A novel miniature cell retainer for correlative high-content analysis of individual untethered non-adherent cells

Abstract

The importance of research involving non-adherent cell lines, primary cells and blood cells is generally undisputed. However, the task of investigating the complexity and heterogeneity of these cells calls for their long-run monitoring at a single-cell resolution. Such a capability is currently unavailable without having to use disruptive cell tethering. The present Cell Retainer (CR) concept enables high-content correlative multi-parametric measurements, from the functional to molecular level, of the same living individual non-adherent cells within a population. Thereby, despite extensive long-term bio-manipulations, the cells preserve their identity without tethering. Several exemplary experiments, using a microscope-slide-based version of the CR, are presented, which could not be performed by other state of the art methods.

---

Introduction

The living cell is the ultimate target of all drugs. In order to understand complex cellular responses to various biological modulators, two fundamental research capabilities are required: a) the ability to trace the temporal behavior of cells at a single-cell resolution for minutes, hours and even days; and b) the ability to reduce the inaccuracy of interpretation resulting from cell heterogeneity – a phenomenon which exists even in synchronized cell lines, and is probably the most frequent and overlooked obstacle in the investigation of critical questions in cell and developmental biology, in the fields of stem cell research, immunology, oncology, and pharmacology.¹

Practically, to accomplish both requirements, individual cell identity must be preserved. In other words, long-term “eye contact” (i.e. during all stages of complex, repetitive experiments on the same sample, up to several hours) must be maintained with each cell in a population. In this way, static and kinetic features of the individual cells can be recorded, even when the cells are being subjected to bio-manipulations. Consequently, cell diversity becomes distinguishable, thus elucidating the differential aspects obscured in average-based measurements. Thus, difficulties originating from sample heterogeneity can be resolved.

Obviously, the need for long-term cell identification cannot be addressed by even the most sophisticated flow-through systems. Undetermined cell identity also hinders the research of non-adherent cells, which include major subgroups of blood cells, primary cells, or cell lines – all of which are heterogeneous.

The tremendous importance of non-adherent primary cells or cell lines in drug discovery and cell therapy is undisputed. This holds true for the non-adherent blood cells as well, with functions ranging from oxygen delivery to immunity. Moreover, both the crucial functions and availability of blood tissue make it an optimal model for personalized medicine. Nevertheless, its use in cell-based assays, whether in low or high throughput systems or high-content screening, is slow to appear, perhaps due to the limitations of current technologies in controlling the spatial distribution of cells without tethering.

Efforts have been made to capture non-adherent cells and control their spatial distribution using suction,^2,3 centrifugal forces,⁴ or by artificial tethering^5–8 using antigens, resins, etc. However, all of these techniques interfere, to different extents, with cell physiology.^9–12 On the other hand, adhesion-independent patterning methods, such as electrophoresis,¹³ dielectrophoresis,¹⁴ and optical techniques¹⁵ are complex, cumbersome and may induce uncharacteristic effects in cells due to active electromagnetic fields.¹⁶ These drawbacks, as well as the complexity of the apparatuses involved, make these methods less attractive. Other techniques are based on manipulation of precise pressure gradients and flow stagnation,¹⁷ however these use extremely diluted solutions, and are slow and consecutive by nature.

Cell locations have also been controlled by microstructures such as clefts, grooves or holes, made on dielectric,¹⁸ metallic or glass^19,20 surfaces and even on optical fiber edges.⁴ However, these arrangements are restricted to fluorescent signal analysis, making transmitted light or fluorescent image acquisition and analysis unfeasible. Moreover, since these microstructures are not tightly packed, a significant portion of cells in the sample settle between the microstructures, and are consequently lost and/or alter their position before and during bio-manipulations. This also occurs in microstructures made of polydimethylsiloxane (PDMS), especially when rinsing the retained objects for bio-modulation purposes.²¹ None of such adhesion-independent technologies enable simultaneous and repetitive monitoring of multiple cellular responses in thousands of individual cells.

