Microﬂuidic Large-Scale Integration- The Evolution of Design Rules for Biological Automation

7/31/2019 Microfluidic Large-Scale Integration- The Evolution of Design Rules for Biological Automation

1/21

Microuidic Large-ScaleIntegration: The Evolutionof Design Rules forBiological Automation Jessica Melin and Stephen R. QuakeDepartment of Bioengineering, Stanford University and Howard Hughes Medical

Institute, Stanford, California, 94305; email: [email protected]; [email protected]

Annu. Rev. Biophys. Biomol. Struct. 2007.

36:21331First published online as a Review in Advance onFebruary 1, 2007

The Annual Review of Biophysics and Biomolecular Structure is online at biophys.annualreviews.org

This articles doi:10.1146/annurev.biophys.36.040306.132646

Copyright c 2007 by Annual Reviews. All rights reserved

1056-8700/07/0609-0213$20.00

Key Wordssoft lithography, biological automation, polydimethylsiloxane

Abstract Microuidic large-scale integration (mLSI) refers to the development of microuidic chips with thousands of integrated micromechanical valves and control components. This technology is utilizein many areas of biology and chemistry and is a candidate to rplace todays conventional automation paradigm, which consists uid-handling robots. We review the basic development of mLSand then discuss design principles of mLSI to assess the capabities and limitations of the current state of the art and to facilitatthe application of mLSI to areas of biology. Many design and pratical issues, including economies of scale, parallelization strategimultiplexing, and multistep biochemical processing, are discusseSeveral microuidic components used as building blocks to creaeffective, complex, and highly integrated microuidic networks aalso highlighted.

213


2/21

Microuidics: thestudy andmanipulation of uidic behavior onthe nanoliter scaleand below

ContentsINTRODUCTION... . . . . . . . . . . . . . . 214 THE MICROMECHANICAL

VALVE . . . . . . . . . . . . . . . . . . . . . . . . . . 215Push-Up versus Push-Down Valves

Using Soft Lithography. . . . . . . . 216HIGHER-LEVEL

COMPONENTS: PUMPS, MIXING, AND METERING. . . . 216Peristaltic Pump. . . . . . . . . . . . . . . . . . 216 Mixing. . . . . . . . . . . . . . . . . . . . . . . . . . . 217 Metering Strategies. . . . . . . . . . . . . . . 218

HIGHER-LEVELCOMPONENTS: MULTIPLEXED ADDRESSING. . . . . . . . . . . . . . . . . . 220Latches: Valves Switching Valves . . 220

Nonlatched Multiplexing . . . . . . . . . 220Latched Multiplexing............. 222

HIGHER-LEVELCOMPONENTS:COMBINATORICECONOMIES OF SCALE. . . . . . . 224

HIGHER-LEVELCOMPONENTS: AFFINITY COLUMNS AND THE SIEVE VALVE . . . . . . . . . . . . . . . . . . . . . . . . . . 225Serial Processing . . . . . . . . . . . . . . . . . 225

PARALLELIZATION:INCREASING THROUGHPUT WITHOUT INCREASINGCONTROL COMPLEXITY . . . . 225

CONCLUSIONS.. . . . . . . . . . . . . . . . . . 227

INTRODUCTION Microuidics refers to the science and tech-nology of systems that manipulate smallamounts of uids, generally on the nanoliterscale and below. Numerous applications of microuidics have been developed in chem-istry, biology, and other elds (6, 7, 37, 49). Many clever technological inventions havebeen developed to control uid behavior insmall channels, often taking advantage of the

physical properties of the uid, the channelsand the contents of the uid (37). In this re view, we focus on the development of a paticular microuidic technology that has madenormous strides in the effort to automatebiology: microuidic large-scale integration(mLSI).

mLSI refers to the development of microuidic chips with hundreds to thousandof integrated micromechanical valves. Thitechnology enables hundreds of assays to bperformed in parallel with multiple reagentin an automated manner and has been usedin applications such as protein crystallography (1214), genetic analysis (26), amino acanalysis (36),high-throughput screening (42)bioreactors (10), chemical synthesis (24, 47and single cell analysis (30). It is a candida

to replace todays conventional biological automation paradigm, which consists of uidhandling robots.

Although mLSI is not the only microuidic approach used in these areas, using mcromechanical valves gives one the abilito create highly complex, integrated design without accounting for the detailed properties of the uids that one is manipulating. Juas the development of digital electronics enabled ever more complex microprocessor de

signs, mLSI allows one to treat microuidicas a design problem in which components o various complexity are stitched together in seamless and transparent fashion.

In electronic LSI, many distinct functional subcomponents such as memory, comparators, counters, multiplexers, and so onare integrated on a single chip to performapplication-specic or generic tasks. Theshigher-level components are assembled fromtransistors, andin digital electronics thetrans-fer of information in these components ibased on the 0 and 1 logic states. Each subcomponent in LSI is well characterized, ana given input produces a reliable and repeatable output. In this way, subcomponents canbe treated essentially as black boxes whedesigning LSI circuits with a limited number of user-dened parameters (i.e., number

214 Melin Quake


3/21

of inputs for a multiplexer, step resolutionfor a counter, threshold voltage for a com-parator, etc.). This enables modular designand the use of automated software designtools.

As tasks and experiments to be performedby microuidics become more and more

complex and with hundreds or thousands of components housed on a single chip, digitalmicrouidics enables a similar strategy. How-ever, until recently, focus has been limitedto optimizing individual microuidic com-ponents that are often not easily integrated with each other, and microuidic systemshave been designed using a bottom-up ap-proach. A top-down approach simplies thedesign of integrated microuidic systems ona chip by providing a library of microu-

idic components, similar to a cell library inLSI design. This would also allow automatedsoftware synthesis tools for both microu-idic architecture-level synthesis (i.e., schedul-ing and resource binding) and geometry-level synthesis (i.e., routing and componentplacement) (40). The development of ex-plicit design rules and strategies allowingmodular top-down design methodology isan exciting frontier in microuidic systemdesign.

