Anal. Chem. 2010, 82, 3183–3190
`
`High-Performance Single Cell Genetic Analysis
`Using Microfluidic Emulsion Generator Arrays
`
`Yong Zeng,† Richard Novak,‡ Joe Shuga,§ Martyn T. Smith,§ and Richard A. Mathies*,†,‡
`
`Department of Chemistry, UCSF/UC Berkeley Joint Bioengineering Graduate Group, and School of Public Health,
`University of California, Berkeley, California 94720
`
`High-throughput genetic and phenotypic analysis at the
`single cell level is critical to advance our understanding
`of the molecular mechanisms underlying cellular function
`and dysfunction. Here we describe a high-performance
`single cell genetic analysis (SCGA) technique that com-
`bines high-throughput microfluidic emulsion generation
`with single cell multiplex polymerase chain reaction
`(PCR). Microfabricated emulsion generator array (MEGA)
`devices containing 4, 32, and 96 channels are developed
`to confer a flexible capability of generating up to 3.4 ×
`106 nanoliter-volume droplets per hour. Hybrid glass-
`polydimethylsiloxane diaphragm micropumps inte-
`grated into the MEGA chips afford uniform droplet
`formation, controlled generation frequency, and effec-
`tive transportation and encapsulation of primer func-
`tionalized microbeads and cells. A multiplex single cell
`PCR method is developed to detect and quantify both
`wild type and mutant/pathogenic cells. In this method,
`microbeads functionalized with multiple forward prim-
`ers targeting specific genes from different cell types are
`used for solid-phase PCR in droplets. Following PCR,
`the droplets are lysed and the beads are pooled and
`rapidly analyzed by multicolor flow cytometry. Using
`Escherichia coli bacterial cells as a model, we show
`that this technique enables digital detection of patho-
`genic E. coli O157 cells in a high background of
`normal K12 cells, with a detection limit on the order
`of 1/105. This result demonstrates that multiplex
`SCGA is a promising tool for high-throughput quantita-
`tive digital analysis of genetic variation in complex
`populations.
`
`Traditional biological analyses probe large ensembles on the
`order of 103-106 cells, thereby revealing only the average
`genotypic and/or phenotypic characterization of the population.
`The advent of single cell analysis has revealed marked cellular
`heterogeneity in gene and protein expression,1-4 genetic/
`
`* Corresponding author. Department of Chemistry, MS 1460, University of
`California, Berkeley, CA, 94720. Phone: (510) 642-4192. Fax: (510) 642-3599.
`E-mail: ramathies@berkeley.edu.
`† Department of Chemistry.
`‡ UCSF/UC Berkeley Joint Bioengineering Graduate Group.
`§ School of Public Health.
`(1) Cai, L.; Friedman, N.; Xie, X. S. Nature 2006, 440, 358–362.
`(2) Elowitz, M. B.; Levine, A. J.; Siggia, E. D.; Swain, P. S. Science 2002, 297,
`1183–1186.
`
`10.1021/ac902683t  2010 American Chemical Society
`Published on Web 03/01/2010
`
`genomic alterations,5-7 and responsiveness to environmental and
`chemotherapeutic stimuli.8,9 A better understanding of cellular
`heterogeneity and the quantitative detection of rare mutants, such
`as circulating tumor cells present at an estimated frequency of 1
`per 106-107 nucleated blood cells,10 demands high-throughput
`single cell analysis techniques to both detect these very
`infrequent members of the population and to provide good
`statistical information regarding these stochastic events.
`Microfluidics offers unprecedented capabilities for precisely
`manipulating small volumes and for enhancing reaction speed
`and efficiency by increasing relative analyte concentration,
`reducing diffusion time, and facilitating large-scale integration
`and automation. These capabilities suggest that microfluidic
`systems will be very valuable for ultrahigh-throughput single
`cell analysis.11-14 Emulsion polymerase chain reaction (ePCR)
`provides another powerful tool for high-throughput genetic
`analysis because one can perform massively parallel single copy
`PCR reactions by partitioning statistically diluted targets (DNA
`or RNA) into small droplets dispersed in an oil phase.15,16
`Conventional ePCR has found many applications, including
`next-generation sequencing,17 detection of rare mutations,18-20
`and quantification of DNA methylation.21 However, ePCR also
`suffers from inherent limitations due to the use of mechanical
`agitation for emulsion generation, which produces high shear
`force and polydisperse droplet sizes. The transition to droplet
`
`(3) Toriello, N. M.; Douglas, E. S.; Thaitrong, N.; Hsiao, S. C.; Francis, M. B.;
`Bertozzi, C. R.; Mathies, R. A. Proc. Natl. Acad. Sci. U.S.A. 2008, 105,
`20173–20178.
