Biomedical Microdevices 6:4, 325–339, 2004C© 2004 Kluwer Academic Publishers. Manufactured in The Netherlands.

NanoLiterBioReactor: Long-Term MammalianCell Culture at Nanofabricated Scale

Ales Prokop,1, 2,∗ Zdenka Prokop,1 David Schaffer,3

Eugene Kozlov,2 John Wikswo,4, 5, 6 David Cliffel,7

and Franz Baudenbacher4

1NanoDelivery, Inc., Nashville, TN 372112Chemical Engineering, Vanderbilt University, Nashville, TN 372353Mechanical Engineering4Biomedical Engineering5Physics & Astronomy6Molecular Physiology & Biophysics7Chemistry, Nashville, TN 37235, USAE-mail: ales.prokop@mcmail.vanderbilt.edu

Abstract. There is a need for microminiaturized cell-culture environ-ments, i.e. NanoLiter BioReactors (NBRs), for growing and main-taining populations of up to several hundred cultured mammaliancells in volumes three orders of magnitude smaller than those con-tained in standard multi-well screening plates. These devices wouldenable the development of a new class of miniature, automatedcell-based bioanalysis arrays for monitoring the immediate environ-ment of multiple cell lines and assessing the effects of drug or toxinexposure.

We fabricated NBR prototypes, each of which incorporates aculture chamber, inlet and outlet ports, and connecting microflu-idic conduits. The fluidic components were molded in polydimethyl-siloxane (PDMS) using soft-lithography techniques, and sealed viaplasma activation against a glass slide, which served as the pri-mary culture substrate in the NBR. The input and outlet ports werepunched into the PDMS block, and enabled the supply and with-drawal of culture medium into/from the culture chamber (10–100nL volume), as well as cell seeding. Because of the intrinsically highoxygen permeability of the PDMS material, no additional CO2/airsupply was necessary.

The developmental process for the NBR typically employed sev-eral iterations of the following steps: Conceptual design, mask gener-ation, photolithography, soft lithography, and proof-of-concept cul-ture assay. We have arrived at several intermediate designs. Oneis termed “circular NBR with a central post (CP-NBR),” another,“perfusion (grid) NBR (PG-NBR),” and a third version, “multitrap(cage) NBR (MT-NBR),” the last two providing total cell retention.

Three cells lines were tested in detail: a fibroblast cell line, CHOcells, and hepatocytes. Prior to the culturing trials, extensive bio-compatibility tests were performed on all materials to be employedin the NBR design. To delineate the effect of cell seeding densityon cell viability and survival, we conducted separate plating ex-periments using standard culture protocols in well-plate dishes. Inboth experiments, PicoGreen assays were used to evaluate the ex-tent of cell growth achieved in 1–5 days following the seeding. Lowseeding densities resulted in the absence of cell proliferation forsome cell lines because of the deficiency of cell-cell and extracellu-lar matrix (ECM)-cell contacts. High viabilities were achieved in alldesigns.

We conclude that an instrumented microfluidics-basedNanoBioReactor (NBR) will represent a dramatic departure fromthe standard culture environment. The employment of NBRs for

mammalian cell culture opens a new paradigm of cell biology, sofar largely neglected in the literature.

Key Words. nanobioreactor, long-term, mammalian culture

Background and Significance

Scaling laws, non-specific screen, and scope of workToday’s pharmaceutical industry is faced with the un-precedented challenge of managing the progress of arapidly expanding pool of molecular targets, novel com-pounds, and biological assays, all of which are needed todiscover and develop new drugs. High-throughput screen-ing (HTS) in a high-density format may provide some re-lief. In addition, there is a clear move within the pharma-ceutical industry towards increased emphasis on the cell-based assay, which requires parallel processing of multi-ple, smaller batches of a wide range of cell lines derivedfrom various tissue origins. The advent of cell robots thatincorporate incubators, laminar air flow, automatic seed-ing, feeding, trypsinizing, harvesting and counting cells,followed by the dispensing of these cells into a measure-ment format suitable for subsequent analysis, is becomingreality (Slater, 2001). High content screening that facili-tates multiple analyses on a solitary sample is anotherdirection that has yet to be developed in detail. New sen-sor concepts with increased chemical and physical sen-sitivity facilitates miniaturization, leading to paramountimprovements in signal/background ratio, reproducibility,simplicity, and cost.

While individual living cells displace picoliter vol-umes, the flasks, dishes, and wells in which cells arecultured in vitro possess volumes ranging from 100 µL

∗Corresponding author.

325

326 Prokop et al.

to 100 mL—between seven and eleven orders of mag-nitude greater than the cellular volume! The scaling lawsthat govern the microminiaturization of silicon-based elec-tronic devices, microfluidics, Micro Electro MechanicalSystems (MEMS), and nanoparticles have been studiedexhaustively, but little is known about how microminia-turization of cell-culture processes to nanoliter volumesmight affect the chemical and physical parameters re-quired for maintaining cells in vitro. There has been sub-stantial work on scaling bioreactors to larger volumes—commercial bioreactors have volumes from 104 to 105

liters, with one system having 106 liters—but not on thescaling in the other direction.

It is well known that in the scaling-up of a fluidic sys-tem, many physical similarities break down because cer-tain properties, such as diffusion and inertial influence,become dominant. When one physical parameter is fixed,its relationship with, and influence on, other parameterscan become considerably distorted during scale-up. An ex-ample can be found in an aerated bioreactor: if the powerinput of its mixer remains constant, such parameters as thelevel of turbulent liquid motion, liquid circulation time,impeller tip speed, and heat transfer, may become rela-tively distorted (Oldshue, 1983). Because of this, scal-ing up a bioreactor process can yield adverse results, inthis case, potential damage to shear-sensitive organismssuch as mammalian cells can occur. Therefore, mixing andliquid homogenization times are necessarily increased toavoid such damage.

The scale-down of a bioreactor for individual mam-malian cells would benefit from the simultaneous scaling,or lack thereof, of a number of associated physical andbiological parameters and phenomena. In particular, thereare a number of biological phenomena that occur quiterapidly (10−3–10−5 seconds) at the cellular dimensionsof 10 µm, but are much slower at the spatial scale of atypical cell culture environment—for example, quiescentmass and heat transfer within the interior of a living cell.The dynamic hierarchy of biological systems and sub-systems, including physical and biological phenomena, ispresented in Table 1 in terms of relaxation time or timeconstant. Those biological processes with comparativelylong time constants, such as DNA replication and mes-senger RNA synthesis (102–104 s), will not benefit fromscale-down. However, many cellular processes occur insub-second, millisecond, and even microsecond intervals.It is for these processes that the scaling down in size ofthe physical environment and its concomitant variablesresults in an analogous biological process, and therebyoffers the greatest potential. The real-time dynamics ofinterest to this scale-down process include the faster re-laxation times of enzymatic systems, enzyme-substrate-inhibitor interactions, and receptor-ligand interactions, all

Table 1. Dynamic hierarchy of physical (reactor) and biologicalsystems (Prokop, 1982, 1995)

System Relaxation(subsystem) time (s)

Mixing time to homogenize liquid in a large-scale 104–108

bioreactor (10–100 m3)Time to exchange liquid volume to 90% level 105–106

(depending on growth rate) in a continuous reactorOxygen transfer (forced, not free diffusion) 102–103

