Integration in a multilayer microfluidic chip of 8 parallel cell sorters with flow control by sol–gel transition of thermoreversible gelation polymer

Abstract

Microfluidic systems have significant implications in the field of cell separation since they could provide platforms with inexpensive, disposable and sterile structures. Here, we present a novel strategy to integrate microfluidic sorters into a single chip for high throughput sorting. Our parallel sorter consists of a microfluidic chip with a three-dimensional channel network that utilizes flow switching by a heat-induced sol–gel transition of thermoreversible gelation polymer. The 8 parallel sheathed sample flows were realized by injecting sample and buffer solutions into only 2 inlets. The sheathed flows enabled disposal of unwanted sample waste without laser irradiation, and collection of wanted sample upon irradiation. As an application of the sorter, two kinds of fluorescent microspheres were separated with recovery ratio and purity of 70% or 90% at throughputs of about 100 or 20 particles per second, respectively. Next, Escherichia coli cells expressing green fluorescent protein were separated from those expressing DsRed with recovery ratio and purity of 90% at a throughput of about 20 cells per second.

---

Introduction

Mammalian and bacteria cell sorting systems, such as fluorescence-activated cell sorters (FACS), have made enormous contributions to clinical diagnosis, cellular therapy, infectious diseases research, and pharmaceutical industry.¹ Although FACS systems allow us to separate cells at speeds of 10 000 cells per second (s⁻¹) with specific fluorescence detection, they have several disadvantages. They are often difficult to sterilize,² and infectious materials need to be treated cautiously, as droplet-based sorters generate aerosols.³ Microfluidic platforms with disposable, sterile, and miniaturized structure devices could allow safe handling of valuable cells and infectious materials. Therefore, there are many interests in developing new microfluidic technologies that overcome issues related to current systems.

Various types of passive and active microfluidic sorters have been reported. Passive microfluidic sorters separate cells according to their intrinsic properties. Passive sorters use flow splitting^4,5 or Dean flow^6,7 to separate cells, while active microfluidic sorting is performed by fluorescence labeling of target cells. Active sorters use on-chip⁸ or off-chip valves,⁹ or thermoreversible gelation polymer (TGP)^10–12 to separate cells. Sorting techniques such as dielectrophretic force,^13–15 ultrasound,^16,17 or optical force^18–20 can be used as both passive and active sorting. Passive sorters can attain throughputs of ∼10⁶ cells s⁻¹. In contrast, throughputs of active sorters are restricted to 1–100 cells s⁻¹, although they could specifically detect cells of interest with fluorescent markers. For practical use, both high throughput and specific detection are required. Active sorters are expected to attain high throughput by integrating many sorters onto a single chip. However, integrations of active sorters have been rarely reported because of the complexity of the accompanying parallelization. All processes, such as creation of sample sheath flow, detection of specific cell fluorescence, and cell separation, need to be carried out at each integrated sorter individually. Therefore, it is difficult to avoid increased complexity in parallel sorting systems.

Three-dimensional (3D) microfluidic systems are useful for parallelization of sorters.²¹ In active sorters, the sample stream needs to be sheathed by two neighboring streams of buffer solution in order to align cells to the center of a channel.^11,19 Thus, parallel sorting channels designed in a two-dimensional plane require multiple inputs to create multiple sheath flows, which causes the increasing complexity of fluidic control. A 3D microfluidic system, in contrast, allows us to design a flexible channel network which reduces the number of inlets and the complexity of fluidic control.²¹

A TGP-based sorting scheme is also suitable for integration of sorters.^10–12 It uses transition of a TGP solution, which is liquid at low temperature and turns to a gel upon heating, to achieve flow switching. The TGP solution is mixed with a sample solution, and then introduced into a chip where the solution locally turns into a gel at a sorting region by heating. We use the gel as a flow switching valve, which means the TGP method does not require any mechanical valve. Therefore, the method could be easily applicable to control flow in a parallel channel or in a more complicated channel structure. The TGP-based sorting has additional advantages. First, cells could be collected without physical contact, since cell sorting is accomplished with flow switching by a gel valve. Second, small particles such as supramolecules or bacteria cells could be separated by a channel with dimension of ∼5 μm. On the other hand, the limitation of this method is the switching time of 3–60 ms^11,12 which is slower than that of optical force methods (2–4 ms),^18,19 or conventional FACS.

