microfluidic sorting of mammalian cells by optical force switching
TRANSCRIPT
Microfluidic sorting of mammalian cells by opticalforce switchingMark M Wang, Eugene Tu, Daniel E Raymond, Joon Mo Yang, Haichuan Zhang, Norbert Hagen, Bob Dees,Elinore M Mercer, Anita H Forster, Ilona Kariv, Philippe J Marchand & William F Butler
Microfluidic-based devices have allowed miniaturization and
increased parallelism of many common functions in biological
assays; however, development of a practical technology for
microfluidic-based fluorescence-activated cell sorting has
proved challenging. Although a variety of different physical
on-chip switch mechanisms have been proposed1–6, none has
satisfied simultaneously the requirements of high throughput,
purity, and recovery of live, unstressed mammalian cells.
Here we show that optical forces can be used for the rapid
(2–4 ms), active control of cell routing on a microfluidic chip.
Optical switch controls reduce the complexity of the chip and
simplify connectivity. Using all-optical switching, we have
implemented a fluorescence-activated microfluidic cell sorter
and evaluated its performance on live, stably transfected
HeLa cells expressing a fused histone–green fluorescent
protein. Recovered populations were verified to be both
viable and unstressed by evaluation of the transcriptional
expression of two genes, HSPA6 and FOS, known indicators
of cellular stress.
Although conventional flow cytometers are still the standard for highspeed, multi-parametric cell sorting7, a microfluidic-based approach isadvantageous for some applications2. In particular, the microfluidicplatform enables handling of small numbers of cells (100–100,000)with high yield—an impractical task with conventional flowcytometers that typically require 4100,000 cells in the startingpopulation to achieve high yield7. This feature is particularly beneficialin applications involving precious cells, such as primary cells thatcannot be expanded to large populations. Assay miniaturization andreduced reagent consumption also become possible. Loading thebiological sample onto an inexpensive, self-contained disposabledevice ensures a sterile and nuclease-free environment, minimizessample carryover and facilitates safe handling of biohazardous materi-als. Most importantly, as has been proposed for other lab-on-a-chipdevices, the potential exists for further functionality integratedon-chip, such as sample preparation1, cell incubation8, chemicalanalysis9,10, PCR11,12 or other assays of the sorted populations.
Microfluidic cell sorters have used several methods for activecontrol of cell movement or flow. Some of the first demonstrationsof the sorting of bacterial cells relied on electrokinetic mobilization of
fluid through a microfluidic network, achieving rates of 1–20 cells/s1–3.Unfortunately, this method is limited by the difficulty of maintainingcell viability under high electric fields, particularly for eukaryoticcells, and by buffer incompatibilities. Dielectrophoretic forces havealso been proposed for cell switching, although this approachsuffers from similar buffer incompatibilities and slow sorting speeds4.Hydrodynamic flow control based on either on-chip or off-chip fluidicvalves has been demonstrated for sorting living cells; in this case,preserving cell viability is less of a problem5–6. By driving fluidthrough the chip either directly or pneumatically, one can obtainhigh throughputs. However, because of the slow cycle time of themechanical switch and the relatively large volume of fluid that isdisplaced in every switch cycle, cell sorting demonstrations usinghydrodynamic switching have been limited to the enrichment of rarecell populations, and purities have been poor.
The use of optical forces for the deflection of particles or living cellsthrough a fluidic channel was first proposed13 not long after the firstdemonstrations of optical trapping of living cells14,15. In the case ofoptical trapping, or optical tweezers, it has been shown that theradiation pressure forces of a focused optical beam can hold and evenlevitate a small particle, or living cell, in a fluidic medium withoutphysical contact16. When using optical forces for microfluidic applica-tions, depending on the configuration of the illumination system,similar radiation pressure forces can either push or pull a particle orcell in a fluidic medium, potentially at high speeds. The force exertedon a particle by an optical beam is a function of the optical power andthe relative optical properties of the particle and its surroundingfluidic medium, and can reach magnitudes on the order of 1 pN/mWfor cells B10 mm in size.