The present paper describes the Cell Retainer (CR) approach that serves as a conceptual basis for a variety of individual cell based assays. The CR concept was embodied in several operative types of CR-bearing devices, including an autonomous, capillary-flow, microscope-slide-based CR (ACR), and a controlled flow CR (FCR) system, each designed for a particular type of cell research.

The ACR version presented here, enables repetitive, high-content signal and image analysis of the same non-adherent, non-tethered individual cells or groups of cells, being subjected to various bio-manipulations and staining procedures, while maintaining cell viability and identity up to several hours. Hence, it uniquely allows the correlation between pre- and post-fixation measurements. The CR approach can be used for the study of adherent cells as well.

Experimental

The cell retainer

The CR is a densely packed 2-D arrangement of hexagonal picolitre wells (PWs), with ∼8 µm depth, 20 µm pitched, made of glass, in a honeycomb-like pattern (Fig. 1). Each PW, designed to contain a single untethered cell, functions as a high-quality concave micro-lens, with a depression of 35–40% of its diameter, resulting in a numerical aperture of 0.3 and a focal length of about 30 µm.

Fig. 1 (A) SEM micrograph of the CR structure, scale bar: 10 µm. (B) A detailed scheme of the ACR. The main components are: (I) two engraved longitudinal grooves, (II) elevated flow conduit, (III) the front basin of the conduit, (IV) cylindrical container with a CR as the bottom, (V) cover slip, (VI) the slide's waste receptacle, (VII) and a pressing static transparent plate, which secures the cover slip movement.

The PW's internal surface is smooth, preventing light scattering; its bottom is nearly flat, minimally diverting the wave front of transmitted light. The edges of the walls are extremely sharp (less than 0.1 µm wide), causing the precipitating cells to settle inside the wells rather than in between. The 6 sides of a PW are semi-lunar (Fig. 1A), allowing for the flow of solutions, while confining the freely moving cell within it (see Supplementary movie S1).†

CR micro-fabrication

Glass masking. A chromium image of honeycomb-like arrays, each containing 10 000 (100 × 100) adjacent 20 µm PW's, was formed on a 0.5 mm thick polished BK7 glass substrate using an image generator (EM-5019, Planar, Minsk, Belarus). Next, a thin (∼1 μm) layer of photoresist (S1813-SP15, Shipley Company, Marlborough, MA, USA) was spin-coated over the chromium layer and soft-baked. The glass substrate, serving as its own photomask, was then exposed using an improved mask-aligner (EM-5006A, Planar, Minsk, Belarus), modified for photoresist exposure on thick glass substrates. The exposed photoresist was developed and removed using a developer (0.9% KOH) mixture.

Glass etching. Adjacent rough cone-shaped wells were etched/drilled at a rate of approximately 0.2–1.0 µm min⁻¹, by concentrated (48%) hydrofluoric acid (Merck, Darmstadt, Germany) in water, yielding negative contacted depressions. Final geometrical and surface quality was obtained by a slow chemical grinding and polishing cycle, utilizing a very diluted etching solution. Etching was terminated by rinsing the substrate with water. A profilometer (ATOS GmbH, Pfungstadt, Germany) was used to inspect surface profile. Additional polishing cycles may be performed until the required profile is achieved. The glass substrate is then cut to pieces of 2 mm × 2 mm discrete CRs.

CR-bearing devices, cell loading and bio-manipulation

The CR is the core of a variety of CR-bearing devices, among which are the FCR, designed for days-long experiments, and the easy-to-use ACR, designed for experiments lasting up to a few hours.

The ACR is shown schematically in Fig. 1B. Embedded under a 0.7 mm deep, 2 mm diameter cylindrical aperture drilled in an engraved plastic plate, the CR offers about 7800 free PWs out of the original 10 000 (due to the relative exposed area of the inscribed 2 mm diameter circle, as the corners are used for adhesion to the plate). The slide's dimensions are similar to those of a standard microscope slide (76 mm × 26 mm), thus it is suitable for both upright and inverted microscopy, utilizing standard microscope stages.

