High-Throughput and High-Resolution Flow Cytometry in Molded

The cytometry system built around the first microfluidic device has fluorescence detection accuracy comparable with that of a commercial flow cytomete...
9 downloads 7 Views 404KB Size
Download PDF
Anal. Chem. 2006, 78, 5653-5663

High-Throughput and High-Resolution Flow Cytometry in Molded Microfluidic Devices Claire Simonnet and Alex Groisman*

Department of Physics, University of California, San Diego, 9500 Gilman Drive, La Jolla, California 92093

We describe the design, fabrication, and operation of two types of flow cytometers based on microfluidic devices made of a single cast of poly(dimethylsiloxane). The stream of particles or cells injected into the devices is hydrodynamically focused in both transverse and lateral directions, has a uniform velocity, and has adjustable diameter and shape. The cytometry system built around the first microfluidic device has fluorescence detection accuracy comparable with that of a commercial flow cytometer and can analyze as many as 17 000 particles/s. This high-throughput microfluidic device could be used in inexpensive stand-alone cytometers or as a part of integrated microanalysis systems. In the second device, a stream of particles is focused to a flow layer of a submicrometer thickness that allows imaging the particles with a high numerical aperture microscope objective. To take long-exposure, low-light fluorescence images of live cells, the device is placed on a moving stage, which accurately balances the translational motion of particles in the flow. The achieved resolution is comparable to that of still micrographs. This high-resolution device could be used for analysis of morphology and fluorescence distribution in cells in continuous flow. Flow cytometry is a common technique for high-throughput characterization of particles and cells in liquid suspension. The core of a standard cytometer is a flow chamber, where the stream carrying particles is hydrodynamically focused to a small region near the middle of the chamber cross section. The spatial confinement of the particles facilitates detection of their optical properties at high sampling rate because it allows high-intensity illumination by low-power lasers and efficient collection of scattered and fluorescence light by high numerical aperture (NA) lenses. The particles in the focused stream move with nearly identical velocities and have a well-defined passage time through the illumination spot, which is another factor facilitating accurate and reliable optical analysis at high throughput. Recent progress in flow cytometry has mostly involved advanced multispectral analysis. In contrast, the geometrical parameters of cytometry chambers and the pattern of flow in them have remained relatively unchanged. The cross-sectional dimensions of a cytometry chamber are typically in the 100-µm range, and the stream carrying the cells is focused to a central region * To whom correspondence should be addressed; electronic mail: agroisman@ ucsd.edu. 10.1021/ac060340o CCC: $33.50 Published on Web 07/13/2006

