High Yield Patterning of Single Cells from Extremely Small

Feb 18, 2013 - A high-patterning yield and low cell loss rate were demonstrated ... Travis W. Murphy , Qiang Zhang , Lynette B. Naler , Sai Ma , Chang...
0 downloads 0 Views 4MB Size
Article pubs.acs.org/ac

High Yield Patterning of Single Cells from Extremely Small Populations Andrea Faenza,*,† Massimo Bocchi,†,‡ Enri Duqi,† Luca Giulianelli,† Nicola Pecorari,† Laura Rambelli,† and Roberto Guerrieri† †

ARCES-University of Bologna, Via Toffano 2, 40125 Bologna, Italy MindSeeds Laboratories s.r.l., Via Fondazza 53, 40125 Bologna, Italy



S Supporting Information *

ABSTRACT: Many biological assays require the ability to isolate and process single cells. Some research fields, such as the characterization of rare cells, the in vitro processing of stem cells, and the study of early stage cell differentiation, call for the additional and typically unmet ability to work with extremely low-count cell populations. In all these cases, efficient single-cell handling must be matched with the ability to work on a limited number of cells with a low cell loss rate. In this paper, we present a platform combining flow-through processing with deterministic (nonstatistical) patterning of cells coming from extremely small cell populations. We describe here modules using dielectrophoresis to control the position of cells flowing in microchannels and to pattern them in open microwells where cells were further analyzed. K562 cells continuously flowing at a speed of up to 100 μm/s were tridimensionally focused, aligned, and patterned inside microwells. A high-patterning yield and low cell loss rate were demonstrated experimentally: 15uL drops, containing an average of 15 cells, were transferred to the microchannel with an 83% yield, and cells were then patterned into microwells with a 100% yield. The deterministic patterning of cells was demonstrated both by isolating single cells in microwells and by creating clusters composed of a predetermined number of cells. Cell proliferation was assessed by easily recovering cells from open microwells, and a growth rate comparable to the control was obtained.

T

processing, analysis, and efficient patterning of single cells within low-count cell populations are features missing from most of the existing platforms where, by contrast, statistical patterning is a commonly adopted approach. In these platforms, only a fraction of cells are singly isolated and large arrays are typically implemented in order to obtain the desired number of single-cell events.3,17 The ability to combine nonstatistical single-cell patterning with a low cell loss rate in the macro-to-micro transfer would enable one to perform complex analyses, even on cells coming from an extremely low-count population or to single out daughter cells for early stage lineage analysis. However, a solution integrating high transfer yield with deterministic single-cell patterning, processing, and live single-cell recovery is still lacking. Microwells are convenient microstructures where cells can be analyzed and processed.10,11,17,18 Recently, we presented a platform19 based on an array of open microwells where cells are trapped, processed, and imaged at the air−fluid interface on the bottom side of an open microwell and then easily recovered

he ability to analyze cells at a single-cell level is important when investigating cell heterogeneity and holds the promise of improving the outcome of many biological and medical assays. Application areas of single-cell analysis range from clinical diagnosis to immunology, drug discovery, stem cell analysis, rare cell isolation, and cancer treatment.1−6 In addition to fluorescence-activated cell sorting (FACS) technology, several solutions have been proposed to isolate and analyze single cells out of large populations by hydro-dynamic, dielectrophoretic, acoustic, or mechanical techniques.7−9 Most of these solutions rely on cell trapping/patterning which is an essential step when it comes to performing complex analyses and manipulations, such as monitoring cell growth and differentiation,10 analyzing dynamic cell response to different kinds of stimuli such as the ones produced by other cells or by interaction with molecules,11−13 and processing cells for PCR and gene expression analysis.4 Various applications call for the additional ability to isolate and process single cells starting from extremely low-count cell populations like, for example, (i) processing of rare cells separated by FACS,14 (ii) in vitro processing of stem cells,15 or (iii) early stage cell lineage tracing.16 When each cell counts, the yield becomes crucial and efficient single-cell handling must be matched with an extremely low cell loss rate. However, © 2013 American Chemical Society

