electronic sensors with living cellular components

Electronic Sensors with Living CellularComponents

GREGORY T. A. KOVACS

Invited Paper

For more than three decades, it has been possible to usemicrolithographically fabricated extracellular electrodes to recordaction potential (AP) signals from electrically active cells suchas neurons of intact organisms. There has also been a steadyevolution of techniques to interface between electronic circuitsand neural or cardiac cells cultured on arrays of such electrodes,mainly directed toward basic science goals. More recently, suchcombinations of living cells and electronics have been harnessedas tools for the detection of chemical and biological toxins,and for screening of pharmacologically active compounds. Thispaper presents a survey of the relevant technologies, cell typesavailable, specific requirements for applications, and discussionof opportunities for future development.

Keywords—Biosensor, cell-based sensor, cardiac culture,microelectrode.

I. INTRODUCTION

The development of biosensors based upon live, electri-cally active cells in culture [cell-based biosensors (CBBs)]represents the convergence of several technologies. Sensorsof this type are by definition two-stage. Live biologicalcells serve as the primary transducers, converting the de-tected molecular signal into signals measured in turn viaa secondary transduction means such as an extracellularelectrode or optical detector. This review is restricted to thecellular signal detection using electrodes, and the interestedreader is directed toward Pancrazioet al. [1], Ziegler [2],and McFadden [3] for more general overviews. It is alsonoteworthy that electronic detection using insulated-gateFET devices (as demonstrated by Fromherzet al. [4], [5]),impedance methods [6]–[10], extracellular pH measure-ments [11]–[15], and micromachined “patch-clamp” devices

Manuscript received November 17, 2002; revised February 19, 2003. Thiswork was supported in part by the DARPA Microflumes Program underContract N66001-96-C-8631 and in part by the Tissue-Based BiosensorsProgram under Contract N66001-99-C-8642.

The author is with the Departments of Electrical Engineering andMedicine, Stanford University, Stanford, CA 94305-4075 USA (e-mail:[email protected]).

Digital Object Identifier 10.1109/JPROC.2003.813580

[16] are beyond the scope of this paper, but quite relevant tothe overall scheme of CBBs.

In a general sense, cell-based sensors are useful becausethey harness highly evolved cellular pathways. As opposedto sensors based on specific molecular detection (such asthose using enzyme reactions [17], binding of antibodies[1], [17]–[19], or hybridization of nucleic acids [20]),cellular responses are physiologic and allow the detectionof unanticipated threat agents (hence without need forpredetermined and fabricated immunoglobulins or DNAprobes). In addition, and importantly, they can be used tostudy the potentially complex cellular effects of mixturesof compounds. These properties make such sensors highlyattractive for detection of chemical and biological toxins,environmental toxins, and for the screening of pharmaceuti-cals. Unlike many of the alternatives mentioned previously,however, the sensitivities of cell-based sensors to biologicaltoxins and chemical agents depend upon the type of cellsused and the nature of the agents.

It is important to note that one cannot directly compareCBBs to the extremely high-gain approaches such as thepolymerase chain reaction (PCR) for DNA [20]–[23]. WithPCR, one might be able to find minute traces of DNA frombiological production of a toxin, but not measure the actualtoxin concentration in a sample. The two techniques maywell be used together, since unknowns to be detected mayinclude infectious agents, for which PCR is well suited if thespecific diagnostic reagents for each organism are availablein advance. The ability of cell-based sensors to detect newtoxins without advance preparation suggests that they willplay a very useful and complementary role to assays basedon molecular recognition.

II. ELECTRONIC COUPLING TOCELLULAR SIGNALS

The recording of extracellular action potentials (APs) iswell established [24], as is the culture of a variety of celltypes that spontaneously generate action potentials that can

0018-9219/03$17.00 © 2003 IEEE

PROCEEDINGS OF THE IEEE, VOL. 91, NO. 6, JUNE 2003 915

Fig. 1. SEM view of a typical microelectrode in crosssection, showing the platinum recording electrode surface.Often, amorphous platinum (“platinum black”—not shown) iselectroplated onto the electrode area to decrease impedance.Courtesy of D. A. Borkholder, Stanford University, Stanford, CA.

be recorded successfully [25]–[27]. The basic interface be-tween the cells and external electronics consists of conduc-tive microelectrodes created by exposing small regions ofthin-film conductors beneath an insulating top layer. Fig. 1shows a typical microelectrode in cross section. In this case,a glass substrate was used, but as can be seen from the priorwork referenced herein, a variety of biocompatible substratesare suitable, such as silicon (potentially including active elec-tronic circuits), plastics, ceramics, etc.

As discussed in several prior publications [24], [28]–[30],an equivalent circuit suitably represents the micro-electrode as a recording interface, with the addition of aparallel nonlinearity when the electrodes are driven for usein stimulation. Tight coupling of an overlying cell to theelectrode will increase the detected AP signal amplitude,as will decreasing the impedance of the electrode (andthereby its thermal noise contribution) [24], [30]. Althoughdifferent physical designs have been demonstrated beyondsimple planar electrodes, including structures to enhancecell-to-electrode coupling (e.g., [31]), the basic conceptsremain the same.

The signal source, action potentials originating at thecell itself, are caused by gated ionic currents traversing thecell membranes. While the specifics are different for neuralversus cardiac cells, the basic concepts are comparable. Thereader is directed to one or more of the literature referencesdiscussing the underlying mechanisms [32]–[34]. Fig. 2illustrates the relationship of the intracellular signal andwhat is actually measured using extracellular electrodes. Themovement of ions in the extracellular solution creates po-tentials that can be detected using a closely spaced electroderelative to a more distant reference electrode. Dependingupon cell type, the signal amplitudes seen range from tensof microvolts through several millivolts, with bandwidthsgenerally between a few hundred hertz and one kilohertz.For neurons, signals tend to be on the lower end of the ampli-tude range. Cardiac cells typically interconnect via cellulargap junctions(effectively resistive couplings) and generateAPs in the form of propagating depolarization fronts of

Fig. 2. Plot of (top trace) intracellular action potential signal(sampled at 20 kHz and postfiltered with a 150-Hz cutoff,eighth-order Butterworth low-pass filter) recorded from an HL-1cell using a whole-cell patch clamp in current-clamp mode, withnormalized-amplitude first, second, and third derivatives shown inthe remaining three traces. Typical extracellular recordings mostclosely resemble the second derivative. Courtesy of K. Gilchrist,Stanford University, Stanford, CA.

considerably larger amplitudes (such a coupled layer ofcells is referred to as asyncitium). Example myocardial APsignals are shown in Fig. 3. In either case, since the signalsare coupled through the electrode capacitance and also sincethe transmembrane current sources active at a given instantare moving relative to the electrode, the recorded waveformsare essentially second time derivatives of the intracellularpotential [24], [34], [35]. This relationship can be seen inFig. 2, but in practice may span a continuum from firstthrough third derivative depending upon how well coupled acell is to an electrode [34], [35].