Heat transfer (forced convection) 103–104

Oxygen uptake rate (mammalian cells) 104–105

Cell proliferation, DNA replication 102–104

Response to environmental changes 103–104

(temperature, oxygen)Messenger RNA synthesis 103–104

Translocation of substances into cells 101–103

(active transport)Protein synthesis 102–101

Allosteric control of enzyme action 1Glycolysis 10−1–101

Oxidative phosphorylation in mitochondria 10−2

Intracellular quiescent mass and heat transfer 10−3–10−5

(dimension 10−5 m)Enzymatic reaction and turnover 10−6–10−3

Bonding between enzyme and substrate, inhibitor 10−6

Receptor-ligand interaction 10−6

Table 2. Physical hierarchy at scale-down for a stagnant spherea

Relaxation time

Subsystem 2000 µm 200 µm

Oxygen quiescent (free) diffusion in/from the 0.5 s 5 msliquid phase within a sphere (Newman, 1931)

Heat transfer by convection into/out of sphere 7 s 70 ms(Crosby, 1961)

aIn practice, gradients at the interface will facilitate higher transfer ratesand shorter relaxation times.

of which are involved in the direct response of the cell toexternal physical or chemical stimuli.

In Table 2 we list results of our calculations for scale-dependence of chemical and thermal diffusion time con-stants for two reactor configurations: a spherical reactionphase (droplet) with 2-mm diameter, and one with 200 µmdiameter (possessing volumes of 24 µl and 24 nL, respec-tively). As seen from this table, the time constants for oxy-gen diffusion from these spheres (to achieve about 50%depletion or saturation) and heat transfer time constants (toheat up these spheres from ambient to 50% of the targeted37◦C temperature; no heat source is considered within thesphere) have a strong dependence upon sphere size. Thisbehavior is reflected in the squared dependency on dropletsize in both cases (e.g., for the oxygen transfer time con-stant t = πR2 E/D, where R is the sphere radius, E is afractional oxygen depletion and D is oxygen diffusivity in

NanoLiterBioReactor 327

Table 3. Time constants of instruments used in this work

Response timeInstrument/probe Size (µm) (90% change) (s)

DO probe 5 (planar) 0.1pH probe 5 (planar) 0.1ORP probe 5 (planar) 0.1

water; for the heat transfer the relationship is similar). Forthis reason, the reduction of the linear scale by a factorof ten (from 2000 to 200 µm) results in a hundred-folddecrease in both relaxation times, and provides the ulti-mate rationale for microminiaturization of the cell cul-ture system: By matching the chemical and thermal timeconstants of a nanoliter cell culture environment to thoseof the cells contained there within, it will be possible touse monitoring of external variables to determine changesin intracellular processes that heretofore have only beenmeasurable with the much longer time constants of muchlarger, conventional cell culture chambers. Consequently,the behavior of such a scaled-down system is determinedby diffusion while other times (transport time, responsetime of sensor) are reduced substantially (Manz et al.,1990; Becker and Gartner, 2001).

Table 3 compares the time constants of electrochemicalsensors that we plan to use to monitor the environmentalstate of the cells. It is apparent that the electrochemicaland fluorescence probes (dissolved oxygen—DO, redoxpotential—ORP, and pH) have very short time constants,hence they should adequately provide insight into the verydynamics of a variety of cellular responses.

In terms of continuous-flow fluid mechanics, we needonly about three volume changes of the NBR content inorder to achieve steady-state conditions from an engineer-ing viewpoint (an analogy to continuous stirred tank re-actor concept in reaction engineering or to a continuousculture in microbial physiology). It is possible to totallyreplace the volume very rapidly when dealing with vol-umes as small as those contained by the NBR. Thus it isfeasible in this system to assess quickly the physiologicalstatus of a culture as soon as steady-state conditions areachieved. Reaction time constants for chemical reactingsystems (and cellular ones) at the nanoliter scale have beenrecently discussed (Bratten et al., 1997; Cooper, 1999).

Reduced NBR volumes would not only shorten the timerequired for diffusive mixing, for achieving thermal equi-librium, and for cells to grow to confluence, but wouldalso simplify accurate cell counting, minimize requiredvolumes of expensive analytical pharmaceuticals or tox-ins, and allow for thousands of culture chambers on asingle instrumented chip.

The ultimate goal is to incorporate biosensor elementsinto the NBR to monitor cell physiology and cell cul-ture conditions in situ and to characterize in a nonspecific

manner the metabolic activity of cells. The biosensor el-ements of the NBR might include planar pH, dissolvedoxygen, and redox potential sensors, or even an isother-mal picocalorimeter to monitor thermodynamic response.Equipped with such sensors, the NBR could be used toperform short- and long-term cultivation of several mam-malian cell lines in a perfused system, and to monitor theirresponse to analytes in a massively parallel format. Thisapproach will enable automated, parallel, and multiphasicmonitoring of multiple cell lines for drug and toxicologyscreening. An added bonus is the possibility of studyingcell populations with low cell counts whose constituentsare completely detached from typical tissue environment,or populations in controlled physical and chemical gradi-ents. The challenge, beyond that of optimizing the NBRphysically, is to detect cellular response, provide appropri-ate control signals, and, eventually, facilitate closed-loopadjustments of the environment—e.g., to control temper-ature, pH, ionic concentration, etc., to maintain homeosta-sis, or to apply drugs or toxins followed by the adaptiveadministration of a selective toxin antidote.

The advantages of miniaturized detection systems arenumerous, and include: (1) reduced assay response times;(2) minimal volume requirement for analyte; (3) reliableand reproducible operation achieved by automated samplehandling; (4) parallel operation of many assays realizedthrough integration of multiple reactors on a single chip;(5) low-cost devices resulting from fabrication of all com-ponents onboard a single, disposable platform.

We first describe the microfabrication techniques em-ployed to create glass/PDMS NBRs with different geome-tries and area/volume ratios (A/V). We then report on thetesting of various NBR designs, their ability to facilitatecontinuous growth of selected cell lines, and on cell vi-ability measurements. In future versions of the NBR weplan to incorporate on-chip pumps for low flow perfusion,and electrochemical sensing of pH, glucose and lactate tomonitor cell metabolism.

Experimental Design and Methods

Fabrication of NBR and integration intoa single functional deviceWe used soft lithography in PDMS to fabricate NBRswith various area/volume ratios. The microfluidic el-ements were sealed to conventional glass microscopeslides, which served as the substrate. As evident from databelow, there was no need to selectively modify the sub-strate to confer a hydrophilic character to its surface, al-though selective patterning may be necessary to confinecell growth to particular regions within the NBR. We testedseveral geometric configurations (shapes), and a range ofcell confinement configurations within the reactor volume.

328 Prokop et al.

We also tested an integrated system for cell feeding andwaste withdrawal. The first generation of NBRs employedoff-chip supply and withdrawal of nutrients and waste.

The NBR devices were fabricated exclusively fromglass and PDMS. Because their manufacture was con-ducted entirely on-site, the culture chamber, inlet and out-let ports, and connecting microfluidics channels could beadapted according to demand. All NBRs were fabricatedwith soft lithography techniques (Whitesides et al., 2001).The following steps outline device production: (1) Maskfabrication: masks were drawn with CAD and the designswere printed 100× larger than the final mask. The maskimage was reduced 100× onto high contrast slide film us-ing a 35 mm camera. Once developed, the film was used di-rectly as a mask in photolithography. Alternatively, customchrome masks were ordered per CAD design. (2) Masterproduction: a thin film of negative-tone photoepoxy wasestablished on a silicon substrate. The photoepoxy wasselectively exposed through the mask and developed. Anegative relief, or master, was thereby established in theshape of the NBR. (3) Device fabrication and assembly:Liquid PDMS was cast against the master and allowed tocure. Individual devices were cut from the PDMS block,punched with plumbing holes, and bonded to their sub-strates with plasma activation.