We review the basic development of mLSIand then discuss design principles of mLSIto assess the capabilities and limitations of the current state of the art and to facili-tate the application of mLSI to other areasof biology. Many design and practical issuesare discussed, such as economies of scale,parallelization strategies, multiplexing, mul-tistep biochemical processing, and meteringstrategies. Several microuidic componentsthat can be used as building blocks to cre-ate effective, complex, and highly integratedmicrouidic networks are also highlighted.Our goal is to go beyond the details of device fabrication and application already dis-cussed in the published literature and to dis-cuss someof the general design rules thathaveevolved.

Large-scaleintegration: theability to integratehundreds tothousands of functionalcomponents on asingle device toexecute complextasks

mLSI: microuidiclarge-scaleintegration

High-throughput screening: theability to screen asample (or reaction)in a large number of

conditions in ahighly parallel andefcient manner

MEMS: microelec-tromechanicalsystems

PDMS:polydimethylsiloxan

MSL: multilayersoft lithography

THE MICROMECHANICAL VALVE The valve is the basic unit of uid-handlingfunctionality andplaysa role analogousto thatof the transistor in semiconductor electronics.Early micromechanical valves with movingparts were developed using silicon microelec-

tromechanical systems (MEMS) technology,butwerechallenging to fabricateand were notincorporated into highly integrated devicesbecause of their complexity (23). They weresucceeded by a series of devices that includeda silicone rubber membrane integrated ontosilicon/glass substrates withchemicallyetchedchannelandpressureports,andwereactivatedeither pneumatically or thermopneumatically (3, 31, 34, 35, 45, 51). More recently, thetechniques of soft lithography (50) have been

used to make monolithic valves from poly-dimethylsiloxane (PDMS) (43). This advancehas led to a number of other valve designs(1, 9, 17), including a seat valve in which theseat is integrated into the PDMS diaphragmby a molded extension, allowing contact witha planar substrate surface (20). Another poly-mer valve includes a Parylene/PDMS double-layer diaphragm resistant to aggressive chem-icals (which otherwise degrade PDMS) (18).Othervalves include a single-layerPDMSmi-

crochannel devicein whichvalving is achievedby compressing the channel via external pres-sure at specic points along the microchannelby using Braille pins (10). Weibel et al. (48)have also developed torque-actuated valves in which screws are embedded into the single-layer PDMS deviceandaremanually adjustedto close valves on-chip.

The rst mLSI was realized using mono-lithic membrane valves made with multilayersoft lithography (MSL). Early devices in-

cluded a microuidic memory that used 3574 valves integrated on a single chip, as wellas a comparator array with 2056 valves (42). These were followed quickly by a number of other devices made with the same fabricationparadigm with diverse applications in biology and chemistry (2, 1216, 24, 26, 30). If the

www.annualreviews.org Microuidic LSI Design 215


4/21

denition of mLSI is expanded to include devices with at least 40 valves, then a number of other basic valve technologies have been usedto create a large degree of microuidic inte-gration. These include a digital microuidiccircuit with mixer and storage cells ( 40 inte-grated valves on-chip) (44), a bioreactor with

elastomeric valves actuated by a Braille writer( 50 valves) (10), a capillary electrophoresisdevice for amino acid analysis ( 60 valves)(36), and a latching pneumatic valve demul-tiplexer ( 70 valves) (8). A microreactor plat-form( 100 valves) (18) has also recently beenshown. The principles of mLSI do not neces-sarily depend on the particular type of valveused, and the majority of this review discusseshigher-level structures and design strategiesthat can be used with many different valve

types.

Push-Up versus Push-Down ValvesUsing Soft Lithography The most densely packed microuidic valvenetworks and complex on-chip integratedfunctionality have been achieved with MSL. The fundamental building block for this plat-form, the monolithic micromechanical valveshown in Figure 1 a, is produced by replica

molding from twomasters andsealing the lay-ers together (32, 43). Two separate molds areproduced:one forthe channelsandone fortheow channels. A membrane is formed wherethecontrol channel andow channel intersectorthogonally. The ow-channel must have arounded prole to enable the valve to closecompletely during actuation. If the controlchannel layer is bonded on top of the ow-channel layer, push-down valves are formed. The geometry (width, height, and thickness)of the membrane determines the valve actu-ation pressure, and the valve experiences lit-tle hysteresis. Actuation pressures for corre-sponding valve dimensions and mechanicalproperties of the valves useful in microu-idic design are discussed in References 21 and39. Holes can be punched in PDMS to accessthe control and ow channels independently.

The push-down valve allows one to combinliquid control with a DNA array or customfunctionalized surface or device.

Push-up valves are better suited for talleow channels in applications, including eukaryotic cell manipulation (seeFigure 1 b) The uniform membrane thickness of this

valve enables lower actuation pressures thaone nds with push-down valves of similar dimensions. By combining push-up and pushdown valves in a single device, as shown Figure 1 c , a high density of valves can bachieved while simplifying channel routing.

For both types of valves, aspect rationot exceeding 1:10 reduce membrane collapses during fabrication; ow-layer supporpillars can be incorporated if channels greatethan 1:10 are desired. The porosity of PDMS

presents certain design advantages such asophisticated device priming and the ability to manipulate the environmental equilib-rium (11, 46). Monolithic membrane valvecompatible with organic solvents can also bmade from photocurable peruoropolyether(33, 19). Further information on PDMS ma-terial properties and biocompatibility can bfound in References 24a, 25, 28, and 50.

