www.elsevier.com/locate/biotechadv

Biotechnology Advances 21 (2003) 431–444

Miniaturized analytical assays in biotechnology

Rosanne A. Guijt-van Duijn a,*, Rob Moerman a, Arthur Kroon a,Gijs W.K. van Dedem a, Richard van den Doel b, Lucas van Vliet b,Ian T. Young b, Frederic Laugere c, Andre Bossche c, Pascalina Sarro c

aKluijver Laboratory for Biotechnology, Delft University of Technology, Julianalaan 67,

2628BC Delft, The NetherlandsbPattern Recognition Group, Department of Applied Physics, Delft University of Technology, Lorenzweg 1,

2628CJ, Delft, The NetherlandscElectronic Instrumentation Group, Faculty of Information Technology and Systems,

Delft University of Technology, Mekelweg 4, 2628CD, Delft, The Netherlands

Abstract

Biotechnology today is a well-established paradigm in many areas of human endeavor, such as

the pharmaceutical industry, agriculture, management of the environment and many others.

Meanwhile, biology is undergoing a spectacular transition: whereas systematic biology was

replaced gradually by molecular biology, the latter is rapidly being transformed into a new systematic

era in which entire genomes are being charted by ever more sophisticated analytical techniques.

In the wake of this onslaught of data, new fields are germinating, such as bioinformatics in an

attempt to find answers to fundamental questions, answers that may be hidden in the massive

amounts of data already available today.

D 2003 Published by Elsevier B.V.

Keywords: Miniaturized analytical assays; Bioinformatics; Biotechnology

1. Introduction

As new fields develop, they attempt to find answers to fundamental questions,

answers that may already be available in the existing massive amounts of data.

Presently, such questions may also be answered using alternative analytical tools. At

0734-9750/03/$ - see front matter D 2003 Published by Elsevier B.V.

doi:10.1016/S0734-9750(03)00059-4

* Corresponding author.

E-mail address: gijs.van.dedem@tip.nl (R.A. Guijt-van Duijn).

R.A. Guijt-van Duijn et al. / Biotechnology Advances 21 (2003) 431–444432

the Kluyver Laboratory, a major thrust is rational strain improvement. As was found in

recent years, improving strains in the straightforward way, e.g., by increasing the

number of genes responsible for the synthesis of a useful secondary metabolite often

results in disappointment. The reason may very well be that the problem is not in the

pathway itself, but in the supply of building blocks or energy from central metabolism.

To get a better understanding of such phenomena, several (analytical) approaches are

being explored. These include:

� Rapid sampling and analysis of metabolites by LC/MS/MS (Visser et al., 2002),� The use of DNA oligonucleotide chips to probe mRNA levels (Piper et al., 2002),� The development of strongly miniaturized analytical devices (Moerman et al., 2001;

Moerman and van Dedem, submitted for publication; Guijt et al., 2001, 2002; Laugere

et al., in press).

The rationale for miniaturization is several-fold: reduction of reagent and—more

importantly—sample consumption, improving analytical speed by shortening analysis

time or by running several analyses in parallel.

The infrastructure of the Delft University, like Waterloo is primarily that of an

engineering school, and is therefore eminently suitable for such developments, thanks

to the presence of (among others) (bio)chemical engineering, electrical engineering and

physics departments. Our efforts have focused on two objectives:

1. Separation and detection of extracellular medium components in fermentation and cell

culture; and

2. Parallel assays for intracellular enzymes and metabolites.

In this paper, an overview is given to some of our developments in this area.

2. Why miniaturized methods?

2.1. Extracellular components

Fermentation technology is still one of the major unit operations in biotechnology,

be it in growing vast amounts of yeast cells for the baking of bread, or the refined

culturing of mammalian cells for the production of valuable biopharmaceuticals. In all

cases, cells are exposed to conditions assumed to be favorable to their growth and/or

product formation. These conditions have largely been the result of trial and error and

proper optimization using statistically valid designs over many different parameters has

often been considered too labor and time consuming.

Over the last 30 years, many attempts have been made to obtain more real-time data

from fermentation processes by on-line and at-line techniques, in addition to the more

conventional off-line mode. These attempts involved the use of Flow Injection Analysis

(FIA) or related techniques among others. These consisted of bleeding small samples

R.A. Guijt-van Duijn et al. / Biotechnology Advances 21 (2003) 431–444 433

from a fermentor and leading them through a system of mixing coils, reaction coils and

detectors to obtain data on extracellular concentrations of carbon and nitrogen sources

as well as on end-products of metabolism such as organic acids (see, e.g., Freitag,

1996].