Received: January 22, 2013 Accepted: February 17, 2013 Published: February 18, 2013 3446

dx.doi.org/10.1021/ac400230d | Anal. Chem. 2013, 85, 3446−3453

Analytical Chemistry

Article

Figure 1. Diagram of the cell isolation method. (a) A droplet containing a single cell is placed on a Petri dish lid and a hanging drop is created by overturning the lid. (b) Cells are optically counted. (c) Microchannels are filled with a buffer not containing cells. (d) A droplet is deposited over the inlet port of a microchannel. (e) Cells are withdrawn into the microchannel and manipulated by the DEP field generated by electrodes arranged in the microchannel. (f) Diagram of a microchannel section with three-dimensional (3D) focusing electrodes TR1, TR2, C1, and C2 used to push cells toward the microchannel floor and align them along a predefined trajectory by means of the DEP force. Cells can be loaded into the microwell by setting electrodes TR3, TR4, C3, C4, and VE1 in the Load mode. On setting the Forward mode, cells are forwarded to the following microwells.

be precisely counted by optical inspection and compared to the number of cells actually transferred into microwells. We describe and characterize here both the procedure for loading cells into our device and the modules for controlled cell sorting and patterning based on DEP and imaging. We compare simulations with experiments to characterize each module and validate it with K562 cancer cells. Finally, we present the transfer yield and patterning results of extremely low-count cell populations and compare overall performance with existing solutions.

onto standard microtiter plates. However, despite the great interest in microwell-based microsystems, no methods have been devised so far to perform controlled delivery of cells into microwells apart from a statistical based approach where cell concentration is the only parameter for controlling the number of cells delivered.17,18,20 In addition, characterization of the macro-to-micro cell transfer efficiency is often left out of the literature. In what follows, we present an innovative platform exploiting dielectrophoresis (DEP) to control the positioning and patterning of single cells within open microwells where further processing and analysis can be carried out. Our solution combines flow-through manipulation with cell patterning in the microwell, avoiding complex macro-to-micro interfacing, minimizing cell loss rate, and allowing for deterministic cell patterning with a high yield and single-cell resolution. The extremely low-count cell populations processed by our system were initially obtained by creating drops where input cells could



EXPERIMENTAL SECTION System Setup and Workflow. The platform used in this work to capture and pattern cells in low cell-count conditions is based on a microsystem integrating an array of open microwells surmounted by microchannels where cells can be made to flow. The device was mounted inside an electro-mechanical package and positioned on the microscope stage over a prefilled microtiter plate. 3447

dx.doi.org/10.1021/ac400230d | Anal. Chem. 2013, 85, 3446−3453

Analytical Chemistry

Article

and at the center of the microchannel. The overall effect was that cells were focused on the xy plane because of DEP and on the xz plane because of both DEP and gravitational force. After being 3D focused, at the patterning stage, cells were controlled by five actuation electrodes surrounding each microwell and acting as a patterning module. Two top-rail electrodes (TR3-TR4) were placed on the sidewalls and two opposing control planar electrodes (C3-C4) were placed on the floor of the microchannel, just before a third electrode featuring a 60° V-shaped cut centered on the microwell. Electrodes were routed so that each patterning stage was independent. TR3 and TR4 were polarized with counter-phase sinusoidal signals and VE1 was tied to the ground. By applying proper signal configurations to C3 and C4 and keeping a constant fluid flow in the microchannel, it was possible to selectively control delivery of the desired number of cells inside each microwell. With C3 and C4 polarized in the Load configuration (see Table 1 and Figure 2), the overall effect was a DEP force pointing

