Research for Evolutional Science and Technology, Japan Science and Technology Agency, Saitama 332-0012, Japan

We have developed a microfabricated fluorescenceactivated cell sorter system using a thermoreversible gelation polymer (TGP) as a switching valve. The glass sorter chip has Y-shaped microchannels with one inlet and two outlets. A biological specimen containing fluorescently labeled cells is mixed with a solution containing a thermoreversible sol-gel polymer. The mixed solution is then introduced into the sorter chip through the inlet. The sol-gel transformation was locally induced by sitedirected infrared laser irradiation to plug one of the outlets. The fluorescently labeled target cells were detected with sensitive fluorescence microscopy. In the absence of a fluorescence signal, the collection channel is plugged through laser irradiation of the TGP and the specimens are directed to the waste channel. Upon detection of a fluorescence signal from the target cells, the laser beam is then used to plug the waste channel, allowing the fluorescent cells to be channeled into the collection reservoir. The response time of the sol-gel transformation was 3 ms, and a flow switching time of 120 ms was achieved. Using this system, we have demonstrated the sorting of fluorescent microspheres and Escherichia coli cells expressing fluorescent proteins. These cells were found to be viable after extraction from the sorting system, indicating no damage to the cells. The technique of sorting specific cells is indispensable in the studies of biology and medical diagnosis. Although conventional fluorescence-activated cell sorters are widely used due to their high performance in cell screening, these devices have certain drawbacks.1 First, these devices are rather expensive, are often difficult to sterilize, and require relatively large sample volumes. Second, there is the issue of cell viability, which is often decreased by exposure to high electric fields. Micro total analysis systems * To whom correspondence should be addressed. Phone: +81-3-5841-4760. Fax: +81-3-5802-3339. E-mail: [email protected]. † Department of Physics, Waseda University. ‡ Department of Electrical Engineering and Bioscience, Waseda University. § Kyoto University. | The University of Tokyo. ⊥ Japan Science and Technology Agency. (1) Shapiro, H. M. Practical flow cytometry, 3rd ed.; Wiley-Liss: New York, 2003. 10.1021/ac0511041 CCC: $33.50 Published on Web 12/16/2005

© 2006 American Chemical Society

attract particular attention since they enable the integration of multiple steps, for example, mixing, reacting, analyzing, separating, and recovering of samples, all together on the one device. Microfabricated fluorescence-activated cell sorters (µFACS) use several methods to control the flow or movement of cells. The separation of biological cells by µFACS initially used electroosmotic or electrophoretic pumping to drive the cell transport within a network of capillary channels.2-5 However, cell viability was difficult to maintain due to the high electric field. In response, hydrodynamic flow-controlled on-chip fluidic valves were developed for sorting living cells. These achieved better recovery of viable cells, albeit with lower throughput.6-8 Recently, a number of optical-based cell sorters have been reported by several groups. MacDonald et al. demonstrated an optical sorter with a threedimensional optical lattice.9 Wang et al. proposed an optical force switching mechanism to sort cells passing or flowing through a fluidic channel.10 Although this last method achieved high throughput, it suffers from dilution of the sample. To circumvent these issues, we developed an on-chip cell sorting system with valves that rely on the sol-gel transformation of a thermoreversible gelation polymer (TGP). A solution containing cells and TGP was introduced into the Y-shaped microchannel of the sorting device via an inlet. The sol-gel transformation is locally induced by site-directed infrared laser irradiation at the junction of the two outlets, essentially plugging one of the two microchannel outlets, as confirmed by the absence or presence of a fluorescence signal detected upstream of the junction. In this manner, the gel functions as a microvalve to switch the microflow. A sensitive fluorescence detector was installed into an epifluorescence microscope to detect the fluorescence signals from individual target cells. These improvements have realized the sorting of (2) Li, P. C.; Harrison, D. J. Anal. Chem. 1997, 69, 1564-1568. (3) Fiedler, S.; Shirley, S. G.; Schnelle, T.; Fuhr, G. Anal. Chem. 1998, 70, 1909-1915. (4) Fu, A. Y.; et al. Nat. Biotechnol. 1999, 17, 1109-1111. (5) Dittrich, P. S.; Schwille, P. Anal. Chem. 2003, 75, 5767-5774. (6) Fu, A. Y.; Chou, H.-P.; Spence, C.; Arnold, F. H.; Quake, S. R. Anal. Chem. 2002, 74, 2451-2457. (7) Kru ¨ ger, J.; et al. J. Micromech. Microeng. 2002, 12, 486-494. (8) Wolff, A.; et al. Lab. Chip 2003, 3, 22-27. (9) MacDonald, M. P.; Spalding, G. C.; Dholakia, K. Nature 2003, 426, 421424. (10) Wang, M. M.; et al. Nat. Biotechnol. 2005, 23, 83-87.