In practice, these methods proved to be too complicated and, to our knowledge, have

not been applied in industrial fermentations.

What remained are the traditional pH- and dissolved oxygen probes that are routinely

applied in almost all fermentations. It dawned upon us that any system meant to gain more

information on medium composition should be as simple as using a pH electrode. This in

turn would involve a large amount of integration of the functions of an analytical system.

These functions include:

1. Sampling,

2. Sample conditioning,

3. Injection and separation,

4. Detection.

To integrate all these functions into a single device and to make it operate on a time

interval compatible with that of a fermentation process, miniaturized devices seemed the

obvious way to go.

2.2. Intracellular components

As in the now widely used DNA chips which are used in analyzing the tran-

scriptome, understanding of the workings of the cell will require further probing into its

metabolism. Since metabolism is brought about by enzymes, enzyme levels are a

worthwhile object of study in cells when exposed to a variety of conditions, such as

steady-state and—more interesting—to sudden changes in external conditions. Another

consideration is that optimization of production of useful metabolites or proteins may

involve channeling central metabolites into other than the usual directions. This may

necessitate the insertion of additional copies of certain key enzymes, which in turn

necessitates to quantify such enzymes.

As in DNA chips, the design of highly parallel methods using miniaturized techniques

for such enzyme assays has several advantages, such as:

1. A complete picture is obtained;

2. More assays can be done e.g., in a rapid sampling mode; and

3. Less sample is needed.

Using ‘‘classical’’ methods, a common measurement principle can be employed, both

for measuring enzyme levels as well as for measuring metabolite levels, i.e., methods

that are based on the formation or consumption of NAD(P)H. This allows for the use of

fluorescence as a common detection method, which can be adapted for a wide range of

analyses.

R.A. Guijt-van Duijn et al. / Biotechnology Advances 21 (2003) 431–444434

3. Tools for extracellular components

3.1. Sampling

A small-scale sampling device was designed and built at the mechanical workshop of

the Kluyver Laboratory. It basically consists of a support with a cross section of 5 mm,

into which tiny holes of 200 Am are drilled. The support can be covered with any filter

material deemed suitable. In our case, we used a 0.22-Am sterilizable filter, which was cut

to size. It was kept fixed by a screwed-on open nut, in combination with an O-ring. The

outlet was connected to a small reservoir, connecting to a piece of 100-Am tubing. An

illustration is shown in Fig. 1.

3.2. Sample conditioning

Sample conditioning is usually a necessary part of the analytical process. In handling

cell extracts, a lot of the work can be done using automatic liquid handling systems that are

commercially available. This is so because the data are considered not to be needed in

quasi real-time.

However, for the analysis of extracellular components, sample pretreatment must be done

in real-time using methods suitable for miniaturization. One of the more promising

approaches is fluid handling by Electro-Osmotic Flow (EOF) and its even more versatile

alternative Electro-Osmotically Induced Pumping (EOIP), which were recently developed

and demonstrated (Guijt et al., 2002). In both cases, the driving force is an electric field

applied between two points along a channel. The phenomenon is based on the fact that the

inside walls of glass channels are negatively charged. These negative charges attract positive

counter-ions from the buffer inside the channel.When a potential is applied between the ends

of the channel, these positive counter-ions are forced to move in the direction of the cathode,

dragging along the liquid. An almost flat velocity profile is achieved, which in theory can

give rise to very high plate numbers.

The simplest form of this arrangement is a channel ending in open reservoirs on

either side, into which the high-voltage electrodes can be placed. Any gas bubbles

emanating at the electrode surface are thus vented to the air.

Fig. 1. Sampling device used for testing long-term, small volume sampling from continuous cultures of yeast.

R.A. Guijt-van Duijn et al. / Biotechnology Advances 21 (2003) 431–444 435

Variations on this theme include the application of indirect EOF, whereby the high

voltage is applied through electrically conducting microchannels connecting to side

channels, which contain the actual electrodes. This configuration expands the use of

EOF into the EOIP regime, since the side channels can be placed anywhere along a

channel and also outside the field, liquid will be forced to move once the field is applied

(see Figs. 2 and 3).

3.3. Injection and separation

Using the EOIP or indirect EOF principles, injection becomes almost trivial. The most

straightforward approach is to have a crossing of channels or a ‘‘double-T’’ configuration,

whereby one channel serves as a sample application channel and the other as the

separation channel proper. Adequate timing and application of voltages allows for filling

the cross or double T, retracting the superfluous sample and injecting it into the separation

channel.

A schematic layout of a testing device is shown in Fig. 4.