`(4) King, K. R.; Wang, S. H.; Irimia, D.; Jayaraman, A.; Toner, M.; Yarmush,
`M. L. Lab Chip 2007, 7, 77–85.
`(5) Hanahan, D.; Weinberg, R. A. Cell 2000, 100, 57–70.
`(6) Klein, C. A.; Blankenstein, T. J. F.; Schmidt-Kittler, O.; Petronio, M.; Polzer,
`B.; Stoecklein, N. H.; Riethmuller, G. Lancet 2002, 360, 683–689.
`(7) Fuhrmann, C.; Schmidt-Kittler, O.; Stoecklein, N. H.; Petat-Dutter, K.; Vay,
`C.; Bockler, K.; Reinhardt, R.; Ragg, T.; Klein, C. A. Nucleic Acids Res. 2008,
`36, e39.
`(8) Batchelor, E.; Loewer, A.; Lahav, G. Nat. Rev. Cancer 2009, 9, 371–377.
`(9) Li, X. J.; Ling, V.; Li, P. C. H. Anal. Chem. 2008, 80, 4095–4102.
`(10) Pantel, K.; Brakenhoff, R. H.; Brandt, B. Nat. Rev. Cancer 2008, 8, 329–
`340.
`(11) Whitesides, G. M. Nature 2006, 442, 368–373.
`(12) Ottesen, E. A.; Hong, J. W.; Quake, S. R.; Leadbetter, J. R. Science 2006,
`314, 1464–1467.
`(13) Sims, C. E.; Allbritton, N. L. Lab Chip 2007, 7, 423–440.
`(14) Marcus, J. S.; Anderson, W. F.; Quake, S. R. Anal. Chem. 2006, 78, 3084–
`3089.
`(15) Nakano, M.; Komatsu, J.; Matsuura, S.; Takashima, K.; Katsura, S.; Mizuno,
`A. J. Biotechnol. 2003, 102, 117–124.
`(16) Dressman, D.; Yan, H.; Traverso, G.; Kinzler, K. W.; Vogelstein, B. Proc.
`Natl. Acad. Sci. U.S.A. 2003, 100, 8817–8822.
`
`Analytical Chemistry, Vol. 82, No. 8, April 15, 2010
`
`3183
`
`1
`
`

`
`microfluidics overcomes these limitations and has substantially
`improved the performance of emulsion-based bioassays.22-27
`Microfluidic droplet generation technology enables the produc-
`tion of monodisperse droplets with precise control over the droplet
`size while maintaining high throughput. This capability not only
`enables uniform efficiency of enzymatic reactions or affinity
`binding23 but also permits digital quantification of the absolute
`number of targets present in the initial sample.26,28 In addition,
`microfluidic encapsulation significantly reduces the damage to
`cells caused by vigorous mechanical agitation. Microfluidic opera-
`tion is normally performed with a shear stress lower than 10 dyn/
`cm2, while vigorous mechanical agitation can potentially gen-
`erate much higher shear stress that can disrupt cells.29 Our
`previous work indicated greater than 90% cell viability following
`microfluidic emulsification versus less than 80% for other
`encapsulation methods.25,30 Integrated microfluidic circuits also
`allow programmable in situ manipulation of droplets, such as
`droplet steering, trapping,31 and fusion,32 leading to a much
`including real time PCR,23
`broader spectrum of applications,
`protein expression studies,27 and drug screening.33
`
`(17) Margulies, M.; Egholm, M.; Altman, W. E.; Attiya, S.; Bader, J. S.; Bemben,
`L. A.; Berka, J.; Braverman, M. S.; Chen, Y. J.; Chen, Z. T.; Dewell, S. B.;
`Du, L.; Fierro, J. M.; Gomes, X. V.; Godwin, B. C.; He, W.; Helgesen, S.;
`Ho, C. H.; Irzyk, G. P.; Jando, S. C.; Alenquer, M. L. I.; Jarvie, T. P.; Jirage,
`K. B.; Kim, J. B.; Knight, J. R.; Lanza, J. R.; Leamon, J. H.; Lefkowitz, S. M.;
`Lei, M.; Li, J.; Lohman, K. L.; Lu, H.; Makhijani, V. B.; McDade, K. E.;
`McKenna, M. P.; Myers, E. W.; Nickerson, E.; Nobile, J. R.; Plant, R.; Puc,
`B. P.; Ronan, M. T.; Roth, G. T.; Sarkis, G. J.; Simons, J. F.; Simpson, J. W.;
`Srinivasan, M.; Tartaro, K. R.; Tomasz, A.; Vogt, K. A.; Volkmer, G. A.;
`Wang, S. H.; Wang, Y.; Weiner, M. P.; Yu, P. G.; Begley, R. F.; Rothberg,
`J. M. Nature 2005, 437, 376–380.