Calcein-stained cells were counted and diluted in order to obtain an average concentration of 15 cells in 15 μL. Droplets with a volume of 15 μL were placed over the lid of a plastic Petri dish (OrDish, Orange Scientifique, Belgium) which was subsequently turned upside down to prompt the formation of hanging drops (Figure 1a). The Petri dish was prefilled with the same liquid as the drops in order to reduce evaporation as in the well-known hanging-drop method, which has already proven to be effective in providing a suitable and viable environment for cell growth and long-term culturing.21−24 The actual number of cells in a drop was measured by optical inspection (Figure 1b) with an inverted microscope (Nikon Eclipse Ti, Nikon). Microchannel outlet ports were connected to a peristaltic pump (Watson Marlow 101U/R) and medium not containing cells was inserted in the microchannels (Figure 1c). When injecting a fluid inside a microchannel, open microwells were filled both by the capillary effect and by pressure, while surface tension prevented fluid leakage from the lower end. The flow rate initially used to fill the microchannels ranged from 2 to 24 μL/min. This last flow rate ensured complete filling of the inlet port, microchannel, and outlet port in less than 10 s. The liquid was made to flow from the outlet to the inlet port and was then partially withdrawn in order to accommodate droplet deposition without any overflow as well as to prewet the port sidewalls so as to reduce cell adhesion. The pump was then stopped and the cell injection process began. A hanging-drop was then manually aligned with a microchannel inlet aperture and brought into contact with the liquid in the port (Figure 1d). The number of cells successfully delivered into the port was counted by means of an upright microscope (Nikon Eclipse 80i, Nikon). The peristaltic pump was then reactivated in withdrawal mode. Acting this way meant that the dead volume was practically zero, and all the cells could be processed by the system. Cells randomly distributed inside the inlet port were drawn into the microchannel (Figure 1e) using an initial flow rate of 100 μL/h. When cells approached the microchannel entrance, the flow rate was lowered. A flow rate of 20 μL/h produced an average cell speed of 100 μm/s. To avoid inlet draining, for each withdrawal of a volume equal to 10 μL, a matching droplet of cell-free liquid was dispensed over the inlet port by means of a standard manual pipet. Once in the microchannel, cells were first three-dimensionally focused using an innovative electrode geometry which combines planar and sidewall electrodes. Then, they proceeded in single file to the subsequent stages where, following fluorescent optical inspection, control electrodes were duly activated (Figure 1f) in order to promote cell isolation inside microwells following a desired pattern. Cell recovery was performed by disconnecting the peristaltic pump, connecting an air impulse injection system to the outlet port, and plugging the inlet port similarly to the method reported by Bocchi et al.19 Module Description. Cells flowing in the microchannel were first aligned by a set of electrodes enabling DEP-based tridimensional cell focusing. The two top-rail electrodes (TR1, TR2) embedded in the sidewalls of the microchannel (Figure 1f) were polarized by counter-phase sinusoidal signals, whereas the two control electrodes (C1, C2) on the floor were both electrically connected to the ground (GND). A negative dielectrophoretic force (nDEP) pushed cells toward the minimal electric field region, which is located on the floor

Table 1. Signals for Load and Forward Configurationsa load top rail left (TR1, TR3) top rail right (TR2, TR4) V-shaped electrode (VE1) 3D-focusing control (C1, C2) control left (C3) control right (C4)

A × cos(ωt + φ) A × cos(ωt − φ) GND GND GND GND

forward A × cos(ωt A × cos(ωt GND GND B × cos(ωt B × cos(ωt

+ φ) − φ)

− φ) + φ)

a

Electrode labels refer to the ones indicated in Figure 1. Electrodes forming the 3D-focusing module and top-rail electrodes remain unchanged in the two configurations. Other microwells are similarly actuated.

Figure 2. (a) Device photograph. For each one of the three microchannels, there is a focusing stage followed by four microwells. (b) Package photograph. (c) Top-view photograph of a focusing stage. (d) Top-view photograph of a microwell and the electrodes dedicated to cell sorting and patterning. Control planar electrodes (denoted by C) are placed on the floor of the microchannel. Top-rail electrodes (denoted by TR) are placed on the sidewalls near the channel ceiling.

from the ceiling to the floor and from the sidewalls to the center of the channel. In this way, 3D focusing, already established by the focusing stage, was here reinforced and cells were pushed toward the microwell entrance. In the Forward configuration, the total force was pointing toward the ceiling of the channel and was able to shield the microwell entrance, causing cells to skip the well and move along the microchannel to the next microwell. The V-shaped electrode has multiple functions: in the Load mode, it helps reinforce cell focusing without adding any hydro-mechanical barrier. In the Forward 3448

dx.doi.org/10.1021/ac400230d | Anal. Chem. 2013, 85, 3446−3453

Analytical Chemistry

Article

Figure 3. Characterization of the focusing and patterning modules. (a−c) Simulation results for focusing and patterning modules represented by slice plots of the dielectrophoretic force and an arrow plot of the sum of DEP, gravitational force, and drag force on a vertical plane orthogonal to the microchannel length. (a) The 3D focusing module and patterning module in the Load configuration on a vertical plane crossing control electrodes. (b) A patterning module in the Forward configuration on a vertical plane crossing control electrodes. (c) A patterning module in the Load configuration on a vertical plane crossing the microwell. (d) An experimental relation between cell speed, amplitude of the signal applied to exert a DEP force, and the width of the cell beam at the output of the focusing stage. (e) The experimental load efficiency varying cell speed and amplitude of the applied signals which were the same for both the focusing module and the patterning ones. The efficiency comprises the fraction of cells correctly patterned into the microwell and, therefore, is also affected by cell−microwell alignment.