Analytical Chemistry, Vol. 78, No. 3, February 1, 2006 695

individual Escherichia coli cells labeled with fluorescent proteins, with switching times of 120 ms. EXPERIMENTAL SECTION Samples. Fluorescent microparticles (carboxylate-modified microspheres, 1.0 and 0.2 µm, red fluorescence (λex 580/λem 605)) were purchased from Molecular Probes (Eugene, OR). A plasmid containing GFP construction (pEGFP, Clontech, CA) or DsRed (pDsRed, Clontech) was transfected into E. coli BL21 (DE3), (Invitrogen). Transfected E. coli was grown at 37 °C in LB medium (5 g/L yeast extract, 10 g/L bactopeptone, 5 g/L NaCl) containing 100 mg/L ampicillin. The growth of E. coli was monitored with a spectrophotometer (V-570, Jasco) by measuring the amount of light (600 nm) scattered by the culture. When the level of absorbance at 600 nm reached 0.6 OD unit, isopropyl β-Dthiogalactopyranoside was added to a final concentration of 0.3 mM and the resultant mixture incubated for 3 h at 37 °C. Additional incubation for 48 h at 27 °C was required for E. coli expressing DsRed. The culture medium containing cells was centrifuged, and fluorescent E. coli was harvested. Thermoreversible Gelation Polymer. Mebiol Gel11 is a block copolymer of poly(N-isopropylacrylamide-co-n-butyl methacrylate) and poly(ethylene glycol). Poly(N-isopropylacrylamide-co-n-butyl methacrylate) is a statistical copolymer of N-isopropylacrylamide and n-butyl methacrylate. The ratio of N-isopropylacrylamide and n-butyl methacrylate changes depending on the critical temperature. Mebiol Gel is an inert biocompatible polymer with a critical solution temperature of 36 °C that is used for culturing mammalian cells (Mebiol Inc., Tokyo, Japan).12 Mebiol Gel was dissolved in phosphate-buffered saline (PBS; T900, Takara Bio Inc., Otsu, Japan) to a final concentration of 15% (w/w). The biological specimen was mixed with the Mebiol Gel solution (final concentration 10%), so that the viscosity of the solution is ∼140 times higher than that of pure water at 23 °C. Care should be taken not to make bubbles, since they sometimes destabilize or block the flow in a microchannel. Fabrication of the Microchannels. Microchips (as shown in Figure 1) were fabricated using the bonded glass technique. Y-Shaped microchannels (30 µm in width, 5 µm in depth) with one inlet and two outlets were etched in the bottom of a 700-µmthick Pyrex plate using standard wet etching processes. Holes (500-µm diameter) for introducing or collecting the biological samples were then drilled through the Pyrex plate at the positions corresponding to the three ends of a Y-shaped microchannel. The bottom of the plate was shielded with a Pyrex glass cover of 110µm thickness to allow microscopic observation and detection of the cells inside of the channels and reservoirs. The tube connectors were attached to the inlet and outlet holes in the top of the Pyrex plate. The inlet was then connected to a syringe pump (CMA/102, CMA Microdialysis AB) via polyethylene tubing, while the tube connectors of waste and collection outlets remained open and they were used as cell reservoirs. Optical Setup. A schematic drawing of the apparatus is shown in Figure 2. An inverted optical microscope (IX70, Olympus Optical, Tokyo, Japan) was modified to install an infrared laser (11) Yoshioka, H.; et al. J. Macromol. Sci.-Pure Appl. Chem. 1994, A31, 113120. (12) Hishikawa, K.; et al. Biochem. Biophys. Res. Commun. 2004, 317, 11031107.