Masters were fabricated using conventional pho-tolithography technology. Mask layouts were first draftedin AutoCAD. The designs were then either sent to Ad-vance Reproductions Corp. (North Andover, MA) wherethey were transferred to a chrome mask, or they were usedin our on-site mask fabrication facilities. In the latter case,a printout of the design was reduced in size onto high-contrast photofilm via a 35 mm camera outfitted with a50 mm lens. Regardless of its origin, the mask was usedin contact photolithography to generate masters in SU-82025, a negative-tone photoepoxy (Micro-Chem, Newton,MA). Fluidic elements were cast in Sylgard 184 PDMS(Dow Corning Corp., Midland, MI) against the SU-8 mas-ters. PDMS was prepared by mixing the prepolymer andthe catalyst at a ratio of 15:1. The mixture was degassedunder vacuum (20–50 mtorr) for one hour and cured for 2 hat 80◦C. The cured PDMS was peeled from the master, cutinto individual blocks (devices), and fitted to a glass sup-port. For supply and withdrawal of culture medium, as wellas for cell seeding, input and outlet ports were punchedinto the PDMS block with a modified 16-gauge syringeneedle. Some access ports were siliconized with Sigma-Coat (Sigma Chemical Company, St. Louis, MO) to inhibitcell attachment in their vicinity. Standard 1“×3” micro-scopic glass (Fisher Scientific, Pittsburg, PA) served as themain cell substrate at the bottom of the NBR. Air-plasmatreatment was used to seal the PDMS blocks to their glasssubstrate. Oxygen-plasma oxidation is known to introducesilanol groups at the surfaces, thereby activating adhesive

properties of PDMS (Wang et al., 2003). Instead of oxy-gen plasma, we employed simple air plasma and achievedvery strong seals in the hybrid glass/PDMS device (Mc-Donald and Whitesides, 2002). Both PDMS device andglass substrate were placed into a plasma sterilizer (Har-rick Scientific Corp., Ossining, NY) and exposed to airplasma for 20 s. They were then promptly removed andplaced in mutual contact, and a strong, irreversible bondformed. Finally, the access ports were plumbed with 22-gauge stainless steel capillary tubes connected to flexiblevinyl tubing. Heat-shrinkable tubing (Advanced PolymersInc., Salem, NH) secured all plumbing connections to en-sure waterproof sealing using Microtorch MT-10 (MasterAppliance Corp., Racine, WI). Automated syringes pumps(WPI, Saratosa, FL) supplied controlled flow to the mi-croscale devices.

We have not yet embarked on topographical or physico-chemical substrate patterning to modulate cell phenotypeand cell behavior and function, but such investigationsmay be beneficial. Topographical patterning refers to theestablishing of shape or texture patterns on the substrate,while physicochemical patterning incorporates chemicaladhesion, or electrical/physical force imposition on cells(Jung et al., 2001). The integration of such functionalmodules into a single device is the goal of the micro to-tal analysis system (µ-TAS), and may offer highly ef-ficient, simultaneous analyses in genomics, proteomics,and metabolomics (Lee and Lee, 2004).

Design. The diameter of the Circular NBR chamber was825 µm and is shown in Figure 1. The chamber incor-porated a 275 µm post, and assumed a volume of ap-proximately 20 nL (at a depth of 45 µm). The Perfu-sion Grid/Sieve NBR (PG-NBR) enclosed a similar vol-ume, and incorporated a sieve with openings rangingfrom 3 (Figure 2) to 8 µm. The Multitrap NBR (MT-NBR; Figure 3) was designed larger to accommodatemany miniature traps that were outfitted with sieves whose

Fig. 1. Schematic of the first generation CP-NBR with the central post,without the seeding channel.

NanoLiterBioReactor 329

Fig. 2. Schematic of PG-NBR with 3 µm sieve with a separate channelfor seeding.

Fig. 3. Schematic of perfused MT-NBR with multiple trapping sieves,capable of generating nutrient gradients.

openings were similar to those of the PG-NBR. By assem-bling the PG-NBR and MT-NBR, we succeeded in fabri-cating a sieve/grid system by means of soft lithography.Initially, the depth of the devices was 45 µm; later we alsotested 25 and 8 µm depths.

Sterilization. To minimize contamination, all NBR de-vices were sterilized under ultra-violet light just prior touse. A UV-TipCleanerTM (Bioforce Nanosciences, Inc.,Ames, IA) provided the necessary UV energy. The as-sembled NBRs were positioned at a distance of 5 cm fromthe instrument’s mercury lamp, and exposed for 10 min atthe maximum power. We have previously established thatthese parameters are effective in bioburden removal by us-ing B. subtilis spores and vegetative cells as a bioassay (seealso Moisan et al., 2001). In addition to device steriliza-tion, the UV treatment was reported to render hydrophilicthe originally hydrophobic PDMS surface (Wang et al.,2003; Efimenko et al., 2002; Kim et al., 2000). We be-lieve that this condition aided in chamber wetting and cellseeding.

Equipment assembly. Cell culture was conducted withina custom-manufactured Plexiglas incubation box simi-lar in design to those available from Buck Scientific(www.bucksci.com). The environment inside the boxwas maintained at a constant air/5% CO2 and 37◦C(AirThermTM, WPI). The box was fitted to the stage ofan inverted fluorescence microscope (CK40F Olympus)equipped with a digital camera (QImaging MicropublisherMP-CLR-10). This configuration allowed direct observa-tion and recording of cell status. PDMS is optically trans-parent from 240 to 1100 nm; 240 UV cutoff, allowingfor fluorescence microscopy (McDonald and Whitesides,2002).

Fluorescence microscopy. Cell viability was determinedvia a Trypan Blue (Sigma) exclusion test, Acridine or-ange/Ethidium bromide (AO/EB; Molecular Probes, Eu-gene, OR) stain, and VybrantTM DiI (Molecular Probes)and by fluorescence microscopy. The pictures wererecorded by means of a CCD camera. DilI stain has beenselected as a result from the testing of several possible can-didates for noncytotoxic staining of cells in situ. It couldbe applied successively in vitro.

PDMS and gas permeability. PDMS is known to behighly permeable to gases (Mekel et al., 2000; De Boet al., 2003; Charati and Stern, 1998). High permeabil-ity for oxygen and carbon dioxide are of benefit for ourapplication; thus it was not necessary to explore methodsto deliver or remove these gases to/from the cell culturechambers.

Physical characterization/shear rate. Shear rates werecalculated for a certain NBR geometry in order to obtain asense for ranges experienced by cells exposed to mediumflow (see Culture Tests).