`(18) Diehl, F.; Li, M.; Dressman, D.; He, Y. P.; Shen, D.; Szabo, S.; Diaz, L. A.;
`Goodman, S. N.; David, K. A.; Juhl, H.; Kinzler, K. W.; Vogelstein, B. Proc.
`Natl. Acad. Sci. U.S.A. 2005, 102, 16368–16373.
`(19) Li, M.; Diehl, F.; Dressman, D.; Vogelstein, B.; Kinzler, K. W. Nat. Methods
`2006, 3, 95–97.
`(20) Diehl, F.; Schmidt, K.; Choti, M. A.; Romans, K.; Goodman, S.; Li, M.;
`Thornton, K.; Agrawal, N.; Sokoll, L.; Szabo, S. A.; Kinzler, K. W.; Vogelstein,
`B.; Diaz, L. A. Nat. Med. 2008, 14, 985–990.
`(21) Li, M.; Chen, W. D.; Papadopoulos, N.; Goodman, S. N.; Bjerregaard, N. C.;
`Laurberg, S.; Levin, B.; Juhl, H.; Arber, N.; Moinova, H.; Durkee, K.;
`Schmidt, K.; He, Y. P.; Diehl, F.; Velculescu, V. E.; Zhou, S. B.; Diaz, L. A.;
`Kinzler, K. W.; Markowitz, S. D.; Vogelstein, B. Nat. Biotechnol. 2009,
`27, 858–U118.
`(22) Teh, S. Y.; Lin, R.; Hung, L. H.; Lee, A. P. Lab Chip 2008, 8, 198–220.
`(23) Beer, N. R.; Hindson, B. J.; Wheeler, E. K.; Hall, S. B.; Rose, K. A.; Kennedy,
`I. M.; Colston, B. W. Anal. Chem. 2007, 79, 8471–8475.
`(24) Beer, N. R.; Wheeler, E. K.; Lee-Houghton, L.; Watkins, N.; Nasarabadi,
`S.; Hebert, N.; Leung, P.; Arnold, D. W.; Bailey, C. G.; Colston, B. W. Anal.
`Chem. 2008, 80, 1854–1858.
`(25) Kumaresan, P.; Yang, C. J.; Cronier, S. A.; Blazei, R. G.; Mathies, R. A.
`Anal. Chem. 2008, 80, 3522–3529.
`(26) Mazutis, L.; Araghi, A. F.; Miller, O. J.; Baret, J. C.; Frenz, L.; Janoshazi,
`A.; Taly, V.; Miller, B. J.; Hutchison, J. B.; Link, D.; Griffiths, A. D.;
`Ryckelynck, M. Anal. Chem. 2009, 81, 4813–4821.
`(27) Huebner, A.; Srisa-Art, M.; Holt, D.; Abell, C.; Hollfelder, F.; Demello, A. J.;
`Edel, J. B. Chem. Commun. 2007, 1218–1220.
`(28) Kiss, M. M.; Ortoleva-Donnelly, L.; Beer, N. R.; Warner, J.; Bailey, C. G.;
`Colston, B. W.; Rothberg, J. M.; Link, D. R.; Leamon, J. H. Anal. Chem.
`2008, 80, 8975–8981.
`(29) Roubos, J. A.; Krabben, P. M; Luiten, R. G.; Verbruggen, H. B.; Heijnen,
`J. J. Biotechnol. Prog. 2008, 17, 336–347.
`(30) Yu, S.-H.; Buchholz, R.; Kim, S.-K. Biotechnol. Tech. 1999, 13, 609–614.
`(31) Huebner, A.; Bratton, D.; Whyte, G.; Yang, M.; deMello, A. J.; Abell, C.;
`Hollfelder, F. Lab Chip 2009, 9, 692–698.
`(32) Mazutis, L.; Baret, J. C.; Griffiths, A. D. Lab Chip 2009, 9, 2665–2672.
`(33) Brouzes, E.; Medkova, M.; Savenelli, N.; Marran, D.; Twardowski, M.;
`Hutchison, J. B.; Rothberg, J. M.; Link, D. R.; Perrimon, N.; Samuels, M. L.
`Proc. Natl. Acad. Sci. U.S.A. 2009, 106, 14195–14200.