In early fluidic channel experiments, optical forces were the soledriving force used to propel cells through the length of the fluidicchannel, which resulted in low cell-handling throughputs13. Theseexperiments demonstrated, however, the convenience of optical forcesfor which the only required interconnection to the fluidic chip is aclear optical window. More recently, it has been suggested thatconfining the optical field to only the critical junctions of a micro-fluidic network is a more efficient use of optical power17. Theseexperiments showed that optical forces can provide an active binaryswitch for beads suspended in flow in a microfluidic channel,although the throughputs demonstrated were on the order of only
Published online 19 December 2004; doi:10.1038/nbt1050
Genoptix, Inc., 3398 Carmel Mountain Road, San Diego, California 92121, USA. Correspondence should be addressed to P.M. ([email protected]).
NATURE BIOTECHNOLOGY VOLUME 23 NUMBER 1 JANUARY 2005 8 3
L E T T E R S©
2005
Nat
ure
Pub
lishi
ng G
roup
ht
tp://
ww
w.n
atur
e.co
m/n
atur
ebio
tech
nolo
gy
a few beads per minute. Passive sorting is also possible and has beenused to demonstrate fractionation of small beads based on refractiveindex and size18.
Here we describe a fast, active, all-optical control switch for live cellsand its implementation in a high-throughput, fluorescence-activatedmicrofluidic cell sorter. The noncontact nature of optical switchingfacilitates a low-cost, passive chip device and the use of physiologicalbuffers such as cell culture media. Because this switch operates bydisplacing the cell within the laminar flow rather than by attemptingto make a transient change to the fluid flow, switch rates can be high.We have demonstrated mammalian cell sorting with high purities andrecovery rates.
There are many possible configurations for the orientation of theoptical beam and the design of the sorting junction. The design usedhere is depicted in Figure 1. Cells flowing in from the sample input arealigned to a narrow stream by hydrodynamic focusing. Cells in thefocused flow pass first through an analysis region and then throughthe optical switching region. Flow to the two output channels isasymmetrically biased such that, with no optical beam present, all cellsflow out to the waste channel. When a cell is detected and determinedto be a target cell, the optical switch is activated and a focused laserspot deflects the cell to the target output channel. The laser spot istranslated at a speed matched to the flow velocity and at a small anglerelative to the axis of the flow to maximize the interaction timebetween the laser and the cell. The resulting lateral displacement of thecell across the flow stream is sufficient to ensure that it will be directedtoward the target output channel.
Our cell sorter uses two lasers: a near-infrared laser for the opticalswitch and a visible wavelength laser for detection and fluorescencemeasurement (Fig. 2a). The optical switch laser is a 20-W CWYtterbium fiber laser with a wavelength of 1,070 nm. Modulationand translation of the beam from this laser is controlled by an
acousto-optic modulator (AOM). The laser is focused with a numer-ical aperture (NA) of 0.2 onto the microfluidic chip. This NA isrelatively low compared to what is used in optical trapping experi-ments (typically, NA 4 1); however, trapping is not required in thiscase, and the lower divergence and longer depth-of-focus of theresulting focused beam allows cells to be moved laterally by theoptical force regardless of their initial depth in the microfluidicchannel. The laser used for cell detection and fluorescence excitationis a 5-mW, 488-nm semiconductor laser that is copropagated with thenear-infrared laser. The fluorescence signal from each cell is measuredby a photomultiplier tube (PMT) and electronically gated. Theremovable microfluidic cartridge used in these experiments wasdesigned to minimize cell losses even for small starting cell popula-tions. Starting sample volumes are in the range of 5–25 ml, whereassheath buffer loaded onto the cartridge is approximately 1 ml. Forstarting cell populations as small as 1,000 cells, typically 490% ofcells loaded into the sample input can be retrieved at the combinedoutputs at the end of a complete run. Figure 2b,c shows the resulting
Waste Collection
Optical switch
Analysis region
BufferBuffer
Sample input
Figure 1 Layout of the microfluidic sorting junction and the optical switch.
After being aligned to the center of the channel by flow focusing, cells are
analyzed and then switched based on their detected fluorescence. Target
cells are directed by the laser to the collection output while all other cells
flow to the waste output.