HIGHER-LEVEL COMPONENTS:PUMPS, MIXING, AND METERING

Peristaltic Pump

A linear array of valves actuated in sequencan be used to create a pump. Peristaltipumpingoccurs whenthreemembranevalvesshown in Figure 1 d , are actuated in the pat-tern 101, 100, 110, 010, 011, 001, wher0 and 1 represent open and closed valverespectively. A maximum pumping rate o2 cm s 1 was reported at 100 Hz (14). In thicase, the control channels were lled with aiand the valve actuation frequency was limiteonly by the maximum frequency of the offchip solenoid control valves. On-chip peristaltic pumping provides the ability to control

216 Melin Quake


5/21

Figure 1(a) A two-layerpolydimethylsiloxan(PDMS) push-downmicrouidic valve. An elastomericmembrane is formed where the owchannel is positionedorthogonal to thecontrol channeldirectly above. Fluidow is out of thepage. (b) A two-layePDMS push-upmicrouidic valve where a controlchannel liesorthogonal to andbelow the owchannel. (c ) A three-layer device with both push-upand push-down valves. (d ) Schematicof a linear peristalticpump using threemembrane valves in series (43).

uid ows without the need for precisely regulated external pressure sources.

Mixing One of the fundamental requirements inbiological automation is to mix reagentsefciently. A number of clever schemesfor continuous-ow mixing have been developed, including chaotic advection in athree-dimensional herringbone-shaped mi-crochannel (38), spiral mixers (4), and theuse of hydrodynamic focusing (22). Thesecontinuous-ow mixers can often be adaptedto batch processing in mLSI by simply adding

peristaltic pumps in a closed-loop geometry with the mixers (27). It is also possible to cre-ate a simple yet powerful batch mixer by com-bining a rotary geometry with a peristalticpump (5) (Figure 2 a,b). Once reagents areloaded into the device, the loop is sealed andthe peristaltic pump is activated. Liquid in thecentral part of the ow channels travels fasterthan liquid close to the channel walls. Thisresults in rapid stretching and an increase of the interface between the two reagents and,consequently, a shorter diffusion distance formixing. Mixing times can be reduced to amatter of seconds compared to several hoursfor passive diffusion. The rotary pump is a



6/21

Figure 2(a) Rotary micromixer, where two colored liquids (aqua, DNA; yellow, cells) are loaded into the mixingloop and completely mixed after actuating the peristaltic pump (b) (15). (c ) Geometric metering via threesets of chambers of different sizes. Yellow channels are membrane-valve control channels. Interface valves are closed, and upper chambers have been lled with colored liquid, while bottom chambers havbeen lled with clear liquid (11). (d ) Interface valves are opened and diffusive mixing occurs. (c , d ) Scalebar is 1 mm.

versatile microuidic component that is use-ful ina wide variety of biological and chemicalassays.

Metering StrategiesPrecise metering of small volumes of liquidcanbeachievedbyanumberofschemes.Geo-metricmeteringinvolvesmixingtworeagents, where the volumes are predened by channelor chamber geometry (13). Chambers of dif-ferent volumes are separated by an interface valve, as shown inFigure2 c ,d . First,the inter-face valve is closed and the chambers are lled with different reagents by dead-end priming.Next, valves at the chamber inlets are closed

to dene a xed volume. As the interface valvis opened, mixing occurs by free interfacdiffusion (13).

Positive displacement cross-injection is aactivemethod of liquid meteringand involvesserially dispensing an arbitrary number oreagents (11). As seen inFigure 3 ad , the re-action junction is rst loaded vertically witreagent A. Next, the peristaltic pump is acti vated for a dened number of cycles, whicallows a predictable volume of reagent B tenter the reaction junction. By increasing odecreasing the number of pump actuation cycles, any predened volume can be metereby calculating the volume of displaced liquid. Finally, the reaction junction is sealed b

218 Melin Quake


7/21

Figure 3(a) Serial positive displacement cross-injection, where valves are shown in red and uid lines in blue.(b) Valves are actuated to allow junction to be loaded with uid (orange). (c ) A metered amount of blueliquid is injected by activating the peristaltic pump in a predetermined number of sequence cycles.(d ) Diffusive mixing occurs in the junction (11). (e) Parallel active metering, where blue liquid is injectedinto the vertical column. ( f ) Compartmentalization valves are actuated, and the metered liquid samplesare ushed horizontally downstream using a peristaltic pump (11). (e, f ) Scale bar is 2 mm.

inlet and outlet valves, and diffusive mixingoccurs. The mixed reagents can then be ex-pelled downstream, and twonewreagents canbe metered and mixed.

An early version of positive displacementcross-injectionwas parallelized usingthecon-

guration shown in Figure 3 e, f (11). The vertical channel is lled with reagent, andthe valves compartmentalize the desired vol-ume. The two horizontal valves of each com-partment are opened, and each reagent vol-umeis simultaneouslypumped downstream in



8/21

individual horizontal channels using a peri-staltic pump. If the pumping rate is slowenough, the amount of liquid injected is in-dependent of uidic microchannel resistanceand liquid viscosity (11).