696 Analytical Chemistry, Vol. 78, No. 3, February 1, 2006

Figure 1. Cell sorter chip. (a) Layouts and dimensions of the sorting device. Dimensions are given in micrometers. (b) An optical micrograph of a microchip fabricated with two microchannels. (c) An optical micrograph revealing the branch point within the microchannel.

Figure 2. Schematic diagram of the cell sorting apparatus. A cell sorter chip was placed on a fluorescence microscope equipped with an infrared laser system operating at 1480 nm. Biological specimens were introduced into the microchannel using a syringe pump. Fluorescence images were split by a half-mirror and projected onto both a CCD camera and a PMT, simultaneously. Based on the detection of a fluorescence signal, the position of the infrared laser beam could be switched, enabling the position of laser illumination to be switched. Mirror 1 is a resonant scanner used to project laser light across the width of the microchannel, and mirror 2 is a scanner mirror that changes the center of the laser illumination window.

(FRL-DC 1,480 nm 3 W, IPF Technology) for heating. Briefly, a second dichroic mirror used to reflect the infrared laser beam was placed in the optical path under the conventional dichroic mirror for epifluorescence microscopy. The optical coating of the windows was modified to transmit the infrared beam more efficiently. Special precautions were required to avoid exposure

to the intense infrared laser irradiation. Biological specimens labeled with GFP and DsRed were observed by epifluorescence microscopy using filter sets of U-MGFPHQ and U-MWG2 (Olympus Optical), respectively. The GFP and DsRed were excited using 488 (Sapphire 488-20, Coherent, Inc.) and 532 nm (Compass 315M-20, Coherent, Inc.) wavelengths, respectively. A UApo /340 20x (Olympus Optical) objective lens was used for the purpose of transmitting infrared light and also used for observing the corresponding fluorescence. The fluorescence from the specimen was split into two paths by a half-mirror. One path was used for imaging the microchip by a CCD video camera (MC681SPD, Texas Instruments Inc.) coupled to an image intensifier (C901601, Hamamatsu Photonics, Hamamatsu, Japan). A high-speed camera (Motion Pro model 1000 Mono, Redlake MASD Inc., San Diego, CA) was used instead of the standard video camera only for evaluating the response speed of the flow switching. The other path was used for detecting the fluorescence signal in the microchannels, amplified with the aid of a photomultiplier tube (PMT; H7421-40, Hamamatsu Photonics). The detection window (40 × 40 µm in dimension) was positioned 60 µm upstream of the junction. The window was implemented with a rectangular aperture used to reduce the background fluorescence emitted from the solution and the glass. The position of the infrared laser beam was deflected by way of two mirrors placed around the position conjugate to the back aperture of the objective lens. The scanner mirror 1 in Figure 2 was a resonant scanner mirror (CRS 4 kHz, GSI Lumonics Japan, Tokyo, Japan) that was used to give a fluctuation of laser light across the width of the microchannel to block the flow completely. The second scanner mirror 2 (VM2000, GSI Lumonics Japan) in Figure 2 is used to switch the beam position between the collection and waste channels. The fluorescence signals emitted from the target cells and detected by the PMT were converted into electric pulses, which were then counted every 5 ms (PHC-2500, Scientex, Hamakita, Japan) and entered automatically into a personal computer via an AD converter (ADM-687APCI, MicroScience, Tokyo, Japan). A Visual Basic program was written to read the signals from the PMT and to digitally control scanner mirror 2. When the fluorescence signals exceeded a predefined threshold, the program transmitted a signal via a DA converter (DA16P4M3-7, Interface, Tokyo, Japan) to deflect scanner mirror 2 during a predefined time to collect the target cells. Procedures. A suspension of microspheres or E. coli cells was mixed with the Mebiol Gel solution (final concentration 10%) and introduced into the microchannel using a syringe pump via polyethylene tubing (Figure 2). The syringe pump controlled the flow rate of the solution of cells. Figure 3 shows the principle of the cell sorting strategy based on TGP. The fluorescence emitted from the biological specimen is detected upstream of the branch point. In the absence of a fluorescence signal from the target particles in the detection area, the infrared laser locally induces a sol-gel transformation in TGP plugging the collection channel, thereby directing the nonfluorescent specimens to the waste channel. Upon detection of a fluorescence signal from the target particles, the software alters the laser position to target the GFP in the waste channel, effectively plugging the waste channel and redirecting the fluorescent samples to the collection channel. After the target particle is accommodated in the collection channel, the