IR laser AOM
488-nm laser
PMT
Photodiode
Filter (525/40)
Filter(488/10)
Sample
Outputwells
cartridgeMicrofluidic
input
reservoirSheath
a
b
c
Figure 2 Cell sorting with an optically switched microfluidic fluorescence-
activated cell sorter. (a) Schematic of the cell sorter instrument and the
microfluidic cartridge. The near-infrared laser and 488-nm laser are focused
through the same lens onto the microfluidic chip. The presence of a cell inthe analysis region is detected by a photodiode and the fluorescence of that
cell is measured by the PMT. Based on a gating of the fluorescence signal
the AOM is triggered to optically switch the cell (GFP-positive cells are
switched to the collection well). Sample is loaded directly onto and off the
cartridge by pipette. Windows on the bottom of the collection wells permit
viewing of the sorted populations. IR, infrared. (b,c) Brightfield (left) and
fluorescence (right) images of the resulting cell populations in the collection
well (b) and waste well (c) are shown for a typical sort of the GFP-expressing
HeLa cells.
8 4 VOLUME 23 NUMBER 1 JANUARY 2005 NATURE BIOTECHNOLOGY
L E T T E R S©
2005
Nat
ure
Pub
lishi
ng G
roup
ht
tp://
ww
w.n
atur
e.co
m/n
atur
ebio
tech
nolo
gy
sorted populations in the output wells of the cartridge after a typicalsorting experiment.
To evaluate the performance of this cell sorter, we sorted cells froma stably transfected HeLa cell line constitutively expressing a fusedhistone–green fluorescent protein (GFP) protein19. GFP-expressingcells were mixed with nonexpressing parental HeLa cells to simulatevarying initial target concentrations. The results of sorting runs of theHeLa cells under varying conditions are summarized in Table 1. Threedifferent sets of data were taken for starting ratios of the targetpopulation of 50%, 10% and 1%. Throughputs were obtained inthe range of 20–100 cells/s by adjusting both the cell density of thesample and the flow parameters. High purity in the collection well wasobtained in most cases. Even for the case of starting target ratios of 1%in which the lowest purities resulted, a 63�–71� enrichment of theoriginal sample was achieved. The recovery rate, which in this case isequivalent to a measure of the efficiency of the optical switch, wasfound to be typically 485% except in the case of the highestthroughput of 106 cells/s. Target recovery can be increased by reducingthe asymmetry of the output flow ratio, although this will also resultin lower purities. Target recovery is also affected by the laser-on time.For experiments in which the target ratio was low, we used a laserswitching time of 4 ms; however, for the samples with B50% targetcells, it was preferable to reduce the laser switching time to 2 ms tominimize the chance of missing a target cell. In the longest sortingrun, nearly 280,000 cells were sorted in 44 min.
In conventional flow cytometer sorting, cell stress and cell viabilityloss are often a concern because of the high hydrodynamic forcesinvolved in the acceleration, droplet formation and collection pro-cesses20. To evaluate cell viability in our experiments, we measuredtrypan blue exclusion for cells that had been run through themicrofluidic chip both with and without exposure to the near-infraredlaser. In both cases, no loss in cell viability was observed (Supple-mentary Table 1). Of the numerous studies in the literature investi-gating the effects of near-infrared exposure on living cells21–25, mostindicate that cellular damage occurs only for exposure energies manyorders of magnitude higher than those used here (SupplementaryTable 2). Although we should therefore not expect to observepermanent cellular damage from this switching technique, we wishedto perform a more sensitive evaluation of possible transitory stress onthe cells. We chose to examine two indicators of cellular stress:transcriptional expression of the HSPA6 (also known as hsp70B¢)and FOS (also known as c-fos) genes as indicators of heat shock andshear stress, respectively.
HSP70B¢ is one of a family of heat shock proteins that play anessential role in cellular responses in both physiological and stress-related conditions by serving as molecular chaperones for other cellu-lar proteins26–28. Because it is upregulated in response to multiple stressconditions including heat shock, we selected it as an indicator of cellu-lar functional response to laser exposure during sorting. FOS isinvolved in the regulation of early signal transduction pathways by
Table 1 Results of sorting GFP-expressing HeLa cells from nonexpressing parental cells for varying starting ratios and throughputs
Sample
Target ratio Cell density (�106 cells/ml) Run time (min) Total cells sorted Throughput (cells/s)a Purity of collection well Recovery rateb Laser-on time (ms)
0.514 3.8 13.9 19,166 22.9 0.985 0.907 2
0.519 7.9 17.5 48,207 45.9 0.980 0.951 2
0.107 5.2 22.0 47,192 35.7 0.970 0.864 4
0.091 12.0 25.5 118,208 77.2 0.964 0.931 4
0.107 12.4 44.0 279,699 105.9 0.886 0.736 4
0.012 7.4 15.9 40,106 42.0 0.849 0.913 4
0.013 7.4 8.9 41,876 78.2 0.825 0.862 4
Flow rates were adjusted for each experiment accordingly. Throughput and total volume loaded onto the cartridge determined the experiment run time.aThroughput is given as an average over the duration of the run. bRecovery rate is calculated as the number of target (GFP-positive) cells recovered in the collection well divided by the number offluorescent events observed at the sorting junction.