© 2006 American Chemical Society

with a diameter of ∼5-10 µm. Typical cell velocity is ∼10 m/s, and sampling rates usually vary between 1000 and 50 000 cells/s, depending on the number of spectral lines to be analyzed and the accuracy required.1,2 Cells usually pass through a sequence of spots illuminated by lasers of different wavelengths, and the power of fluorescence and light scattering by the cells is detected by a set of photomultipliers (PMT) and digitized at a high rate. A notable recent development in flow cytometry is a high-resolution multispectral platform ImageStream by Amnis.3 It uses a standard cytometry chamber operated at low flow speed (20 mm/s) and cell throughput (∼100 cell/s), and it can image cells in up to six spectral modes using a special CCD array. In general, modern commercial flow cytometers are sophisticated, expensive, and voluminous machines, usually requiring dedicated personnel for operation and maintenance. Flow cytometry chambers used in most commercial systems are based on coaxial tubing.1 As photolithography fabrication methods, initially developed for microelectronics, became increasingly advanced and accessible, they were also adapted for machining flow cytometry chambers. Microfabricated flow cytometry chambers can substantially reduce the sample volume required for an assay. In addition, they can be combined with optical4-8 and electronic9,10 components on the same chip to implement alternative particle detection schemes. More importantly, an on-chip flow cytometry chamber that is part of a larger microfabricated channel network can be a useful element for integrated lab-on-a-chip and micro total analysis systems (µTAS).11 Many types of microfabricated flow cytometry chambers, including some early models,12,13 were machined in quartz, metal, (1) Van Dilla, M. A. In Flow Cytometry: Instrumentation and Data Analysis.; Academic Press Inc.: London, 1985; pp 1-20. (2) Leary, J. F. Cytometry 2005, 67A, 76-85. (3) George, T. C.; Basiji, D. A.; Hall, B. E.; Lynch, D. H.; Ortyn, W. E.; Perry, D. J.; Seo, M. J.; Zimmerman, C. A.; Morrissey, P. J. Cytometry 2004, 59A, 237-245. (4) Lin, C. H.; Lee, G. B.; Fu, L. M.; Hwey, B. H. J. Microelectromech. Syst. 2004, 13, 923-932. (5) Tung, Y. C.; Zhang, M.; Lin, C. T.; Kurabayashi, K.; Skerlos, S. J. Sens. Actuators, B 2004, 98, 356-367. (6) Lien, V.; Zhao, K.; Lo, Y. H. Appl. Phys. Lett. 2005, 87, Art. No. 194106. (7) Chabinyc, M. L.; Chiu, D. T.; McDonald, J. C.; Stroock, A. D.; Christian, J. F.; Karger, A. M.; Whitesides, G. M. Anal. Chem. 2001, 73, 4491-4498. (8) Wang, Z.; El-Ali, J.; Engelund, M.; Gotsaed, T.; Perch-Nielsen, I. R.; Mogensen, K. B.; Snakenborg, D.; Kutter, J. P.; Wolff, A. Lab Chip 2004, 4, 372-377. (9) Cheung, K.; Gawad, S.; Renaud, P. Cytometry 2005, 65A, 124-132. (10) Chun, H. G.; Chung, T. D.; Kim, H. C. Anal. Chem. 2005, 77, 2490-2495. (11) Huh, D.; Gu, W.; Kamotani, Y.; Grotberg, J. B.; Takayama, S. Physiol. Meas. 2005, 26, R73-R98.