`
`3184 Analytical Chemistry, Vol. 82, No. 8, April 15, 2010
`
`Figure 1. Multiplex single cell/copy genetic analysis (SCGA).
`Statistically dilute beads and templates are encapsulated into uniform
`nanoliter volume PCR-mix-in-oil droplets, which are then thermally
`cycled for PCR amplification. Each bead is functionalized with forward
`primers for all targets. PCR mix contains reverse primers each labeled
`with a unique fluorescent dye. Each bead in a droplet containing only
`a single target will carry one type of fluorescent amplicon after PCR,
`while a bead compartmentalized with two different templates or cells
`will be linked with multiple dye-labeled products. Following emulsion
`PCR, the droplets are broken and the beads are recovered and
`analyzed by flow cytometry for multicolor detection of the bound
`amplicons.
`
`Our previous work established a high-throughput single copy
`genetic amplification (SCGA) technique based on the use of a
`hybrid glass-poly(dimethyl siloxane) (PDMS)-glass microdroplet
`generator (µDG) integrated with a three-valve diaphragm micro-
`pump for effective transport and encapsulation of large microbeads
`and cells into uniform nanoliter-volume droplets.25 To enable
`detection of extremely low-frequency events in a vast population,
`herein we scale up the single-channel µDG to 4, 32, and 96-channel
`microfluidic emulsion generator array (MEGA) systems, further
`increasing the generation throughput up to 3.4 × 106 droplets per
`hour. In these MEGA systems, on-chip micropumps provide
`sufficient power to drive multiple droplet generators in parallel
`while the symmetrically designed microfluidic networks ensure
`even fluidic transport which is crucial for uniform droplet
`encapsulation. A new compact micropump composed of three
`coaxial ring-shaped valves is designed to enable the imple-
`mentation of a 96-channel MEGA on a 4 in. wafer. This
`improvement simplifies device fabrication and operation and
`substantially reduces the dead volume.
`The integration of multiplex PCR with high-throughput emul-
`sion generation is valuable for the analysis of genomic deletion,
`forensic genotyping, mutation and polymorphism analysis, and
`identification of pathogens.34 To detect and quantify both normal
`and mutant/pathogenic cells, herein we describe a multiplex
`single cell PCR approach, as illustrated in Figure 1, which allows
`efficient high-throughput PCR amplification of multiple target
`genes specific to different cell types. In this process, primer-linked
`beads and cells are diluted in the PCR mix such that isolated
`
`(34) Markoulatos, P.; Siafakas, N.; Moncany, M. J. Clin. Lab. Anal. 2002, 16,
`47–51.
`
`2
`
`

`
`Figure 2. Microfluidic emulsion generator array (MEGA) devices. (A) Layout of a glass/PDMS/glass hybrid four-channel MEGA device with a
`pneumatically controlled three-valve micropump integrated to drive four nozzles for droplet generation. (B) Design of a 32-channel MEGA device
`using an array of eight identical micropumps to operate 32 nozzles simultaneously. Two adjacent emulsion channels are combined to increase
`device density. (C) Layout of 96-channel MEGA on a 4 in. wafer composed of a single ring pump and 96 droplet generators. Inset: close-ups
`of a single repeating unit composed of four T-shaped nozzles (left) and the pump structure schematically showing three pairs of coaxial ring-
`shaped valves and displacement trenches (right). (D) Exploded view of the complete four-layer 96-channel MEGA device and the plexiglass
`assembly module used to infuse oil and to collect the generated emulsion.
`
`individual beads or cells are encapsulated into individual uniform
`reaction droplets dispersed in the carrier oil. Statistically, a small
`fraction of droplets will contain both one bead and one or more
`cells. Every bead is functionalized with forward primers for all
`targets, and the PCR mix contains reverse primers each labeled
`with a unique fluorescent dye. Thousands of such droplets,
`generated by MEGA chips within minutes, are collected in
`standard PCR tubes and thermally cycled in parallel. Each bead
`in a droplet containing only a single cell will carry one type of
`dye-labeled double-stranded amplicons after PCR, while a bead
`compartmentalized with different types of target cells will be
`labeled with multiple dyes. Post-PCR beads are recovered from
`the emulsion and rapidly analyzed by flow cytometry for multicolor
`fluorescent digital counting of each single cell detection event.
`To demonstrate the utility of the multiplex single cell/copy
`genetic analysis (SCGA) technique for high-throughput single cell
`analysis of complex sample mixtures, we perform the detection
`and quantification of a major foodborne bacterial pathogen,
`Escherichia coli O157:H7, which alone causes an estimated 73 000
`infections and 61 deaths annually in the United States.35,36 High-
`throughput digital multiplex SCGA allows us to detect and quantify
`pathogenic E. coli O157:H7 cells in a background of normal E.