HIGHER-LEVEL COMPONENTS: MULTIPLEXED ADDRESSING

Latches: Valves Switching Valves

A microuidic latch is a powerful tool thatenables valves to be actuated and remain ac-tuated without the need for continuous ex-ternal pressure. Latches are often used to de-crease the number of external control portsneeded while increasing functional complex-ity and integration on-chip. Figure 4 a shows

suchalatch,whichusesmonolithicmembrane valves (41, 42). Pressure is applied throughport A to close valve 2. Next, pressure is ap-plied through port B to close valve 1. Then,external pressure can be released from port A,and valve 2 remains closed because of built-up pressure. This pressure can be sustainedfor > 10 min until depressurization occurs viadissipation through pores in PDMS. Hence,control valves are used to actuate other con-trol valves that subsequently restrict uid ow

on-chip.Grover et al. (8) have also demonstrated aset of latchingvalvesbased ona three-or four- valve network. Normally closed seat valvesform a vacuum-latching or pressure-latching valve (Figure 4 b). In the case of the vacuum-latchingvalveshowninFigure4 c ,thevacuum valve is responsible for holding the latching valve open by sealing a vacuum on-chip, andthe pressure valve is responsible for expellingthebuilt-up vacuum uponcommand. Initially,a pulse of vacuum is applied to the input, andthevacuum valveandlatchvalveopen.Shortly thereafter (120 ms), the vacuum valve closesagain because of depressurization of the pres-sure valve and vacuum valve outputs. Discon-necting the vacuum control input from thechip at this point allows the latch valve to re-main open. To close the latch valve, a pulse of

pressure is applied to the input, thereby open-ing the pressure valve and closing the latc valve for up to 10 min, regardless of wheththechip is connected toan external controller.

Nonlatched Multiplexing

Noteverymemorycell in a semiconductor in-tegratedcircuit is connected to a chip pin-out. A variety of schemes are used to reduce thnumber of pin-outs, including multiplexedaddressing. The microuidic equivalent of multiplexer is a powerful tool that addresselarge numbers of valves with a small number of connections from the chip to the out-side world. Figure 5 shows a schematic of nonlatched microuidic multiplexer (42), i.epressurizing individual control lines directl

actuatevalves controlling uidow. N verticaow channels are controlled by 2log2N hor-izontal control channels. Because threshold valve pressure is highly dependent on membrane geometry, valvesareformedonly where wider sections of the control lines intersecow lines, and control lines are grouped intsets of complementary (binary) valve pairas seen in Figure 5 . Therefore, pressurizingcontrol lines corresponding to bit 1, 2, 3 (i.e1110) allows uid from channel 1110= 14 to

be chosen. Although this multiplexer addresses owchannels effectively, cross-contamination betweenowchannelscan occurbecause ofdead volume at the outlet (Figure 5 b). A modied multiplexer based on a binary tree design eliminates this problem, as shown iFigure 6 a,b (11). Each ow channel incorporates N consecutive bifurcations, allowin2N inlet channels to be connected throughequivalent uidic paths. At each bifurcation valves control the direction of liquid movement. Each ow channel lled with reagenhas a corresponding ush channel connectedto deionized water. After reagent A is chosen and ows through the outlet of the multiplexer, the corresponding ushing chan-nel is opened. This allows deionized wateto remove any reagent A remaining in the

220 Melin Quake


9/21

Figure 4(a) A microuidic latch that uses monolithic membrane valves (red , control channels; blue, ow channels).Port A is pressurized and closes valve 2. Port B is pressurized and closes valve 1. Pressure supplied to port A can then be disconnected, and valve 2 remains closed. (b) A normally closed seat valve. When vacuum isapplied to the bottom channel, the diaphragm deects and uid can ow freely from inlet to outlet. (c ) A three-valve network, creating a vacuum-latching valve (8). (d ) Demultiplexer using vacuum-latched valves, where the input port at the top of the device is supplied by a pressure/vacuum pulse. The uppersection of the device houses four rows of valves, where each alternate valve (on each row) is connected to vacuum or pressure off-chip and comprises the demultiplexer. The bottom of the device contains 16 vacuum latch valves activated via the demultiplexer (8).

multiplexer, leaving the multiplexer ready foranother reagent to be addressed.

A multiplexer with even higher efciency than the binary multiplexer described above was recently demonstrated. This device usesa combinatorial scheme in which all possible

combinations of addressing valves are used. This strategy allows N!/(N/2)!2 chambers tobeaddressedusingNcontrollines(for16con-trol lines, this corresponds to 12,870 address-able chambers, compared with 256 address-able chambers using the binary method) (18).



10/21

Figure 5(a) Microuidic multiplexer, where N vertical ow channels can be individually addressed by 2log2Nhorizontal control lines. Valves are created only where a wide control channel (red ) intersects a owchannel. Narrow channels depict passive crossovers where no valve is created. X represents an actuated valve. (b) When each ow line contains different reagents, cross-contamination can occur because of dea volume at the output of the multiplexer. Figure adapted from Reference 11.

For the 16-channel multiplexer in Figure 6 c ,pressurizing any three of the six control linesresults in only one open ow channel. For ex-ample, pressurizingcontrol channels 2, 3, and5 addresses ow channel 7.

Latched Multiplexing The latching mechanism shown in Figure 6 was used to realize a microuidic analogy toa random access memory. This device utilizestwo multiplexers based on monolithic mem-

brane valves (row and column multiplexerand a 25 40 = 1000 compartment matrixthat includes 3574 valves (42). Contents fromindividual chambers can be recovered usinlatched multiplexing. Each ow channel ineighbored by two other parallel ow channels. To recover the contents of an individual chamber, the row and column multiplexers control uid pressure to create a purgingow path via these two parallel ow channelEach multiplexer can be likened to a memoryaddress registerin randomaccess memory

222 Melin Quake


11/21

Figure 6Binary tree formatmultiplexer thateliminatescross-contamination(a) Green sample isselected. (b) Greensample is ushedusing adjacent buffechannel (11).(c ) Combinatorialmultiplexer, where Ncontrol channelsaddress N!/(N/2)!2ow channels.

The row multiplexer is responsible for uidpurging and replenishment, and the columnmultiplexer controls the vertical input/output valves.