Here, we propose a new strategy to integrate active sorters into a single chip by combination of a chip with a 3D channel network and a flow control based on sol–gel transition of TGP. First, we describe the fabrication of a poly-dimethylsiloxane (PDMS)-glass chip with 8 sheath-flow channels using a replica molding technique and a PDMS dry-etching process; we show that sheathing the sample stream in 8 sorting channels was accomplished using as few as two inlets for sample and buffer solution injection. Next, we demonstrated that fluorescence microspheres were successfully sorted at each sorting channel using flow control based on a TGP method. Finally, we applied this parallel sorting system to separate bacteria cells expressing fluorescent proteins.

Experimental

Operating principle of separation

In this study, we applied the TGP method to flow switching in 8 parallel sorters. Fig. 1 shows the operating principle of the sorting chip. Both sample and buffer solutions contain TGP so that they change from sol to gel upon heating. The sample flow containing target particles is focused by two buffer solution flows and runs into the waste channel in the waste mode (Fig. 1a). Fluorescence signals from 8 sorters are measured with a 1 × 16 arrayed photomultiplier tube (PMT). When a fluorescence signal from the target particle is detected at the detection area in each sorter, the waste channel of the sorter is locally heated with an infrared (IR) laser deflected by an acousto-optical deflector (AOD) to plug the flow (Fig. 1b and c). The target particle is separated into the unheated collection channels. In this method, flow switching in each integrated sorters could be performed with a single IR laser.

Fig. 1 Illustrations of the operating principle of the 8 parallel sorters by flow control of TGP. Sample and buffer solutions contain TGP, which is liquid at low temperature and turns into a gel upon heating. The window for detecting the fluorescence signals is located before the junction. (a) In the absence of fluorescence signal from the target particle, the sample flows into the waste channel. (b) Upon detection of fluorescence signal, the waste channel is irradiated with an IR laser. As a result, the entrance of the waste channel is plugged by TGP and the target particle is collected to the collection channel. (c) The IR laser is deflected by an AOD to heat the sorters from which the fluorescence signals of target particles are detected.

Chip design

We designed a multilayer PDMS-glass chip that has a 3D channel network and 8 sheath flow channels. Fig. 2 shows the design and images of 8 parallel sorters integrated in a chip. The chip consists of two PDMS layers and one glass plate. The top PDMS layer has one inlet for sample injection (Inlet 1), another inlet for buffer solution injection (Inlet 2), one outlet for waste (Outlet 1), and another outlet for collection of the target particles (Outlet 2). The second PDMS layer has through-hole structures and a sorting area (Fig. 2b). Sample and buffer solutions are injected from the top PDMS layer, and flow into the second layer via the through-holes. Each inlet channel has a pre-filter structure after its through-hole to prevent particles larger than 5 μm from flowing into the sorting area (Fig. 2d). The glass plate has metal dot patterns which absorb IR laser beam and function as heaters upon sorting (Fig. 2e). A sample solution injected from Inlet 1 splits into 8 flows and a buffer solution injected from Inlet 2 splits into 16 flows. Each sample flows is sandwiched and focused by 2 buffer solution flows at the sorting area (Fig. 2f). The fluidic resistance of a waste channel is designed to be two-thirds of that of a collection channel so that the sample automatically flows into the waste channel in the absence of IR irradiation. The depth of the sorting area is 5 μm and the depths of other areas are 12 or 20 μm to reduce fluidic resistance. The through-hole structure is about 80 μm in height. The depth and width of the pre-filter is 5 μm × 5 μm. The metal dot patterns consist of 80-nm-thick Cr and 40-nm-thick Au. The size of square dots is 10 μm at 3 μm intervals.