Figure 3. Principle of the cell sorting method. (a) The window for detecting the fluorescence signal is located upstream of the junction. In the absence of a fluorescence signal, the collection channel continues to be plugged and flow directed to the waste channel. (b, c) Upon detection of a fluorescence signal, the entrance to the waste channel is plugged by switching the position of laser illumination, directing flow to the collection channel.

Figure 4. Distribution of temperature elevation in a microchannel. A PBS solution containing 1.6 µM TMR, 10% Mebiol Gel was introduced into the microchannel. Fluorescence images before and during the infrared illumination were taken, and the temperature elevation at each image pixel was obtained from the relative fluorescence intensity. This figure shows the distribution of temperature elevation when the entrance of the lower outlet is illuminated. The temperature of the glass was 18 °C.

infrared laser is switched back to the default position to plug the collection channel. Measurement of Local Temperature in a Microfluidic Channel. The temperature distribution in a microchannel was determined by the fluorescence intensity distribution of 5-carboxytetramethylrhodamine succinimidyl ester (TMR) (Molecular Probes) in solution.13 The fluorescence intensity of PBS containing 1.6 µM TMR was measured at various temperatures using a fluorometer (FP-6500, Jasco), and the dependency of the fluorescence intensity on temperature was determined (data not shown). First, PBS containing 1.6 µM TMR, 10% Mebiol Gels was introduced into the microchannel. The fluorescence images before and during the infrared laser illumination were taken with a CCD video camera (MC681SPD, Texas Instruments Inc.) and recorded on videotape for subsequent analysis. The average intensities of four video frames were measured using Scion Image (Scion Corp.). The distribution of temperature elevation at each image pixel was obtained from the relative fluorescence intensity, as shown in Figure 4. RESULTS AND DISCUSSION Local Heating by Infrared Laser. To precisely control the flow in the microchannel using temperature-sensitive sol-gel (13) Kato, H.; et al. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 9602-9606.

Analytical Chemistry, Vol. 78, No. 3, February 1, 2006

697

transformation requires localized heating with a micrometer spatial resolution. Here, an infrared laser of wavelength 1480 nm was used, since it corresponds to the resonant frequency associated with the stretching and bending vibrations of water molecules, therefore enabling efficient heating of the surrounding water. The temperature distribution in the microchannel was measured using fluorescence quenching upon heating.13 Here, a solution containing TMR and Mebiol Gel was introduced into the Y-shaped microchannel, and an infrared laser with a power of 615 mW positioned at the specimen plane was used to illuminate one of the outlet channels (Figure 4). In these illumination conditions, the flow in the dammed channel was completely halted through gel polymerization when the flow velocity of the inlet channel was less than 1 mm/s. The laser illumination could not stop the flow completely at faster velocity. The temperature distribution was then determined pixel by pixel by comparing the fluorescence intensities before and during infrared illumination. As shown in Figure 4, a gel plug 10 µm wide was formed within the channel. A temperature above 80 °C was observed at the center of the plug, while the width of the area where temperature exceeded 40 °C was less than 30 µm. The illumination area is so small that the temperature reaches a steady state reversibly within 30 ms upon switching between the on and off states13 Thus, the restricted and temporal heating prevented the cells from lysing, so that the majority of cells was viable after separation. Response Time of the Gel Valve. To evaluate the response time of the valve, i.e., sol-gel transformation of the Mebiol Gel, we measured the flow velocity in both the collection and waste channels simultaneously, while plugging only one of the channels and then immediately switching to the other channel. The flow velocity in both channels was determined by measuring the velocities of the flowing microspheres at 20-80 µm downstream of the laser illumination areas. The velocities of the microspheres were obtained from displacement of spheres detected every 1 ms using a high-speed video camera. We examined video footage 10 ms prior to switching the laser illumination from one channel to the other, and 25 ms after switching, and selected a series of video frames in which one microsphere was observed stopping downstream of the plug and a second microsphere was observed flowing within the other open reservoir. Figure 5 shows the relative velocities of the microspheres when the laser illumination was changed from on to off and off to on at time 0. The results demonstrate that the transformation from gel to sol and vice versa is complete within 3 ms. The sol-gel transformation of Mebiol Gel showed no hysteresis, which is in sharp contrast to that of methylcellulose.11 The response time of the described method is better than that reported for other devices, including those with active control of particle movements such as electrokinetic force3,5,14,15 and hydrodynamic flow control.6,8 We have successfully operated the cell sorter for more than 6 h. Our device is more robust than others since it does not contain any mechanical valves in a chip and does not require high voltage. Sorting of Particles and Cells. To evaluate the performance of the sorter, microspheres with diameters of 1 µm were extracted from a mixture of microspheres with diameters of 1 and 0.2 µm. The detection volume was typically ∼8 pL, which reduces the (14) Takahashi, K.; et al. J. Nanobiotechnol. 2004, 2, 5. (15) Morgan, H.; Hughes, M. P.; Green, N. G. Biophys. J. 1999, 77, 516-525.