This testing device was used for injection experiments with fluorescent dyes, the

movement of which could be monitored using a fluorescence microscope. A typical

injection procedure using EOIP is shown in Fig. 5.

The separation mode we have explored so far is electrophoresis, i.e., separation based

on charge and size. Molecules move with the EOF, but are retarded or accelerated when

negatively or positively charged, respectively.

3.4. Detection

In the most widely studied device so far, separation and detection were combined into a

single microdevice with dimensions 10� 20 mm. It contained a 6-cm separation channel

Fig. 2. Example of indirect Electro-osmotic flow (EOF) and electro-osmotically induced hydraulic pumping

(EOIHP), using side channels which are electrically connected to the main channel through a thin liquid

channel.

Fig. 3. Electrically conducting microchannel between main (lower channel) channel and side channel (upper

channel).

R.A. Guijt-van Duijn et al. / Biotechnology Advances 21 (2003) 431–444436

with open reservoirs on either end. Therefore, cross injection could not be applied.

Furthermore, it contained an integrated 4-electrode nongalvanic conductivity detector of a

new design.

Conductivity is a generally applicable detection mode, which does not depend on the

presence of chromophores or other special chemical properties. It is a very attractive

detection technique for miniaturized devices because its sensitivity does not depend on, for

example, (optical) path length.

The detector is operated by applying an AC voltage across the outermost electrodes

and detecting the current flowing between the innermost ones. This arrangement

eliminates the effects of stray capacitances between the electrodes and hence allows

for a strongly improved signal to noise ratio, specifically since the signal is fed through

a lock-in amplifier which is tuned to the applied AC input frequency. The advantage of

nongalvanic over galvanic electrodes is the elimination of reactions taking place at the

Fig. 4. Schematic of testing device. S and Sw are sample loading and sample waste reservoirs, B and Bw are buffer

reservoirs. Side channels are numbered 1 through 6. On each reservoir and on each side channel, high voltages

can be applied from independent power supplies. The injection cross is marked with a square. The sample channel

runs between reservoirs S and Sw, whereas the separation channel runs from the injection cross to reservoir Bw.

Fig. 6. Separation and detection device fabricated in glass. This device does not contain a cross or double-T

injection. The buffer reservoirs on the left side of the serpentine-like separation channel serves also as a sample

application reservoir. Buffer is sucked out of it and is replaced with sample. After a short loading period, sample is

sucked out again and is replaced with buffer. Separation occurs along the serpentine channel. The detector is

inside the square and consists of 4 aluminum electrodes, covered with a layer of silicium carbide to isolate them

from direct contact with the liquid.

Fig. 5. Example of injection with concomitant sample dilution. Top frame: sample is drawn from sample reservoir

S into the injection cross by applying � 3 kV on side channel 4, while all other side channels and reservoirs are

kept at 0 V. Buffer is drawn from the buffer reservoirs along with the fluorescent dye coming from reservoir S,

thus diluting the sample. Middle and bottom frames: sample is injected into the separation channel by applying

+ 3 kVon side channel 5, while keeping all other side channels at 0 V. A sample plug is drawn into the separation

channel and superfluous dye is drawn into the sample channels on both sides of the injection cross.

R.A. Guijt-van Duijn et al. / Biotechnology Advances 21 (2003) 431–444 437

R.A. Guijt-van Duijn et al. / Biotechnology Advances 21 (2003) 431–444438

electrode surface and possibly causing changes in surface quality or the formation of

air bubbles.

A typical example of such a device is shown in Fig. 6.

4. Tools for intracellular components

Components such as metabolites and proteins need to be extracted from the cells in

such a way that the actual sampling does not affect their levels. This is probably not a very

big problem for proteins. However, for mRNA, this issue is unresolved and for

metabolites, it is certain that they will be affected by any sampling procedure. Several

sampling procedures have been described in the literature, and for yeast, we have adopted

the following:

1. Collect cells in a methanol–buffer mixture at � 40 jC in such a way that the

temperature never exceeds � 30 jC;2. Remove the supernatant by centrifugation at � 30 jC;3. Extract metabolites in boiling ethanol at 60 jC;4. Evaporate solvent and redissolve the residue in water or buffer for analysis.

This method has been applied to follow the dynamic response of sudden jumps in

glucose supply in a steady-state yeast culture.

For proteins, a simpler protocol was followed, involving breaking-up cells with glass

beads and extracting the proteins in aqueous medium (Overkamp et al., 2002).

Fig. 7. Size comparison between a classical 96-well microtiter plate and a 5� 5 nanowell array with circular wells

of dimensions of 300� 50 Am.