Fig. 2 Layout of the 8 parallel sorters integrated in a microfluidic chip which consists of two PDMS layers and a glass plate. (a) A photograph of the chip. (b) Cross-sectional view of the chip. (c) Three-dimensional view of the chip. Sample and buffer solutions are injected from the inlets and flow into the second PDMS layer via through-holes. The sample flow branches into 8 flows and buffer solution flow branches into 16 flows. (d) A scanning electron micrograph of a through hole and a pre-filter in a second-layer mold. (e) Layout of metal dos on the glass plate which absorb an infrared laser and function as a heater. (f) A scanning electron micrograph of the sorting area in a second-layer mold. Each sample flow is sheathed with two buffer solution flows at this area. (g) An optical micrograph of the sorting area in the chip. Metal dots are deposited on the glass plate.

Fabrication process

The 3D channel network was fabricated with a replica molding technique and a PDMS dry etching process. The fabrication process is shown in Fig. 3. The mold of the second-layer structures was formed on a Si substrate by photoresist (Fig. 3a). The molds of the 5-μm structures, the 20-μm structures, and the through-holes were fabricated with SU-8 3005, SU-8 3035, and CA 3000, respectively (SU-8 3005 and SU-8 3035, Kayaku Microchem, Co. Ltd., Japan; PMER-N CA 3000, Tokyo Ohka Kogyo, Tokyo, Japan). The PDMS were spin-coated on the mold. To ensure that the top of the through-hole structure was completely devoid of PDMS, the PDMS was then etched down (Fig. 3b). The PDMS dry-etching process was performed in capacitive-coupled plasma ion etching (CCP-RIE, RIE-10NR; Samco, Japan) to make through-holes. Although PDMS dry etching with CF₄ and O₂ gases has been reported,^22,23 CF₄ gas was not available for our reaction-ion etching (RIE) equipment. Instead, we performed PDMS dry etching with CHF₃ and O₂ gases in this fabrication process. Considering etching rate and surface roughness, the etching condition was determined to be 25% O₂ and 75% CHF₃, which resulted in an etching rate of about 9.5 μm per hour and surface roughness of about 27 nm (see Table S1 in ESI for more information about PDMS dry etching†). The molds of the top-layer structures (5 and 12 μm in height) were also formed on a Si substrate by photoresist (SU-8 3005); the PDMS structure was formed by casting (Fig. 3c). The top PDMS structure was bonded to the second PDMS layer after O₂ plasma treatment. The PDMS stack was subsequently bonded to a glass plate which had metal dots on its surface (Fig. 3d).

Fig. 3 Schematic overview of the process steps of the device. The second layer process is shown in (a, b). The top layer process is shown in (c). (a) The second layer mold consists of three-layer structure formed on a Si substrate. PDMS layer is formed on the mold. (b) The PDMS which covers the through hole is removed by dry etching. (c) The top layer mold consists of two-layer structure formed on a Si substrate. PDMS layer is formed on the mold. (d) The top PDMS layer is bonded to the second PDMS layer, and the PDMS block is bonded to a glass substrate.

Optical setup

Fig. 4 shows the optical setup for the device. A microfluidic chip containing 8 parallel sorters was placed on an inverted microscope (IX71, Olympus Optical, Tokyo, Japan) equipped with an IR laser (Millennia IR, Spectra-Physics, USA) operating at 1064 nm. The IR laser was used to heat the metal dots to raise the temperature of solution in the microchannels. An AOD (LS110-1.06XY, Isomet Corporation, USA) was placed around the position conjugated to the back focal point of the objective (UAPO/340 20x NA 0.75, Olympus Optical). The laser beam was deflected by the AOD and focused on the specimen plane. The power of the IR power was 43 mW at the specimen plane.