When the cover slip (170 µm thick) is closed, the gap between the slip and the conduit is 0.15 mm. When a drop of fluid is introduced to the basin, capillary forces impel the fluid into the conduit, toward the CR at a constant flow rate of 20 µL s⁻¹. This flow rate is practically independent of the volume introduced, and enables rinsing of the cells in the CR without dislodging them. Capillary forces and fluid surface tension keep the fluid in contact with the hydrophilic glass slide, thus eliminating spilling from the conduit even when the glass slide is moved.

Cell loading into the ACR is performed simply by sliding back the glass cover slip, exposing the cylinder opening, pouring into it 2 µL of a solution containing 3.5×10⁶ cells per mL (∼7000 cells), sliding forward the cover slip (minimizing evaporation), and waiting about 4 minutes for completion of cell sedimentation. The cells in the CR can be subjected to bio-modulations by introducing a drop of a reagent or a dye on the intake basin when the cover slip is closed. The reagents travel toward the cells, located at the bottom of the container, by diffusion and local micro-turbulence. A total volume of ∼200 µL can be applied and stored in the ACR's waste receptacle without compromising sterility.

Cells retained in the ACR can thus be easily treated and studied in any desired bioassay, while preserving their identity during observations at the interrogation point.

For further representation of cell loading, staining and bio-modulation in the ACR, the reader is referred to Supplementary movie S2.†

Cell conditions for measurement

Unless otherwise specified, a cell concentration of 3.5×10⁶ cells per mL was used. The experiments were conducted at room temperature, using PBS or HEPES buffers (pH 7.4), or media supplemented with 10 mM HEPES buffer solution under controlled conditions within the incubator, as described in the text. In all the experiments shown here, cells were treated and measured within PWs for up to 4 hours. For the typical observations described here, a 20× objective was used. At least three non-overlapping regions were imaged, each including about 300 individual cells.

Light microscopy

Olympus' upright BX61 and inverted IX81 microscopes (Tokyo, Japan) were equipped with a sub-micron Marzhauser–Wetzlar motorized stage (types SCAN and SCAN-IM, respectively, with a Lstep controller, Wetzlar–Steindorf, Germany), a 100 W Hg fluorescence excitation lamp, a 14-bit ORCA II C4742-98 camera (Hamamatsu, Japan), wire grid linear polarizers (Edmund Industrial Optics, Barrington, NJ, USA), a filter wheel including fluorescence cubes (excitation filters, dichroic mirrors and emission filters) – for Fluorescein: 470–490 nm, 505 nm long pass, and 510–530 nm respectively (all obtained from Chroma Technology Corporation, Brattleboro, VT, USA). For confocal microscopy, a BioRad MRC-1024 system (Hercules, CA, USA) was fitted with an Axioskop 50 upright microscope (Carl Zeiss AG, Germany).

Software

Under an Image-Pro Plus (IPP) environment (Media Cybernetics, Inc. Silver Spring, MD, USA), a variety of in-house developed software packages were used to run experiments according to a user pre-determined menu, automatically controlling the microscope stage, camera, filters, polarizers, the formation of a ‘detector-per-well-per-cell-array’, image acquisition, management, processing and analysis, including background and offset noise filtering and calculation of fluorescence intensity (FI) and fluorescence polarization (FP) per pixel and/or per cell. Deconvolution was performed using Olympus SIS software.

Reagents and cell treatment

Details on reagents, cell culture, cell treatment and measurement procedures are given in the Supplementary materials and methods.†

Results and discussion

Optical characteristics of the CR

Optical inertness. The fact that each PW, in addition to its retaining capabilities, acts as a concave microlens, might raise the question of a possible distortions of light wave-fronts passing through the CR. Extensive experiments show that the CR is actually optically inert and no image distortion can be detected under common microscope optics, whether in upright, inverted, wide-field, or laser-scanning microscopy (LSM).