mode, it pulls down to the floor any cell previously lifted by the two control electrodes, promoting delivery to a subsequent microwell in the channel. Device Description and Fabrication. A lab-on-a-chip composed of a 3 × 4 microwell array was fabricated on a 2-layer flexible-PCB polyimide substrate with copper metal layers metalized by a Ni−Au galvanic process to guarantee biocompatibility. The three microchannels were obtained by creating through cuts by a laser in the top polyimide/copper layer, which is then laminated onto the other two polyimide/ copper layers where each microwell is obtained by drilling an 80 μm diameter hole. Channels are sealed by bonding a 750 μm thick top polycarbonate cover, using an optically clear adhesive laminated at 100 °C and cured at 70 °C for 2 h. The top cover also provides a fluid inlet and outlet: the former is obtained by drilling a 4 mm diameter hole, the latter by means of a 5 mm long tip of peristaltic Teflon tube (Watson-Marlow, Wilmington, Massachusetts) with an internal bore of 500 μm. Microchannels have a height of 150 μm, a width of 350 μm, and a length of 2.7 cm. The microwell pitch (4.5 mm) and microchannel mutual distance were designed to match with those of a standard 384-well microtiter plate. After fabrication, microchannels are filled with 1 mM of bovine serum albumin (BSA) for 5 min, rinsed with DI water, and then dried by injecting air in the microchannels. The BSA coating prevents cell adhesion along microchannel walls and electrode surfaces. Dielectrophoresis. The dielectrophoretic effect and its contribution to the overall system behavior were simulated by creating a 3D model of the system and solving the Poisson equation in order to obtain the electric field distribution. The

former was used to determine the expression of the timeaverage dielectrophoretic force for a spherical particle, according to the well-established theory widely described in the literature:25 2 ⟨ FDEP(t )⟩ = 2πεmR3[fCM (ω)]∇Erms

(1)

where Erms is the root-mean-square magnitude of the electric field, εm is the medium permittivity, R is the particle radius, ω = 2πf where f is the frequency of the applied signals while f CM is the Clausius−Mossotti factor given by fCM =

εp̃ − εm̃ εp̃ + 2εm̃

(2)

with ε̃p and ε̃m the complex permittivity (ε̃ = ε + σ/jω) of particles and medium, respectively. Details about the effect of dielectrophoresis over temperature and trans-membrane potential are reported in the Supporting Information. Cell Culture and Preparation. We tested our system with K562 myeloid leukemia cells cultured in RPMI 1640 and 10% FBS. Cells were initially washed to remove the culture medium and resuspended in PBS with the addition of 1 μM green fluorescent calcein for 40 min, in order to obtain fluorescent staining. Subsequent washing was performed to remove calcein and suspend cells in an isotonic PBS solution with electrical conductivity adjusted to 190 mS/m by the addition of glycerol. The same buffer was used to prefill the microchannels before cell processing. The effect of this modified buffer on cell viability and growth was assessed and is reported in the Supporting Information. After the final resuspension, cell 3449

dx.doi.org/10.1021/ac400230d | Anal. Chem. 2013, 85, 3446−3453

Analytical Chemistry

Article

arrow plots of the sum of DEP, gravitational force, and drag force for both the Load and Forward configurations. When a cell processed in the Load configuration reaches the boundary of the microwell, it starts falling at a vertical velocity reinforced by DEP (see the Supporting Information). In the absence of DEP, considering a fluid density of 1000 kg/m3, a viscosity of 1 e−3 kg m−1 s−1), a particle density of 1050 kg/m3, a cell radius of 8 μm, and a microwell diameter of 80 μm, equation SE3 (reported in the Supporting Information) gives a theoretical maximum velocity value of approximately 60 μm/s for performing proper cell loading. This value was confirmed by experimental results. DEP increases the vertical force pointing down inside the microwell (Figure 3c) and allowed us to deliver cells flowing up to vmax = 100 μm/s with a yield higher than 80%. Figure 3e reports the experimental Load efficiency of cell patterning for a varying cell speed and different signal amplitudes. Figure 4b shows an example of cell loading inside a microwell. With 0.7 Vrms applied to the control electrodes and maintaining a constant flow, the Forward mode is always successful (efficiency results not shown). Figure 4 (panels c−f) show an example of cell forwarding. Video V1 in the Supporting Information shows an example of cell loading and forwarding into a microwell. High Yield Patterning. Twenty hanging drops with an average of 15 fluorescent cells per drop were processed into our system in order to estimate the drop-to-microwell patterning yield and the percentage of cells lost. Optical inspection of both the hanging drop (Figure 5a) and the inlet port (Figure 5b)

concentration was adjusted in order to obtain the desired number of cells per hanging drop.