Fig. 4 Schematic diagram of the optical setup for the device. A chip is placed on a fluorescence microscope equipped with an IR laser at 1064 nm. The specimen is illuminated with a 488 nm laser or a mercury lamp. Fluorescence is split by a half mirror, and the fluorescence image is captured by a CCD camera and the fluorescence signal is detected by a 1 × 16 arrayed PMT, simultaneously. Depending on the fluorescence signals from the 8 parallel sorters, the position of the IR laser beam is deflected by an AOD.

Yellow-green and crimson microspheres were excited by a mercury lamp through an excitation filter (HQ470/40X, Chroma Technology, USA). Excitation light was reflected by a dichroic mirror (FF497/554, Semrock, USA) and introduced to the objective lens. The fluorescence from microspheres passed through the dichroic mirror and an emission filter (XF3086 510ALP, OMEGA OPTICAL, USA). E. coli cells expressing green fluorescent protein (GFP) and DsRed were excited by a 488-nm laser line (Sapphire 488-20, Coherent Inc., USA) without the excitation filter. A notch filter (Stop Line Notch filter 488, Semrock) was used in addition to the emission filter. The fluorescence from the specimen was split by a half mirror and the fluorescence image was projected onto both a charge-coupled device (CCD) video camera (MC681SPD, Texas Instruments Inc., USA) and a 1 × 16 arrayed PMT (H9531-01, Hamamatsu Photonics, Japan). The 1 × 16 arrayed PMT consisted of 16 independent channels of PMTs and 8 channels were used to detect fluorescence of target particles. Fluorescence intensities at 8 channels were converted to 8-channel electric pulses which were acquired every 2 ms with a counter (C9945, Hamamatsu Photonics). When observing or detecting yellow-green microspheres and E. coli cells expressing GFP, a band pass filter (HQ535/50M, Chroma Technology) was placed in front of the CCD camera and the 1 × 16 arrayed PMT; when observing crimson microspheres and E. coli cells expressing DsRed, a band pass filter (635DF55, OMEGA OPTICAL) was placed in front of the CCD camera.

The detection window was 20 × 400 μm at the specimen plane; it was positioned 30 μm upstream of the sorting junctions, before the 1 × 16 arrayed PMT, and used to reduce background fluorescence emitted from the solution, PDMS, and glass. The fluorescence signal was acquired every 2 ms with a photon counter (C9945, Hamamatsu Photonics) and converted by the 1 × 16 arrayed PMT to 16-channel electric pulses. A Visual Basic program was written to read the signals from the counter and to digitally control the AOD. When a target particle was detected in a sorter, the AOD automatically deflected the IR laser to the waste channel of the sorter where the fluorescence was detected. As a result, the waste channel was plugged with a gel, and the target was separated into the collection channel.

Materials and reagents

Yellow-green (Ex.505/Em.515) and crimson (Ex.488/Em.645) microspheres with 1-μm diameters were purchased from Molecular Probes (FluoSpheres carboxylate-modified microspheres, Eugene, OR, USA). E. coli cells expressing fluorescent proteins were prepared as described in ESI.† A thermoreversible gelation polymer, Mebiol Gel,²⁴ with a critical solution temperature of 36 °C, was purchased from Mebiol Inc, Tokyo, Japan; this is a block copolymer of poly (N-isopropylacrylamide-co-n-butyl methacrylate) and polyethylene glycol. Mebiol Gel is an inert biocompatible polymer used for culturing mammalian cells.²⁵ Mebiol Gel powder was dissolved in phosphate-buffered saline (PBS; T900, Takara Bio Inc., Otsu, Japan) to a final concentration of 20% (w/v) and stored at −80 °C until used. Mebiol gel can be used in the presence of phosphate buffered saline, Tris-buffered solution, and cell culture medium. The microspheres or E. coli cells were mixed with the Mebiol Gel solution (final concentration of 10%) so that the viscosity of the solution was about 55 times higher than that of pure water at 23 °C. Care was taken to avoid making bubbles, which sometimes destabilize or block the flow in a microchannel.