`coli K12 cells with a detection limit on the order of 1/105 within
`a 30 min microdroplet generation time. Such sensitivity is
`critical for many applications, such as food safety, where
`microbial pathogen detection needs to meet a zero tolerance
`
`policy for many foods.37 The results presented here also
`suggest
`that our technique has the potential
`for high-
`throughput single cell genotyping and quantitative detection
`of rare mutations in circulating fluids.
`
`MATERIALS AND METHODS
`MEGA Fabrication and Preparation. The four-layer MEGA
`chip shown schematically in Figure 2 is constructed from three
`100 mm-diameter glass wafers and a thin PDMS membrane
`following a process similar to that described by Grover et al.38
`The glass channels were coated with octadecyltrichlorosilane
`(OTS, Sigma-Aldrich) to render the surface hydrophobic. The
`devices were assembled with a custom plexiglass manifold (Figure
`2D) which provides fluidic connections for oil
`infusion and
`emulsion collection.
`Cell Preparation. All cell culture and preparation were
`performed in a class II biosafety cabinet (Labconco, Kansas City,
`MO) to avoid contamination. Two types of cells, E. coli K12 (ATCC
`no. 700926) and nontoxigenic E. coli O157 (ATCC 700728), were
`grown separately in tryptic soy broth medium (TSB, Hardy
`Diagnostics K131, CA) overnight at 37 °C. E. coli K12 cells were
`transformed with a 3.9-kb pCR 2.1-TOPO vector (Invitrogen,
`Carlsbad, CA) to confer ampicillin resistance and grown with 1
`mg/mL ampicillin added to the media. E. coli O157 was untreated
`and frequently tested for contamination by PCR. Cells were
`washed three times in 1× PBS, and the final cell density was
`determined by using a hemacytometer.
`
`(35) Mead, P. S.; Slutsker, L.; Dietz, V.; McCaig, L. F.; Bresee, J. S.; Shapiro,
`C.; Griffin, P. M.; Tauxe, R. V. Emerg. Infect. Dis. 1999, 5, 607–625.
`(36) Nugen, S. R.; Baeumner, A. J. Anal. Bioanal. Chem. 2008, 391, 451–454.
`
`(37) Batt, C. A. Science 2007, 316, 1579–1580.
`(38) Grover, W. H.; Skelley, A. M.; Liu, C. N.; Lagally, E. T.; Mathies, R. A.
`Sens. Actuators, B: Chem. 2003, 89, 315–323.
`
`Analytical Chemistry, Vol. 82, No. 8, April 15, 2010
`
`3185
`
`3
`
`

`
`Bead and PCR Preparation. All samples were handled in a
`UV-treated laminar flow hood (UVP, Upland, CA). Primers specific
`to the KI #128 island on the K12 genome and the OI #43 island
`on the O157 genome were designed to prevent cross-amplification
`between strains.39 Reverse primers were labeled with 6-carboxy-
`fluorescein (6-FAM) or Cy5 dye on the 5′ end. The 5′ amine
`modified forward primers were linked to agarose beads (34 µm
`mean diameter, Amersham Biosciences, NJ) via amine-NHS
`conjugation chemistry.25 The coupling reaction was performed
`at a ratio of ∼1.5 µmol of oligos per gram of beads for the pUC18
`target. For E. coli cells, equimolar forward primers were used at
`a concentration of ∼0.3 µmol of oligo per gram beads. PCR mixes
`contain forward primer functionalized beads (40 beads/µL) and
`varied amounts of freshly prepared template DNA or cells.
`Device Operation and PCR Quantification. A homemade
`pneumatic system controlled by Labview was used to operate the
`on-chip micropump. The microchip was prerun with a coating
`solution to minimize nonspecific adsorption on glass, PDMS, and
`tubing surfaces. PCR-mix-in-oil droplets were then generated at
`5-8 Hz and collected into individual PCR tubes containing 40 µL
`of microfine solution17 for subsequent thermal cycling. After
`emulsion PCR, beads were recovered from the droplets and
`analyzed using a multicolor flow cytometer (FC-500, Beckman
`Coulter). (See the Supporting Information for detailed methods.)