Cell lines and media. Standard media were used to cul-tivate selected cell lines: (1) Mouse fibroblasts (ATCC# CRL-10225); DMEM supplemented with 4 mM L-glutamine and adjusted to contain 1.5 g/L sodium bicar-bonate and 4.5 g/L glucose, FBS, 10%. (2) CHO Epithelialcells (ATCC # CRL-1981); Hem’s F12 medium with 0.05to 0.1 mg/ml G418, FBS 10%. (3) Hepatocytes (ATCC# CRL-2254); DMEM/Hem’s F12 medium 1:1 supple-mented with 2.5 mM L-glutamine and adjusted to contain1.2 g/L sodium bicarbonate, 15 mM HEPES and 0.5 mMsodium pyruvate, 5 µg/ml insulin, 5 µg/ml transferrin,5 ng/ml selenium, 40 ng/ml dexamethasone and FBS10%. (4) Hybridoma cells (ATCC CRL-16060); serum-free, hydrolysate-free IMDM formulation, comprising

330 Prokop et al.

glutamine-free IMDM basal medium, 4.0 mM glutamine,10 mg/L insulin, 5 mg/L holo-transferrin, 2.44 µL/L 2-aminoethanol, 3.5 µL/L 2-mercaptoethanol, and 10 U/mLpenicillin-10 µg/mL streptomycin.

NBR seeding. Gravity seeding (Powers et al., 2002) wasroutinely used to introduce cells into the NBR. Prior toseeding, cell suspension, obtained via trypsinization of astock T-flask, was filtered to 30 µm to remove cell aggre-gates. This procedure prevented the NBR channels andintegrated filters from clogging with aggregated cells.

Sensing. Integrated planar array sensors for monitor-ing cell culture within the NBR were fabricated us-ing an immobilized enzyme technology, Nafion coat-ing, and electrochemical principles (Moussy et al., 1994;Steele et al., 1991; Gerritsen et al., 2000). The sensorswere intended specifically for monitoring pH and glucoseactivity.

Biocompatibility tests of materials usedfor NBR fabricationBefore cell culture was attempted in the NBR, biocompati-bility tests were conducted on all compositional materials.The subjects of this testing included the glass substrate,cured PDMS polymer, and all potential extracellular ma-trix (ECM) components, including collagen, gelatin, fi-bronectin, laminin and poly-lysine. Standard 12-well mi-croplates were used for PDMS testing. PDMS was intro-duced as a layer at the bottom of plates, sterilized withhelp of 70% isopropyl alcohol, extensively (5×) washedwith sterile water and coated with ECM components. Thecoating densities applied were 8.25, 1.67, 2.1, 1.38 and2.1 µg/cm2, respectively, in line with standard coating(Sigma). The three cell lines were tested for viabilityand proliferation over extended periods of time rangingfrom 3 to 5 days. Viability and proliferation were eval-uated based on a DNA assay via fluorescent PicoGreenassays. Glass substrate tests were carried out in eight-chamber CC2 Glass slides (Lab-Tek R© Chamber SlideTM

System, Nalge Nunc Int., Naperville, IL) and standard tis-sue culture polystyrene dishes were employed as controlsubstrates.

Because we anticipate the incorporation of processedsilicon in future NBR sensors, additional biocompatibil-ity tests were conducted on silicon nitride and oxide ma-terials. Silicon (100) wafers with a layer of low pressurechemical vapor deposited (LPCVD) silicon nitride, andthose with thermally grown silicon oxide, were the sub-jects of this study. The wafers with LPCVD silicon nitridelayers were provided by Motorola, Inc., while those withthermal oxide were prepared at North Carolina State Uni-

versity. 10 mm by 10 mm samples were sectioned fromthe wafers using a diamond tip scribe.

Plating experiments (for glass substrate) were con-ducted in standard culture setting to delineate the ef-fect of cell seeding density on cell viability and survival.The PicoGreen DNA test was used to evaluate the ex-tent of the cell growth achieved in 1–5 days followingseeding.

Testing of NanoliterBioReactor designDemonstration of batch and continuous culture growthof selected cell lines within the NBR. Three cells lineswere selected for detailed testing: fibroblasts, CHO cells,and hepatocytes. In all cases, a serial dilution of a standardculture was used to seed the NBR content. Cells wereintroduced into the NBR via a static mode to achieve thedesired number of cells per reactor. This population wasverified with microscopy. In all cases, the cells proliferateduntil the substrate was covered by a monolayer, at whichpoint contact inhibition became influential, and growthceased. Cells were allowed to populate the NBR contentas attached cells. During all experiments, the NBR wasenclosed within a cell culture incubator, which maintainedenvironmental temperature and gas levels. Untreated glasswas used as the substrate for this set of tests.

Medium perfusion was facilitated by an UltraMicroP-ump II nanoinjector actuated with a Micro 4 controller(both by WPI). The pump was outfitted with a sterile1 mL B-D syringe, which served as the medium reser-voir. Medium fluid was drawn into the supply line andthen delivered to the NBR chamber at the required rate.This system allowed for batch, fed-batch, and continu-ous feed configurations, as well as in situ standard andfluorescence microscopy of the culture progress. Mediawere vacuum de-gassed to remove dissolved gas and toprevent the formation of bubbles. The controllable exter-nal pumps provided automated, periodic culture feedingand waste withdrawal in an overflow mode. Media andanalytes were supplied to the cells in the NBR chambervia microchannels with cross-sectional area of approxi-mately 250 µm2. Waste was withdrawn through channelsof similar dimension. The fluidic system could easily beadapted to address and control individually each chamberin a multiple-NBR arrangement.

In addition to a direct microscopic count, standardcell viability assays were performed to monitor exper-imental progress, namely cell growth kinetics and to-tal cell count. At the terminus of each experiment, cellswere stained with AO/EB OT Trypan blue, and viabilitywas thereby assessed. The end-point criteria for evalu-ation of culture growth and proliferation characteristicswere determined by cell viability and surface coveragerates.

NanoLiterBioReactor 331

Protocol and system parameters for fed-batch andcontinuous growth configurations. Cultures were testedunder two basic conditions: actively growing populations,and quiescent cultures whose growth was limited by con-tact inhibition or lack of nutrients. The actively grow-ing cultures were perfused frequently with fresh medium.Volumetric flow rates for the medium ranged from 5–50nL/min. The cell longevity may be important for routineuse in a mass-screening program. The maintenance statewas induced by supplying media at lower rates, therebyretarding cell growth to a stagnant status (Tolbert, 1985).Typically, growth can be arrested by removal of growthfactors (serum).

Results and Discussion

Biocompatibility testing of different materialsGlass, PDMS, polystyrene (PS), silicon nitride (Si3N4),and silicon oxide (SiO2) were tested for their suitabilityas substrates in the NBR. Fibroblasts, CHO cells, and hep-atocytes were first tested on bare glass substrates, and glasscoated with ECM proteins (Figures 4–6). For all threecell lines, plain glass appeared to be ample for cell cul-turing, although some improvement was noted with ECMcoated glass. This finding has significant consequence forNBR design because the main substrate material is in-variably glass. Next, the biocompatibilities of PDMS andPS were considered. Figures 7–9 compare the resultsobtained from PDMS filled wells and plain PS. The in-fluence of additional ECM coatings was also considered.For each cell line tested, non-treated PDMS proved tobe an inferior substrate compared to polystyrene. Whilesome growth is noted for CHO cells on plain PDMS, ECMcoating improves its biocompatibility. Finally, the effectof plasma exposure on PDMS was investigated (Figures 7

Fig. 4. Proliferation of fibroblasts on glass (and coated glass) ascompared to PS( legend: PS—standard tissue cell culture qualitypolystyrene, COL—collagen coating, FIB—fibronectin coating,LAM—laminin coating, GEL—gelatin coating, PLL—poly(lysine)coating. (Reproduced with permission of Materials Research Society.)