Flow control

Sample solution was delivered to Inlet 1 and buffer solution was delivered to Inlet 2 (Fig. 2) via polyetheretherketone (PEEK) tubes by the pressure of precisely controlled compressed air (DPP-AYAD, KOGANEI) via an electro-pneumatic transducing regulator (ETR010, KOGANEI, Tokyo, Japan). The linear flow velocity around the detection area was 9 mm s⁻¹.

Data analysis

Performance of the 8 parallel sorters was characterized by three parameters: recovery ratio, purity, and throughput. Recovery ratio was defined by the ratio of the number of recovered target particles divided by the total number of target particles. Purity is the ratio of the number of target particles to the total number of target and unwanted particles collected. Throughput is the number of target particles sorted per second. These parameters were calculated by analyzing movies captured by a CCD camera. First, we recorded target particles being sorted using a CCD camera with a band pass filter of HQ535/50M. We then changed to a band pass filter of 635DF55, and recorded unwanted particles flowing into waste or collection channels. Target and unwanted particles that flowed into the collection or waste channel were then counted manually.

Viability of E. coli cells

Viability of E. coli cells before and after the sorting was assessed using a BD Cell Viability Kit (BD Biosciences, San Jose, CA, USA). The kit contains thiazole orange and propidium iodide. All cells were stained with thiazole orange; only dead cells were stained with propidium iodide. Viability was determined by fluorescence microscopy.

Results and discussion

Sheath flows in 8 parallel sorters

Sheathed sample flow in 8 channels was demonstrated by injecting a sample solution from Inlet 1 and a buffer solution from Inlet 2 (Fig. 2). Sample solution included yellow-green microspheres and 10% TGP solution; buffer solution included 10% TGP solution only. Fig. 5a shows that sheathed flow of the sample was realized at each sorter, indicating that multi-sheath flows can be created by introducing sample and buffer solutions from just two inlets, and that samples flowed into waste channels in the absence of laser-induced heating.

Fig. 5 Fluorescence micrographs of the flows of microspheres in a microfluidic chip. (a) Waste mode: Yellow-green microspheres were collected in the waste channels (the left channels) without sorting operation. Stream lines indicate the flows of the microspheres. Broken lines indicate the outlines of microchannels. Exposure time was 10 s. Scale bar indicates 50 μm. (b) Collection mode: A microsphere flowed into the right channel for collection with a sorting operation. The bright spot is the reflection of an IR laser which heated metal dots in the waste channel to plug the flow. Fluorescence images were captured every 33 ms. Scale bar indicates 50 μm. The supplemental movie 1 is available in ESI.†

Separation of fluorescent microspheres

We demonstrated microsphere sorting with 8 parallel sorters. Yellow-green and crimson microspheres were suspended in a 10% (w/v) TGP solution and introduced from Inlet 1, while a buffer solution containing 10% TGP was introduced from Inlet 2 using compressed air. All microspheres flowed into the waste channels without IR irradiation. Upon detection, the yellow-green microspheres were sorted by irradiating the waste channel by IR laser for 10 ms (Fig. 5b). A movie of this sorting (Movie 1) is available in ESI.† These results show that the 8 parallel sorters by microfluidic control using sol–gel transition of TGP worked well. When all the channels were illuminated by IR laser for long time, flow resistance increased due to insufficient heat dispersion. Thus the concentration of target particles was adjusted so that the number of channels illuminated by IR laser simultaneously was no more than 4.