698 Analytical Chemistry, Vol. 78, No. 3, February 1, 2006

Figure 5. Time course of sol-gel transformation in a microchannel in response to switching the infrared laser between on-off and offon. Fluorescence microspheres of 1-µm diameter were added to a solution containing 10% Mebiol Gel, and their movement were recorded using a high-speed video camera. The average relative velocity of the microspheres was measured every 1 ms, where the relative velocity is the ratio of a microsphere to a constant velocity without plugging. The infrared laser was switched on or off at time 0, and the transformation from sol to gel (closed circles) or gel to sol (open circles) was completed within 3 ms. Error bars indicate standard deviations.

background fluorescence from the solution and chamber glass. The planar geometry of the chip allows the use of high numerical aperture optics, which are sensitive enough to visualize single fluorescence molecules by epifluorescence microscopy.16 When a microsphere of 1-µm diameter passes through the detection window, the fluorescence intensity exceeds the predefined threshold triggering the device to switch the position of the infrared laser beam to the waste channel to capture the microbeads in the collection channel. The sorting routine was run in an automated mode. After sorting, the total number of microspheres in the collection and waste reservoirs was determined by image analysis. The microsphere concentration, throughput, and the sorting accuracy calculated from the fractions of large and small spheres found in both collection and waste reservoirs are listed in Table 1. Next, to evaluate the feasibility of the device for biological applications, E. coli bacterial cells expressing GFP were separated from those expressing DsRed. A mixture of E. coli bacterial cells expressing GFP and DsRed were suspended in PBS containing 10% Mebiol Gel and separated by the method described above. Figure 6 shows a typical sorting process for isolating E. coli cells expressing GFP. Once the target cell had been detected, the positioning of the infrared laser beam was switched to the waste channel for ∼120 ms and then switched back to the collection channel. The target cells were directed into the collection channel, where they remained ∼60 µm downstream from the junction. In the sorting device, the cells must travel 100 µm from the detection site to the collection channel. Thus, it took more than 100 ms to collect cell because the limit of flow velocity was 1 mm/s. Figure 7 shows fluorescence micrographs of a suspension of E. coli cells expressing GFP and DsRed before and after the sorting operation. (16) Funatsu, T.; et al. Nature 1995, 374, 555-559.

Table 1. Results Obtained from Sorting Different-Sized Microspheres and E. coli Cells Expressing Different Fluorescent Proteins concn run through- fractions of target cells run (× 106 time cells put input collection waste recovery no. cells/mL) (min) sorted (cells/s) well well well ratioa 1 2

8.5 218

1 2 3 4 5 6

1.0 4.5 7.9 37.8 26.9 208.6

40.5 2.7 360 120 30 41 30 30

949 2156 5154 7924 2915 12437 9243 45,50

Microspheresb 0.39 0.53 13.3 0.08 E. colic 0.24 0.27 1.1 0.44 1.6 0.49 5.0 0.080 5.1 0.17 25.4 0.026

0.94 0.20

0.14 0.03

0.86 0.79

0.97 0.96 0.87 0.67 0.66 0.34

0.12 0.22 0.15 0.025 0.031 0.004

0.65 0.64 0.83 0.72 0.86 0.87

a Recovery ratio is the number of target cells in the collection well divided by the number in the input well. b Microspheres of 1-µm diameter were separated from spheres of 0.2-µm diameter. c E. coli cells expressing GFP were separated from those expressing DsRed.