A 4-bit (N) binary demultiplexer address-ing 16 (2N ) independent vacuum-latching valves and outputting pressure or vac-

uum pulses has also been demonstrated (8)(Figure 4 d ). In this case, each row of valves iscontrolled by a four-connection, two-positionsolenoid valve and connected in an alternat-ing manner to the valves in each demulti-plexer row. Thereby, vacuum or pressure isapplied to the even- or odd-numbered valves,



12/21

Figure 7N N = 400 reaction chamber matrix requires only 41 pipetting steps. Enlargement depicts onereaction chamber: White valves are used as peristaltic pumps and green valves are used forcompartmentalizing reagents. Two differently sized green valves are used to compartmentalize reagentsat two different pressures during the reagent-loading sequence. This reduces the number of individualcontrol channels needed. Scale bar = 6.4 mm (26).

respectively, and the input signal can berouted to the correct output. This devicecan be latched every 120 ms, and all valvesin the device are in their nal chosen stateafter 2 s.

HIGHER-LEVEL COMPONENTS:COMBINATORIC ECONOMIESOF SCALE Microuidic devices offer new economies of scale in biological automation. One exam-ple of this is the ability to perform combi-natorial experiments with a minimum num-ber of pipetting steps. An N N (N = 20)matrix of reaction chambers has been demon-strated for 400 PCR reactions, in which only 2N + 1 = 41 pipetting steps were required to

set up the experiment (compared with the3N2 = 1200 pipetting steps required whenusing conventional pipetting into a microtiterplate) (26). This allows 2 L of reagent tobe amortized over as many as 400 chambers. Figure 7 shows the full microuidimatrix and an enlargement of one individual reaction chamber (microchannel loop) The chip contains 2860 valves controlleby two independent pressure supply portsSets of two differently sized valves (actated at different threshold pressures) are connected together and used to compartmental-ize the reaction chamber when lling withthe DNA template, primers, and polymerase The large valves (270 m wide) actuate a110 kPa, and the small valves (96 m wideat 260 kPa. Crossover routing (42- m-wide

224 Melin Quake


13/21

tapered channels) also enables a singlecontrolline to be used for both sets of valves. Thisdevice is composed of three-layers, wherethe ow layer is sandwiched between up-per control lines (push-down valves used forcompartmentalizing reagents) and lowercon-trol lines (push-up valves used for peristaltic

pumping). A 294-bp segment of human -actincDNA fragment was amplied using this chipand the output uorescence emission moni-tored (519 nm and 570 nm). Over multipleexperiments, 98% of 3200 reaction chambersproduced the expected result, and the detec-tion limit for this experiment was 60 templatecopies/reactors. This chip also enabled a setof all possible forward and reverse primersto be tested on the same DNA template,

where the three-component reaction included(a) a cDNA template and polymerase, (b) forward primers, and (c ) reverse primers. Otherplausible applications for this chip includegenetically screening N patients for N mu-tations, and multistep reactions where in-termediate products can be contained inone compartment of the chamber while theremaining compartments are ushed andloaded with reagents needed for a subsequentreaction (26).

HIGHER-LEVEL COMPONENTS: AFFINITY COLUMNS AND THESIEVE VALVE An on-chip afnity column enables severalsequential processes to be performed andadds yet another dimension of functional-ity to mLSI. Microbeads, i.e., functionalizedpolystyrene microspheres, can be trapped up-stream of a partially closed valve (16). How-ever, slight variations in pressure can causelarge variations in column properties (29). By replacing the variable pressure valve with asieve valve, a more reproducible column isformed. This sieve valve comprises a rectan-gular cross-section ow channel, where leak-age occurs along the edges of the channel when the valve is closed, while the movement

of beads or even cells is blocked. Submicron-sized leakage channels can be created, en-abling beads ofonly a few microns indiameterto be stacked. Figure 8 a,b shows an actu-ated sieve valve and a stacked afnity columnupstream of the valve. Another advantage of this sieve valve is that a temporary column

or lter can be created and the contentslater ushed farther downstream by open-ing the valve completely (30). These columnscan also be used for ion-exchange chro-matography (24) and solid-phase synthesis(21a).

Serial Processing The microuidic afnity column can be usedas a component in a serial process, such as cell

isolation, cell lysis, mRNA or DNA purica-tion, andrecovery of theproduct. Figure 8 c , g showsa step-by-stepschematic ofa singlebio-processing unit for such a microuidic device, which has a loading mode and processingmode of operation (16). Initially, an afnity columnof magneticbeads is stacked upstreamof a sieve valve. Next, bacterial cell cultureis introduced upstream, followed by load-ing of dilution buffer in a compartment di-rectly downstream. Lysis buffer is introduced

to yet another compartment, and the valvesare opened and mixed in a rotary pump mixer.Once mixing is complete, the lysate is ushedover the afnity column and the remaindergoes to a waste outlet. The nucleic acid boundon the afnity column can be recovered by ushing elution buffer over the column or by opening the sieve valve and recovering thebeads.

PARALLELIZATION:INCREASING THROUGHPUT WITHOUT INCREASINGCONTROL COMPLEXITY Effective parallelization is an important com-ponent of mLSI. The serial processingscheme described above lends itself to par-allelization by using common control lines



14/21

Figure 8(a) Leakage occurs along the edges of a rectangular proled ow channel when the valve is actuated. (b) Abirds-eye view of a stacked afnity column of microspheres upstream of a sieve valve (30). (c ) A microuidic device for serial biochemical processing, where a microsphere column is stacked upstreamthe sieve valve. Lysis buffer, dilution buffer, and bacterial cell solution are loaded. (d ) Cell solution isloaded into rotary pump. (e) The pump is activated and cells are lysed. ( f ) Wash buffer and lysate areushed over the columns. ( g ) The sieve valve is opened, and puried genomic DNA on beads is ushedout (16).

to allow loading and processing of multiplereactors simultaneously without increasingthe number of control lines needed. Sample volumes can also be geometrically customizedin an array of such processing units to allow various dilution factors to be tested. This is arelatively generic processing system that canbe used for a variety of biochemical processes(15, 16).