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Media37 °C
Cellbuffer
Sorterlaser on
Sorterlaser off
CHX37 °C
FO
S/G
AP
DH
0.0
0.2
0.4
0.6
0.8
1.0
Media37 °C
Cellbuffer
Sorterlaser on
Sorterlaser off
Media42 °C
HS
PA
6/G
AP
DH
ba
Figure 3 Evaluation of stress levels of optically switched cells. (a,b) HSPA6 (a) and FOS (b) gene expression normalized to the GAPDH housekeepinggene for transfected HeLa cells under different conditions. Cells passed through the cell sorter with and without the switching laser are compared to
cells incubated at 37 1C in cell culture media, or at 22 1C in the sorting buffer. Upregulation of HSPA6 and FOS was evident in response to 42 1C and
cycloheximide (CHX) treatments, respectively, which were included as positive controls. Gene expression was determined by RT-PCR, with mRNA extracted
immediately after cell harvesting for FOS, and after 1 additional h of incubation at 37 1C for HSPA6. Data are presented as an average of a duplicate for
FOS and triplicate for HSPA6 experiment 7 s.e.m.
NATURE BIOTECHNOLOGY VOLUME 23 NUMBER 1 JANUARY 2005 8 5
L E T T E R S©
2005
Nat
ure
Pub
lishi
ng G
roup
ht
tp://
ww
w.n
atur
e.co
m/n
atur
ebio
tech
nolo
gy
modulating expression of multiple genes as a part of the AP-1 trans-criptional activator complex29. This gene can be superinduced inresponse to many different physiological and stress conditions includ-ing heat shock30 and also mechanical stress caused by fluidic shear31,32.
Expression of these genes was evaluated on cell populations that hadbeen run through the cell sorter with and without the near-infraredlaser to resolve laser effects from flow effects. The results (Fig. 3)indicate that neither the laser nor the fluidic flow induces increasedexpression of HSPA6 as compared with the two negative controls: cellsincubated in culture media at 37 1C and cells in sorting buffer at22 1C. Similarly, expression levels of FOS are equivalent to those of the22 1C cell buffer control, although all are slightly higher than that ofthe 37 1C media control. Although not comprehensive, these findingsindicate that sorting using this all-optical switch does not perturbcellular homeostasis for the stress-mediated pathways to which thesegenes are sensitive and suggest that microfluidic cell sorting by thistechnique should result in viable and healthy cell populations.
On-chip cell routing by optical controls as demonstrated hereovercomes the throughput and recovery issues that have limitedother microfluidic switching mechanisms. Although the throughputsare lower than those of conventional flow cytometers, we have shownthat sorting runs of cell populations ranging from as few as 1,000 cellsup to 280,000 cells can be completed in less than an hour, which isreasonable for the small cell number/high yield applications for whichsuch a microfluidic device is most suited. Optical control switching isscalable to fluidic networks of greater complexity or parallelism eitherby scaling the number of optical beams or multiplexing a single opticalsource to multiple locations on the chip. In general, the planar natureof most microfluidic chip designs lends itself to insertion of opticalswitching controls at any desired location on the chip, and theselocations are arbitrarily reconfigurable. We have also demonstratedother variations of all-optical controls including completely passiveoptical routers and configurations of the optical switch for multi-layerchip designs (data not shown). Given the flexibility and relative ease ofimplementation of optical forces for active and passive routingcontrols, we believe that this approach will facilitate a variety ofmore integrated and complex cell handling functions.