RESULTS DEP-Controlled 3D Focusing Module. The purpose of this module is to align cells at the center of the microchannel along the floor. Simulations show that the total force points are from the ceiling to the floor and from the lateral sidewalls to the center of the microchannel, as confirmed by Figure 3a which reports the slice plot of the DEP force and the arrow plot of the sum of DEP, gravitational force, and drag force. Figure 3d shows the experimental relation between cell velocity, amplitude of the signal applied to exert a DEP force, and the width of the cell beam at the output of the focusing stage (all signals have a frequency of 100 kHz so that the Clausius− Mossotti factor is negative). Figure 4a shows the experimental

Figure 4. Focusing and isolation: experimental examples. (a) Cells seen at the focusing stage. A high cell concentration is used to emphasize focusing efficiency. (b) Focalized cells flowing toward a microwell polarized in the Load configuration. (c) A single cell has already been patterned inside a microwell (white circle) and control electrodes are polarized in the Forward configuration. The approaching cell (green fluorescent dot inside the red circle) skips the microwell and moves along the microchannel to the next microwell as shown in (d), (e), and (f). All these images refer to K562 cells.

results of focusing calcein stained K562 cells. The cell beam width (CBW) must be smaller than the diameter of a microwell in order to promote downstream cell isolation. Upon application of an amplitude of 2.8 Vrms, the CBW proved to be smaller than 100 μm up to a cell speed of 100 μm/s. With the use of signals with amplitudes up to 2.8 Vrms, temperature variations due to Joule heating remain within an acceptable range with respect to cell viability (see the Supporting Information). DEP Controlled Patterning Module. Cells flow in single file until the first microwell where the five actuation electrodes regulate cell isolation and patterning. Electrodes are normally set to the Forward mode so that approaching cells are lifted from the floor, skip the microwell, and then descend again to the floor after the microwell. The amplitudes of the applied signals A and B (see Table 1) are respectively 2.8 Vrms and 0.7 Vrms. Cell alignment is preserved so that cell isolation in a subsequent microwell is not hampered. To perform the delivery of a cell inside an open microwell, electrodes were switched to the Load configuration. Figure 3, panels a−c show the slice plots of the electric field and the

Figure 5. Low-count cell population experimental results. (a) Calceinlabeled K562 cells seen inside a hanging drop. (b) Calcein-labeled K562 cells after the drop has been transferred into the inlet port of a microchannel. (c) Cell patterning of 16 input cells in the 4 microwells of a microchannel in the array: in each microwell, it is possible to isolate the desired number of cells. In this example, we patterned cells in order to have them equally distributed between the four microwells of a microchannel.

demonstrated that more than 91% of the cells can be successfully transferred from the drop to the device inlet. The other cells remain in the residual hanging drop portion left on the Petri. Once in the inlet port, cells are drawn into the microchannel. On optically observing the port−microchannel junction, we were able to determine that 90.9% of cells are successfully transferred from the inlet port to the microchannel, while the rest remain stuck to the inlet sidewalls. During cell loading inside the microchannel, some drops of cell-free liquid were dispensed over the inlet port by using a manual pipet to avoid air aspiration inside the microchannel. 3450

dx.doi.org/10.1021/ac400230d | Anal. Chem. 2013, 85, 3446−3453

Analytical Chemistry

Article

As a last step, 100% of cells entering the microchannel were successfully patterned into the microwells. In the end, by composing the partial transfer results, we achieved a total yield of 82.83%. Additional details are included in Table S1 of the Supporting Information. On average, of the 15 cells processed, 12 were actually transferred to the microwells by actuating the control electrodes, according to a pattern scheme that can be deliberately chosen. The potential of distributed patterning is that it allows for creating inside each microwell clusters composed of a programmable number of cells as confirmed by Figure 5c, which shows an example of cell patterning in the four microwells of a microchannel in the array. Moreover, every microwell can be filled so that, unlike statistical patterning approaches, it is possible to avoid both empty and overfilled microsites. Viability. Temperature rise due to Joule heating was assessed by performing simulations which enabled viable conditions to be maintained using the signals previously indicated (see the Supporting Information). Particle-tracing simulations allowed us to make sure that cell paths did not cross regions characterized by unsafe trans-membrane potential values (see Supporting Information). After patterning cells in microwells, electrodes were turned off, and PBS was used to rinse the microchannels and restore a standard medium in place of the conductivity-adjusted PBS previously used. In this way, cell exposure to nonstandard buffer can be reduced and microwells can be used for longer-term cell analysis and imaging. Experimentally, cell viability and growth were assessed, after cells had been patterned in microwells, by turning off electrodes and rinsing the microchannels with PBS. Cells were then recovered into a 384-well microtiter plate prefilled with RPMI + 10% FCS + 25 mM Hepes + pen-strep, and cell growth was measured daily for three days. Growth rate, both for single cells and for cell clusters (number of cells per microwell in the range 1 to 6 with single cell cases representing 50% of the total), was compared to the rate obtained by delivering cells passively into microwells (no electric field applied) and not exposing them to the low conductivity buffer. Experimental results, shown in the Supporting Information, demonstrated the growth of all cells recovered, the growth rate being comparable to controls for both single cells and cell clusters.