R.A. Guijt-van Duijn et al. / Biotechnology Advances 21 (2003) 431–444 439

Metabolites were analyzed by a combination of liquid chromatography (LC) and mass

spectrometry (MS). This method has been extensively described in Visser et al. (2002).

(Glycolytic) enzymes were analyzed using NAD(P)H-dependent methods, basically as

described by Bergmeyer et al. (1984). Arrays of 25 wells, 400� 400� 50 Am in size (see

Fig. 7 for a size comparison with a conventional 96-well microtiter plate) were prefilled

with the appropriate mixture(s) of substrate(s) and ancillary enzymes.

At the time of measurement, appropriately diluted cell-free extract was applied to the

wells and these were then immediately closed with a cover slide and placed in the

measuring system. This consisted of an x–y table, the settings of which were programmed

to allow each well to be exposed to the detector (vide infra) in turn for e.g., 1 s. When all

wells had been scanned, the process was repeated until sufficient information had been

gained (typically 2–5 min).

The detector proper was an adapted fluorescence microscope fitted with a CCD camera

and the appropriate read-out electronics. In this way, progress curves could be collected for

each individual well.

5. Results

5.1. Extracellular components

An example of a typical separation is shown in Fig. 8. All components are baseline

separated within a little over 2 min, which is fine on the scale of any imaginable

fermentation.

Fig. 8. Separation of 1 mM (each) fumaric (1), citric (2), succinic (3), pyruvic (4), acetic (5), and lactic acid (6) in

a 20 mM MES/His (pH 5.8) buffer. 0.2 mM TTAB was added to the buffer in order to reverse the electro-osmotic

flow. Separation at 1000 V or 167 V/cm.

Fig. 9. Electropherogram of a separation of a 5 mM (each) mixture of K+, Na+ and Li+ in 20 mM MES/His (pH

6.0). Injection for 3 s at 300 V, separation at 300 V (50 V/cm).

R.A. Guijt-van Duijn et al. / Biotechnology Advances 21 (2003) 431–444440

Another example is shown in Fig. 9, where cations are separated in the device

described in the previous section. In both examples, the integrated conductivity detector

was used.

The sensitivity of the conductivity detector was tested by injecting a dilution series of

Na+ ions as shown in Fig. 10. Clearly, concentrations down to 10 AM can still be detected.

The last example shows a separation of fluorescently labeled amino acids on the device

shown in the previous section which was used for testing of EOIP as an alternative

injection tool. In this example, laser-induced fluorescence was used for detection (see Fig.

11). A baseline separation is achieved within 20 s.

Fig. 10. Calibration curve for Na+-ions on the integrated device.

Fig. 11. Separation of FITC-labeled amino acids. Combination of EOIP and EOP for the separation of 10 mM

FITC-labeled Arg, Phe and Gly in buffer. Sample loading took 30 s using EOIP, with a potential difference

applied between channels 3 and 4 (4 kV). Injection was performed using a potential difference applied between

reservoirs 5 and 6 (4 kV), and pushback voltages applied to side channels 1 and 3 (2.3 kV). Separation length

(LD): 11 mm, buffer: 32 mM carbonate buffer, pH 9.5.

R.A. Guijt-van Duijn et al. / Biotechnology Advances 21 (2003) 431–444 441

It is concluded that separations on integrated devices are possible and that sensitivities

down to the micromolar level can be attained in these integrated, miniaturized separation

systems.

5.2. Intracellular enzymes

Several couples or single enzyme systems were tested, first in microtiter plates and

subsequently in chips containing wells with nominally 8 nl volumes. As an example, we

show a typical progress curve of the reaction:

Glucoseþ NADþ ! Gluconolactoneþ NADH

recorded in a single well in Fig. 12.

Fig. 12. Progress curve of glucose-6-phosphate dehydrogenase in a single 8 nl well. The ‘‘camera response’’

represents arbitrary fluorescence units.

R.A. Guijt-van Duijn et al. / Biotechnology Advances 21 (2003) 431–444442

Similarly, an entire 5� 5 array can be recorded and typical progress curves for the

reaction of pyruvate kinase, coupled to lactate dehydrogenase as shown in the frame

below.

The progress curves recorded on the chip are shown in Fig. 13.

Fig. 13. Readings of pyruvate kinase off an array of 8 nl wells. In each case, the x-axis (time) ranges from 0 to 400

s. The y-axis (fluorescence) is in arbitrary units.

Fig. 14. Yeast cell ‘‘colonies’’ grown in an 8 nm well.