As mentioned previously, the ability to record APs fromindividual cells requires microelectrodes comparable in sizeto the cells themselves. This calls for electrode dimensionson the order of tens of micrometers or smaller unless thecells are interlinked as in a myocardial syncitium. Electrodesmuch larger than cells will not yield recordable signals dueto decreasing “seal” impedance (voltage division of thesignal), whereas electrodes that are too small will have elec-trical outputs dominated by Johnson noise of the electrodeimpedance’s real component [29], [30].

The ability to produce micrometer-scale electrodes wasan offshoot of the explosive growth of IC technology in the1960s. A wide variety of lithographic, thin-film deposition,etching, and other tools soon became available. These toolswere initially used to make digital logic and analog ICs,but soon were harnessed to fabricate a variety of mechan-ical, chemical, and optical sensors [36], [37]. During thisperiod, researchers began exploring the microfabrication ofelectrodes for recording APs from neurons. Wiseet al. [38],Starr et al. [39], and Wise and Angell [40] demonstratedneedlelike silicon probes with on-board microelectrodes thatwere used to record signals from the brains of intact animals.Similarly, work began on cell- and tissue-culture substrates

916 PROCEEDINGS OF THE IEEE, VOL. 91, NO. 6, JUNE 2003

(a) (b)

Fig. 3. (a) Cultured chick myocardial cells on a microelectrode array. (b) Typical AP recordingsfrom such an array. Courtesy of D. A. Borkholder [24].

equipped with such electrodes, with the earliest work ap-pearing to be that of Thomaset al. [41]. They demonstratedAP recordings from dissociated chick myocardial cells. In1974, Shtarket al. [42] demonstrated the first recordingsfrom neuronal (rat brain) cells on such an array. Neuronalsignals from snail ganglia were demonstrated by Grosset al. [43], [44] using snail ganglia, and from dissociated ratcervical ganglia by Pine [45]. A wide variety of cell typesand microelectrode designs were subsequently demonstrated[46]–[50]. For more detailed surveys, the reader is referredto Stenger and McKenna [51] and Pancrazioet al. [1].

Deliberate stimulation of APs has also been demonstratedin dissociated chick cardiac cells by Israelet al.[52], [53] andConnolly et al. [35], Heliosoma(snail) neurons by Regehret al. [31], mouse and chick dorsal root ganglion cells byJimbo and Kawana [54], [55], and hamster spinal cord slicesby Borkholderet al.[56]. These techniques have not yet beenbroadly applied to CBBs, but can readily be employed usingthe same microelectrode arrays used for AP recording.

An additional noteworthy area of research has been thatof patterning of the cultured cells relative to the microelec-trodes using a variety of methods such as physical or surfacechemical [57]–[60]. It should be noted, however, that for thesensor applications described later, these methods have thusfar only succeeded for relatively short periods (days to per-haps weeks) and do not appear to be critical to the realizationof practical CBBs.

While it is thus clear that much the background literatureon this subject dates back several decades, it was only morerecently that researchers began exploring the development ofrealistic sensors using these approaches. With basic sciencedriving the initial efforts, it was realized that such deviceshad the potential to act as broad-spectrum detectors andscreening tools. Grosset al., published some of the earliestlaboratory work with this clearly in mind [61], [62] andcontinued development of such sensors [63]. (It is noteworthythat Keese and Giaever [10] carried out early work with thesame goals, but using impedance rather than action potentialmethods.) Subsequent interest in detectors for chemicaland biological warfare agents spurred the development offieldable instrumentation based on this principle [64]–[66],as discussed later.

III. CELLULAR ELEMENTS

The preceding discussion presented the basics of op-eration of cell-based sensors. The underlying transducerplatform—an array of microelectrodes—is for the mostpart identical for any cell types used. However, the majordeterminant of sensor function, reproducibility, and sensi-tivity is the choice of the primary transduction element—thecultured cells themselves.

To date, work on electrical detection methods has focusedon neuronal and cardiac cell types since they are primarytargets of compounds of interest and because they naturallysupport microelectrode interfaces through their anchoragedependent nature. Within the scope of this review, the can-didate cells discussed generate spontaneous action poten-tials (or, potentially, generate them in response to a “pacing”stimulus). As mentioned previously, other methods such asimpedance measurements can be used with nonelectricallyactive cell types, as can optical detection and indirect (chem-ical) metabolic sensing.

Regardless of a cell type’s organ-specific origins, one canemploy primary (derived from fetal or adult animals) cellsor immortalized cell lines [26], [27], [67]. The latter are notsubject to the Hayflick limit [68] (although older culturescan show marked changes in characteristics), which deter-mines the maximum number of times a cell can divide priorto death, as are tissues we think of as “normal” (in fact,cell lines are akin to or derived from cancers). The poten-tial benefits of using immortalized lines include the abilityto avoid the use of animals, the possibility of more consis-tent cultures, and the potential to utilize various methodsof genetic engineering that are difficult with primary cells[21], [27], [67]. However, genetically engineered primarycells have been demonstrated in CBBs [69].

Primary cells generally are not of a single genetic makeup(an important property of cell lines) but are more likely toclosely represent normal cell targets. Also, primary neuronscan organize themselves more readily into functional net-works of cells with complex behaviors similar to those inan intact organism. At this time, the bulk of the work oncell-based sensors has been done with primary cells, but con-siderable exploration is under way to utilize cell lines forsuch sensors. The search for suitable neuronal cell lines has

KOVACS: ELECTRONIC SENSORS WITH LIVING CELLULAR COMPONENTS 917

Fig. 4. Typical dissociated neuron (mouse spinal) culture atop a microelectrode array. Insets at rightillustrate the variety of morphologies seen in the culture. Courtesy of Prof. G. Gross, University ofNorth Texas, Denton.

yet to yield candidates with strong, reproducible action po-tentials recordable with microelectrodes, but a useful murine(mouse), atrial-derived cardiac cell line, HL-1 cells has beendeveloped [70].

For neurons, a wide variety of tissues have been used asthe source material. The bulk of the sensor work to date hasbeen done using dissociated primary neurons. Such culturedcells form spontaneously active networks, from which fairlycomplex behaviors such as bursting and oscillation emerge(a typical dissociated neuron culture atop a microelectrodearray is shown in Fig. 4). These patterns of activity can pro-vide subtle or dramatic indications of physiologic actions ofexternal agents. As reviewed in Pancrazioet al. [1], dissoci-ated neurons from a variety of invertebrate and mammaliansources have been used successfully. Tissue slices seem lesslikely to be applied in practical sensors due to the complexityof preparing them and the difficulty in maintaining stablecoupling to the electrodes with movement of the sensor.