As an example, parallel process lines fcell isolation, cell lysis, mRNA puricatiocDNA synthesis, andcDNA purication wereintegrated on a microuidic device (30). Thsensitivity is sufcient for detecting low anmedium copy number transcripts from single NIH3T3 cells. This device uses architecture similar to that of Figure 8 ; howevereach reagent sample is ushed directly int

226 Melin Quake


15/21

Figure 9 A 50-plexcDNA-synthesismicrouidic device with insets of (a) acell lysis module,(b) buffer inputs,(c ) bead and lysisinputs, (d ) capturedNIH3T3 cell,(e) multiplexer andsieve channels,( f ) stacked beadcolumns, and( g ) output, waste,and collection ports(29).

the rotary pump mixer through a sequence of valveactuations, and single cells are trapped inone compartment of the rotary loop. A dilutesolution of cells is injected into the ro-tary loop, and the valves closed when asingle cell is seen to enter. The afnity col-

umn is stacked with oligo-dT functional-ized beads, lysis buffer is loaded into the ro-tary mixer, and the individual cells and lysisbuffer are mixed. Next, the lysate is ushedover the stacked bead column, leaving mRNA bound to the beads. Reverse transcriptaseand dNTPs are then ushed over the col-umn, thereby synthesizing cDNA. This con-cept has been successfully scaled up to a50-plex device (Figure 9 ). This mRNA iso-lation and cDNA-synthesis technique is re-liable for templates that span four orders of magnitude in copy number and for which

the presence of GAPDH and HPRT mRNAsfrom 0.1 to 1 pg of mRNA can be detected(29).

CONCLUSIONS

From theshortlist of examplesdiscussedhere,onecansee that mLSI designcomponents en-able a virtually unlimited set of tools for bio-logical automation. This technology is nowat a stage at which design can be decoupledfrom fabrication, which allows users to fo-cus on applications without requiring themto become skilled in the details of fabrica-tion. Going forward, there is a need for com-puter CAD tools that incorporate mLSI de-sign rules. Soon, designing microuidic tools

for experiments may be as common as design-ing DNA oligonucleotides.

SUMMARY POINTS

1. mLSI demonstrates excellent economies of scale and is a strong candidate to replaceconventional methods of uidic automation such as pipetting robots.



16/21

2. mLSI functional components enable a virtually unlimited set of tools for biologi-cal automation, and we are now at a stage where microuidic system design can bedecoupled from fabrication.

3. mLSI based on an MSL platform is excellent for biological and biochemical appli-cations such as screening conditions for protein crystallography, highly parallel and

combinatorial PCR, andparallelizationof multistepbiochemical processessuch as cellcapture, cell lysis, mRNA purication, and cDNA synthesis for multiple and singlecell analysis.

4. There is a need for computer design tools that incorporate mLSI design rules. Soon,designing microuidic tools for experiments may be as common as designing DNA oligonucleotides.

FUTURE ISSUES

1. mLSI should be made more accessible to researchers in order to apply the technology to an even wider range of biological and biochemical applications. This includesmaking the MSL mLSI platform readily usable for nonaqueous-based applicationssuch as synthetic chemistry.

2. Researchers should develop an even more mature and specic set of design rules formLSI.

3. A truly portable microuidic design and end-user software allowing high-level archi-tecture to be designed without consideration to detailed microuidic behavior andthe specics of the microuidic platform to be used should be developed.

ACKNOWLEDGMENTS This work was supported by NIH 1RO1 HG002644-01A1.

DISCLOSURE STATEMENT SRQ founded a company that operates in the microuidics eld.

LITERATURE CITED1. Baechi D, Buser R, Dual J. 2002. A high density microchannel network with integrat

valves and photodiodes.Sens. Actuators A Phys . 95:77832. Balagadde FK, You L, Hansen CL, Arnold FH, Quake SR. 2005. Long-term monitorinof bacteria undergoing programmed population control in a microchemostat. Scienc309:13740

3. Bohm S, Olthuis W, Bergveld P. 1999. A plastic micropump constructed with conventional techniques and materials.Sens. Actuators A Phys . 77:22328

4. Chen H, Meiners JC. 2004. Topological mixing on a microuidic chip.Appl. Phys. Le84:219395

228 Melin Quake


17/21

5. Chou HP, Unger MA, Quake SR. 2001. A microfabricated rotary pump. Biomed. Microdevices 3:32330

6. Dittrich PS, Tachikawa K, Manz A. 2006. Micro total analysis systems. Latest advance-ments and trends. Anal. Chem. 78:3887907

7. Dittrich PS, Manz A. 2006. Lab-on-chip: microuidics in drug discovery.Nat. Rev. Drug Discov. 5:21018

8. Grover WH, Ivester RHC, Jensen EC, Mathies RA. 2006. Development and multi-plexed control of latching pneumatic valves using microuidic logical structures.LabChip 6(5):62331

9. GroverWH, Skelley AM,Liu CN,LagallyET,MathiesRA.2003.Monolithicmembrane valves and diaphragm pumps for practical large-scale integration into glass microuidicdevices.Sens. Actuators B Chem. 89(3):31523

10. Gu W, Zhu XY, Futai N, Cho B, Takayama S. 2004. Computerized microuidiccell culture using elastomeric channels and Braille displays.Proc. Natl. Acad. Sci. USA101(45):1586166

11. HansenC. 2004. Microuidic technologies for structural biology. PhDthesis.Calif. Inst. Tech.217 pp.

12. Hansen CL, Classen S, Berger JM, Quake SR. 2006. A microuidic device for kineticoptimization of protein crystallization and in situ structure determination. J. Am. Chem.Soc . 128:314243