of analysis and overcome the typical drawbacks of DEP-based devices.15 The workflow presented enables the reiterated analysis and tracing of cell lineage, opening up new opportunities for cell differentiation analysis. For instance, being able to process 10− 15 cells at a time makes it possible to analyze the cell lineage every 3−4 replication cycles. The interest in this subject is shown, for instance, by the work of Kobel et al.26 who demonstrated isolation of single stem cells flowing in microchannels as well as automated retrapping of daughter cells, thus paving the way for cell lineage tracing. However, in that approach, the analysis is confined to cell imaging only and no means are offered either of studying cell−cell and cell− molecule interactions or of recovering cells for gene expression analysis or reiterated cell differentiation analysis. The method we describe relies on the use of hanging drops to transfer cells into the device. This enabled us to count cells accurately and determine the macro-to-micro transfer yield by means of optical observation. The use of hanging drops obtained by turning the drops upside down not only provides a suitable environment for cell counting but also is a well-known method of culturing cells and inducing the formation of cell spheroids and embryoid bodies and is therefore widely found in the literature on stem cell processing.23 Hence, by coupling the hanging drop method with our platform, we envision the opportunity of performing reiterated cell lineage tracing where single cells recovered by open microwells can be further expanded in hanging drops. Despite the potential interest in combining the use of hanging drops with our system, we point out that the workflow here presented is not bound to this technique: the use of other cell delivery methods (e.g., standard pipetting) and recovery substrates is not hampered. It is therefore possible to process cells either taken from an Eppendorf tube or from a microtiter plate so that our system could be used, for example, as a second-step processing once a rare-cell sorting process has been made by FACS. Compared to existing works describing cell patterning/trapping solutions,5 which usually refer to handling cells already loaded into a device, this work enlarges the scope and covers the often evaded issue of macro-to-micro transfer yield, which is fundamental when processing extremely low-count cell populations. Among the few existing solutions aimed at processing singlecells starting from relatively small cell population, Fludigm (www.fluidigm.com) offers a highly integrated commercial platform (C1 Single-Cell Auto Prep System, Fluidigm, San Francisco, CA), allowing for single-cell trapping with integrated RT-qPCR. Despite these extremely interesting features, this platform is specifically targeted to molecular analysis alone and no means are offered of investigating cell subcloning or performing different analyses such as imaging, high content screening (HCS), or cell recovery. Moreover, the optimal input cell number (≈1000) and cell trapping yield (around 10% at the optimal input concentration) make this platform unsuitable for processing extremely low-count cell populations. By comparison, the solution we present achieves a much higher cell trapping yield (83%) and, at the same time, reduces the minimum number of input cells required by 2 orders of magnitudes, thus enabling the processing of rare cells and the study of early stage cell differentiation. Moreover, the flexible macro-to-micro interfacing offered by our platform makes it possible to integrate this system in a standard laboratory

DISCUSSION

The results we have presented show that our solution, based on DEP-activated microchannels and open microwells, provides a means for high-yield controlled patterning of cells even when coming from extremely small populations, which can subsequently be processed, analyzed, imaged, and recovered as described by Bocchi et al.19 Our solution combines the advantages of a flow-through approach with those of platforms suitable for longer-term analysis of patterned cells. The key features and advantages of the solution proposed are that single-cell patterning is deterministic, issues due to macro-tomicro interfaces are reduced, subsequent droplets containing different kinds of cells could be used to create clusters composed of different cell types, and seamless live-cell recovery is made possible. In addition, the innovative capability of onchip buffer substitution and, consequently, cell environment modification, thus enabling both cell−cell and cell−molecule interaction analysis while preserving cell position without the continuous need for electric fields, may widen the feasible range 3451