Analytical Chemistry, Vol. 78, No. 16, August 15, 2006 5653

or glass.4,9,10,14-16 Recently, there has been interest in cytometry chambers made of plastic materials, machined by either laser cutting17 or casting with lithographically fabricated master molds.5,6,8,11,18-22 Molded microfluidic devices (which are often made of a silicone elastomer poly(dimethylsiloxane), PDMS) can be made to tight tolerances and are simple and inexpensive to manufacture. A cytometry chamber machined in plastic could serve as a disposable cartridge that could be loaded with a cell sample to be analyzed and inserted into an advanced flow cytometry system. Alternatively, a plastic cytometry chamber could be used in inexpensive compact systems intended for limited tasks, such as cell (or particle) counting and sizing and singleline fluorescence detection. The hydrodynamic focusing of the stream carrying particles to the middle of the cytometry chamber cross section is achieved by enclosing the stream of particles from all sides with a sheath flow. In the cytometry literature, this type of flow focusing is usually called three-dimensional (3D).11 The 3D flow focusing is relatively straightforward to realize in standard cytometry chambers with coaxial ducts.1 In contrast, lithographically fabricated cytometry chambers normally have planar microchannel networks and allow flow focusing only in plane of the device,5,14,16,21-23 which is usually called two-dimensional (2D) focusing. To prevent excessive spreading of the particles in the out-of-plane direction, some lithographically fabricated devices had cytometry chambers of small depths (∼10 µm).23 In other microfabricated devices, both cross-sectional dimensions of the cytometry chambers were made small, and no flow focusing was used.18,24,25 However, since the flow normally has a parabolic velocity profile, particle velocities in streams confined by channel walls (even in one direction) are highly nonuniform. In addition, the designs with shallow cytometry chambers cannot reach high flow speeds and throughputs typical for commercial cytometers, since achieving those high speeds would require prohibitively high driving pressures and lead to exposure of cells to high mechanical stresses. Partial flow focusing in the out-of-plane direction was demonstrated in microfluidic devices with variable channel depth20 and (12) Sobek, D.; Young, A. M.; Gray, M. L.; Senturia, S. D. Proceedings of IEEE Micro Electro Mechanical Systems Workshop; Fort Lauderdale, FL, 1993; 219224. (13) Miyake, R.; Ohki, H.; Yamazaki, I.; Takagi, T. JSME Int. J., Ser. B 1997, 40, 106-113. (14) Schrum, D. P.; Culbertson, C. T.; Jacobson, S. C.; Ramsey, J. M. Anal. Chem. 1999, 71, 4173-4177. (15) McClain, M. A.; Culbertson, C. T.; Jacobson, S. C.; Ramsey, J. M. Anal. Chem. 2001, 73, 5334-5338. (16) Lee, G. B.; Lin, C. H.; Chang, S. C. J. Micromech. Microeng. 2005, 15, 447454. (17) Weigl, B. H.; Bardell, R.; Schulte, T.; Battrell, F.; Hayenga, J. Biomed. Microdevices 2001, 3, 267-274. (18) Chou, H. P.; Spence, C.; Scherer, A.; Quake, S. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 11-13. (19) Eyal, S.; Quake, S. R. Electrophoresis 2002, 23, 2653-2657. (20) Chung, S.; Park, S. J.; Kim, J. K.; Chung, C.; Han, D. C.; Chang, J. K. Microsyst. Technol. 2003, 9, 525-533. (21) Huh, D.; Tkaczyk, A. H.; Bahng, J. H.; Chang, Y.; Wei, H. H.; Grotberg, J. B.; Kim, C. J.; Kurabayashi, K.; Takayama, S. J. Am. Chem. Soc. 2003, 125, 14678-14679. (22) Dittrich, P. S.; Schwille, P. Anal. Chem. 2003, 75, 5767-5774. (23) Kruger, J.; Singh, K.; O’Neill, A.; Jackson, C.; Morrison, A.; O’Brien, P. J. Micromech. Microeng. 2002, 12, 486-494. (24) Altendorf, E.; Zebert, D.; Holl, M.; Yager, P. Proceedings of International Conference Transducers ’97; Chicago, Illinois, 1997; pp 531-534. (25) Fu, A. Y.; Spence, C.; Scherer, A.; Arnold, F. H.; Quake, S. R. Nat. Biotechnol. 1999, 17, 1109-1111.