13. Hansen CL, Skordalakes E, Berger JM, Quake SR. 2002. A robust and scalable microu-idic metering method that allows protein crystal growth by free interface diffusion.Proc. Natl. Acad. Sci. USA99:1653136

14. Hansen CL, Sommer MOA, Quake SR. 2004. Systematic investigation of protein phasebehavior with a microuidic formulator.Proc. Natl. Acad. Sci. USA101(40):1443136

15. Hong JW, Chen Y, Anderson WF, Quake SR. 2006. Molecular biology on a microuidicchip. J. Phys. Condens. Matter 18(18):S691701

16. Hong JW, Studer V, Hang G, Anderson WF, Quake SR. 2004. A nanoliter-scale nucleicacid processor with parallel architecture.Nat. Biotechnol . 22(4):43539

17. Hosokawa K, Maeda R. 2000. A pneumatically-actuated three-way microvalve fabri-

cated with polydimethylsiloxane using the membrane transfer technique.J. Micromech. Microeng . 10(3):41520

18. Hua ZS, Xia YM, Srivannavit O, Rouillard JM, Zhou X, et al. 2006. A versatile microre-actor platform featuring a chemical-resistant microvalve array for addressable multiplexsyntheses and assays.J. Micromech. Microeng.16(8):143343

19. Huang Y, Castrataro P, Lee CC, Quake SR. 2007. Solvent resistant microuidic DNA synthesizer. Lab Chip7:2426

20. Irimia D, Liu SY, Tharp WG, Samadani A, Toner M, et al. 2006. Microuidic systemfor measuring neutrophil migratory responses to fast switches of chemical gradients.LabChip 6(2):19198

21. Kartalov EP, Scherer A, Quake SR, Taylor CR, Anderson WF. 2007. Experimentally-

validated quantitative linear model for the device physics of elastomeric microuidic valves.J. Appl. Phys . In press22. Knight JB, Vishwanath A, Brody JP, Austin RH. 1998. Hydrodynamic focusing on a

silicon chip: mixing nanoliters in microseconds.Phys. Rev. Lett . 80:38636623. Kovacs GTA. 1998. Micromachined transducers sourcebook. In Valves , pp. 82337.

New York: McGraw-Hill24. Lee CC, Elizarov A, Shu CJ, Shin YS, Dooley AN, et al. 2005. Multistep synthesis of a

radiolabeled imaging probe using integrated microuidics.Science310:179396



18/21

24a. Lee JN, Jiang X, Ryan D, Whitesides GM. 2004. Compatibility of mammalian cells surfaces of poly(dimethylsiloxane).Langmuir 20:1168491

25. Lee JN, Park C, Whitesides GM. 2003. Solvent compatibility of poly(dimethylsiloxanebased microuidic devices.Anal. Chem. 75:654454

26. Liu J, Hansen C, Quake SR. 2003. Solving the world-to-chip interface problem withmicrouidic matrix.Anal. Chem. 75(18):471823

27. Liu J, Williams BA, Gwirtz RM, Wold BJ, Quake SR. 2006. Enhanced signals and fnucleic acid hybridization by microuidic chaotic mixing.Angew. Chem. Int. Ed. Eng45:361823

28. Makamba H, Kim JH, Lim K, Park N, Hahn JH. 2003. Surface modication opoly(dimethylsiloxane) microchannels.Electrophoresis 24:360719

29. Marcus J. 2006.Single cell gene expression analysis using microuidics . PhD thesis. Calif. Inst. Tech. 165 pp.

30. Marcus JS,Anderson WF, Quake SR. 2006. Microuidic single-cell mRNA isolation ananalysis.Anal. Chem. 78(9):308489

31. Ohori T, Shoji S, Miura K, Yotsumoto A. 1998. Partly disposable three-way microvalfor a medical micro total analysis system (uTAS).Sens. Actuators A Phys . 64:5762

32. Quake SR, Scherer A. 2000. From micro to nano fabrication with soft materials.Scienc290:153640

33. Rolland JP, Van Dam RM, Schorzman DA, Quake SR, DeSimone JM. 2004. Solvenresistant photocurable liquid teon for microuidic device fabrication.J. Am. ChemSoc . 126(8):232223

34. Schomburg WK, Ahrens R, Bacher W, Martin J, Saile V. 1999. AMANDA-surface mcromachining, molding, and diaphragm transfer. Sens. Actuators A Phys . 76:34348

35. Sjolander S, Urbaniczky C. 1991. Integrated uid handling system for biomoleculinteraction analysis.Anal. Chem. 63:233845

36. Skelley AM,SchererJR,AubreyAD, GroverWH, Ivester RHC, etal.2005.Developmentandevaluation of a microdevice for amino acid biomarkerdetectionandanalysis on Mars Proc. Natl. Acad. Sci. USA102(4):104146

37. Squires TM, Quake SR. 2005. Microuidics: uid physics at the nanoliter scale.Rev

Mod. Phys . 77:977102638. Stroock AD, Dertinger SKW, Ajdari A, Mezic I, Stone HA, Whitesides GM. 2002

Chaotic mixer for microchannels.Science295:6475139. Studer V, Hang G, Pandol A, Ortiz M, Anderson WF, Quake SR. 2004. Scaling prop

erties of a low-actuation pressure microuidic valve.J. Appl. Phys . 95(1):3939840. Su F, Chakrabarty K, Fair RB. 2006. Microuidics-based biochips: technology issu

implementation platforms, and design-automation challenges. IEEE Trans. Comp. AideDes. Integr. Circuits Sys . 25(2):21123

41. Thorsen T. 2003. Microuidic technologies for high-throughput screening application. PhDthesis. Calif. Inst. Tech. 155 pp.