METHODSMicrofluidic cartridge. The microfluidic channels were fabricated in Schott
D263 glass by standard isotropic etching techniques33. To achieve deep channel
depths, we used a 2,000 A-thick chrome layer and 2 mm-thick photoresist layer
to pattern the glass. The glass was then etched in a 1:5 solution of nitric
acid:buffered oxide etch (Shipley BOE) for 105 s at 22 1C and rinsed with
deionized water. The etched bottom glass was subsequently thermally bonded
at 600 1C to a top glass coverplate with 400-mm diameter, ultrasonically-drilled
access holes. The main sorting channel was 50-mm deep and 150-mm wide. The
microfluidic chip was attached to a plastic cartridge that had reservoirs for
input, both outputs and sheath buffer while leaving a clear window for the
analysis and sorting lasers. The sample input reservoir contained a polypropy-
lene insert to minimize cell adhesion and maximize yield. Flow through the
microfluidic chip was driven pneumatically by four pneumatic regulators
(Pneutronics) that allow control over the flow rate at each of the reservoirs
independently. In this way the hydrodynamic focusing and output flow split
ratio were optimized for each run. Flow focusing in the main sorting channel
was in the range of 8�–15�, and the flow rate in the main channel had a peak
linear velocity of 25 mm/s.
System setup. The near-infrared fiber laser (IPG Photonics PYL-20-LP) and
blue semiconductor laser (Novalux Protera 488-5) were focused at the sorting
junction of the microfluidic cartridge by an f ¼ 20 mm gradient index
achromat (Lightpath Technologies). A cylindrical lens in front of the blue laser
provided anamorphic imaging at the chip. Steering of the near-infrared beam
was achieved by the AOM with driver (IntraAction). Losses along the path of
the near-infrared laser reduced the power at the chip to approximately 13.2 W.
Fluorescence was detected by a Hamamatsu PMT with a 525/40 bandpass filter
(Chroma). Forward scattered light was detected by a photodiode with a 488/10
bandpass filter (Chroma). Dichroic beamsplitters were used to combine and
separate the sorting laser, fluorescence excitation and fluorescence emission
wavelengths. Custom electronic boards performed the real-time signal detec-
tion and thresholding, sent control signals to the AOM, and provided event
counters to an external computer. A CCD camera (not shown in the figure) was
used to monitor the sorting junction and for alignment.
Cell lines and sorting. Parental HeLa cell line was purchased from the
American Type Culture Collection. A stable HeLa cell line transfected with
the genes encoding human histone H2B–green fluorescent protein (GFP) was
obtained from Geoffrey Wahl at the Salk Institute in La Jolla, California19. Both
cell lines were maintained in Dulbecco’s Modified Eagle Medium (Invitrogen)
supplemented with 10% fetal bovine serum (Invitrogen). For sorting experi-
ments, parental and GFP-expressing HeLa cells were mixed in the ratios given
for each experiment and resuspended at the given densities in the sorting
buffer: PBS, pH 7.4, containing 1% bovine serum albumin (BSA, Fraction V,
Sigma-Aldrich) and 14.4% OptiPrep density gradient medium (Sigma-Aldrich)
to increase buoyancy. The sheath buffer was PBS/1%-BSA only. Quantification
of the sorted populations was performed on-chip using a Leica inverted
epifluorescent microscope.
Quantification of mRNA expression. For gene expression studies, HeLa-GFP
cells were pelleted and resuspended at a concentration of 1 � 107 cells/ml. Cells
were either incubated at 37 1C, 5% CO2 in cell culture media or at 22 1C in the
sorting buffer and subjected to cell sorting in the microfluidic cell sorter.
Incubation for 1 h at 42 1C or treatment with 10 mg/ml cycloheximide (Sigma-
Aldrich) at 37 1C, 5% CO2 served as positive controls for HSPA6 and FOS
mRNA expression, respectively30,34. All samples were harvested, washed with
culture media and resuspended at approximately 1 � 106 cells/ml. For the FOS
studies, mRNA was extracted immediately after the cell harvesting, and for
HSPA6 studies, cells were incubated for an additional 1 h at 37 1C, 5% CO2 for
maximum mRNA expression. These time points were chosen based on kinetic
studies (data not shown).
RNA was extracted after recovery from B5 � 104 cells using the RNeasy
Mini Kit (Qiagen) according to the manufacturer’s instructions. Resulting RNA
was transcribed into cDNA in a total volume of 60 ml using the SuperScript III
First-Strand Synthesis Kit (Invitrogen) according to the manufacturer’s instruc-
tions. Levels of mRNA were estimated by real-time PCR performed on an ABI
PRISM 7000 Sequence Detection System using the Sequence Detection System
Software 1.1 (Applied Biosystems). The PCR-primers and TaqMan probes were
purchased from Applied Biosystems as follows: hsp70B¢ (HSPA6): TaqMan
Assays on Demand for Gene Expression; Assay ID: Hs00275682_s1. c-fos
(FOS): TaqMan Assays on Demand for Gene Expression; Assay ID:
Hs00170630_m1. GAPD: Human GAPD (GAPDH) Endogenous Control.