`
`RESULTS
`MEGA Design. The four-channel MEGA illustrated in Figure
`2A is a multilayer device consisting of a bonded glass-glass
`microfluidic chip, a PDMS membrane, and a microfabricated
`manifold wafer. The glass-glass fluidic chip contains a microfab-
`ricated pattern of valve seats exposed on the top surface (blue),
`which is connected through a “via” hole to an all-glass microfluidic
`network enclosed between two thermally bonded wafers. The
`assembly is then completed by contact bonding a manifold wafer
`(magenta) onto the fluidic chip with a PDMS membrane to form
`the micropump structure. The microfluidic network consists of
`the symmetrically bifurcated channels for the aqueous phase
`(black) and oil phase (red), which form four parallel crosses or
`droplet nozzles at the junctions. The 4-channel array defines a
`basic unit that is used to build up multiplexed MEGAs, such as
`the 32-channel MEGA with eight such arrays integrated onto a 4
`in. wafer (Figure 2B). Further multiplexing of the MEGA on a 4
`in. wafer is restricted by the number of individual pumps that can
`be symmetrically arranged in a circle. To achieve a higher density,
`we designed a new ring micropump composed of three pairs of
`coaxial ring-shaped valve seats connected by offset channels, as
`well as corresponding circular displacement trenches (Figure 2C,
`top right). This compact micropump, along with the T-shaped
`nozzle design (Figure 2C, top left) enables the implementation of
`a 96-channel MEGA on a 4 in. wafer (Figure 2C, bottom). In 32-
`and 96-channel MEGAs, oil channels are connected to the oil inlet
`holes drilled on the bottom substrate of the device.
`A manifold module is employed to support MEGA devices, as
`sketched in Figure 2D, providing oil infusion and routing gener-
`ated droplets into PCR tubes for thermal cycling. The bottom part
`of the manifold is designed to form a circular oil reservoir when
`
`(39) Beyor, N.; Yi, L. N.; Seo, T. S.; Mathies, R. A. Anal. Chem. 2009, 81, 3523–
`3528.
`
`3186 Analytical Chemistry, Vol. 82, No. 8, April 15, 2010
`
`Figure 3. Characterization of droplet generation and PCR amplifica-
`tion using MEGAs. (A) 32-channel MEGA chip. Left: photo of the
`cross-shaped nozzles each generating ∼4 nL droplets at 5.6 Hz (6.4
`× 105 droplets per hour). Middle: size distribution of uniform droplets
`collected from eight nozzles in a device. RSD: relative standard
`deviation. Right: a representative flow cytometric analysis of 624 bp
`amplicons on beads amplified from pUC18 template (1 copy/droplet)
`in ∼3 nL droplets generated from four nozzles in a device. (B) 96-
`channel MEGA. Left: image of droplet production at the T-shaped
`nozzles with a total throughput of 2.4 × 106 droplets per hour. Middle:
`size distribution of uniform ∼2 nL droplets collected from 16 nozzles
`in a device. Right: a representative flow cytometric histogram of beads
`carrying the FAM labeled PCR product from E. coli K12 at 0.2 cells
`per droplet (cpd) in 2.5 nL droplets. For all tests of droplet generation,
`the mock PCR mix containing ∼100 beads per µL was used.
`sealed against the MEGA chip. The oil pressure in the circular
`reservoir is uniformly distributed across the area on the bottom
`side of the chip, evenly infusing oil into channels via the symmetric
`inlet holes. This design minimizes the number of syringe pumps
`and tubing connections required for oil infusion.
`Droplet PCR Performance of MEGAs. The symmetric
`design of the MEGA assures uniform fluidic transport across the
`array, leading to the generation of monodisperse droplets from
`all channels. A microphotograph in Figure 3A (left) demonstrates
`the generation of uniform 4 nL droplets by flow focusing at the
`cross-injectors of a 32-channel MEGA device. The droplets were
`collected from eight nozzles for the size measurement revealing
`a mean diameter of 198 µm with a size deviation of only 2.5%
`(Figure 3A, middle). Because of the pulsatile nature of the on-
`chip diaphragm pump, the droplet formation rate corresponds
`precisely with the pumping frequency of 5.6 Hz. At this speed,
`the 32-channel device offers a total throughput of 6.4 × 105 droplets
`per hour (dph).
`To characterize the performance of bead-based droplet PCR
`using the 32-channel MEGA, pUC18 DNA molecules were
`compartmentalized into ∼3 nL droplets at 1 copy/droplet as the
`templates, from which a 624 bp product was amplified. The post-
`PCR beads recovered from each PCR tube were randomly mixed
`prior to flow cytometric analysis. The result obtained from four
`droplet generators is presented in Figure 3A (right), showing that
`60.5% of the bead population (459 events) gains FAM signal due
`to the bead-bound amplicons, in line with the value expected from
`Poisson statistics (63.2%). The fluorescence intensity of the positive
`beads indicates an averaged PCR yield of ∼150 amol of DNA
`product per bead, consistent with the yield obtained from our
`previous single-channel device.25 The results from other channels
`
`4
`
`

`
`confirm that the 32-channel MEGA offers constant digital emulsion
`PCR performance as a result of
`the uniformity of droplet
`compartmentalization.