42. Thorsen T, Maerkl SJ, Quake SR. 2002. Microuidic large scale integration.Scienc

298:5808443. Unger MA, Chou HP, Thorsen T, Scherer A, Quake SR. 2000. Monolithic microfabri-cated valves and pumps by multilayer soft lithography.Science288:11316

44. Urbanski JP,ThiesW,RhodesC,AmarasingheS, ThorsenT.2006.Digital microuidicsusing soft lithography. Lab Chip6:96104

45. Vieider C, Ohman O, Elderstig H. 1995. A pneumatically actuated micro valve withsilicone rubber membrane for integration with uid-handling systems.Proc. Eighth InConf. Solid-State Sens. Actuators Eurosens. IX , pp. 28486, Stockholm, Sweden

230 Melin Quake


19/21

46. Vollmer AP, Probstein RF, Gilbert R, Thorsen T. 2005. Development of an integratedmicrouidic platform for dynamic oxygen sensing and delivery in a owing medium.LabChip 5:105966

47. Wang J, Sui GD, Mocharla VP, Lin RJ, Phelps ME, et al. 2006. Integrated microuidicsfor parallel screening of an in situ click chemistry library.Angew. Chem. Int. Ed. Engl .45:527681

48. Weibel DB, Kruithof M, Potenta S, Sia SK, Lee A, Whitesides GM. 2005. Torque-

actuated valves for microuidics.Anal. Chem. 77:47263349. Whitesides GM. 2006. The origins and the future of microuidics.Nature 442:3687350. Xia Y, Whitesides GM. 1998. Soft lithography.Angew. Chem. Int. Ed. Engl . 37:5507551. Yang X, Grosjean C, Tai YC, Ho CM. 1998. A MEMS thermopneumatic silicone rubber

membrane valve.Sens. Actuators A Phys . 64:1018

RELATED RESOURCEShttp://thebigone.stanford.edu/foundry/ http://www.kni.caltech.edu/foundry/



20/21

Annual Revof BiophysBiomoleculStructure

Volume 35,Contents

Frontispiece Martin Karplus p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p xii

Spinach on the Ceiling: A Theoretical Chemists Return to Biology Martin Karplus p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p1

Computer-Based Design of Novel Protein StructuresGlenn L. Butterfoss and Brian Kuhlmanp p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p49

Lessons from Lactose Permease Lan Guan and H. Ronald Kabackp p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 67

Evolutionary Relationships and Structural Mechanisms of AAA +Proteins Jan P. Erzberger and James M. Berger p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p93

Symmetry, Form, and Shape: Guiding Principles for Robustness in Macromolecular Machines Florence Tama and Charles L. Brooks, III p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 115

Fusion Pores and Fusion Machines in Ca 2+ -Triggered Exocytosis Meyer B. Jackson and Edwin R. Chapmanp p p p p p p p p p p p p p p p p p p p p p p p p p p p p p135

RNA Folding During TranscriptionTao Pan and Tobin Sosnickp p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p161

Roles of Bilayer Material Properties in Function and Distribution of Membrane ProteinsThomas J. McIntosh and Sidney A. Simonp p p p p p p p p p p p p p p p p p p p p p p p p p p p p p177

Electron Tomography of Membrane-Bound Cellular OrganellesTerrence G. Frey, Guy A. Perkins, and Mark H. Ellismanp p p p p p p p p p p p p p p p p p p p p p p199

Expanding the Genetic Code Lei Wang, Jianming Xie, and Peter G. Schultzp p p p p p p p p p p p p p p p p p p p p p p p p p p p225

Radiolytic Protein Footprinting with Mass Spectrometry to Probe theStructure of Macromolecular Complexes Keiji Takamoto and Mark R. Chancep p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 251

v


21/21

The ESCRT Complexes: Structure and Mechanism of a Membrane-Trafcking Network James H. Hurley and Scott D. Emr p p p p p p p p p p p p p p p p p p p p p p p p p p277

Ribosome Dynamics: Insights from Atomic Structure Modeling intoCryo-Electron Microscopy Maps Kakoli Mitra and Joachim Frankp p p p p p p p p p p p p p p p p p p p p p p p p p p299

NMR Techniques for Very Large Proteins and RNAs in Solution Andreas G. Tzakos, Christy R.R. Grace, Peter J. Lukavsky, and Roland Riekp p p p p p p319

Single-Molecule Analysis of RNA Polymerase Transcription Lu Bai, Thomas J. Santangelo, and Michelle D. Wang p p p p p p p p p p p p p p p p p343

Quantitative Fluorescent Speckle Microscopy of CytoskeletonDynamicsGaudenz Danuser and Clare M. Waterman-Storer p p p p p p p p p p p p p p p p p p 361

Water Mediation in Protein Folding and Molecular RecognitionYaakov Levy and Jos N. Onuchic p p p p p p p p p p p p p p p p p p p p p p p p p p p389

Continuous Membrane-Cytoskeleton Adhesion Requires Continuous Accommodation to Lipid and Cytoskeleton Dynamics Michael P. Sheetz, Julia E. Sable, and Hans-Gnther Dbereiner p p p p p p p p p p p p417

Cryo-Electron Microscopy of Spliceosomal Components Holger Stark and Reinhard Lhrmannp p p p p p p p p p p p p p p p p p p p p p p p 435

Mechanotransduction Involving Multimodular Proteins: ConvertingForce into Biochemical SignalsViola Vogel p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 459

INDEX

Subject Index p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 489

Cumulative Index of Contributing Authors, Volumes 3135 p p p p p p p p p p p p p p509

Cumulative Index of Chapter Titles, Volumes 3135 p p p p p p p p p p p p p p p p p p512

ERRATA

An online log of corrections to Annual Review of Biophysics and Biomolecular Structuchapters (if any, 1997 to the present) may be found athttp://biophys.annualreviews.org/errata.shtml

Documents

Microﬂuidic Large-Scale Integration- The Evolution of Design Rules for Biological Automation