Next, we evaluated the performance of the 8 parallel sorters by sorting yellow-green microspheres under different microsphere concentrations. The yellow-green and crimson microspheres were mixed at a ratio of 1 : 10 and suspended in a 10% TGP solution. The final concentration of yellow-green microspheres was from 2 × 10⁶ to 3 × 10⁸ particles mL⁻¹. Recovery ratio, purity before and after sorting, and throughput are shown in Table 1. Throughput increased as the concentration of yellow-green microsphere increased. A 90% recovery ratio and 90% purity were attained at a low throughput (0.9, 1.9, and 16.8 particles s⁻¹), while they deceased at a high throughput (61.6, 101.6, and 107.8 particles s⁻¹). The decreases in recovery ratio and purity are ascribed to the successive detection of microspheres in the same channel. Upon detection of the wanted particle, IR laser was irradiated for 10 ms at the corresponding channel. If the second wanted particle flowed into the detection area of the channel during the irradiation, the particle was wasted by our separation algorithm. Thus the flow rate affected both recovery ratio and purity. The difference in throughput under the concentration of 3 × 10⁸ particles mL⁻¹ as shown in Table 1 was ascribed to the different flow rates of sample and buffer solutions.

Table 1 Results of sorting microspheres at different concentrations^b

Concentrations^a [particles mL^−1]	Number of target microspheres			Recovery ratio	Purity		Throughput [particles s^−1]
	Detected	Wasted	Collected		Before	After
a Concentrations of yellow-green microspheres are shown. b Irradiation time of the IR laser was 10 ms.
2 × 10^6	297	34	263	0.89	0.09	0.94	0.9
2 × 10^6	335	22	313	0.93	0.09	0.94	1.0
6 × 10^6	620	37	583	0.94	0.09	0.97	1.9
6 × 10^6	703	53	650	0.92	0.09	0.93	2.2
6 × 10^7	5333	308	5025	0.94	0.09	0.88	16.8
3 × 10^8	6024	2330	3694	0.61	0.09	0.69	61.6
3 × 10^8	5955	1445	4510	0.76	0.09	0.71	75.2
3 × 10^8	9142	3048	6094	0.67	0.09	0.60	101.6
3 × 10^8	8928	2459	6469	0.72	0.09	0.62	107.8

---

Separation of E. coli cells expressing fluorescent proteins

As an application of the sorting system, we separated E. coli cells expressing GFP from those expressing DsRed. E. coli cells expressing GFP or DsRed were mixed at a ratio of 1 : 7, or 1 : 1 with final concentrations of E. coli cells expressing GFP being from 8 × 10⁶ to 4 × 10⁷ cells mL⁻¹. Recovery ratio, purity before and after sorting, and throughput are summarized in Table 2. When the ratio of E. coli cells expressing GFP to those expressing DsRed was 1 : 7, E. coli cells expressing GFP were sorted with a 95% recovery ratio and 76% purity at a throughput ranging from 2.0–4.4 cells s⁻¹. A 94% recovery ratio and 88% purity were attained at throughputs ranging from 13.1 to 18.4 cells s⁻¹. Higher throughput, as seen in fluorescent microspheres, was not realized, because pre-filter structures were clogged with E. coli cells under concentration of 1 × 10⁸ cells mL⁻¹, which was higher than that handled with other sorters (10²–10⁷ cells mL⁻¹).^4–19 We expect that a low flow-resistance device would allow us to sort E. coli cells at higher throughput, because flow rate would increase.

Table 2 Results of sorting E. coli cells at different concentrations.^b

Concentrations^a [cells mL^−1]	Recovery ratio	Purity		Throughput [cells s^−1]
		Before	After
a Concentrations of E. coli cells expressing GFP are shown. b Irradiation time of the IR laser was 10 ms. Run time was 5 min.
8 × 10^6	0.94	0.15	0.85	2.0
8 × 10^6	0.96	0.16	0.81	2.9
8 × 10^6	0.96	0.14	0.79	3.0
8 × 10^6	0.95	0.16	0.73	4.4
4 × 10^7	0.91	0.50	0.82	6.8
4 × 10^7	0.95	0.53	0.89	13.1
4 × 10^7	0.95	0.53	0.89	14.6
4 × 10^7	0.94	0.56	0.87	18.4

---

To assess E. coli viability, we sorted E. coli cells expressing GFP for an hour; their viability was determined using a BD cell viability kit by fluorescence microscopy. Cell viability before and after sorting were 87 ± 2% and 85 ± 3%, respectively (mean values ± SD of three experiments), indicating that our sorting device imposes little damage on cells.