5654

Analytical Chemistry, Vol. 78, No. 16, August 15, 2006

with two layers of channels,26 but it did not completely resolve the problems of velocity dispersion and exceedingly high stresses at high flow speeds. Full out-of-plane flow focusing was implemented in a microfluidic device for studies of protein folding.27 However, this device would be difficult to use for flow cytometry, because the channel with the focused flow had a small depth. Focusing of particles in the out-of-plane direction was achieved using dielectrophoretic forces, but the success of this method depended on the polarizability and size of the particles, and the microfluidic devices required complex fabrication.4,28 Full 3D hydrodynamic focusing was realized in several types of microfluidic devices.17,29-31 Fabrication of each of these devices, however, was an involved multistep process. Moreover, the cytometry chambers in these devices were not readily compatible with standard, high numerical aperture objective lenses. Recently, we have constructed and tested a molded microfluidic chip that allows tight hydrodynamic focusing in both in-plane and out-of-plane directions.32 (It was coined 2D flow focusing in ref 32, but here, we call it 3D flow focusing, as is common in the literature.) In this paper, we describe the design, fabrication, and operation of two microfluidic flow cytometry devices that are based on this flow focusing chip, as well as the setup and performance of the detection and imaging systems built around the two devices. Both devices are made of a single cast of PDMS sealed with a cover glass and are disposable and compatible with standard shortworking-distance microscope objectives. The high-throughput device (HTD) is designed to emulate the performance of a standard flow cytometry chamber. The flow of particles is confined to a region with a diameter of 5-10 µm near the central axis of a cytometry channel, which has a cross section of 300 × 110 µm. The HTD can operate at particle flow velocities of up to 8 m/s and reaches throughputs of up to 17 000 particles/s with high accuracy in the fluorescence detection. In the high-resolution device (HRD), the particles typically move at 0.5 mm/s in a 750 × 110 µm cytometry channel, and their stream is focused to a thin flow layer with a width of ∼250 µm and a thickness that becomes as small as 0.5 µm. The HRD permits low-light, highresolution imaging of particles and cells in a continuous flow under brightfield and fluorescent illumination using a standard microscope and CCD camera. The lateral resolution of the images, estimated as ∼0.5 µm, is comparable with the resolution obtained for the particles and cells immobilized on a substrate. EXPERIMENTAL SECTION Layout of the Microfluidic Devices. The high-throughput device (Figure 1) has four inlets (ports A, B, C, and E), one outlet, and channels of three different depths, 20, 110, and 200 µm. The 3D focusing element (Figure 1b, c) is composed of 20- and 200µm-deep channels, and the cytometry channel is 110 µm deep. (26) Hofmann, O.; Niedermann, P.; Manz, A. Lab Chip 2001, 1, 108-114. (27) Kauffmann, E.; Darnton, N. C.; Austin, R. H.; Batt, C.; Gerwert, K. Proc. Natl. Acad. Sci. U.S.A. 2001, 98, 6646-6649. (28) Yu, C. H.; Vykoukal, J.; Vykoukal, D. M.; Schwartz, J. A.; Shi, L.; Gascoyne, P. R. C. J. Microelectromech. Syst. 2005, 14, 480-487. (29) Yang, R.; Feeback, D. L.; Wang, W. J. Sens. Actuators, A 2005, 118, 259267. (30) Sundararajan, N.; Pio, M. S.; Lee, L. P.; Berlin, A. A. J. Microelectromech. Syst. 2004, 13, 559-567. (31) Takeuchi, S.; Garstecki, P.; Weibel, D. B.; Whitesides, G. M. Adv. Mater. 2005, 17, 1067-1072. (32) Simonnet, C.; Groisman, A. Appl. Phys. Lett. 2005, 87, Art. No. 114104.

Figure 1. High-throughput device. Arrows show direction of flow. Capital letters indicate various ports and channels connected to these ports. (a) Schematic drawing of the microchannel network; the dashed line box indicates the cytometry channel. Gray scale coding of the channel depths is shown below the drawing. Suspension of particles is injected into port B; flow focusing in the out-of-plane (z) direction is provided by the liquids injected into ports A (focusing from the top) and C (focusing from the bottom); flow focusing in the in-plane direction is provided by the liquid injected into port E. Port F is the device outlet. The 200-µm-deep channels connecting the inlets A, B, and C with the 3D focusing element are 150 µm wide. The resistances of the two 200-µm-deep channel lines connecting inlet E with the 3D focusing element are set by their narrowest segments, which are 200 µm wide and 6 mm long. (b) Micrograph of the 3D focusing element. The channels are 200 µm deep (darker boundaries) and 20 µm deep (fainter boundaries). The width × length dimensions of the 20-µmdeep channels are as follows: A, 120 µm × 150 µm; B, 40 µm × 1 mm (two channels); C, 60 µm × 200 µm (two channels). (c) Schematic diagram showing structure of the flow in the device, from the 3D focusing element to the cytometry channel. The liquid injected into port B appears in a dark color.