Successful demonstrations of cardiac-cell CBBs haveutilized chick [35], [52], [53], [64] and mouse primarymyocardial cells [69], and the mouse-derived HL-1 atrialcell line [65], [66]. As mentioned previously, dissociatedmyocardial cells generally form contiguous, physicallyand electrically connected sheets—syncitia—over the elec-

trodes. Fig. 5 shows such a sheet of HL-1 cells culturedatop a CMOS microelectrode array. The gap junctionsthat form between the cells allow them to synchronizetheir activity, resulting in waves of depolarization (andphysical contraction) that propagate across the electrodearray. While the mechanical movement of the cells canpotentially introduce signal artifacts not seen with neurons,the generally synchronous behavior of the cardiac cellsproduces compound action potentials with larger amplitudes(neuronal APs derive from individual cells).

It is worth noting that there are other cell types that couldpotentially produce spontaneous APs. Specifically, neuroen-docrine cells, such as pancreatic islet cells are known to ex-hibit such behavior [71], [72], but have yet to yield successfulAP recordings in microelectrode-based biosensors.

As mentioned previously, it is possible to genetically en-gineer the cells to obtain highly specific response capabili-ties. For example, it is possible toknock out(remove) genesthat code for certain receptors, as long as their absence isnot lethal to the organism or cell. With this approach, dualcell-based sensors, one with the knockout cells and one withnormal, orwild-type, cells can be used, and differences inresponse to a test agent applied to both can be attributed tothe specific receptor that has been removed. Aravaniset al.


Fig. 5. Micrograph of HL-1 myocardial cells growing (eight days after initial seeding) on a CMOSmicroelectrode array [66]. Each electrode is provided with a source follower, and the IC containson-board multiplexing for electrode selection as well as an integrated temperature controller.Courtesy of B. DeBusschere, Stanford University, Stanford, CA.

[69] demonstrated this approach using primary mouse cellsin which the -adrenergic receptors had been knocked out,and showed that they could obtain a truly differential re-sponse between knockout and wild-type cell populations onelectrode arrays. This approach may prove particularly usefulfor the screening and testing of pharmaceutical agents, andcan be scaled up to diverse arrays of engineered cells to becompared to a wild-type population.

While both neuronal and cardiac cell types may prove tohave significant overlaps in terms of the types of agents theycan detect, it is likely that having the complementary celltypes will provide a superior overall capability. In fact, sinceculture conditions for mammalian (for example) tissues tendto be fairly uniform, considerable cell-type diversity is pos-sible within CBB systems, allowing parallel screening of un-knowns using several tissue types in separate culture wells.

For some applications (such as lengthy sensor deploy-ments in military settings), longevity of the cultured cells ona scale of weeks or even months is of great importance. With

appropriate choices of cells and methods, this is achievable[73]. For pharmaceutical screening applications, only a dayor two of operational lifetime may be adequate. However,for both applications, it may be desirable to prepare CBBsubstrates with cells in advance and place them into astorage or “hibernation” state, with procedures availablefor rapidly preparing them for use. This approach hasbeen demonstrated for both neuronal [74] and cardiac cells[75], and could likely be broadly applied for CBBs. Fig. 6shows results of a hibernation experiment using the HL-1myocardial cell line, wherein the media change requirementwas reduced from daily to weekly changes.

As discussed later, the primary source of inconsistencybetween cell-based sensors of any type is variation in thecellular components. For primary cells, variations in geneticsbetween source animals and even differences in dissectiontechniques can create such variations. For cell lines, despitethe promise of more uniform cultures at the expense ofsome differences from wild-type cells, significant variation


Fig. 6. Example of extracellular AP recordings from HL-1 cellsthat were maintained for one week in hibernation medium withonce-per-week media changes (versus daily) and restored by returnto normal medium. Courtesy of T. Doukas, Stanford University,Stanford, CA.

in batch-to-batch and sensor-to-sensor performance is oftenseen (but rarely published). To date, a detailed study hasnot been carried out to compare the variations betweenprimary and immortalized cell types, nor have sufficientresources yet been applied to improving the net yield fromthe viewpoint of controlling the biological processes.

IV. SIGNAL ANALYSIS AND INTERPRETATION

As mentioned previously, early work with cultured cellsand tissues on microelectrode arrays focused on basic sciencegoals. Initially, simply obtaining a believable AP recordingwas an achievement in itself. As the field matured, this be-came routine, and some researchers turned their attention to-ward developing practical sensors based on the technology.To date, several demonstrations indicate that this can be done,with a fair number of compounds proven to be detectableover useful concentration ranges (for example, [63]). In orderto develop robust CBBs for general use and for unknownagents, it is necessary to demonstrate reproducible sensorswith statistically defined sensitivities (both in terms of whichagents they respond to, but also what concentration rangesthey can detect). A further complication is that realistic sam-ples may contain a number of different agents and interferingchemicals.

In comparison to molecular detection methods, currentcell-based sensors are much more variable in their nature.Not only is the physical coupling of cells to electrodesdifferent on a case-by-case basis, but variations frombatch-to-batch, chip-to-chip, and electrode-to-electrode canbe significant. Thus, robust signal analysis algorithms areneeded, hopefully which can deliver useful and relativelyconsistent sensor responses in the face of biological varia-tion. Rather than providing some sort of absolute calibrationprior to use, as is common with most chemical sensors,methods for using CBBs tend to be based on relative changeswith respect to a locally established baseline. Further, it isultimately necessary for most sensor applications to outputclear, numerical responses or even “red-light/green-light”binary outputs based on well-defined relative thresholds.This must be done automatically if sensors are to be used

by nonspecialist personnel or in high-throughput situationssuch as pharmaceutical screening.

In practice, algorithms fall into two categories, withincreasing complexity. First, and most simply,detectionalgorithms respond to statistically significant changes insome property or properties of AP signal behavior acrossone or more electrodes. Since such changes are often seenvia simple metrics such as rate changes up or down, onecan begin toclassifycompounds asexcitatoryor inhibitoryon such a basis. For the example of pyrolysis products ofsynthetic ester turbine engine lubricants known to causeneurologic problems, Keeferet al. [76] showed in a blindedexperiment that it is possible to predict the physiologic man-ifestations using CBBs consistently with animal exposures.Full classification, however, would entail something moreakin to comparing multiparameter fingerprints (some direct,such as rate, and others computed, such as dose/response re-lationships with straight and diluted samples) against storedcalibrations. In a practical system, it is likely that potentiallymatching compounds would be indicated to a user ranked byprobability [77]. Again, since the biological components ofCBBs can be variable, such algorithms represent significantchallenges. Another approach to classification involvesso-called “challenge assays,” in which known competitiveor “antidote” compounds are added to the CBB mixed withthe unknown (in addition to controls: unknown alone andcompetitive compound alone), potentially providing usefulinformation about the nature of the unknown.