R.A. Guijt-van Duijn et al. / Biotechnology Advances 21 (2003) 431–444 443

The same reactions could be followed starting from cell-free yeast extracts. In the same

platform, metabolites can be measured, again using the Bergmeyer methodology (Berg-

meyer and Bermeyer, 1984). So far, we have not applied this technology for assaying

metabolite levels.

We calculated the number of cells from which the enzyme contents measurable in a

single 8-nl well corresponded. This turned out to be approximately 3000 cells. Further

miniaturization by a factor of 1000 would bring measuring enzyme levels in single cells

very close. Such a further down scaling might seem difficult, but it is entirely feasible in

terms of microfabrication.

A first step into this direction was to see if single cells could grow in these small wells.

Very dilute suspensions of yeast cells in growth medium were bulk filled using the co-

verslip method (see Moerman and van Dedem, submitted for publication) into wells and

incubated in a moist environment. These cells contain a GFP fusion protein, and therefore

show a nice green fluorescence. After one day, colonies could be seen as illustrated in

Fig. 14.

6. Conclusion and outlook

We have demonstrated that miniaturized methods are feasible and can lead to

significant savings in time, reagent and sample volumes.

In the third section, elements for a real-time monitoring system for extracellular

components in (bio)processes were discussed, ranging from the sampling system through

a fluid manipulation system to a separation and detection system. All these ingredients will

need to be integrated into a robust and reliable instrument that can be used with the same

R.A. Guijt-van Duijn et al. / Biotechnology Advances 21 (2003) 431–444444

ease as a pH electrode. This indeed is the subject matter of a follow-up project. In the

fourth section, miniaturized tools for functional testing of intracellular enzymes and

metabolites were described. Although their realization has in fact been demonstrated, a lot

of technical improvements are still needed, as well as standardization. However, we feel

that such developments are more suitably pursued by an instrument manufacturer than by

academic research. Finally, the yeast experiments show interesting possibilities for

studying single cells at the level of the transcriptome, the proteome and the metabolome.

This line of research will be actively pursued in the near future.

References

Bergmeyer HU, Bermeyer J, editors. Metabolites 1: carbohydrates. 3rd ed. Methods of enzymatic analysis, vol. 6.

New York, USA: John Wiley and Sons; 1984.

Bergmeyer HU, Bermeyer J, Bral M, editors. Methods of enzymatic analysis, enzymes I, oxido-reductases,

transferases. New York, USA: John Wiley and Sons; 1984.

Freitag R. Biosensors in analytical biotechnology. Austin Texas, USA: R.G. Landes Company; 1996.

Guijt RM, Baltussen E, van der Steen G, Frank J, Billiet HAH, Schalkhammer T, et al. Capillary electrophoresis

with on-chip four-electrode capacitively coupled conductivity detection for application in bioanalysis. Electro-

phoresis 2001;22:2537–41.

Guijt RM, Lichtenberg J, Baltussen E, van Dedem GWK, de Rooij NF, Verpoorte E. Indirect electro-osmotic

pumping. JALA 2002;7(3):62–4.

Laugere F, Guijt RM, Bastemeijer J, van der Steen G, Berthold A, Baltussen E, et al. On-chip contactless four-

electrode conductivity detection for capillary electrophoresis devices. Anal Chem 2003;75:306–12.

Moerman R, van Dedem GWK. A cover slide method for rapid filling and sealing of (sub)nanoliter wells to

perform reactions in defined volumes; 2002 [in press].

Moerman R, Frank J, Marijnissen JCM, Schalkhammer TGM, van Dedem GWK. Miniaturized electrospraying as

a technique for the production of microarrays of reproducible micrometer sized protein spots. Anal Chem

2001;73:2183–9.

Overkamp KM, Bakker BM, Kotter P, Luttik MA, Van Dijken JP, Pronk JT. Metabolic engineering of glycerol

production in Saccharomyces cerevisiae. Appl Environ Microbiol 2002;68(6):2814–21 [Jun].

Piper MD, Daran-Lapujade P, Bro C, Regenberg B, Knudsen S, Nielsen J, et al. Reproducibility of oligonucleo-

tide microarray transcriptome analyses. An interlaboratory comparison using chemostat cultures of Saccha-

romyces cerevisiae. J Biol Chem 2002;277(40):37001–8 [Oct. 4].

Visser D, van Zuylen GA, van Dam JC, Oudshoorn A, Eman MR, Ras C, et al. Rapid sampling for analysis of in

vivo kinetics using the BioScope: a system for continuous-pulse experiments. Biotechnol Bioeng 2002;79(6):

674–81 [Sep 20].