The high-resolution device (Figure 2) has four inlets (ports A, B, C, and E), two outlets (ports D and F), and channels of two depths, 8 and 110 µm. The 3D focusing element (Figure 2b, c) is composed of 8- and 110-µm-deep channels, and the cytometry channel is 110 µm deep. Fabrication of the Microfluidic Devices. The devices were cast of PDMS (Sylgard 184 by Dow Corning, Midland, MI) using UV-lithography-machined master molds. The master for the HTD was fabricated in a three-step procedure using two different formulations of a UV-curable epoxy SU8 (MicroChem, Newton, MA). First, a 20-µm layer of SU8 2015 was spin-coated onto a 4-in. silicon wafer, exposed to UV-light through a specially designed photomask (a laser-printed film with a resolution of 20 000 dpi), and baked on a hot plate to cure the SU8. In the second step, a 90-µm-thick layer of SU8 2100 was spin-coated on top of the patterned 20-µm-thick layer, a second photomask was aligned with respect to the existing cured SU8 pattern, and the wafer was again exposed to UV-light and baked on a hot plate. In the third step, this procedure was repeated with coating by an additional 90-µmthick layer of SU8 2100, exposure through yet another photomask, and another hot plate baking. The developed wafer was a master mold with a three-level relief. The master mold for the HRD was fabricated in a two-step procedure that involved coating a 4-in. wafer consecutively with 8- and ∼100-µm-thick layers of SU8 2005 and SU8 2100, respectively, and exposing it to UV light through two different photomasks.

Figure 2. High-resolution device. Arrows show direction of flow. Capital letters indicate various ports and channels connected to these ports. (a) Schematic drawing of the microchannel network; the dashed line box indicates the cytometry channel. Gray scale coding of the channel depths is shown below the drawing. The respective roles of the ports A-C, E, and F are the same as in Figure 1 (e.g., suspension of particles is injected into the inlet B). Port D serves as an auxiliary outlet to improve the flow focusing. The 110-µm-deep channels connecting the inlets and outlets with the 3D focusing element are mostly 120 µm wide. The width × length dimensions of the 8-µmdeep channels in the lines connecting different ports with the 3D focusing element are as follows: A, 120 µm × 400 µm; B, 40 µm × 2 mm (two channels); C, 40 µm × 400 µm (two channels); D, 25 µm × 160 µm (two channels); E, 80 µm × 300 µm (two channels). (b) Micrograph of the 3D focusing element. The 110-µm-deep channels have darker boundaries than the 8-µm-deep channels. (c) Schematic diagram showing structure of the flow in the device, from the 3D focusing element to the imaging region. The liquid injected from port B appears in a dark color.

To facilitate release of the PDMS replicas, the molds were silanized by exposure to a vapor of chlorotrimethylsilane (SigmaAldrich, St. Louis, MO) at room temperature for 90 s. The PDMS casts (∼5 mm thick) were cut into individual chips, which were trimmed to size, punched with luer stubs to make inlet and outlet holes, and bonded to no. 1.5 microscope cover glasses by overnight baking in an 80 °C oven. Flow Control. The flow in the microchannels was driven by setting differences in pressure between the inlets and outlets. The pressures were generated hydrostatically using vertical rails with precise rulers and sliding stages.33 The liquids used in the experiments were held in transparent plastic vessels of different sizes that all had a luer connector at the bottom and were normally open to the atmosphere at the top. Each vessel was attached to an individual stage and connected to a microfluidic device through a blunt luer needle, a long piece of Tygon tubing with an internal diameter of 0.76 mm (1 mm for tubing connected to ports E and F in the HTD) and a short piece of hypodermic steel tubing (inserted into a port of the device). The suspensions of particles and cells fed to ports B of both devices (Figures 1, 2) were kept in 1-mL plastic syringes (held upright). The rest of the liquids were kept in 60-mL syringes, except for the liquids fed to port E and drawn off from port F of the HTD (Figure 1) that were kept in wide containers (∼40 cm2 in cross section) to reduce their level variation during the experiments. The difference in liquid elevation (33) Groisman, A.; Enzelberger, M.; Quake, S. R. Science 2003, 300, 955-958.