A variety of algorithmic approaches have been explored,and work in this area is ongoing. For both cell types,counting AP rates over time is a starting point. Moresubtle time-domain measures, such as interspike phase,computed conduction velocity using multiple electrodesapply primarily to cardiac tissues, while burst patterns,rates, and more complex signal relationships are morerelevant to neural signals. Frequency domain approaches arealso promising, and are based on the observation that themorphology of the recorded APs can change significantlywhen agents act on the cells [78]. Naturally, time-domainmorphologic change measures can also be used [79],such as waveform templating (approximation of matchedfiltering), thresholding (oscilloscope-like triggering), andmore complex approaches such as wavelet analysis [80]. Itis likely that parallel combinations of algorithms—perhapsdifferent ones for different classes of agents–will be requiredto obtain optimal detection and classification. It should benoted that counted parameters such as AP and burst rateshave accuracies directly proportional to the length of timeover which they are acquired, and nontemporal methodshave the potential to provide much more rapid results ifsufficiently high signal-to-noise ratios (SNRs) are possible.

In any such CBB system, the first step in analyzing thesignals is electrode selection, which is carried out to limitdata collection and analysis to those electrode sites producingsignals of suitable form and SNR. An example approach isillustrated in Fig. 7, showing a manual selection process basedon waveform inspection. An example automated electrode-selection approach based on direct SNR measurement is


Fig. 7. Illustration of (left) a manual electrode selection process based on individual site signalquality and (right) resulting analyses of individual AP rising slope (top) and group beat rate (bottom).Courtesy of L. Giovangrandi, Stanford University, Stanford, CA.

Fig. 8. Example of parameter computation relative using minute AP rate parameters averagedover a population of cells, as used for neuronal cultures. The previous example represents theapplication of tetrodotoxin in various concentrations. Courtesy of Prof. G. Gross, University ofNorth Texas, Denton.

described in [66]. In any case, one needs sufficient numbersof useable electrodes per array to maintain statistical validityof any detection/classification algorithms. It is noteworthythat this issue is far from having been dealt with fully foreither neural nor cardiac CBBs.

For neuron-based sensors, the typical approach to de-tection algorithms is to use time-domain methods. As anexample, the method employed by Grosset al. [82]–[85]using networks of dissociated mouse spinal neurons is tocompute, on a minute-to-minute basis, the following param-


(a)

(b)

Fig. 9. Example of computed parameters for a myocardial culture upon exposure to quinidine. (a)Illustration of some of the parameters extracted from an action potential. (b) Raw data outputsfor ten electrodes, as well as population averages and standard deviations (over 60 actionpotentials) for different doses ranging from 1 nM to 500�M. Courtesy of L. Giovangrandi,Stanford University, Stanford, CA.

eters (for each electrode): 1) spike number; 2) burst number;3) mean burst duration; 4) mean integrated burst amplitude;and 5) mean burst period. These minute parameters are av-eraged over the entire neuronal network to obtain the globalequivalents, as shown in the example in Fig. 8. To establish

a statistical baseline for a network that has not been exposedto any agents, these parameters are averaged over severalminutes to obtain a reference mean and standard deviation.Detection is thereafter indicated by events that cause one ormore parameters to exceed the standard deviation in either


Fig. 10. Example of population-derived dose-response curves for HL-1 myocardial cells inresponse to quinidine and nifedipine. Courtesy of K. Gilchrist, Stanford University, Stanford, CA.

direction. As agent concentrations are varied, the parame-ters are logged, and can be used to derive a dose/responsecurve. Classification can be carried out based on the tem-poral dynamics of the parameters, allowing the effects tobe categorized as excitatory, inhibitory, or disinhibitory. Inaddition, other parameters can be considered such as delaytimes, activity decay slopes, systematic changes in burstvariables, and reversibility versus irreversibility.

For cardiac cells, the general approach is similar, butmuch more inclusive of AP morphology and phase changesas output parameters. In addition, since a syncitium isformed, it is possible to evaluate propagation (“conduction”)velocity, which can be a sensitive indicator of cellulardamage. As illustrated in Fig. 9, a variety of parameters canbe computed in parallel (real-time) and used to generatepopulation means and standard deviations. As describedpreviously, changes in these parameters can be used tocategorize the nature of the detected agent. If various knownconcentrations can be applied, dose-response curves can bederived [65], as shown in Fig. 10.

V. APPLICATION DOMAINS

With practical cell-based sensors on the horizon, it is im-portant to consider the necessary supporting technologiesthat allow their adaptation to specific application domains.The primary drivers for their development are the need tosense chemical and biological toxins in air and water, andthe desire to bring relatively high throughput physiologic-function-based screening to the pharmaceutical marketplace.While both ultimately represent significant markets, the po-tential volume in the latter makes it an attractive entry andfocal point for commercial development.

A. Fieldable Systems

To realize field-portable detectors for chemical and bio-logical agents, considerable changes must be made relativeto the equipment used in typical laboratory-based situations.First, the entire system must be robust, relatively compact,and able to source or self-power for a reasonable operationalperiod. Second, the cells must be packaged in such a way asto maintain sterility and proper environmental conditions forculture, yet allow the introduction of unknown agents. Typ-ically, the cells, microelectrode arrays, and potentially somefluidic interfaces are combined into a user-replaceable car-tridge. The cells must also be maintained at constant pH, os-molarity, and temperature, typically via chemical buffering,humidification, and closed-loop thermal control, respectively[65], [66], [86]. The high sensitivities of cells to changes inthese parameters can lead to false “signals” if they are nottightly controlled. In addition, sample preparation, which istypically done manually in the laboratory, must be automatedso that it is capable of being carried out in realistic environ-mental conditions.

Beginning with the first such work presented in Pan-crazio et al. [64], there has been steady progress towardrealizing such practical, field-portable systems with the ad-vent of a DARPA program in the area. Gilchristet al.[65] demonstrated a field-portable incubator and CBB datarecording system in field tests that included ground andair transportation and were able to acquire and pharma-cologically modulate AP signals in a desert setting (seeFig. 11). A hand-held recording system with integrated cellcartridges incorporating closed-loop thermal control andfront-end amplification and multiplexing electronics wasalso demonstrated [66], as shown in Fig. 12. For cell typesthat require temperature regulation, shrinking the controlled


(a)

(b)

Fig. 11. (a) Field-portable incubator. (b) Recording system. The incubator included storage capacityfor ten electrode arrays with cells, and provided closed-loop control of temperature and COlevels.The recording system used a laptop computer to capture data for later analysis, which currently canbe done in real-time. Both systems were tested in realistic field conditions [65]. Courtesy of V.Barker (a) and K. Gilchrist (b), both of Stanford University, Stanford, CA.


Fig. 12. Prototype hand-held CBB instrument [66]. At lower left, the disposable cartridgecan be seen. It contains a CMOS-based microelectrode array die and has two independent cellculture wells (one for the unknown and one to serve as a control). Courtesy of B. DeBusschere,Stanford University, Stanford, CA.

volume has definite advantages in terms of power consump-tion and hence weight and volume.

More recent work has addressed the need for automatedsample preparation. For practical application in military andfirst-responder situations, cumbersome measurement, titra-tion, and other steps must be carried out without signifi-cant user intervention. Basic requirements for CBB samplehandling include filtration, balancing of pH and osmolarity,mixing with appropriate cell culture media, and warming tothe culture’s nominal operating temperature. Fig. 13 is a pho-tograph of such a system, which processes water-based sam-ples (from potential drinking water sources or air samplerswith aqueous outputs) for application to CBBs.