`Incorporation of a compact ring-shaped micropump in the 96-
`channel MEGA maintains the generation performance while
`increasing throughput. Uniform droplets can be formed from a
`mock PCR mix containing agarose beads by flow shearing at the
`T-shaped injectors (Figure 3B, left, and the video in the Supporting
`Information). The mean diameter of the droplets collected from
`16 nozzles via 4 randomly chosen outlets was determined to be
`162 µm (2.2 nL) with a deviation of only 3.8% (Figure 3B, middle).
`Furthermore, such size uniformity is preserved across a range of
`droplet volumes (1-5 nL; RSD < 5%). When the on-chip pump is
`operated at 7 Hz, the device produces up to 2.4 × 106 dph. To
`assess the encapsulation performance of the 96-channel MEGA,
`single cell emulsion PCR was performed targeting the KI #128
`island on the K12 genome. Bacterial E. coli K12 cells and the
`forward primer linked beads were introduced at a statistical
`dilution of 0.2 cells and 0.1 beads per 2.5 nL droplet, respec-
`tively. After PCR and isolation, flow cytometric analysis (Figure
`3B, right) shows that 15.1% of the total bead population (4423
`beads) is strongly fluorescent, corresponding well to a theoretical
`value of 18.1% predicted by the Poisson distribution. This good
`agreement in yield indicates that successful single copy genetic
`amplification resulted from uniformly distributed cells and beads.
`Pathogen Detection Using Multiplex Single Cell Genetic
`Analysis (SCGA). Multiplex SCGA is demonstrated here by
`detecting pathogenic E. coli O157 cells in a background of normal
`E. coli K12 cells. Unique genes on the K12 genome (KI #128
`island) and on the O157 genome (OI #43 island) are targeted by
`primers labeled with different fluorophores so that these two
`strains can be identified. For all pathogen detection, ∼2.5 nL
`droplets were used and beads were introduced at ∼0.1 beads per
`droplet (bpd). A four-channel MEGA device was operated for ∼18
`min to obtain ∼3000 beads for flow cytometric analysis. A mixed
`E. coli bacterial sample containing 50% O157 cells was first
`analyzed at an average cell concentration (Cavg) of 0.2 cells per
`droplet (cpd) using a four-channel MEGA device. As seen in
`Figure 4A, the flow cytometry profile shows four distinct bead
`populations: 212 FAM positive beads due to the amplicons from
`K12 (7.93%, green), 198 Cy5 positive beads for O157 (7.40%, red),
`45 double positive beads due to coexistence of both cell types in
`a single droplet (1.68%, orange), and 2220 negative beads (82.99%,
`blue). The O157 cell ratio (O157 positive beads/total positive
`beads) is then determined to be 0.48, in good accordance with
`the input O157 cell fraction of 0.5. Figure 4B presents the detection
`of a lower input O157 ratio of 1/10 at Cavg ) 0.07 cpd: 5.20% of
`the beads are positive for K12, 0.66% are positive for O157, 0%
`are double positive beads. The experimental O157/K12 ratio
`of 1.1/10 is consistent with that expected.
`Improving Throughput and Sensitivity. The multiplex SCGA
`process discussed above uses droplets/beads inefficiently because
`cells are highly dilute so that most droplets are empty. The
`detection of
`low-frequency genetic variations requires high
`analysis throughput in order to obtain the statistically significant
`population for the target. One way to increase the process
`efficiency is to perform multiplex SCGA at elevated cell density,
`while still keeping the cells of interest statistically dilute. This is
`
`Figure 4. High-throughput digital multiplex detection of E. coli O157
`in a background of E. coli K12 with various input ratios and average cell
`concentrations (Cavg) using MEGAs. (A, B) With Cavg e 0.2 cpd, flow
`cytometry profiles of beads show four distinct populations: negative
`(blue), FAM positive beads specific for K12 (green), Cy5 positive beads
`specific for O157 (red), and double positive beads for both cells (orange).