The PCR mix contained 1�Universal Master Mix, 1� primer/probe mix,
and approximately 400 ng of the cDNA reaction in a total volume of 25 ml. The
PCR cycler conditions were 50 1C for 2 min, 95 1C for 10 min, and 45 cycles at
95 1C for 15 s and 60 1C for 1 min. For quantification, HSPA6 and FOS
expression were normalized to GAPD levels. For the construction of standard
curves, twofold serial dilutions of a positive control were prepared for each
experiment. Each experiment represents triplicate RT-PCR determination of
gene expression for each sample.
Note: Supplementary information is available on the Nature Biotechnology website.
ACKNOWLEDGMENTSWe wish to thank Pamela Rose for preparation of the cells and MirianasChachisvillis for helpful discussions. H2B-GFP transfected HeLa cells wereobtained from the Salk Institute. This work was supported in part by DefenseAdvanced Research Projects Agency (DARPA) contract No. DAAH01-03-C-R184.
COMPETING INTERESTS STATEMENTThe authors declare competing financial interests (see the Nature Biotechnologywebsite for details).
8 6 VOLUME 23 NUMBER 1 JANUARY 2005 NATURE BIOTECHNOLOGY
L E T T E R S©
2005
Nat
ure
Pub
lishi
ng G
roup
ht
tp://
ww
w.n
atur
e.co
m/n
atur
ebio
tech
nolo
gy
Received 6 July; accepted 22 October 2004
Published online at http://www.nature.com/naturebiotechnology/
1. Li, P.C.H. & Harrison, D.J. Transport, manipulation, and reaction of biological cellson-chip using electrokinetic effects. Anal. Chem. 69, 1564–1568 (1997).
2. Fu, A.Y., Spence, C., Scherer, A., Arnold, F.H. & Quake, S.R. A microfabricatedfluorescence-activated cell sorter. Nat. Biotechnol. 17, 1109–1111 (1999).
3. Dittrich, P.S. & Schwille, P. An integrated microfluidic system for reaction, high-sensitivity detection, and sorting of fluorescent cells and particles. Anal. Chem. 75,5767–5774 (2003).
4. Fiedler, S., Shirley, S.G., Schnelle, T. & Fuhr, G. Dielectrophoretic sorting of particlesand cells in a microsystem. Anal. Chem. 70, 1909–1915 (1998).
5. Fu, A.Y., Chou, H.-P., Spence, C., Arnold, F.H. & Quake, S.R. An integrated micro-fabricated cell sorter. Anal. Chem. 74, 2451–2457 (2002).
6. Wolff, A. et al. Integrating advanced functionality in a microfabricated high-throughputfluorescent-activated cell sorter. Lab Chip 3, 22–27 (2003).
7. Shapiro, H.M. Practical flow cytometry (Wiley-Liss, New York, 2003).8. Inoue, I., Wakamoto, Y., Moriguchi, H., Okano, K. & Yasuda, K. On-chip culture system
for observation of isolated individual cells. Lab Chip 1, 50–55 (2001).9. Schilling, E.A., Kamholz, A.E. & Yager, P. Cell lysis and protein extraction in a
microfluidic device with detection by a fluorogenic enzyme assay. Anal. Chem. 74,1798–1804 (2002).
10. McClain, M.A. et al. Microfluidic devices for the high-throughput chemical analysisof cells. Anal. Chem. 75, 5646–5655 (2003).
11. Waters, L.C. et al. Microchip device for cell lysis, multiplex PCR amplification, andelectrophoretic sizing. Anal. Chem. 70, 158–162 (1998).
12. Kopp, M.U., de Mello, A.J. & Manz, A. Chemical amplification: continuous-flow PCRon a chip. Science 280, 1046–1048 (1998).
13. Buican, T.N. et al. Automated single-cell manipulation and sorting by light trapping.Appl. Opt. 26, 5311–5316 (1987).
14. Ashkin, A. & Dziedzic, J.M. Optical trapping and manipulation of viruses and bacteria.Science 235, 1517–1520 (1987).
15. Ashkin, A., Dziedzic, J.M. & Yamane, T. Optical trapping and manipulation of singlecells using infrared laser beams. Nature 330, 769–771 (1987).