For such sensors, longevity of the cellular componentscould be a limiting factor, depending upon operational ex-pectations. As noted previously, dissociated primary neuroncultures have been maintained on electrode arrays for manymonths. Primary and immortalized cardiac cells tend to offeron-chip operational lifetimes on the order of a week, but canbe metabolically slowed greatly using special culture media

and revived when needed [75] (similar “hibernation” mediahave been demonstrated with neuronal cells [74]).

Ultimately, it is likely that CBB systems will be fullyintegrated, with end-to-end processing, analysis, archiving,and display integrated into a single unit. A mature field-able system would be rugged, compact, transportable, andcapable of long standby times between uses. In addition,the user interface would require considerable developmentbeyond current prototypes, such that an appropriate levelof information was made available to the user, and withextremely high reliability and accuracy.

B. Pharmaceutical Screening

The requirements for pharmaceutical screening systemsbased on CBBs are naturally very different from those formilitary and civil defense functions. In the former case,moderate-to-high throughput of parallel samples is required,with a high degree of automation and a full suite of datainterpretation and management tools. Typical formats for


(a)

(b)

Fig. 13. Photographs of a prototype sample preparation system for CBB applications. In (a), theentire system is shown in a top view. In (b), the sample processing module is raised to its operationalposition and the sample cup can be seen. This system filters particulates from the sample and thenautomatically measures and adjusts its pH, osmolarity, and temperature. Courtesy of V. Barker,Stanford University, Stanford, CA.

molecular screens are 96- and 384-well plates, and suchformats should be achievable with CBB technologies.Robotic systems for handling the dispensing of fluids andmovement of such plates are commercially available, as aredatabases and archiving tools (useful review articles include[87]–[89]). The massive parallelism possible with such anapproach would allow not only for rapid screening of the

vast compound libraries pharmaceutical companies possess,but also allow matrix experiments to examine dose/responsebehavior and drug synergies, for example.

In order to realize such systems, one would need to de-velop low-cost, highly parallel cartridges for use in culturingthe cells and in the screening experiments themselves. Ascurrently envisioned, such a cartridge would consist of a


low-cost electrode substrate, a plastic well-forming plate,and potentially a removable lid. If one were using cardiaccells, with synchronous beating in each well, a fairly smallnumber of electrodes per well would be required, with asomewhat higher number if networks of cultured neuronswere used in order to guarantee an acceptable yield of“usable” sites. Assuming a few electrodes for each well ofa 96-well cartridge, some form of electronic multiplexingwill almost certainly be required. This could be achieved ata reasonable cost for a disposable cartridge. In this applica-tion, cell longevity may not be a significant issue, and localcell culture capability is typical in laboratories engaged inpharmaceutical research.

It seems quite possible to not only meet the physicaland electronic requirements for such a system, but with theaddition of reproducible cell cultures and computationallyoptimized algorithms, high-throughput operation appearsrealistic in the near future.

VI. DISCUSSION

This review has presented an overview of background andcurrent issues in the development of practical, cell-basedsensors for biologically active compounds. By harnessingthe natural physiologic pathways of live cells, these devicesoffer the potential for broad-spectrum detection of chemicaland biological warfare agents, naturally occurring environ-mental toxins, and potentially beneficial pharmaceuticals inscreening applications.

Research in this field to date has been encouraging, yetthere are several significant barriers to commercializationand more widespread use. First, for all systems demonstratedto date (based on primary cells or lines), variability betweenbatches, chips, and electrodes is relatively large. Althoughnot well documented in the literature, this issue still requiresconsiderable effort, and is tied into the nature of the cellsused.

While biocompatible materials are commonplace and rel-atively simple to screen for, cell culture techniques tend tobe fairly empirical, and, if serum-containing media are used,depend on pooled animal products and their potential vari-ations. Surfaces upon which cells are plated may also havea wide range of effects, although far easier to standardize.In all of these cases, results may be cell-type specific and,making matters worse, exact details are seldom published.However, successes to date suggest that even with a less-than-perfect yield, successful single-shot (fieldable) and par-allel cell-based screening tools could be realized.

Another vital requirement for such systems (from whichmilitary systems are not entirely exempt) is that of a prac-tical and profitable business model to spur development intorealistic products. While beyond the scope of this review,this factor must be considered as it can be even more crit-ical than technology capability in terms of what ultimatelyreaches end-users.

Overall, the vast potential of this approach will likely drivewider adoption. Key in this process are the abilities to scaleup to parallel, high-throughput systems, to develop better and

better signal interpretation algorithms, and to engineer theprimary transducer—the cell—to provide otherwise unavail-able capabilities.

ACKNOWLEDGMENT

The author would like to thank Prof. G. Gross, Dr. D.Borkholder, Dr. D. DeBusschere, Dr. L. Giovangrandi, Dr. J.Pancrazio, K. Gilchrist, V. Barker, T. Doukas, and S. Plewafor their generous help in the preparation of this paper.

REFERENCES

[1] J. J. Pancrazio, J. P. Whelan, D. A. Borkholder, and D. A. Stenger,“Development and application of cell-based biosensors,”Ann.Biomed. Eng., vol. 27, pp. 697–711, 1999.

[2] C. Ziegler, “Cell-based biosensors,”Fresenius J. Anal. Chem., vol.366, pp. 552–559, 2000.

[3] P. McFadden, “Broadband detection: Holmes on a chip,”Science,vol. 297, pp. 2075–2076, Sept. 20, 2002.

[4] P. Fromherz, A. Offenhäusser, T. Vetter, and J. Weis, “A neuron-silicon junction: A retzius cell of the leech on an insulated-gate field-effect transistor,”Science, vol. 252, pp. 1290–1293, 1991.

[5] P. Fromherz, “Extracellular recording with transistors and the distri-bution of ionic conductances in a cell membrane,”Eur. Biophys. J.,vol. 28, pp. 254–258, 1999.

[6] I. Giaever and C. R. Keese, “Use of electric fields to monitor thedynamical aspect of cell behavior in tissue culture,”IEEE Trans.Biomed. Eng., vol. 33, pp. 242–247, Feb. 1986.

[7] C. R. Keese and I. Giaever, “A whole cell biosensor based oncell-substrate interactions,” inAnnu. Int. Conf. IEEE EngineeringMedicine and Biology Society, vol. 12, 1990, pp. 500–501.

[8] I. Giaever and C. R. Keese, “Micromotion of mammalian cells mea-sured electrically,”Proc. Nat. Acad. Sci., vol. 88, pp. 7896–7900,1991.

[9] , “A morphological biosensor for mammalian cells,”Nature,vol. 366, pp. 591–592, 1993.

[10] C. R. Keese and I. Giaever, “A biosensor that monitors cell mor-phology with electric fields,”IEEE Eng. Med. Biol. Mag., vol. 13,pp. 402–408, June–July 1994.