Fig. 5. Proliferation of CHO cells on glass (and coated glass) ascompared to PS (legend same as in Figure 4).

Fig. 6. Proliferation of hepatocytes on glass (and coated glass) ascompared to PS (legend same as in Figure 4).

and 10). The results clearly show considerable improve-ment for the plasma-treated PDMS. We can therefore con-clude that plasma treatment of our fluidic elements doesnot adversely affect cellular response.

Finally, preliminary tests were performed on siliconnitride and oxide supports (data not shown). Fibroblastsand CHO cells were seeded at a density of 10,000/cm2

and evaluated against the glass and PS controls. At theconclusion of day 3, fibroblasts exhibited 75 and 65% ofthe glass control (evaluated as DNA); CHO cells 18 and10% of the control.

Seeding densityLow seeding densities retard proliferation in some celllines because of the absence of suitable cell-cell andECM-cell contacts. Seeding density experiments for threebasic cell lines were conducted in standard culture en-vironments on glass substrates (Figures 11–13). Verylow seed densities (100–200 cells/cm2) prevented vigor-ous growth and proliferation in some cases. For exam-ple, the growth of hepatocytes was particularly sensitiveto low seed numbers (Michalopoulos et al., 1982). Re-sults clearly show that a minimum critical density is ini-tially required for some cell lines to commence healthy

332 Prokop et al.

Fig. 7. Proliferation of fibroblasts on PDMS (legend same as in Figure 4).

Fig. 8. Proliferation of CHO cells on PDMS (legend same as inFigure 4). (Reproduced with permission of Materials Research Society.)

growth. On the other hand, some cell lines (e.g., CHO)are capable of developing a colony from a single seededcell (Konrad et al., 1977; Park et al., 1987; Pomp et al.,1996; Michalopoulos et al., 1982). These observationsare of great consequence for initiating the cell growthin the NBR environment. One way of improving cellplating efficiency is to employ ECM, growth factors,and conditioned media. Summarizing, the employmentof the NBRs for mammalian cell culturing opens a newparadigm of cell biology, so far largely neglected in theliterature.

The above results serve as models for cell behaviorwithin the NBR. For CP-NBR (see below), 5, 20 and50 cells seed per NBR corresponded to about 140, 280and 1,400 cells/cm2 of the NBR area. Typically, we em-ployed between 10–20 cells per seed, somewhat on the lowside.

Fig. 9. Proliferation of hepatocytes on PDMS (legend same as inFigure 4).

NBR design: CP-NBR, PG-NBR, MT-NBRSeveral NBR designs were conceived, manufactured, andtested. One is called the “circular NBR with central post”(CP-NBR), another, the “perfusion grid NBR” (PG-NBR),and third version, the “multitrap NBR” (MT-NBR). TheCP-NBR involved an 825 µm diameter culture chamberof 40 µm height, fitted with a central post of diameter275 µm. The addition of the post improved fluid distri-bution by eliminating dead zones populated with non-perfused cells. The chamber was connected to inlet andoutlet ports by 100 µm channels (also of 40 µm height).The fluidic elements were sealed against a glass sub-strate on which the cells were growing. The CP-NBRchamber contains a net volume of 20 nL. The perfusionPG-NBR was difficult to design without the cell leak-age between the sieve and the substrate. Lowering downthe chamber height to 8 µm, some success was noted.

NanoLiterBioReactor 333

Fig. 10. Proliferation of fibroblasts on coated PDMS and plasma-treated PDMS (legend same as in Figure 4).

Fig. 11. Effect of plating density on proliferation of fibroblasts (notethe lower detection limit is as low as 10 cells).

Reproducibility of using such design was not assured,however. Only when the design of sieve posts was changedto a more robust one, with more material attached to theglass bottom, we were able to fabricate such NBRs withoutany leaking. The MT-NBR would allow a dynamic seed-ing avoiding the cell attachment to coated areas of thesubstrates.

Culture testsBatch growth within the NBR is not a viable possibilitybecause the medium volume is extremely small, and cellscould therefore suffer from a nutrient limitation even inearly culture stages. (See contrast with a static T -flask cul-ture environment with huge media layer above the attached

Fig. 12. Effect of plating density on proliferation of CHO cells.

Fig. 13. Effect of seeding density on proliferation of hepatocytes.

334 Prokop et al.

cells.) However, we employed static seeding just prior tothe start of the feeding process to allow cell attachment.Typically, fibroblasts required a 2–3 hr attachment period,while the other cell lines required close to 16 h. The attach-ment period was not necessary in the case of the PG-NBRbecause its design offers total cell retention. For this rea-son, only fed-batch and continuous modes were used inthe PG-NBR. Typically, cells were perfused at a rate cor-responding to one-fourth to one-half of the chamber vol-ume per minute. Higher perfusion rates were also tested,including a continuous removal of mitotic cells by shear.This way a quiescence state of the remaining cells (forCHO cells) was obtained. Flow rates conducive to cell qui-escence were observed to be on the order of 5 nL/min forall cells lines we considered, but feeding rates lower thanthis led to poor viability. For all three cell lines, viabilityrates averaged 95% under a continuous feed regime. Veryfew viability measurements of mammalian cells on glassor PDMS substrates have been reported (e.g., van Kootenet al., 1998). Nevertheless, such experiments, in line withour results, demonstrated that the glass/PDMS chamber isa satisfactorily biocompatible environment. It should bementioned that the shear stresses cells experienced in theNBRs ranged from 5·10−5 to 1·10−2 dynes/cm2 for flowrates of 5–50 nL/min. These values are too low to affectcell physiology (Prokop and Bajpai, 1992). For fibroblasts,higher flow rates yielded elongated cell morphology andlining up with the fluid flow stream. Typical examples offibroblast and CHO cultures (in CP-NBR) are illustratedin Figures 14 and 15, exemplary employment of MT-NBRis in Figure 16.

Fig. 14. Fibroblasts in CP-NBR, 2 day culture, low seed (20 cells), with focal adhesion points.

Hepatocytes were cultured 10 days (Figure 17). Wedid not make any attempt to follow their function anddid not establish the maintenance of differentiated phe-notype. We have reserved these studies for the MT-NBR,in which nutrient gradients can easily be established. Wealso intentionally avoided the use of hepatocyte aggre-gates (Parsons-Wingerter and Salzman, 1993) as a seedmaterial in order to minimize clogging of NBR chan-nels. Aggregate use is reserved for the MT-NBR designonly.

A few experiments were performed on hybridomas.The PG-NBR was particularly adequate in retaining thesecells and maintaining reasonable viabilities in a perfusedstate lasting up to 5 days. Beyond that time, cell vi-abilities were much lower as accessed by Trypan bluestain.

Special consideration of cell deformability is relevantto the sieve design in both PG-NBRs and MT-NBRs.We observed that some rounded-off freshly-trypsinizedcells (resulting from T -flask culture), which were usedas seed, could penetrate the sieve perforations and com-pletely squeeze through in some instances, although thefiltering device appeared perfect. The mean diametersof cells right after the trypsinization were thus assessedand found to be in the range of 11–18 µm for fibrob-lasts, 11–25 µm for CHO and 16–30 µm for hepa-tocytes. We thereby concluded that sieve openings of3, 5 and 8 µm allowed some penetration due to celldeformation.