Analytical Chemistry, Vol. 78, No. 16, August 15, 2006

5655

between the syringes (and containers) was measured with a precision of ∼0.5 mm, corresponding to 5 Pa in pressure. The differential pressure applied in the cytometry assays did not exceed ∼11 kPa, corresponding to a difference of 43 in. between the liquid levels in the vessels connected to the inlets and outlets. Chemicals, Particles, and Cells. We used fluoresceinconjugated dextran (FITCD) with a molecular mass of 2 MDa (Sigma, St. Louis, MO) as a low-diffusivity fluorescent marker for flow visualization. Nonfluorescent dextran with a lower molecular mass (Mw = 2.5‚105, by Polysciences, Warrington, PA) was used to prepare aqueous solutions with increased viscosity. There were three types of beads used in the study: plain 0.75-µm polystyrene beads (by Polysciences), 1.9-µm green fluorescent beads (Bangs Labs, Fishers, IN), and 2.5-µm green fluorescent beads of various fluorescence intensities (LinearFlow Green flow cytometry intensity calibration kit, by Invitrogen, Carlsbad, CA). Yeast (Saccharomyces cerevisiae) cells used in testing of the HRD were a kind gift from Prof. Randy Hampton (Division of Biological Sciences, UCSD). We used three strains of S. cerevisiae that had the same genotype (ade2-101, met2, lys2-801, his3∆200, ura3-52) but carried different plasmids. The “ER-labeled” strain had a Hmg2p-GFP expressing URA3 plasmid (pRH671), the “cytoplasm-labeled” strain had a soluble GFP expressing URA3 plasmid (pRH1742), and the “non-fluorescent” strain had an empty URA3 vector (pRH313). For each strain, a culture was grown overnight in YPD medium from a single colony. An aliquot of the suspension was then diluted in 3 mL of fresh YPD and incubated for an additional 3 h to yield log-phase growth. The cells were then gently spun down, and the supernatant was removed. Next, the cells were resuspended in deionized water and spun down again. This rinsing procedure was repeated twice to remove the background fluorescence coming from YPD. The cells were finally resuspended at a density of ∼109/mL in ∼1 mL of Percoll (suspension of colloidal silica particles supplied by MP Biomedicals, Irvine, CA) diluted with a 1.5 M solution of NaCl. Confocal Microscopy. Profiles of fluorescence across the cytometry channels in both the HTD and the HRD were obtained with a laser scanning confocal microscope (BioRad 1024) and a 60×/1.2 WI objective. The green fluorescence of FITCD was excited with the argon laser line at 488 nm. High-Throughput Detection Setup. The HTD was mounted on an inverted fluorescence microscope, Nikon Diaphot TMD, with removed fluorescence collimation optics and lamp house. The fluorescence illumination was derived from an air-cooled 15-mW argon ion laser (Spectra-Physics, Mountain View, CA) with a single-line emission at 488 nm. The laser beam was introduced through the epifluorescence optical train of the microscope and aligned along the central axis of the optical train using two adjustable mirrors. Epifluorescence illumination and light collection were performed with a Zeiss 63×/1.25 NA oil immersion objective, using a fluorescence filter cube with a dichroic mirror (495DRLP by Omega, Brattleboro, VT; with 50% transmission at 505 nm) and an emission filter (503AFLP by Omega, with a cutoff wavelength of 503 ( 4 nm). To expand the diameter of the laser beam in the focal plane of the objective, an auxiliary lens with a focal distance of 110 mm was placed immediately in front of the fluorescence filter compartment. The laser illumination spot (visualized with a CCD camera attached to the microscope and a 5656