`Small populations are expanded along the event axis for better visualiza-
`tion, as indicated. The gray regions mark the population gating. The
`measured O157 ratios (O157 positive beads/total positive beads) are
`(A) 0.48 (expected, 0.5) and (B) 0.11 (expected, 0.1). (C, D) When Cavg
`g 10 cpd while keeping O157 cells at 0.01 cpd, double positive beads
`quantify O157 since a single O157 cell is cocompartmentalized into a
`droplet with multiple K12 cells. In this case, the ratios of double positive
`beads divided by Cavg give the measured O157 ratios: (C) 0.92/103
`(expected, 1/103) and (D) 0.98/104 (expected, 1/104). Up to 3000 beads
`can be analyzed using a four-channel device for ∼25 min run time.
`Increasing Cavg to 100 cpd reduces the number of beads required,
`improving the detection sensitivity to 1/104 without excessively extending
`the run time. (E, F) O157 detection using a 96-channel MEGA shows
`the measured O157 ratios consistent with the inputs: (E) 0.94/103 vs
`1/103 and (F) 0.85/104 vs 1/104. Up to 104 events were processed within
`5 min run time. The capability offered by 96-channel MEGA (up to 3.4
`× 106 droplets per hour) greatly increased analysis throughput and
`decreased processing time necessary for detecting a statistically sig-
`nificant population of a low-frequency sample. Bead concentration was
`0.1 bpd for all cases.
`
`realistic because the target pathogen or mutant cells are typically
`present at very low relative concentrations compared to normal
`
`Analytical Chemistry, Vol. 82, No. 8, April 15, 2010
`
`3187
`
`5
`
`

`
`cells. Figure 4C demonstrates the use of a four-channel MEGA
`device to detect E. coli O157 cells at a frequency of 1/103. With
`an increase of Cavg from 0.2 cpd (Figure 4A) to 10 cpd, the
`effective density of O157 cells is raised from 0.0005 to 0.01 cpd,
`reducing the number of droplets/beads that must be processed
`by 50-fold. In this case, all beads should be FAM fluorescent due
`to coencapsulated K12 cells, and the presence of O157 cells in
`the droplets will modify only a fraction of beads with the Cy5 dye.
`As expected, a vast majority of beads (1846 out of 1961 events,
`96.13%) are FAM positive and a small fraction of double positive
`beads (102 events, 0.88%) are observed. The measured O157 ratio
`is determined to be 0.92/103 (the ratio of double positive beads
`over all positive beads divided by Cavg), close to the input, 1/103.
`Multiplex SCGA preserves the quantitative performance for
`pathogen detection even when the average concentration is up
`to 100 E. coli cells per 2.5 nL droplet. As seen in Figure 4D, an
`experiment carried out at Cavg ) 100 cpd records 11 O157
`positive events in a total of 1167 events, giving an output of
`0.98/104 in response to the input O157 ratio of 1/104.
`When Cavg is higher than 100 cells per 2.5 nL droplet,
`nonspecific amplification was found to be significant in the
`multiplex PCR assay, causing false positive scores for O157.
`To further improve the detection sensitivity, therefore, it is
`necessary to use highly multiplexed MEGA devices for high-
`throughput droplet generation. A 96-channel MEGA device was
`evaluated for multiplex PCR detection of E. coli O157 cells,
`where the device was operated at ∼7 Hz, allowing us to
`encapsulate up to 104 cells within a 5 min microdroplet
`generation time. As shown in parts E and F of Figure 4, the
`experiments at input ratios of 1/103 (10 cpd) and 1/104 (100 cpd)
`detect 88 and 21 O157 positive events, giving output fractions
`of 0.94/103 and 0.85/104, respectively. A small number of
`negative events were observed in Figure 4C-F, which may be
`due to cell debris aggregates or unlysed cells. We have optimized
`the bead cleaning protocol for the 96-channel trials to remove cell
`debris and other interfering species. The high percentage of
`positive events (>99%) in Figure 4E,F results from the combination
`of high-efficiency droplet PCR and improved bead cleanup.
`To verify that the observed performance is the result of digital
`quantification of each strain, we compare the percentage of
`positive beads obtained with various input ratios and average cell
`concentrations to that predicted by the Poisson distribution
`(Figure 5). The multiplex detection is seen to follow Poisson
`statistics even when individual O157 cells were detected within a
`high background of 100 cells per droplet. The good correspon-
`dence indicates successful single cell emulsion PCR, which allows
`digital quantification of
`the absolute cell concentration. For
`instance, with Cavg ) 100 cpd and O157 cells diluted to 1/104 in
`a K12 background, the O157 cell density is 0.01 cpd and the
`detection resulted in 0.91% ± 0.04% double positive beads,
`consistent with the predicted ratio of 0.995%. Because of the
`presence of negative events, the average percentage of O157
`positive beads is corrected to be 0.93% ± 0.05% (double positive
`beads divided by total positive beads), f