Figure 6. Sequential fluorescence micrographs obtained during GFP-expressed E. coli cell sorting. (a) Fluorescence of E. coli expressing GFP was detected using a PMT. (b) Plugging of the waste channel under the detection of the fluorescence signal forces E. coli to flow to the collection channel. (c) After the collection of the E. coli, the position of the infrared laser beam was switched back to the collection channel. (d) E. coli remained in the collection channel because the infrared laser plugged the collection channel blocking flow in to and out of the channel. Optical micrographs were taken using a standard video camera with exposure time of 33 ms.

Figure 7a shows a mixture of E. coli cells expressing GFP and DsRed in an inlet well before sorting. E. coli cells expressing GFP were successfully collected and enriched in the collection well (Figure 7b), while those expressing DsRed were collected in the waste well (Figure 7c). It should be noted that the concentration of target cells in the collection well was higher than that in the inlet well. Our sorting system has the advantage of concentrating target cells during separation, which is superior to other on-chip cell sorters using hydrodynamic focusing. The optical switching cell sorter described here is robust and capable of running without interruption for more than 6 h. The problems with cell damage or cell death caused by the exposure to the electric fields in other systems are avoided in our device. To confirm cell viability after cell sorting, the E. coli cells in the collection reservoir were spread on a culture medium plate, from which several colonies were grown.

Figure 7. Fluorescence micrographs of E. coli expressing GFP (green) and DsRed (orange) before and after cell sorting. The concentration of total cells before sorting was ∼1 × 107 cells/mL, and the sorting was performed at the rate of 0.7 cells/s for 6 h. (a) Cells in an inlet reservoir before sorting. (b) Cells in a collection reservoir after sorting. (c) Cells in a waste reservoir after sorting. Yellow dots were dusts in the preparation.

The variability of the results from different runs shown in Table 1 can be attributed to certain factors such as the initial cell concentration of cells, the initial ratio of target cells in the input well, and the flow velocity of the fluid. Here, there was a tradeoff between high-speed separation and purity of the sorted cells. As shown in Table 1, the higher the throughput (cells/s), the lower the fraction of target cells in the collection well. The throughput depends on both the flow velocity and the concentration of cells. In our sorting system, a switching time of 3 ms is fast enough as compared with the time required for directing a target cell to the collection channel. In fact, Figure 8a shows that the flow velocity has little effect on the ratio of target cells in the collect reservoir. On one hand, the purity after cell sorting depends on the concentration of cells, i.e., the higher the concentration of cells, Analytical Chemistry, Vol. 78, No. 3, February 1, 2006

699

Figure 8. Dependence of the purity of collected cells on the flow velocity at the inlet channel or on the concentration of cells. (a) E. coli cells expressing GFP were separated from those expressing DsRed at various flow rates. The concentration of total cells before sorting was 1 × 107 cells/mL. The black and white bars represent the ratio of target GFP cells before and after sorting. Each sorting was performed for 10 min. Error bars indicate standard deviations (n ) 3). (b) E. coli cells expressing GFP were separated from those expressing DsRed at various concentrations. The flow velocity at the inlet channel was 790 µm/s.

Figure 9. Dependence of the purity of collected cells on the concentrations of unwanted cells. E. coli cells expressing GFP were collected in the presence of various concentrations of those expressing DsRed. The concentration of wanted cells (E. coli expressing GFP) was different from sorting to sorting within a range of (0.3-5) × 106 cells/mL. The data were fitted to a theoretical curve using the leastsquares method.

the lower the fraction of target cells in the collection well (Figure 8b). Next, we examined the purity of the collected cells in the presence of various concentrations of unwanted cells. E. coli cells expressing GFP were collected in the presence of various concentrations of those expressing DsRed, and the ratio of the GFP cells in the collection reservoir was plotted against the concentration of DsRed cells (Figure 9). The ratio of wanted cells after the sorting was about unity when the initial concentration of unwanted cells was less than 107 cells/mL. Thus, the enrichment factor was inversely proportional to the ratio of the concentration of wanted and unwanted cells (data not shown). Then, the data of Figure 9 were fitted to eq 1 using the leastsquares method.

