Profile. The development of integrated microfluidic systems at GeSiM

Background

GeSiM¹ mbH (short for Gesellschaft für Silizium-Mikrosysteme mbH, i.e., ‘Silicon Microsystems Company Ltd.’) has its origins in research projects performed at the Forschungszentrum Rossendorf,² a major research facility in the former East Germany where materials research, the basis of GeSiM, has become one of the focal points after the German reunification. GeSiM’s aim is to develop integrated microfluidic systems to aid the miniaturisation of biotechnological processes. GeSiM is a world leader in non-contact subnanolitre dosage and microarraying and maintains a cleanroom facility in which microsystems can be built according to customers’ requests. As a small company, GeSiM cannot offer the latest, fanciest, and most expensive technology, but it does offer a cost-effective and fast one-stop service which includes processes as diverse as the development and production of microfluidic devices, chip packaging, and the manufacturing of entire instruments.

The past few years there has been much hype around lab-on-a-chip technologies; many people have entered the field and maybe more have been writing about it. But with the possible exception of microarrays, the market has reacted indifferently to these developments. This attitude is about to change, because so many achievements have been made. What then are the advantages of ‘micro’ in biotechnology? The biggest point is parallel sample processing, which by definition enables high throughput to screen either nucleic acids, proteins, whole cells, or chemicals, thus saving an enormous amount of time while keeping the workforce away from tedious repetitive tasks. The second point is the decrease of sample volume and thus a dramatic reduction of costs (of expensive enzymes, nucleic acids, etc.). GeSiM, as a company not dependent on venture capital, have taken a pragmatic, evolutionary approach to microsystems technology in the biotech in that they introduce novelties one at a time and test their marketability. This, however, requires customers who know what lab-on-a-chip devices they need.

Microfluidic pipetting and microarraying

GeSiM's first achievement was the design of a microfluidic pipette that dispenses tiny volumes, awarded the ‘Innovation Prize of the German Industry’ by the German government in 1997. The pipette, shown in Fig. 1, was invented on the basis of classical microsystems technology: it consists of a thin film of silicon on which a piezoelectric crystal is mounted.^3,4 On the opposite side, a mould and a thin nozzle are etched into the silicon. The cavity is closed by a glass lid through anodic bonding and the pipette is cut out by a wafer saw (Fig. 1A). In order to work, the microcavity of the pipette (‘pump chamber’) must be filled with fluid. If a voltage is applied, thus contracting the piezo crystal, the silicon film contacting the pump chamber is bent and a microdroplet is ejected from the nozzle. A constant fluid flow can be achieved by applying periodic voltage pulses (Fig. 1B).

Fig. 1 The microfluidic piezoelectric pipette. (A) Exploded view of the standard pipette with its thin silicon film. The pump chamber is produced by etching the silicon from both sides, producing a very thin layer of silicon which is able to oscillate. A piezoelectric actuator is mounted on the back side of the silicon. The fluidic system is closed by anodic bonding of the glass lid. (B) Stroboscopic picture of a firing pipette. It can be seen that homogeneous droplets are dispensed. (C) Micropipette as in (A), mounted and electrically connected, above a nanotitre plate (produced by GeSiM) used for mixing sub-microlitre assays.

GeSiM have constantly optimised the pipette design, resulting in a range of pipettes dispensing as little as 15 pl and reaching dispensing rates of up to 600 μl min⁻¹. Pipettes to mix fluids by, e.g., firing droplets from two nozzles against each other were also constructed.^3,4