16. Ashkin, A. Acceleration and trapping of particles by radiation pressure. Phys. Rev. Lett.24, 156–159 (1970).
17. Wang, M.M. et al. Integration of optoelectronic array devices for cell transport andsorting. Proc. SPIE 4260, 68–73 (2001).
18. MacDonald, M.P., Spalding, G.C. & Dholakia, K. Microfluidic sorting in an opticallattice. Nature 426, 421–424 (2003).
19. Kanda, T., Sullivan, K.F. & Wahl, G.M. Histone-GFP fusion protein enables sensitiveanalysis of chromosome dynamics in living mammalian cells. Curr. Biol. 8, 377–385(1998).
20. Seidl, J., Knuechel, R. & Kunz-Schughart, L.A. Evaluation of membrane physiologyfollowing fluorescence activated or magnetic cell separation. Cytometry 36, 102–111(1999).
21. Liu, Y., Sonek, G.J., Berns, M.W. & Tromberg, B.J. Physiological monitoring of opticallytrapped cells: assessing the effects of confinement by 1064-nm laser tweezers usingmicrofluorometry. Biophys. J. 71, 2158–2167 (1996).
22. Liang, H. et al. Wavelength dependence of cell cloning efficiency after optical trapping.Biophys. J. 70, 1529–1533 (1996).
23. Neuman, K.C., Chadd, E.H., Liou, G.F., Bergman, K. & Block, S.M. Characterization ofphotodamage to Escherichia coli in optical traps. Biophys. J. 77, 2856–2863 (1999).
24. Mohanty, S.K., Rapp, A., Monajembashi, S., Gupta, P.K. & Greulich, K.O. Comet assaymeasurements of DNA damage in cells by laser microbeams and trapping beams withwavelengths spanning a range of 308 nm to 1064 nm. Radiat. Res. 157, 378–385(2002).
25. Leitz, G., Fallman, E., Tuck, S. & Axner, O. Stress response in Caenorhabditis eleganscaused by optical tweezers: wavelength, power, and time dependence. Biophys. J. 82,2224–2231 (2002).
26. Kregel, K.C. Heat shock proteins: modifying factors in physiological stress responsesand acquired thermotolerance. J. Appl. Physiol. 92, 2177–2186 (2002).
27. Sonna, L.A. et al. Selected contribution: effect of acute heat shock on gene expressionby human peripheral blood mononuclear cells. J. Appl. Physiol. 92, 2208–2220(2002).
28. Gabai, V.L. & Sherman, M.Y. Interplay between molecular chaperones and signalingpathways in survival of heat shock. J. Appl. Physiol. 92, 1743–1748 (2002).
29. Verma, I.M. & Sassone-Corsi, P. Proto-oncogene fos: complex but versatile regulation.Cell 51, 513–514 (1987).
30. Bukh, A., Martinez-Valdez, H., Freedman, S.J., Freedman, M.H. & Cohen, A. TheExpression of c-fos, c-jun, and c-myc genes is regulated by heat shock in humanlymphoid cells. J. Immunol. 144, 4835–4840 (1990).
31. Hsieh, H.J., Li, N.Q. & Frangos, J.A. Pulsatile and steady flow induces c-fos expressionin human endothelial cells. J. Cell. Physiol. 154, 143–151 (1993).
32. Nollert, M.U., Panaro, N.J. & McIntire, L.V. Regulation of genetic expression in shearstress-stimulated endothelial cells. Ann. NY Acad. Sci. 665, 94–104 (1992).
33. Fan, Z.H. & Harrison, D.J. Micromachining of capillary electrophoresis injectors andseparators on glass chips and evaluation of flow at capillary intersections. Anal. Chem.66, 177–184 (1994).
34. Zinck, R. et al. Protein synthesis inhibitors reveal differential regulation of mitogen-activated protein kinase and stress-activated protein kinase pathways that converge onElk-1. Mol. Cell. Biol. 15, 4930–4938 (1995).
NATURE BIOTECHNOLOGY VOLUME 23 NUMBER 1 JANUARY 2005 8 7
L E T T E R S©
2005
Nat
ure
Pub
lishi
ng G
roup
ht
tp://
ww
w.n
atur
e.co
m/n
atur
ebio
tech
nolo
gy