[11] J. W. Parce, J. C. Owicki, K. M. Kercso, G. B. Sigal, H. G. Wada, V.C. Muir, L. J. Bousse, K. L. Ross, B. I. Sikic, and H. M. McConnell,“Detection of cell-affecting agents with a silicon biosensor,”Sci-ence, vol. 246, pp. 243–247, 1989.

[12] J. C. Owicki, J. W. Parce, K. M. Kercso, G. B. Sigal, V. C. Muir, J. C.Venter, C. M. Fraser, and H. M. McConnell, “Continuous monitoringof receptor-mediated changes in the metabolic rates of living cells,”Proc. Nat. Acad. Sci., vol. 87, pp. 4007–4011, 1990.

[13] J. C. Owicki and J. W. Parce, “Biosensors based on energy metabo-lism of living cells: The physical chemistry and cell biology of ex-tracellular acidification,”Biosens. Bioelectron., vol. 7, pp. 255–272,1992.

[14] G. T. Baxter, L. J. Bousse, T. D. Dawes, J. M. Libby, D. N. Modlin,J. C. Owicki, and J. W. Parce, “Microfabrication in silicon micro-physiometry,”Clin. Chem., vol. 40, no. 9, pp. 1800–1804, 1994.

[15] L. Bousse, “Whole cell biosensors,” inProc. 8th Int. Conf. Solid-State Sensors and Actuators, vol. 2, 1995, pp. 483–486.

[16] K. G. Klemic, J. F. Klemic, M. A. Reed, and F. J. Sigworth, “Mi-cromolded PDMS planar electrode allows patch clamp electricalrecordings from cells,”Biosens. Bioelectron., vol. 17, no. 6–7, pp.597–604, June 2002.

[17] F. W. Scheller, R. Hintsche, D. Pfeiffer, F. Schubert, K. Riedel, and R.Kindervater, “Biosensors: Fundamentals, applications, and trends,”Sens. Actuators B, Chem., vol. B4, no. 1–2, pp. 197–206, May 1991.

[18] A. L. Ghindilis, P. Atanasov, M. Wilkins, and E. Wilkins, “Im-munosensors: Electrochemical sensing and other engineeringapproaches,”Biosens. Bioelectron., vol. 13, no. 1, pp. 113–131,Jan. 1, 1998.

[19] S. Y. Rabbany, B. L. Donner, and F. S. Ligler, “Optical immunosen-sors,”Crit. Rev. Biomed. Eng., vol. 22, no. 5–6, pp. 307–346, 1994.

[20] K. E. Petersen, “The future of DNA diagnostic testing,”Proc. SPIE,vol. 4560, pp. 1–7, 2001.


[21] W. K. Purves, D. Sadava, G. H. Orians, and H. C. Heller,Life: TheScience of Biology, 6th ed. Sunderland, MA: Sinauer, 2001.

[22] Y. Kim, T. R. Flynn, R. B. Donoff, D. T. Wong, and R. Todd, “Thegene: The polymerase chain reaction and its clinical application,”J.Maxillofacial Surg., vol. 60, no. 7, pp. 808–815, 2002.

[23] D. G. Ginzinger, “Gene quantification using real-time quantitativePCR: An emerging technology hits the mainstream,”Exp. Hematol.,vol. 30, no. 6, pp. 503–512, 2002.

[24] D. A. Borkholder, “Cell-based biosensors using microelectrodes,”Ph.D. dissertation, Stanford Univ., Stanford, CA, 1998.

[25] A. Carrell, “On the permanent life of tissues outside of the or-ganism,”J. Exp. Med., vol. 15, pp. 516–528, 1912.

[26] A. A. Boulton, G. B. Baker, and W. Walz, Eds.,Practical Cell Cul-ture Techniques: Neuromethods. Totowa, NJ: Humana, 1992, vol.23.

[27] R. I. Freshney,Culture of Animal Cells. New York: Wiley-Liss,1994.

[28] R. C. Gesteland, B. Howland, J. Y. Lettvin, and W. H. Pitts, “Com-ments on microelectrodes,”Proc. IRE, vol. 47, pp. 1856–1862, Nov.1959.

[29] D. A. Robinson, “The electrical properties of metal microelec-trodes,”Proc. IEEE, vol. 56, pp. 1065–1071, 1969.

[30] G. T. A. Kovacs, “Introduction to the theory, design and modelingof thin-film microelectrodes for neural interfaces,” inEnablingTechnologies for Cultured Neural Networks, D. A. Stenger and T.McKenna, Eds. San Diego, CA: Academic, 1994, pp. 121–166.

[31] W. G. Regehr, J. Pine, and D. B. Rutledge, “A long-termin vitrosilicon-based microelectrode-neuron connection,”IEEE Trans.Biomed. Eng., vol. 35, pp. 1023–1032, Dec. 1998.

[32] A. L. Hodgkin and A. F. Huxley, “A quantitative description of mem-brane current and its application to conduction and excitation innerve,”J. Physiol., vol. 117, pp. 500–544, 1952.

[33] T. F. Weiss,Cellular Biophysics—Volume 2: Electrical Proper-ties. Cambridge, MA: MIT Press, 1996.

[34] R. Plonsey, “Action potential sources and their volume conductorfields,” Proc. IEEE, vol. 65, pp. 601–611, 1977.

[35] P. Connolly, P. Clark, A. S. G. Curtis, J. A. T. Dow, and C. D. W.Wilkinson, “An extracellular microelectrode array for monitoringelectrogenic cells in culture,”Biosens. Bioelectron., vol. 5, no. 3,pp. 223–234, 1990.

[36] K. E. Petersen, “Silicon as a mechanical material,”Proc. IEEE, vol.70, pp. 420–457, May 1982.

[37] G. T. A. Kovacs,Micromachined Transducer Sourcebook. NewYork: WCB/McGraw-Hill, 1998.

[38] K. D. Wise, J. B. Angell, and A. Starr, “An integrated-circuit ap-proach to extracellular microelectrodes,”IEEE Trans. Biomed. Eng.,vol. BME-17, pp. 238–246, July 1970.

[39] A. Starr, K. D. Wise, and J. Csongradi, “An evaluation of pho-toetched microelectrodes for extracellular single-unit recording,”IEEE Trans. Biomed. Eng., vol. BME-20, pp. 291–293, July 1973.

[40] K. D. Wise and J. B. Angell, “A low capacitance multimicroelec-trode probe for use in extracellular neurophysiology,”IEEE Trans.Biomed. Eng., vol. BME-22, pp. 212–219, May 1975.

[41] C. A. Thomas Jr., P. A. Springer, B. E. Loeb, Y. Berwald-Netter, andL. M. Okun, “A miniature microelectrode array to monitor bioelec-tric activity of cultured cells,”Exp. Cell Res., vol. 74, pp. 61–66,1972.