Perspective of this study

We have shown that it is possible to sheath 8 sample streams and control the flows in 8 channels based on a three-dimensional channel network and flow switching by TGP method. Since the sorting system requires only two inlets and a single IR laser for flow control, the complexity of the parallelization system is effectively minimized. Although we demonstrated only 8 gel valves simultaneously formed with a single laser, the TGP method could be applied to sorting system with more than 8 channels. The laser beam can be deflected to more than 8 channels repeatedly during the dwell time of the heating (10 ms), which is much longer than the AOD switching time (15 μs). The maximum number of parallel sorters will be 30, because the diameter of observation area was 2 mm and the distance of waste and collection channels must be larger than 30 μm due to temperature distribution. In spite of multiplexing using thermoreversible gels, our system is still slow compared to conventional FACS. But our system is suitable for sorting of small amount of specimen, sorting of infectious materials, sorting without cross contamination, and sorting of rare cells.

The microfluidic structure and the flow control technique are not limited to integration of microfluidic sorters. The 3D structure could be useful for other sheathing applications such as integrations of flow cytometry or drop-based microfluidics. The flow control by TGP could also be applicable to other channel structures that require multivalve control.

In this study we integrated 8 sorters in a chip of about 400 μm × 400 μm; each channel size was 5 μm deep × 20 μm wide. Space and channel size are comparatively small, given that channel sizes in other cell sorters are usually at least 100 μm. For this reason, the parallel sorter would be especially suited to handle bacterial cells rather than mammalian cells, as bacterial cells are smaller than mammalian cells. We envision further applications as a purification tool in organelle proteomics, separating organelles of interest with specific fluorescence detection. As these organelles are usually no larger than bacterial cells, the TGP method with a microfluidic chip would be particularly appropriate for such investigations.

Conclusions

We integrated 8 parallel sorters on a single chip with a 3D channel network. These sorters were operated by microfluidic control based on TGP sol–gel transition induced by IR laser irradiation. This sorting system is simple, as 8 sheath flows were created with only two inlets; flow switching at every channel was controlled with a single IR laser. Fluorescent microspheres and E. coli cells were successfully separated by the 8 parallel sorters. Throughput of several tens of particles per second was achieved by integration of the sorters.

Acknowledgements

The authors thank Mai Yamagishi (RIKEN) for preparation of E. coli cells expressing fluorescent proteins and assistance in data analysis, and Osamu Ohara (RIKEN) for encouragement. This study was partly supported by SENTAN, JST, and by a Grant-in-Aid for Scientific Research (A) 17201031 (to T.F.) from the Ministry of Education, Culture, Sports, Science and Technology of Japan. H. S. (20-9830) and T. A. (20-10635) are the recipients of a Research Fellowship from the Japan Society for the Promotion of Science for Young Scientists.