We have not yet addressed the option of cell patterning.It may not be so critical at this stage of the project, although

NanoLiterBioReactor 335

Fig. 15. CHO cells in CP-NBR, 5 day culture, spindle shape cells and few rounded mitotic cells. (Reproduced with permission of Materials ResearchSociety.)

some advantages could be visualized (e.g., provided thatchannels in and out of the NBR are free of cells, wecould access a measurement of several metabolic up-take rates across the NBR). Unfortunately, no pattern-ing effort will ever produce a complete absence of cells

without difficulty. A partial solution was achieved withthe silicone coating as previously mentioned. The im-portance of micropatterning is discussed in more de-tail elsewhere (Folch and Toner, 2000; Jung et al.,2001).

Fig. 16. Freshly seeded CHO cells in MT-NBR (cells captured in small traps; some visible outside of traps).

336 Prokop et al.

Fig. 17. Hepatocytes in PG-NBR (4d culture growing in aggregates; few cells penetrated through the sieve channel on the right).

Growth in confined spaceAn observation on the contact inhibition of cells in con-fined spaces resulted from NBR sieve design work. Asmentioned, some attempts were made to limit the sieveleakage by lowering the chamber height to 8 µm. Cells,especially larger ones such as hepatocytes, featured a “flat”morphology while they were squeezed between the NBR’sceiling and floor, and accordingly grew in size asymmetri-cally. Under these conditions, all three cells lines featuredvery limited proliferation and quiescence over period of5 days. While our original consensus was that the util-ity of the NBR would be limited when the chamber ispacked full with cells, we now realize that this configura-tion provides an interesting opportunity for the study ofphysiology in confined spaces. In fact, a confined spacemay help to mimic the three-dimensional cell-cell con-tact behavior. At this time, no attempts were exerted tocharacterize the quiescent state of cells by a molecularmeans.

NBR reuseSpecial attention was given to cell removal and lysis fol-lowing the employment of the NBR. In an effort to achievea complete lysis of cells and of the ECM components fromthe NBR interior, we tested several gentle lysis reagentsdesigned to release nucleic acids and proteins. Trypsin it-self was partially effective. However, cell debris and ECMcomponents are typically left behind and contaminate theNBR space. A successful solution would enable a reuse

of the NBR, although the low cost of fabrication rendersit an easily disposable device.

Sensing in NBR and outlookEvaluation of NanoBioReactor temporal physiologic re-sponses and determination of the sensor sensitivity to in-terventions that alter the homeostatic state of three celllines in continuous culture is an interesting proposition.Future research efforts will concentrate on expanding theculture experiments to include longer-term assays. We alsointend to explore the effects of shear stress and gradientson cell physiology within the multitrap and grid NBRs.The effect of complete cell retention in the PG-NBR willbe further investigated, as will mitotic cell removal in thecircular NBR with open inlets and outlets. We also in-tend to perfect our sensing capabilities (particularly forgradient studies) via NBRs boasting integrated sensors.Further studies will also include considerations such ascell differentiation (expression of specific function) vs.proliferation. These efforts are aimed at obtaining an im-proved fundamental understanding of cell physiology atthe small-scale number density that the NBR affords. Wehope to enable extension of this concept into a rationaltool for studying fundamental cell biology phenomena.

General discussionMicro- and nano-fabrication is considered an enablingtechnology that is expected to generate significant new

NanoLiterBioReactor 337

product/market opportunities through integration of sci-ence and technology. The scope of this field is to en-abling novel capabilities in various applications by cre-ating objects with dimensions in the range of nanometersand micrometers to millimeters (Voldman et al., 1999).These objects can be stationary or moving structures,namely growth chambers, channels, pumps, sensors, etc.The ultimate objective is to design and control experimentsat the micrometer scale. Possible applications include ar-eas of molecular biology, biochemistry, cell biology, med-ical devices and biosensors. The latter three aspects arebriefly reviewed in this section. Substrate materials (fromboth fabrication and cell culture points of view) often in-

Table 4. Status of micro- and nanobafricated cell culture devices

Device function Volume In situ sensing Comments Reference

Perfusion mimic ofpharmacokinetic animalbehavior

4.7–22.3 mL DO Rat hepatocyte, L2 rat lungcells, adipocytes, 24 hculture

Ghanem and Shuller, 2000;Viravaidya and Shuller,2004; Sin et al., 2004

Well-plate adapted for CO2

measurement1.1 mL CO2 production Mouse fibroblasts,

polystyrene, 6 dYang and Balcarcel, 2004;

Balcarcel and Clark, 2003Well-plate adapted

microbioreactor750 µL Biomass, DO, pH E. coli, plastic, 16 h Maharbiz et al., 2004

Bioartificial liver in gradientreactor

0.12 mL None Primary rat hepatocytes,glass/polycarbonate, 8 d

Allen and Bhatia, 2003

CytosensorMicrophysiometer

2.8 µL Acidification rate; lactateproduction, oxygen andglucose consumption

Mammalian cells, membranechamber, few hours

Hafner, 2000; Eklund et al.,2004

Biomedical diagnostic systemfor epithelial cells

N/A Impedance (growth) Epithelial cells,polycarbonate/PDMS, 10 d

Hediger et al., 2001

First microreactor liver mimic 27 nL None Mouse L cells, primary rathepatocytes, 14 d

Weibezahn et al., 1995

High-throughputmembrane-aeratedmicrobioreactor

5–50 µL DO, pH and optidaldensity

Bacterial fermentations Zanzotto et al., 2004

Perfused 3-D liver culture 21 nL Light & fluorescencemicroscopy

Primary rat perfusedhepatocytes, silicon, 14 d,not individuallyaddressable

Powers et al., 2002

Microfluidic single-cellanalysis

0.5 nL Ca flux Jurkat, U937 cells, PDMS, nogrowth

Wheeler et al., 2003

Single cell clinic 0.2 nL Impedance (growth) Xenopus, silicon, 6 h Jager et al., 2002Perfused 3-D liver culture 16 nL None HepG2, PDMS, 10 h Leclerc et al., 2002 and 2004Bacterial culture 5.4 nL Impedance (growth) Microbial, PDMS/silicon,

22 hChang et al., 2003

Cell-based silicon sensor 6 nL Acidification Colorectal carcinoma, silicon,several days

Brischwein et al., 2003

Capillary cultivation system 20 nL None Recombinant cells, silicon,short-term

Grodrian et al., 2002

Picocalorimetry at single celllevel

0.7 nL Picocalorimetry (pC) Adipocytes, cardiomyocytes,polyimide, short-term

Johannessen et al., 2002

Microvascular mimic N/A None Endothelial cells, PDMS,28 d

Borenstein et al., 2002

Single-cell monitoring 0.05 nL Amperometric PC12 cells, glass/PDMS,short-term measurement

Huang et al., 2004

NanoBioReactor 10–100 nL pH, glucose, pC Fibroblasts, CHO, Hep2G,glass/PDMS, 14 d

This paper

clude biocompatible materials such as silicon dioxide andnitride, glass, PDMS and other polymers.