Despite PWs dimensions and structure, no distorting effects of diffraction are observed either when using wide field or confocal microscopy imaging (Fig. 2A, B), or differential interference contrast (DIC). Fig. 2C,D shows images of the same field of interest, including two DIC images of the same promonocytic U937 cells. In Fig. 2C, the focus is on cells positioned within the PWs, while in Fig. 2D, the focus is on cells positioned approximately 8 µm higher, on the plane surface adjacent to the CR. Although the cells on the plane resemble those on a regular microscope slide, their DIC images are similar to the images of cells inside a PW, thus showing no detectable optical influences of the CR structure upon the acquired DIC image. Interestingly, by carefully viewing Fig. 2C, one may discern the 3-D structural aspect of the PWs beside that of the resident cells.

Fig. 2 Cells in the CR were optically examined utilizing a variety of exemplary observation modes, including (A) wide-field transmitted light and (B) confocal microscopy imaging (here the same two R123-stained MOLT-4 cells are shown). In panels (C) and (D), the DIC images of the same U937 cells are shown: one focused on cells within the CR (C), the other on cells outside the CR (D). There is no observable image distortion due to possible optical interference of the CR structure with the DIC images of cells in the PWs.

No evidence of optical distortion was found when images of cells in the CR were acquired using two-photon excitation (TPE) microscopy (not shown). In addition, transmitted light images as well as fluorescence images of cells retained in PWs can be deconvolved (see Supplementary Fig. S1).† Therefore, it can be concluded that the CR structure does not interfere with the light wave front passing through it. Consequently, images of cells in the CR are not distorted.

Cell occupancy in the ACR

Using a 20× objective, 17 non-overlapping areas were imaged (20 × 20 PWs each), covering a total of 6800 PWs. With the cell concentration of 3.5×10⁶ mL⁻¹, single-cell occupancy rates were estimated to be 93 ± 2.8% of all the cells situated in the PWs. The average percentage of the occupied PWs was about 81 ± 4.2% when 7000 cells were loaded. The distribution of cell occupancy throughout the CR area and cell-to-cell spacings were homogenous, and no clusters in occupied PWs were observed. Furthermore, no cell clumping was observed when using non-adherent cells. The coefficient of variance (CV) of cell occupancy measured on 17 different areas within a single CR was 5.1%. The filling efficiency of PWs in the ACR, can be controlled by changing cell concentrations in solution. Notably, with the FCR, saturated loading can be performed. In that case, after sedimentation, the excess cells are tangentially rinsed out of the chamber, and loading efficiency can reach up to 99%. A representative example of highly dense loading is shown in Fig. 3, using K-562 leukemic cell line.

Fig. 3 Highly dense loading of K-562 myelogenous leukemia cells.

Cell retention in the ACR

Cell retaining performance was examined by comparing pairs of images (20 × 20 PWs each) of Jurkat T cells in the ACR, taken before and after introduction of aliquots of 10 µL buffer solution. Comparison of 20 such pairs of images, taken on different locations in the CR, indicated that, on average, more than 95% of the cells retained their original locations.

Cell viability and apoptosis assessment in the ACR

Two non-adherent cell lines were tested: the U937 promonocyte cell line, and Jurkat T cell line. Cells were loaded into the ACR and maintained in complete RPMI medium in 100% humidified atmosphere, containing 5% CO₂, at 37 °C. Cell viability and apoptosis were assessed at different time points by double staining with FITC-Annexin V and propidium iodide (PI). About 500 individual cells loaded into the ACR were analyzed at each time point. The percentage of apoptotic cells (Annexin V positive) and dead cells (PI positive) did not change after 1 h, 2 h or 4 h incubation in the ACR (3.4 ± 1.1 and 2.3 ± 0.9% apoptotic and dead cells, respectively). For both cell types, these levels of viability and apoptosis were the same in control cells that were maintained under the same experimental conditions in a standard tissue culture plate. It should be emphasized that utilizing the FCR, cells were viable for 4 days, when kept at 37 °C under 5% CO₂ in 100% humidified atmosphere. Moreover, when cultured in an open system, e.g. in a Petri-dish-embedded CR, cells were maintained unlimitedly (data not shown).