References

1. W. A. Bonner, H. R. Hulett, R. G. Sweet and L. A. Herzenberg, Rev. Sci. Instrum., 1972, 43, 404–409.
2. H. M. Shapiro, Practical flow cytometry, 4th edn, Wiley-Liss, New York, 2003.
3. I. Schmid, C. Lambert, D. Ambrozak, G. E. Marti, D. M. Moss and S. P. Perfetto, Cytometry, Part A, 2007, 71a, 414–437.
4. M. Yamada and M. Seki, Lab Chip, 2005, 5, 1233–1239.
5. M. Yamada and M. Seki, Anal. Chem., 2006, 78, 1357–1362.
6. D. Di Carlo, J. F. Edd, D. Irimia, R. G. Tompkins and M. Toner, Anal. Chem., 2008, 80, 2204–2211.
7. A. A. S. Bhagat, S. S. Kuntaegowdanahalli and I. Papautsky, Lab Chip, 2008, 8, 1906–1914.
8. A. Y. Fu, H. P. Chou, C. Spence, F. H. Arnold and S. R. Quake, Anal. Chem., 2002, 74, 2451–2457.
9. A. Wolff, I. R. Perch-Nielsen, U. D. Larsen, P. Friis, G. Goranovic, C. R. Poulsen, J. P. Kutter and P. Telleman, Lab Chip, 2003, 3, 22–27.
10. Y. Shirasaki, J. Tanaka, H. Makazu, K. Tashiro, S. Shoji, S. Tsukita and T. Funatsu, Anal. Chem., 2006, 78, 695–701.
11. T. Arakawa, Y. Shirasaki, T. Aoki, T. Funatsu and S. Shoji, Sens. Actuators, A, 2007, 135, 99–105.
12. H. Sugino, K. Ozaki, Y. Shirasaki, T. Arakawa, S. Shoji and T. Funatsu, Lab Chip, 2009, 9, 1254–1260.
13. X. Y. Hu, P. H. Bessette, J. R. Qian, C. D. Meinhart, P. S. Daugherty and H. T. Soh, Proc. Natl. Acad. Sci. U. S. A., 2005, 102, 15757–15761.
14. M. S. Pommer, Y. Zhang, N. Keerthi, D. Chen, J. A. Thomson, C. D. Meinhart and H. T. Soh, Electrophoresis, 2008, 29, 1213–1218.
15. L. Wang, L. A. Flanagan, N. L. Jeon, E. Monuki and A. P. Lee, Lab Chip, 2007, 7, 1114–1120.
16. F. Petersson, L. Aberg, A.-M. Sward-Nilsson and T. Laurell, Anal. Chem., 2007, 79, 5117–5123.
17. T. Franke, A. R. Abate, D. A. Weitz and A. Wixforth, Lab Chip, 2009, 9, 2625–2627.
18. M. M. Wang, E. Tu, D. E. Raymond, J. M. Yang, H. Zhang, N. Hagen, B. Dees, E. M. Mercer, A. H. Forster, I. Kariv, P. J. Marchand and W. F. Butler, Nat. Biotechnol., 2005, 23, 83–87.
19. T. D. Perroud, J. N. Kaiser, J. C. Sy, T. W. Lane, C. S. Branda, A. K. Singh and K. D. Patel, Anal. Chem., 2008, 80, 6365–6372.
20. M. P. MacDonald, G. C. Spalding and K. Dholakia, Nature, 2003, 426, 421–424.
21. H. Wu, T. W. Odom, D. T. Chiu and G. M. Whitesides, J. Am. Chem. Soc., 2003, 125, 554–559.
22. J. Garra, T. Long, J. Currie, T. Schneider, R. White and M. Paranjape, J. Vac. Sci. Technol., 2002, A20(3), 975–982.
23. S. R. Oh, J. Micromech. Microeng., 2008, 18, 115025.
24. H. Yoshioka, M. Mikami, Y. Mori and E. Tsuchida, J. Macromol. Sci., Pure Appl. Chem., 1994, A31, 113–120.
25. K. Hishikawa, S. Miura, T. Marumo, H. Yoshioka, Y. Mori, T. Takato and T. Fujita, Biochem. Biophys. Res. Commun., 2004, 317, 1103–1107.

---

Footnote

† Electronic supplementary information (ESI) available: Dry etching of PDMS, preparation of E. coli cells and separation of fluorescent microspheres including Table S1 and Movie 1. See DOI: 10.1039/c004192k

---