The enabling aspects of micro/nano technologies inthe field of biology are best defined as study of biologi-cal phenomena at the micro- and nano-scales. Some nov-elties introduced via nanotechnology will be discussedlater. Manipulations of the physical world at this scalewith the intent of detecting, separating, manipulating andcharacterizing cells include cellular adhesion, signal trans-duction, motility, deformability, metabolism and secretion(Zieziulewicz et al., 2003). Morphological changes, sur-face detachment, apoptosis, necrotic growth, and parti-cle gathering are additional physiologic activities relevant

338 Prokop et al.

to this scale. Special designs accommodate the genera-tion of complex gradients (e.g., of substrate, chemokines,etc.) with spatial and temporal control. Biosensors canbe embedded within such devices to allow for molecule-,cell-, and tissue-based devices. Physiologic functions thatcan be readily monitored via multicomponent analyte sys-tems include pH fluctuations, oxygen consumption, lac-tate production, glucose utilization, and redox potential(Zieziulewicz et al., 2003). Table 4 lists various micro-,milli-, and nanoliter scale culture devices, constructedwith the aim to monitor short- and long-term cultures inconfined spaces. Only very few devices listed in this tableallow for in situ physiologic measurements or for perfu-sion mode of operation. The emergence of such devices,however, has been unprecedented in last few years.

Future iterations of the NBR should incorporate waysto (1) Automate cell loading into the NBR; (2) Developsingle and multi-parameter control loops for NBR systemvariables; (3) Optimize control parameters for each cellline; (4) Add additional sensors and control loops to NBR;(5) Extend the number of cell lines that can be culturedstably; (6) Determine the maximum duration of stable cellculture in the NBR; (7) Enhance automation capabilitiesto operate multiple NBRs; (8) Incorporate highly-versatilemicrofluidics into the NBR, including on-chip pumps andvalves; (9) Conduct comprehensive metabolic flux anal-yses using the optimized, fully instrumented NBR; and(10) Determine the requirements for construction of NBRarrays. We anticipate that the NBR will demonstrate itselfas a technology suitable for incorporation into massivelyparallel, multiphasic, high-content, toxicology screeningsystems.

Acknowledgments

This work was supported by in part by a National Insti-tutes of Health grant 5 R43 RR016124-02 to NanoDeliv-ery, Inc., and by the Vanderbilt Institute for IntegrativeBiosystems Research and Education. Silicon wafers werekindly provided by Bridget Rogers, Chemical EngineeringDepartment.

References

J.W. Allen and S.N. Bhatia, Formation of steady-state oxygen gradientsin vitro. Biotechnol. Bioeng. 82, 253 (2003).

R.R. Balcarcel and L.M. Clark, Metabolic screening of mammalian cellcultures using well-plates. Biotechnol. Prog. 19, 98 (2003).

H. Becker and C. Gartner, Polymer based micro-reactors. Revs. Molec.Biotechnol. 82, 89 (2001).

J.T. Borenstein, H. Terai, K.R. King, E.J. Weinberg, M.R. Kaazempur-Mofrad, and J.P. Vacanti, Microfabrication technology for vascular-ized tissue engineering. Biomed. Microdev. 4, 167 (2002).

C.T. Bratten, P.H. Cobbold, and J.M. Cooper, Micromachining sensorsfor electrochemical measurement in subnanoliter volumes. Analyt.Chem. 69, 253 (1997).

M. Brischwein, E.R. Motrescu, E. Cabala, A.M. Otto, H. Grothe, and B.Wolf, Functional cellular assays with multiparametric silicon sensor.Lab. Chip. 3, 234 (2003).

S.C. Cahrati and S.A. Stern, Diffusion of gases in silicone polymers:Molecular dynamics simulations. Macromol. 31, 5529 (1998).

W.-J. Chang, D. Akin, M. Sedlak, M.R. Ladisch, and R. Bashir,Poly(dimethylsiloxane) (PDMS) and silicon hybrid biochip for bac-terial culture. Biomed. Microdev. 6, 281 (2003).

J.M. Cooper, Towards electronic Petri dishes and picolitre-scale single-cell techniques. Trends Biotechnol. 17, 226 (1999).

E.J. Crosby, Experiments in Transport Phenomena (Wiley, New York,1961).

I. De Bo, H. Van Langenhove, P. Pruuost, J. De Neve, J. Pieters, I.F.J.Vakelecom, and E. Dick, Investigation of the permeability and selec-tivity of gases and volatile organic compounds for polydimethylsilox-ane membranes. J. Membrane. Sci. 215, 303 (2003).

K. Efimenko, W.E. Wallace, and J. Genzer, Surface modification ofSylgard-184 Poly(dimethyl siloxane) networks by ultraviolet and ul-traviolet/ozone treatment. J. Coll. Interface Sci. 254, 306 (2002).

S.E. Eklund, D. Taylor, E. Kozlov, A. Prokop, and D.E. Cliffel, A micro-physiometer for simultaneous measurement of changes in extracellu-lar glucose, lactate, oxygen, and acidification rate. Anal. Chem. 76,519 (2004).

A. Folch and M. Toner, Microengineering of cellular interactions. AnnRev Biomed Eng 2, 227 (2000).

M. Gerritsen, A. Kros, V. Sprakel, L.A. Lutterman, R.J.M. Nolte, andJ.A. Jansen, Biocompatibility evaluation of sol-gel coatings for sub-cutaneously implantable glucose sensors. Biomat. 21, 71 (2000).

A. Ghanem and M.L. Shuler, Characterization of a perfusion reactorutilizing mammalian cells on microcarrier beads. Biotechnol. Progr.16, 471 (2000).

A. Grodrian, J. Metze, Th. Henkel, M. Roth, and J.M. Kohler, Segmentedflow generation by chip reactors for highly parallelized cell cultivation.In Biomedical Applications of Micro- and Nanoengineering, edited byD.V. Nicolau and A.P. Lee, Proc. SPIE 4937, 174 (2002).

F. Hafner, Cytosensor microphysiometer: Technology and recent appli-cations. Biosensors Bioelectr. 15, 149 (2000).

S. Hediger, A. Sayah, J.D. Horisberger, and M.A.M. Hijs, Modular mi-crosystem for epithelial cell culture and electrical characterization.Biosens. Bioelectr. 16, 689 (2001).

W.-H. Huang, W. Chang, Z. Zhang, D.-W. Pang, Z.-L. Wang, J.-K.Cheng, and D.-F. Cui, Transport, location, and quantal release mon-itoring of single cells on a microfluidic device. Anal. Chem. 76, 483(2004).

E.W.H. Jager, C. Immerstrand, K.H. Peterson, K.-E. Magnusson, I.Lundstrom, and O. Inganas, The cell clinic: Closable microvials forsingle cell studies. Biomed Microdev 4, 177 (2002).

E.A. Johannessen, J.M. Weaver, L. Bourova, P. Svoboda, P.H.Gobbold, and J.M. Cooper, Micromachined nanocalorimetric sen-sor for ultra-low-volume cell-based assays. Analyt. Chem. 74, 2190(2002).

D.R. Jung, R. Kapur, T. Adams, K.A. Giuliano, M. Mrksich, H.G.Craighead, and D.L. Taylor, Topographical and physicochemical mod-ification of material surface to enable patterning of living cells. Crit.Rev. Biotechnol. 21, 111 (2001).

J. Kim, M.K. Chaudhury, and M.J. Owen, Hydrophobic recovery of poly-dimethylsiloxane elastomer exposed to partial electrical discharge. J.Coll. Interface. Sci. 226, 231 (2000).