[42] M. B. Shtark, A. S. Ratushnyak, L. V. Voskresenskaya, and S. N.Olenen, “A multielectrode perfusion chamber for tissue culture re-search,”Bull. Exp. Biol. Med., vol. 78, no. 9, pp. 122–124, 1974.

[43] G. W. Gross, E. Rieske, H. W. Kreutzberg, and A. Meyer, “A newfixed-array multi-microelectrode system designed for long-termmonitoring of extracellular single unit neuronal activityin vitro,”Neurosci. Lett., vol. 6, pp. 101–105, 1977.

[44] G. W. Gross, “Simultaneous single unit recordingin vitro with a pho-toetched laser deinsulated gold multimicroelectrode array,”IEEETrans. Biomed. Eng., vol. BME-26, no. 5, pp. 273–279, May 1979.

[45] J. Pine, “Recording action potentials from cultured neurons with ex-tracellular microcircuit electrodes,”J. Neurosci. Methods, vol. 2, pp.19–31, 1980.

[46] G. W. Gross, A. N. Williams, and J. H. Lucas, “Recording of sponta-neous activity with photoetched microelectrode surfaces from mousespinal neurons in culture,”J. Neurosci. Methods, vol. 5, pp. 13–22,1982.

[47] G. W. Gross, W. Y. Wen, and J. W. Lin, “Transparent indium-tinoxide electrode patterns for extracellular, multisite recording in neu-ronal cultures,”J. Neurosci. Methods, vol. 15, pp. 243–252, 1985.

[48] J. L. Novak and B. C. Wheeler, “Recording from the aplysia ab-dominal ganglion with a planar microelectrode array,”IEEE Trans.Biomed. Eng., vol. BME-33, pp. 196–202, Feb. 1986.

[49] M. D. Eggers, D. K. Astolfi, S. Liu, H. E. Zeuli, S. S. Doeleman,R. McKay, T. S. Khuon, and D. J. Ehrlich, “Electronically wiredpetri dish: A microfabricated interface to the biological neuronal net-work,” J. Vacuum Sci., vol. B8, no. 6, pp. 1392–1398, Nov.–Dec.1990.

[50] S. Martinoia, M. Bove, G. Carlini, C. Ciccarelli, M. Grattarola, C.Storment, and G. Kovacs, “A general-purpose system for long-termrecording from a microelectrode array coupled to excitable cells,”J.Neurosci. Methods, vol. 48, no. 12, pp. 115–121, June 1993.

[51] D. A. Stenger and T. M. McKenna, Eds.,Enabling Technologies forCultured Neural Networks. San Diego, CA: Academic, 1994.

[52] D. A. Israel, W. H. Barry, D. J. Edell, and R. G. Mark, “An arrayof microelectrodes to stimulate and record from cardiac cells in cul-ture,” Amer. J. Physiol., vol. 247, pp. H669–H674, 1984.

[53] D. A. Israel, D. J. Edell, and R. G. Mark, “Time delays in propagationof cardiac action potential,”Amer. J. Physiol., vol. 27, no. 6, pp.H1906–H1917, 1990.

[54] Y. Jimbo and A. Kawana, “Electrical stimulation of cultured neuralcells by planar electrode array,” inProc. Annu. Int. Conf. IEEEEngineering in Medicine and Biology Society, vol. 12, 1990, pp.1741–1742.

[55] , “Electrical stimulation and recording from cultured neuronsusing a planar electrode array,”Bioelectrochem. Bioenergetics, vol.29, pp. 193–204, 1992.

[56] D. A. Borkholder, J. Bao, N. I. Maluf, E. R. Perl, and G. T. A. Kovacs,“Microelectrode arrays for stimulation of neural slice preparations,”J. Neurosci. Methods, vol. 77, no. 1, pp. 61–66, Nov. 7, 1997.

[57] D. Kleinfeld, K. Kahler, and P. E. Hockberger, “Controlled out-growth of dissociated neurons on patterned substrates,”J. Neurosci.,vol. 8, no. 11, pp. 4098–4120, 1988.

[58] D. A. Stenger and J. J. Hickman, “Lithographic definition of neu-ronal microcircuits,” inEnabling Technologies for Cultured NeuralNetworks, D. A. Stenger and T. M. McKenna, Eds. San Diego, CA:Academic, 1994, pp. 77–98.

[59] B. C. Wheeler and G. J. Brewer, “Multineuron patterning andrecording,” in Enabling Technologies for Cultured Neural Net-works, D. A. Stenger and T. M. McKenna, Eds. San Diego, CA:Academic, 1994, pp. 167–186.

[60] Y. Nam, B. C. Wheeler, and G. Brewer, “Gold surface microelec-trode arrays permit geometrical control of neuronal networks viathiol linking chemistry,” in Proc. 2nd Int. IEEE-EMBS SpecialTopic Conf. Microtechnologies in Medicine and Biology, 2002, pp.132–135.

[61] M. H. Drodge, G. W. Gross, M. H. Hightower, and L. E. Czisny,“Multielectrode analysis of coordinated, multisite, rhythmic burstingin cultured CNS monolayer networks,”J. Neurosci. Methods, vol. 6,no. 6, pp. 1583–1592, 1986.

[62] G. W. Gross, B. K. Rhoades, and R. J. Jordan, “Neuronal networksfor biochemical sensing,”Sens. Actuators, vol. 6, pp. 1–8, 1992.

[63] G. W. Gross, H. M. E. Azzazy, M.-C. Wu, and B. K. Rhoades, “Theuse of neuronal networks on multielectrode arrays as biosensors,”Biosens. Bioelectron., vol. 10, pp. 553–567, 1995.

[64] J. J. Pancrazio, P. P. Bey Jr., D. S. Cuttino, J. K. Kusel, D. A.Borkholder, K. M. Shaffer, G. T. A. Kovacs, and D. A. Stenger,“Portable cell-based biosensor system for toxin detection,”Sens.Actuators B, Chem., vol. 53, no. 3, pp. 179–185, Dec. 1998.

[65] K. H. Gilchrist, V. N. Barker, L. E. Fletcher, B. D. DeBusschere, P.Ghanouni, L. Giovangrandi, and G. T. A. Kovacs, “General purpose,field-portable cell-based biosensor platform,”Biosens. Bioelectron.,vol. 16, pp. 557–564, Sept. 2001.

[66] B. D. DeBusschere and G. T. A. Kovacs, “Portable cell-basedbiosensor system using integrated CMOS cell-cartridges,”Biosens.Bioelectron., vol. 16, pp. 543–556, Sept. 2001.

[67] G. Banker and K. Goslin,Culturing Nerve Cells. Cambridge, MA:MIT Press, 1991.

[68] L. Hayflick and P. S. Moorhead, “The serial cultivation of humandiploid cell strains,”Exp. Cell Res., vol. 25, pp. 585–621, 1961.

[69] A. M. Aravanis, B. D. DeBusschere, A. J. Chruscinski, K. H.Gilchrist, B. K. Kobilka, and G. T. A. Kovacs, “A geneticallyengineered cell-based biosensor for functional classification ofagents,”Biosens. Bioelectron., vol. 16, pp. 571–577, Sept. 2001.