dx.doi.org/10.1021/ac400230d | Anal. Chem. 2013, 85, 3446−3453

Analytical Chemistry



workflow, allowing for further analysis and processing after cells recovery. An additional key feature of the platform presented here is the deterministic (nonstatistical) cell patterning. When devices are designed to accommodate cells in microsites larger than the cell size, patterning techniques typically rely on a statistical approach, and the input cell concentration is tuned in order to maximize the frequency of the desired number of cells per microsite. However, the Poisson distribution predicts that only a small fraction of microsites (typically around 30%) will be properly filled,3,17 while the majority of them are empty/underfilled (which reduces the active area of the device) or overfilled (which is not compatible with single-cell subcloning and lineage tracing). Higher patterning yields can be obtained by using triangular-shaped microwells27 or devices featuring cell-sized traps,12,26,28−31 which are designed to fit the specific cell dimension and number. Yields offered by these approaches may be applicable to low-count cell populations, but to our knowledge, no solution has so far allowed recovery of patterned/processed cells. The deterministic patterning approach presented here offers the possibility to both fill all the microwells in the array and precisely control the number of cells loaded with up to 100% precision, while not requiring ad hoc microsite sizing. In the Load and Forward configurations, switching depends on the specific application: if, for example, the aim is to single out cells from a small population, the switching criterion is the empty/filled state of a microwell and the optical detection of a cell approaching a microwell. In consideration of a cell beam width of 50 μm, a microscope equipped with a 20× lens (observed area of 503 × 377 μm) and a 2MPx camera, the electrode configurations can be switched according to the mere monitoring of a 1 × 160 pixels region and the setting of a signal intensity threshold. As a final remark, thanks to standard Flex-PCB microfabrication technology, the 3 × 4 microwell array presented here can easily and inexpensively be scaled up so as to increase the processing parallelism.



Article

ASSOCIATED CONTENT

S Supporting Information *

Further details and data are provided as Supporting Information (SI). This material is available free of charge via the Internet at http://pubs.acs.org.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Tel: +39 0512093810. Fax: +39 0512093822. Notes

The authors declare no competing financial interest.



REFERENCES

(1) Chao, T. C.; Ros, A. J. R. Soc., Interface 2008, 5, S139−S150. (2) Kovac, J. R.; Voldman, J. Anal. Chem. 2007, 79, 9321−9330. (3) Lecault, V.; VanInsberghe, M.; Sekulovic, S.; Knapp, D. J. H. F.; Wohrer, S.; Bowden, W.; Viel, F.; McLaughlin, T.; Jarandehei, A.; Miller, M.; Falconnet, D.; White, A. K.; Kent, D. G.; Copley, M. R.; Taghipour, F.; Eaves, C. J.; Humphries, R. K.; Piret, J. M.; Hansen, C. L. Nat. Methods 2011, 8 (7), 581−586. (4) Marcy, Y.; Ouverney, C.; Bik, E. M.; Losekann, T.; Ivanova, N.; Martin, H. G.; Szeto, E.; Platt, D.; Hugenholtz, P.; Relman, D. A.; Quake, S. R. Proc. Natl. Acad. Sci. U.S.A. 2007, 104 (29), 11889− 11894. (5) Pratt, E. D.; Huang, C.; Hawkins, B. G.; Gleghorn, J. P.; Kirby, B. J. Chem. Eng. Sci. 2011, 66, 1508−1522. (6) Yin, H.; Marshall, D. Curr. Opin. Biotechnol. 2012, 23, 110−119. (7) Bhagat, A. A. S.; Bow, H.; Hou, H. W.; Tan, S. J.; Han, J.; Lim, C. T. Med. Biol. Eng. Comput. 2010, 48, 999−1014. (8) Hsiung, L. C.; Yang, C. H.; Chiu, C. L.; Chen, C. L.; Wang, Y.; Lee, H.; Cheng, J. Y.; Ho, M. C.; Wo, A. M. Biosens. Bioelectron. 2008, 24, 869−875. (9) Tsutsui, H.; Ho, C. M. Mech. Res. Commun. 2009, 36, 92−103. (10) Charnley, M.; Textor, M.; Khademhosseini, A.; Lutolf, M. P. Integr. Biol. 2009, 1, 625−634. (11) Lindstrom, S.; Andersson-Svahn, H. Lab Chip 2010, 10, 3363− 3372. (12) Skelley, A. M.; Kirak, O.; Suh, H.; Jaenisch, R.; Voldman, J. Nat. Methods 2009, 6, 147−152. (13) Zurgil, N.; Afrimzon, E.; Deutsch, A.; Namer, Y.; Shafran, Y.; Sobolev, M.; Tauber, Y.; Ravid-Hermesh, O.; Deutsch, M. Biomaterials 2010, 31, 5022−5029. (14) Khan, S. S.; Solomon, M. A.; McCoy, J. P., Jr. Cytometry, Part B 2005, 64B, 1−8. (15) Kobel, S.; Lutolf, M. BioTechniques 2010, 48 (4), ix−xxii. (16) Kretzschmar, K.; Watt, F. M. Cell 2012, 148, 1. (17) Love, J. C.; Ronan, J. L.; Grotenbreg, G. M.; van der Veen, A. G.; Ploegh, H. L. Nat. Biotechnol. 2006, 24, 703−707. (18) Lindstrom, S.; Andersson-Svahn, H. Biochim. Biophys. Acta 2011, 1810, 308−316. (19) Bocchi, M.; Rambelli, L.; Faenza, A.; Giulianelli, L.; Pecorari, N.; Duqi, E.; Gallois, J. C.; Guerrieri, R. Lab Chip 2012, 12, 3168−3176. (20) Faenza, A.; Bocchi, M.; Pecorari, N.; Franchi, E.; Guerrieri, R. Lab Chip 2012, 12, 2046−2052. (21) Bartosh, T. J.; Ylöstalo, J. H.; Mohammadipoor, A.; Bazhanov, N.; Coble, K.; Claypool, K.; Lee, R. H.; Choi, H.; Prockop, D. J. Proc. Natl. Acad. Sci. U.S.A. 2010, 107 (31), 13724−13729. (22) Guan, K.; Nayernia, K.; Maier, L. S.; Wagner, S.; Dressel, R.; Lee, J. H.; Nolte, J.; Wolf, F.; Li, M.; Engel, W.; Hasenfuss, G. Nature 2006, 440, 1199−1203. (23) Schulz, J. C.; Stumpf, P. S.; Katsen-Globa, A.; Sachinidis, A.; Hescheler, J.; Zimmermann, H. Eng. Life Sci. 2012, 12 (5), 1−4. (24) Sheridan, S. D.; Gil, S.; Wilgo, M.; Pitt, A. Methods Cell Biol. 2008, 86, 29−57. (25) Pohl, H. A. Dielectrophoresis; Cambridge University Press: Cambridge, U.K., 1978.