M.W. Konrad, B. Storrie, D.A. Glaser, and L.H. Thompson, Clonal vari-ation in colony morphology and growth of CHO cells cultured on agar.Cell 10, 305 (1977).

NanoLiterBioReactor 339

E. Leclerc, Y. Sakai, and T. Fujii, Cell culture in 3-dimensional microflu-idic structure of PDMS (polydimethylsiloxane). Biomed. Microdev.5, 109 (2002).

E. Leclerc, Y. Sakai, and T. Fujii, Microfluidic PDMS (Polydimethyl-siloxane) bioreactor for large-scale culture of hepatocytes. Biotechnol.Progr. 20, 750 (2004).

S.J. Lee and S.Y. Lee, Micro total analysis system (µ-TAS) in biotech-nology. Appl. Microbiol. Biotechnol. 64, 289 (2004).

M.M. Maharbiz, W.J. Holtz, R.T. Howe, and J.D. Keasling, Microbiore-actor arrays with parametric control for high-throughput experimen-tation. Biotechnol. Bioeng. 85, 376 (2004).

A. Manz, N. Graber, and H.M. Widmer, Miniaturized total chemicalanalysis systems: A novel concept for chemical screening. SensorsAct B1, 244 (1990).

J.C. McDonald and G.M. Whitesides, Poly(dimethylsiloxane) as a mate-rial for fabricating microfluidic devices. Account Chem. Res. 35, 491(2002).

T.C. Mekel, V.I. Bondar, K. Nagai, B.D. Freeman, and I. Pinnau, Gassorption, diffusion, and permeation in poly(dimethylsiloxane). J. Poly-mer. Sci., Polymer. Phys. B 38, 415 (2000).

G. Michalopoulos, H.D. Cianciulli, A.R. Novotny, A.D. Kligerman, S.C.Strom, and R.L. Jirtle, Liver regeneration studies with rat hepatocytesin primary culture. Cancer Res 42, 4673 (1982).

M. Moisan, J. Barbeau, S. Moreau, J. Pelletier, M. Tabrizian, and L.H.Yahia, Low-temperature sterilization using gas plasmas: A review ofthe experiments and an analysis of the inactivation mechanisms. Int.J. Phamaceut. 226, 1 (2001).

F. Moussy, D.J. Harrison, and R.V. Rajotte, A miniaturized Nafion-basedglucose sensor: In vitro and in vivo evaluation in dogs. Int. J. Artif.Organs. 17, 88 (1994).

A.B. Newman, Trans. Amer. Inst. Chem. Eng. 27, 203 (1931).J.Y. Oldshue, Fluid Mixing Technology (McGraw-Hill, New York, 1983).C.H. Park, B.F. Kimler, and T.K. Smith, Clonogenic assay combined with

flow cytometric cell sorting for cell-cycle analysis of human leukemiccolony-forming cells. Anticancer. Res. 7, 129 (1987).

P.A. Parsons-Wingerter and W.M. Saltzman, Growth versus function inthe three-dimensional culture of single and aggregated hepatocyteswithin collagen gels. Biotechnol. Progr. 9, 600 (1993).

J. Pomp, J.L. Wike, I.J.M. Ouwerkerk, C. Hoogstraten, J. Dvelaar, P.I.Schrier, J.W.H. Leer, H.D. Thames, and W.A. Brock, Cell density de-pendent plating efficiency affects outcome and interpretation of colonyforming assays. Radiother. Oncol. 40, 121 (1996).

M.J. Powers, K. Domansky, M.R. Kaazempur-Mofrad, A. Kalezi, A.Capitano, A. Upadhyaya, P. Kurzawski, K.E. Wack, D.B. Stolz, R.Kamm, and L.G. Griffith, A microfabricated array bioreactor for per-fused 3D liver culture. Biotechnol. Bioeng. 78, 257 (2002).

A. Prokop and R.K. Bajpai, The sensitivity of biocatalysts to hydro-dynamic shear stress. In Advances in Applied Microbiology, editedby A. Laskin (Academic Press, New York, 1992), vol. 37, pp. 165–232.

A. Prokop, Systems analysis and synthesis in biology and biotechnology.Int. J. Gen. Syst. 8, 1 (1982).

A. Prokop, Challenges in commercial biotechnology. Part I: Product,process and market discovery. Advan. Appl. Microbiol. 40, 95 (1995).

A. Sin, K.C. Chin, M.H. Jamil, Y. Kostov, G. Rao, and M.L. Shuller,The design and fabrication of three-chamber microscale cell culturedevices with integrated dissolved oxygen sensors. Biotechnol. Progr.20, 338 (2004).

K. Slater, Cytotoxicity tests for high-throughput drug discovery. Curr.Opin. Biotechnol. 12, 70 (2001).

J.G. Steele, G. Johnson, W.D. Norris, and P.A. Underwood, Adhesion andgrowth of cultured human endothelial cells on perfluorosulphonate:Role of vitronectin and fibronectin in cell attachment. Biomat. 12, 531(1991).

W.R. Tolbert, Perfusion culture systems for production of mammaliancell biomolecules. In Large Scale Mammalian Culture, edited by J.Feder and W.R. Tolbert (Academic Press, New York, 1985), pp. 97–123.

T.G. van Kooten, J.F. Whitesides, and A.F. von Recum, Influence ofsilicone (PDMS) surface texture on human skin fibroblast proliferationas determined by cell cycle analysis. J. Biomed. Mater. Res. (Appl.Biomater.) 43, 1 (1998).

K. Viravaidya and M.L. Shuller, Incorporation of 3T3-L1 cells to mimicbioaccumulation in a microscale cell culture analog device for toxicitystudies. Biotechnol. Progr. 20, 590 (2004).

J. Voldman, M.L. Gray, and M.A. Schmidt, Microfabrication in biologyand medicine. Annu. Rev. Biomed. Eng. 1, 401 (1999).

B. Wang, Z. Abdulali-Kanji, E. Dodwell, J.H. Horton, and R.D.Oleschuk, Surface characterization using chemical force microscopyand the flow performance of modified polydimethylsiloxane for mi-crofluidic device applications. Eletrophoresis 24, 1442 (2003).

K.F. Weibezahn, G. Knedlitschek, H. Dertinger, W. Bier, Th. Schiller,and K. Schubert, Reconstruction of tissue layers in mechanically pro-cessed microstructures. J. Exp. Clin. Cancer. Res. 14 (Suppl. 1), S41(1995).

G.M. Whitesides, E. Ostuni, S. Takayama, X. Jiang, and D.E. Ingber.Soft lithography in biology and biochemistry. Annu. Rev. Biomed.Engr. 3, 335 (2001).

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, Microfluidicdevice for single-cell analysis. Anal Chem 75, 3249 (2003).

Y. Yang and R.R. Balcarcel, Determination of carbon dioxide productionrates for mammalian cells in 24-well plates. Bio. Techniques. 36, 286(2004).

A. Zanzotto, N. Szita, P. Boccazzi, P. Lessard, A.J. Sinskey, and K.F.Jensen, Membrane-aerated microbioreactor for high-throughput bio-processing. Biotechnol. Bioeng. 87, 243 (2004).

T.J. Zieziulewicz, D.W. Unfricht, N. Hadjout, M.A. Lynes, and D.A.Lawrence, Shrinking the biologic world—Nanobiotechnologies fortoxicology. Toxicol. Sci. 74, 235 (2003).