[70] W. C. Claycomb, N. A. Lanson Jr., B. S. Stallworth, D. B. Egeland, J.B. Delcarpio, A. Bahinski, and N. I. Izzo Jr., “HL-1 cells: A cardiacmuscle cell line that contracts and retains phenotypic characteris-tics of the adult cardiomyocyte,”Proc. Nat. Acad. Sci., vol. 95, pp.2979–2984, 1998.


[71] Y. Palti, G. Ben David, E. Lachov, Y. H. Mika, G. Omri, and R.Schatzeberger, “Islets of langerhans generate wavelike electric ac-tivity modulated by glucose concentration,”Diabetes, vol. 45, pp.596–601, 1996.

[72] F. Rivera and T. Herve, “Extracellular electrical activity recording ofpancreatic islet,” inProc. 2nd Int. IEEE-EMBS Special Topic Conf.Microtechnologies in Medicine and Biology, 2002, pp. 490–495.

[73] G. W. Gross, “Internal dynamics of randomized mammalian neu-ronal networks in culture,” inEnabling Technologies for CulturedNeural Networks, D. A. Stenger and T. M. McKenna, Eds. NewYork: Academic, 1994, pp. 277–317.

[74] G. J. Brewer and P. J. Price, “Viable cultured neurons in ambientcarbon dioxide and hibernation storage for a month,”NeuroReport,vol. 7, pp. 1509–1515, 1996.

[75] T. Doukas, private communication, Apr. 2002.[76] E. W. Keefer, A. Gramowski, D. A. Stenger, J. J. Pancrazio, and G.

W. Gross, “Characterization of acute neurotoxic effects of trimethy-lolpropane phosphate via neuronal network biosensors,”Biosens.Bioelectron., vol. 16, pp. 513–525, 2001.

[77] J. Pancrazio, private communication, Au. 2002.[78] D. A. Borkholder, B. D. DeBusschere, and G. T. A. Kovacs, “An

approach to the classification of unknown biological agents with cellbased sensors,” inProc. Solid-State Sensor and Actuator Workshop,1998, pp. 178–182.

[79] M. S. Lewicki, “A review of methods for spike sorting: The detectionand classification of neural action potentials,”Network, vol. 9, no.4, pp. 53–78, 1998.

[80] J. C. Letelier and P. P. Weber, “Spike sorting based on discretewavelet transform coefficients,”J. Neurosci. Methods, vol. 101, no.2, pp. 93–106, 2000.

[81] E. Hulata, R. Segev, and E. Ben-Jacob, “A method for spike sortingand detection based on wavelet packets and Shannon’s mutual infor-mation,”J. Neurosci. Methods, vol. 117, no. 1, pp. 1–12, 2002.

[82] S. I. Morefield, E. W. Keefer, K. D. Chapman, and G. W. Gross,“Drug evaluations using neuronal networks cultured on microelec-trode arrays,”Biosens. Bioelectron., vol. 15, pp. 383–396, 2000.

[83] E. W. Keefer, S. J. Norton, N. A. J. Boyle, V. Talesa, and G. W. Gross,“Acute toxicity screening of novel AChE inhibitors using neuronalnetworks on microelectrode arrays,”Neurotoxicology, vol. 14, pp.1–10, 2000.

[84] E. W. Keefer, A. Gramowski, D. A. Stenger, J. J. Pancrazio, and G.W. Gross, “Characterization of acute neurotoxic effects of trimethy-lolpropane phosphate via neuronal network biosensors,”Biosens.Bioelectron., vol. 16, pp. 513–525, 2001.

[85] E. W. Keefer, A. Gramowski, and G. W. Gross, “NMDA receptor-dependent periodic oscillations in cultured spinal cord networks,”J.Neurophysiol., vol. 86, pp. 3030–3042, 2001.

[86] N. J. Mourlas, K. H. Gilchrist, L. Giovangrandi, N. I. Maluf, and G.T. A. Kovacs, “An in-line osmometer for application to a cell-basedbiosensor system,”Sens. Actuators B, Chem., vol. 83, pp. 41–47,2002.

[87] G. S. Sittampalam, S. D. Kahl, and W. P. Janzen, “High-throughputscreening: Advances in assay technologies,”Curr. Opin. Chem.Biol., vol. 1, no. 3, pp. 384–391, 1997.

[88] J. G. Houston and M. Banks, “The chemical-biological interface:Developments in automated and miniaturised screening tech-nology,” Curr. Opin. Chem. Biol., vol. 8, no. 6, pp. 734–740, 1997.

[89] C. M. Masimirembwa, R. Thompson, and T. B. Andersson, “Invitro high throughput screening of compounds for favorablemetabolic properties in drug discovery,”Combinatorial Chem. HighThroughput Screening, vol. 4, no. 3, pp. 245–263, 2001.

Gregory T. A. Kovacs received the B.A.Sc. de-gree in electrical engineering from the Universityof British Columbia, Vancouver, BC, Canada, in1984, the M.S. degree in bioengineering from theUniversity of California, Berkeley, in 1985, thePh.D. degree in electrical engineering from Stan-ford University, Stanford, CA, in 1990, and theM.D. degree from Stanford University in 1992.

His industry experience includes the design ofhigh-speed data acquisition systems, the designof instruments for IC fabrication, commercial

and consumer product design, extensive patent law consulting, and thecofounding of several companies (most recently Cepheid, Inc. [CPHD], inSunnyvale, CA). He currently serves on the scientific advisory boards ofseveral companies.

In 1991, he joined the faculty of Stanford University and is currentlyan Associate Professor of Electrical Engineering with an appointment inthe Department of Medicine, by courtesy. He teaches courses in electroniccircuits and micromachined transducers. His present research interestsinclude biomedical instruments and sensors, miniaturized spaceflighthardware, biotechnology, and microfabricated devices, all with emphasison solving practical problems. As a collaborator at NASA Ames, he helpsdirect a variety of projects spanning wearable crew physiologic monitors,instruments for space biology and free-flyer experiments. Since February2003, he has served as the Investigation Scientist for the Columbia Acci-dent Investigation Board’s debris analysis team at Kennedy Space Center,FL. He has published extensively in the technical literature, includingauthorship of a popular engineering textbook. He currently has 14 issuedpatents (several pending).

From 1992 to 1994, Dr. Kovacs held the Robert N. Noyce Family Fac-ulty Scholar Chair. In 1993, he received an NSF Young Investigator Award.From 1994 to 1997, he was a Terman Fellow. From 1996 to 1998, he was aUniversity Fellow. In 2002, he was elected a Fellow of the American Insti-tute for Medical and Biological Engineering. He was appointed to the De-fense Sciences Research Council (DSRC) in 1995, and served as AssociateChair (1998–2000) and Chairman (2000–2002). Through the DSRC, he hasparticipated in numerous military activities. He is a Fellow National of theExplorer’s Club.


electronic sensors with living cellular components

Documents