CONCLUSION

We have presented a microsystem based on DEP-activated microchannels and open microwells able to perform controlled isolation and deterministic (nonstatistical) patterning of lowcount cell populations inside open microwells where, as has already been demonstrated, cells can further be analyzed, imaged, made to interact with other cells or molecules, and finally recovered onto standard microtiter plates. Experimentally, we used the hanging-drop technique (with an average of 15 cells per drop) to accurately characterize the input cell transfer efficiency and demonstrated the deterministic patterning of cells both by isolating single K562 cells and by creating clusters composed of a predetermined number of cells. Our solution showed a cell transfer efficiency of 83% from hanging drops to the chip inlet and 100% from the chip inlet to microwells, the microwell filling rate being 100%. The device and approach we have presented fill a need unaddressed by traditional cell analysis techniques and could be used to improve existing and enable new cell analysis assays especially in the fields of rare-cell or early-stage cell differentiation analysis. 3452

dx.doi.org/10.1021/ac400230d | Anal. Chem. 2013, 85, 3446−3453

Analytical Chemistry

Article

(26) Kobel, S.; Valero, A.; Latt, J.; Renaud, P.; Lutolf, M. Lab Chip 2010, 10, 857−863. (27) Park, J. Y.; Morgan, M.; Sachs, A. N.; Samorezov, J.; Teller, R.; Shen, Y.; Pienta, K. J.; Takayama, S. Microfluid. Nanofluid. 2010, 8, 263−268. (28) Chung, K.; Rivet, C. A.; Kemp, M. L.; Lu, H. Anal. Chem. 2011, 83, 7044−7052. (29) Eyer, K.; Kuhn, P.; Hanke, C.; Dittrich, P. S. Lab Chip 2012, 12, 765−772. (30) Rettig, J. R.; Folch, A. Anal. Chem. 2005, 77, 5628−5634. (31) Yamamura, S.; Kishi, H.; Tokimitsu, Y.; Kondo, S.; Honda, R.; Rao, S. R.; Omori, M.; Tamiya, E.; Muraguchi, A. Anal. Chem. 2005, 77, 8050−8056.

3453

dx.doi.org/10.1021/ac400230d | Anal. Chem. 2013, 85, 3446−3453