Assessment of the duration of solution exchange

Following cell sedimentation in the PWs, while cells were in a PBS solution, a drop of 10 µL fluorescein diacetate (FDA) (1.2 µM) staining solution was introduced to the intake basin. Next, the time duration from FDA introduction until the first fluorescence signal, was detected and found to be 5 ± 2 s. Obviously, the time it takes for the reagent to affect the cells might be dependent on various parameters, including the device structure, cell responsiveness to the reagent, reagent concentrations, detector sensitivity, etc. Yet, the given results should be, of course, considered as limited and not necessarily representative.

Functional measurements in living cells

The unique capabilities of the ACR are best exemplified by performing experiments correlating static and/or kinetic functional measurements with post-fixation parameters in the same non-adherent untethered cells. These kinds of investigations would be difficult, if not impossible, to conduct by other state of the art techniques, due to their inability to preserve cell identity without tethering.

Enzyme activity in individual cells. One of the elite capabilities of the CR is cell discrimination according to intracellular enzyme activity, which requires serial in situ manipulation of the same cells. 24 fluorescent images of MOLT-4 cells were acquired within 800 s, during which time cells were rinsed in situ with 4 increasing concentrations (0.6, 1.2, 2.4, and 3.6 μM) of the non-fluorescent substrate FDA. Within the cells, FDA is hydrolyzed into fluorescent fluorescein by esterase,²² yielding 4 slopes (showing enzymatic reaction velocity) per cell, each comprised of 6 FI points at different times (Fig. 4). Consequently, cellular Michaelis–Menten constants (K_m and V_max) can be assessed at a single-cell resolution, before and after bio-modulation.

Fig. 4 The same MOLT-4 cells were exposed to 4 increasing FDA substrate concentrations, with measurement intervals of about 30 s, yielding 4 clusters of FI(t). Every cell is attributed an individual slope in a cluster (per given substrate concentration), each comprised of 6 FI data points, altogether 4 slopes per cell. The slopes for the same exemplary individual cell are shown by blue lines. From the slopes, the cells' Michaelis–Menten constants (K_m and V_max) can be estimated.

The average of the K_m and V_max values of the individual cells, measured with the CR, were 1.62 ± 0.91 µM and 822 ± 562 FI s⁻¹, respectively. Both distributions indicated cell heterogeneity with a CV of over 100%. However, these enzymatic parameters were in agreement with bulk measurements, which yield a K_m of 1.73 µM and V_max of 962 FI s⁻¹. In both analyses, FI values were normalized. Nevertheless, the present CR technology enabled the assessment of these constants in specific cells or sub-populations of cells before and after bio-modulation.

Live versus post-fixation observations of the same individual cells

This section shows a variety of examples, demonstrating the vast possibilities of correlative, live vs. post-fixation studies in the ACR. Regardless of the experiment performed on intact cells, upon completion of functional/live measurements, the same cells can be fixed and permeabilized in their PWs for further post-fixation examination. An ensemble of post-fixation procedures on different cell types is shown in Fig. 5 and Supplementary Figs S2 and S3,† including differential chromatic staining and scanning electron microscopy (SEM) (Fig. 5A and B, respectively). Thanks to gentle cell treatments in the CR, the SEM shows the distinct surface ultra-structure of each cell (Fig. 5B).

Fig. 5 Following functional measurements, post-fixation observations of the same cells can be performed, including chromatic or SEM imaging. (A) Peripheral blood leukocytes stained with Giemsa (note the distinction between neutrophils and lymphocytes morphology). (B) An SEM image of Jurkat T cells. Scale bar: 20 μm.