Analytical Chemistry, Vol. 78, No. 16, August 15, 2006

thin fluorescent specimen) had a circular shape with a diameter ∼12 µm. The collected fluorescence light was relayed to a photomultiplier (Hamamatsu R1516), which was reverse-biased at ∼450 V, corresponding to an electron gain of ∼104. (This relatively low bias voltage was chosen to ensure linear amplification without saturation for all fluorescent particles used in the tests.) The PMT was connected to a homemade current-voltage converter providing a gain of 1 mV/µΑ and low-pass filtering the signal with a cutoff frequency of 5 MHz. The signal was digitized at fs ) 5 MS/s using a computer equipped with a NI PCI-6110 data acquisition device (National Instrument, Austin, TX) and a Matlab program. High-Resolution Imaging Setup. The HRD was mounted on a Nikon TE2000 inverted fluorescence microscope, equipped with the standard Nikon mechanical xy-stage. The stage was driven in the x-direction by a high-resolution linear actuator, Newport 850F, controlled by an ESP-100 driver (Newport, Irvine, CA) connected to a computer. The actuator was attached to the x-axis slide of the stage and spring-loaded. The cytometry channel of the HRD (Figure 2) was aligned parallel to the x-axis of the stage. During cytometry assays, the y-axis slide was clamped to prevent occasional motion along the y-axis during the x-axis translation. The images were taken with a Nikon 60×/1.2 WI objective and a Spot RT-SE18 camera (Diagnostic Instruments, Sterling Heights, MI) with a 1360 × 1024 pixel 2/3-in. CCD array and maximal frame rate of 9.7 s-1 (at 0.04-ms exposure). To enable fast switching between the brightfield and fluorescence illumination, the microscope was equipped with two electronic shutters (Uniblitz CS35S by Vincent Associates, Rochester, NY), which were synchronized with the camera. RESULTS High-Throughput Detection and Fluorescence Measurements. A central component of the microfluidic device for highthroughput flow cytometry (HTD) is the 3D hydrodynamic focusing element (Figure 1b, c).32 Flow-focusing in the plane of the device (in the y-direction) is achieved by squeezing the stream carrying particles between the streams emerging from the two channels, E. Flow focusing in the out-of-plane direction occurs in the region where two pairs of shallow channels (B and C in Figure 1) are connected to opposite sides of the central tall-and-narrow channel (A in Figure 1).32 Channel A has a depth of 200 µm and a width of only 24 µm, whereas channels B and C are both 20 µm deep and 60 µm wide. Because of the high ratio between the depth and width of channel A, the liquid flowing into it from the shallow channels B and C is effectively injected from the bottom and pushes the liquid in channel A toward the top (Figure 1b). The stream carrying particles, which emerges from channels B, is thus sandwiched in the z-direction between the streams from channel A and channels C (Figure 1c). The z-axis extension and position of the particle-carrying stream in the cytometry channel (Figure 1a, c) are controlled independently by pressure settings at ports A, B, and C. The dimensions of the channels in the flow focusing element were made about twice as large as those in ref 32 so as to make the HTD compatible with larger particles and cells. In addition, the flow resistance of channels connecting the 3D focusing element with the inlets was substantially reduced to reach high flow rates at moderate driving pressures. The interrogation region,

where flow cytometry assays were carried out, was in the central area of the cytometry channel (Figure 1a). This channel had length, width, and depth (dimensions along the x-, y-, and z-axes) measurements of lc ) 1.3 mm, wc ) 300 µm, and hc ) 110 µm, respectively. These dimensions were chosen to ensure uniform flow velocity within the focused stream and favorable imaging conditions, while limiting consumption of liquid and Reynolds number, Re. The stream carrying particles was focused near the axis of symmetry of the cytometry channel, which was parallel to the x-axis. The velocity profile of the stream was close to parabolic in the x-z plane and almost uniform in the x-y plane (assuming a fully developed laminar flow in the channel; see below). Therefore, for a typical diameter of the focused stream (