R ) 1/(1 + VC)

(1)

where R is the ratio of wanted cells in the collection reservoir, V 700

Analytical Chemistry, Vol. 78, No. 3, February 1, 2006

is the volume of solution that enters the collection channel per sorting operation, and C is the concentration of unwanted cells. V was determined from the fitting as 100 µm × cross-sectional area of the channel (30 × 5 µm2). These analyses indicated that it was important to reduce V to improve the purity of samples and suggested that the gating time must be optimized to improve both the purity and recovery ratio of wanted cells. The optimal gating time was expected to be 100 µm divided by the average flow velocity. But inhomogeneous flow velocity distributions in microchannel required 20% margin of the time. Shorter gating times were sometimes used to improve the purity, but they brought about a lower recovery ratio of wanted cells. For instance, E. coli Runs 1 and 2 in Table 1 were performed with 80% of the standard gating time. In such cases, the purity of the wanted cell was close to unity while the recovery ratio was reduced to ∼0.6. Our sorting system should be improved so that the gating time is automatically adjusted even when the flow rate fluctuates. The gating time is expected to become shorter by optimizing the shape of the microchannels and by introducing hydrodynamic focusing by sheath flow. Our device using reversible sol-gel transformation is less costly than conventional fluorescence-activated cell sorters, yet it is more expensive than on-chip sorters using hydrodynamic focusing. We are trying to modify our chip by coating a metal that efficiently absorbs infrared light at the entrances of the collection and waste channels. We are also trying to change the material from glass to poly(dimethylsiloxane) to simplify the device microfabrication procedure. These modifications will reduce the cost of the infrared laser and microchip to less than a tenth of the original cost. The on-chip cell sorter using sol-gel transformation described here has proved to be useful for the separation of two kinds of cells. We are now trying to sort several types of cells using multichannels. Multiplexing the cell sorting channels will also improve the overall throughput. Our sorting method with high specificity and sensitivity has the potential for sorting single molecules tagged with a fluorescent dye.

CONCLUSION A novel microflow system for cell sorters, free from any mechanical valves, has been developed. The sol-gel transformation of the thermoreversible gelation polymer was locally induced in a microchannel by site-directed infrared laser irradiation. The reversible sol-gel transformation functions as an in-channel valve and controls the flow direction of the cells. The corresponding response time of the sol-gel transformation was 3 ms. On-chip cell sorting by sol-gel transformation has overcome the issues of low-throughput rates and cell recovery that are associated with other cell sorting devices. Our device is more robust and capable of performing longer runs because it does not include any mechanical valves in the chip. In fact, our sorter has been successfully operated continuously for up to 6 h. The problem of cell death in other commercial devices is often caused by exposure to high electric fields. This drawback is avoided in our system, and a majority of cells remain viable after separation. Given the flexibility and relative ease of microfluidic control with sol-gel transformation by infrared laser, this method will be applicable to a variety of more integrated and complex function. As our device

is capable of separating submicrometer fluorescent samples, it is feasible to apply this device for the sorting of biomolecules. ACKNOWLEDGMENT We thank Hiroshi Yoshioka and Yuichi Mori for kindly donating Mebiol Gel with a critical solution temperature of 36 °C. We also thank Hirokazu Sugino and Masayasu Tatsuoka for technical assistance, Eiichi Kanaumi for helpful discussion, and Kaori Watanabe for critically reading the manuscript. This work was partly supported by SENTAN, JST, and by Grant ′H14-nano018′ of the Ministry of Health, Labor and Welfare of Japan. This work was also supported by Grants-in-Aid for COE Research of “Molecular Nano-Engineering”, Specially Promoted Research, Scientific Research Priority Area (B) 13124209, Scientific research (B) 12450167 and 13558088 from the Ministry of Education, Culture, Sports, Science and Technology of Japan. Received for review June 22, 2005. Accepted November 11, 2005. AC0511041

Analytical Chemistry, Vol. 78, No. 3, February 1, 2006

701