Investigation of sub-cellular organelles in correlation to cell phenotype. With the CR, it is possible to directly investigate specific sub-cellular organelles of a degraded cell, in correlation to the entire original cell. In situ rupture of Jurkat T cells is demonstrated in Supplementary Fig. S2.† Initially, intact cells were exposed to PI to exclude dead cells. Then, by rinsing with a hypotonic solution, osmolytic bursting of the cells was induced. However, each cell's nucleus remains in its respective PW, allowing for its identification. Thus, in addition to correlation between functional and post-fixation parameters in the same individual entire cell, the ability to relate separate sub-cellular components (the nuclei in particular) to the entire cell features may allow differential studies, such as fluorescence in situ hybridization (FISH), for localizing specific nucleic acid sequences in the native cell environment. This will permit the correlation between specific gene expression and a particular cellular phenotype, at a single-cell resolution.

Fluorescence intensity and polarization versus apoptosis-related Bcl-2 protein expression. It is well documented that the relative expression of pro- and anti-apoptotic Bcl-2 protein family regulates a cell's susceptibility to apoptotic inducers.²³ The present approach was used to correlate functional apoptotic indicators and the expression of apoptosis-related proteins. In the current example, fluorescence intensity (FI) and fluorescence polarization (FP) of FDA-stained human peripheral lymphocytes were measured. These early functional indicators of apoptosis^23,24 were correlated with the intracellular level of Bcl-2 protein, at a single-cell resolution (Supplementary Fig. S3).† To allow intracellular introduction of anti-Bcl-2 mAb, following FDA FI and FP measurements (Fig. S3A), the intact cells were fixed in situ and permeabilized, enabling efflux of fluorescein molecules. As a result, the FI of intracellular fluorescein decreased to levels undetectable by the imaging system (Fig. S3B). A scatter diagram of the levels of Bcl-2 protein expression in individual cells is shown in Fig. S3C, and their correlation with intracellular esterase activity is shown in the scattergram in Fig. S3D.

Conclusions

The CR approach addresses investigations where dissimilarity among cells cannot or should not be disregarded, particularly with non-adherent cells. It preserves cell identity and integrity even during numerous long-term bio-manipulations and incubation far from the interrogation area, with no external forces perturbing the cell. Consequently, the essential knowledge of the spatial distribution of the cells is retained, without interference with normal cellular function and/or imaging capabilities.

The CR design, combining a honeycomb arrangement for maximum possible density in a 2-D structure with sub-micron proximity between the PW's, renders the present device the most dense in vitro cell-based information array available. As a result, sample size and reagent volume are minimal. Furthermore, in contrast to DNA and protein arrays that only yield binary information, the CR facilitates generation of multidimensional databases including qualitative and quantitative functional cellomics down to molecular parameters, thus challenging the present bioinformatic data-mining capabilities.

Even the most advanced large-sample multi-parametric automated cytometric technologies, employed in high-content analysis and screening, lack the crucial ability to investigate individual, untethered, non-adherent cells. Specifically, state of the art methods do not enable repetitive dynamic measurements of the same individual non-adherent cells within a population as well as correlations between functional, post-fixation, and molecular data on the same individual cell. All these capabilities are provided, for the first time, by the proposed methodology. Moreover, the CR concept can be used for the study of adherent cells as well.

The present seemingly “cell segregating” approach does not prevent investigation of cell–cell interaction and communication using the CR. By choosing suitable PW's dimensions and cell concentration of a heterogeneous cell suspension, more than one cell can be settled in a single PW, which enables monitoring of cellular interactions. At least 4 occupation combinations of cells in a PW can be produced (data not shown): a cell couple of the same type (interaction control), a couple of different types (interaction signal), and 2 different single-cells, each in a separate PW (control).

The suggested CR approach may have several significant impacts: shifting from binary bulk toward individual-cell-based investigations; opening new vistas for non-adherent-cell-based drug discovery, personalized and advanced cell therapy; and focusing on the cell, the ultimate target of all drugs, as a source of new biological knowledge, in order to address complex unresolved issues of the post-genomic era. Furthermore, a competent implementation of the CR approach with the current multi-well microplate concept will enable, for the first time, the introduction of assays based on non-adherent untethered cells into high-throughput screening and high-content screening, opening new possibilities for drug discovery and personalized therapy.

Acknowledgements

This research was supported by the Horowitz Foundation. We would like to thank Dr Larissa Guejes, Pnina Lebovich and Sergei Moshkov for their valuable technical assistance, and Ilia Stambler for manuscript editing.

References

1. A. D. Weston and L. Hood, J. Proteome Res., 2004, 3, 179–196.
2. M. Deutsch and A. Weinreb, Cytometry, 1994, 16, 214–226.
3. M. Khine, A. Lau, C. Ionescu-Zanetti, J. Seo and L. P. Lee, Lab Chip, 2005, 5, 38–43.
4. I. Biran and D. R. Walt, Anal. Chem., 2002, 74, 3046–3054.
5. J. A. Shapiro and C. Hsu, J. Bacteriol., 1989, 171, 5963–5974.
6. K. C. Neuman, E. H. Chadd, G. F. Liou, K. Bergman and S. M. Block, Biophys. J., 1999, 77, 2856–2863.
7. W. D. Donachie and K. J. Begg, Nature, 1970, 227, 1224–1223.
8. E. Ostuni, C. S. Chen, D. E. Ingber and G. M. Whitesides, Langmuir, 2001, 17, 2828–2834.
9. T. R. Malek, T. J. Fleming and E. K. Codias, Semin. Immunol., 1994, 6, 105–113.
10. F. Manca, Ann. Ist. Super. Sanita., 1991, 27, 15–19.
11. T. Zell, W. J. Kivens, S. A. Kellermann and Y. Shimizu, Immunol. Res., 1999, 20, 127–145.
12. G. Wennemuth, S. Eisoldt, H. P. Bode, H. Renneberg, P. J. Schiemann and G. Aumuller, Andrologia, 1998, 30, 141–146.
13. J. Gao, X. F. Yin and Z. L. Fang, Lab Chip, 2004, 4, 47–52.
14. J. Voldman, M. L. Gray, M. Toner and M. A. Schmidt, Anal. Chem., 2002, 74, 3984–3990.
15. R. A. Flynn, A. L. Birkbeck, M. Gross, M. Ozkan, S. Bing, M. M. Wang and S. C. Esener, Sens. Actuators, B, 2002, 87, 239–243.
16. V. I. Chin, P. Taupin, S. Sanga, J. Scheel, F. H. Gage and S. N. Bhatia, Biotechnol. Bioeng., 2004, 88, 399–415.
17. A. R. Wheeler, W. R. Throndset, R. J. Whelan, A. M. Leach, R. N. Zare, Y. H. Liao, K. Farrell, I. D. Manger and A. Daridon, Anal. Chem., 2003, 75, 3581–3586.
18. J. Tan, H. Shen and W. M. Saltzman, Biophys. J., 2001, 81, 2569–2579.
19. I. Inoue, Y. Wakamoto, H. Moriguchi, K. Okanob and K. Yasuda, Lab Chip, 2001, 1, 50–55.
20. C. D. T. Bratten, P. H. Cobbold and J. M. Cooper, Anal. Chem., 1997, 69, 253–258.
21. J. R. Rettig and A. Folch, Anal. Chem., 2005, 77, 5628–5634.
22. E. Afrimzon, N. Zurgil, Y. Shafran, J. Sandbank, R. Orda, S. Lalchuk and M. Deutsch, Cancer Epidemiol. Biomarkers Prev., 2004, 13, 235–241.
23. M. O. Hengartner, Nature, 2000, 407, 770–776.
24. N. Zurgil, Y. Shafran, D. Fixler and M. Deutsch, Biochem. Biophys. Res. Commun., 2002, 290, 1573–1582.

---

Footnote

† Electronic supplementary information (ESI) available: Supplementary materials and methods, Supplementary Figs. S1, S2 and S3, and Supplementary movies S1 and S2. See DOI: 10.1039/b603961h

---