With the construction of the world’s first microfluidic pipette, GeSiM entered the biotechnology market, a way that had to be taken because a standalone pipette is not of much use. Two possible applications for the pipette were conceived: to dose a defined amount of fluid into narrow cavities, which is described later, and to produce microarrays. The development of DNA microarrays in the mid to late nineties⁵ required the dosage of nanolitre volumes and thus paved the way for the marketing of microfluidic piezoelectric pipets, which GeSiM has built into XY stages since 1997. The newest generation of this instrument, the Nano-Plotter 1.2,⁶ has evolved to an instrument ideally suited for fast, versatile, and dependable microarraying (Fig. 2). It is equipped with two types of pipettes dispensing either ∼0.1 or ∼0.4 nl per droplet. As a piezoelectric pipette can only eject, but not aspirate fluid, the pipettes are connected to syringe pumps (Fig. 2A), which are used both to take up sample and to rinse the pipettes, and as the size of the pump chamber is slightly more than 0.5 μl, the minimal amount of sample used is about 1 μl and thus slightly more than in pin tool instruments. Although the pipettes are cut from silicon, they are only slightly thicker than steel pins and thus suited to pipette into deep-well microtitre plates, which is interesting for the production of low-tech, low-cost arrays for mass screening. Piezoelectric non-contact dosage has several advantages over contact printing with pins:

Fig. 2 The sub-nanolitre arraying and dispensing system ‘Nano-Plotter’, based on microfluidic piezo pipettes. (A) Setup of the instrument, without the controlling computer. The instrument with XY stage, pipettes and fluidic system inside a dust cover is shown in the left part and the monitor to view either the pipette head or the stroboscope and the syringe pumps to aspirate the samples are seen on the right side. (B) Droplet sensor. This chip is mounted in the back of the XY stage and consists of an array of 8 × 4 interdigitated electrodes (whose fine structure is not resolved on the picture) structured on a glass substrate. The glass chip sits on a printed circuit board on which the connector pins are mounted. If a droplet hits the rectangular electrode pad, an electric current is generated and monitored. (C) Array of drops on a flat surface demonstrating that the instrument is also able to pipette and mix larger volumes.

1. Reproducibility: since the droplets have always the same size, the long-term dispensing error is around 2%. Due to the regular spherical shape (Fig. 1B) of the droplets and the defined and predictable mode of action of the piezo, a more uniform and superior spotting quality is achieved.
2. Smaller droplet size: pin tool instruments (but also non-contact dispensing methods using high-speed micro-solenoid valves⁷) typically deliver volumes from slightly under a nanolitre to several nanolitres. The diameters of spherical droplets of 0.1 and 0.4 nl, as produced in the Nano-Plotter, are about 60 and 90 μm, respectively. Therefore, dense arrays with a spot pitch of less than 200 μm can only be produced with piezoelectric pipets in conjunction with hydrophobic target surfaces where the droplets only marginally spread.
3. Dispensing of samples of different composition: whereas pin tools require a proper choice of fluid composition in order to achieve proper adhesion and dispensing, piezoelectric pipettes can handle a range of samples, including protein, detergents, solvents, and small particles, because the piezo parameters (voltage and pulse length) can be adjusted accordingly. The viscosity should be less than 5 mPas, which means that solutions containing 30–40% glycerol can still be handled. Protein solutions, for instance, have different adhesion characteristics than diluted DNA solutions, because they are more viscous and may contain detergent, so when it comes to the production of protein microarrays, customers tend to look for a non-contact arrayer, even when they own a pin tool machine for DNA arrays.
4. Arbitrary spot pattern: pin instruments have a fixed geometry because all pins touch down simultaneously. If more than one pin is used, the spotting geometry is given by the spatial arrangement of the print head (which depends on the microtitre plate used). In a non-contact arrayer, all pipettes can also fire at the same time, thus saving time while having the same geometric limitations as pin tool instruments. However, arbitrary spot patterns (a scrambled order of the samples, circular or radial patterns, etc.) are also possible, because the pipettes can dispense samples sequentially, i.e., one at a time. This gives the operator more freedom for the production of microarrays, especially in the process of chip design, without the need to rearrange the probes in the microtitre plate. The Nano-Plotter is shipped with software which allows the development of spotting programs of unlimited complexity by using a Pascal-like computing language. If a program is set up, however, no further programming is necessary; i.e., microtitre plates and spottable targets are chosen graphically and the user is guided through standard dialogs asking for certain numerical values such as numer of droplets, number of blocks to be spotted, number of replicas in the X and Y directions, etc. The instrument ships with two preset programs for simultaneous and sequential spotting and can integrate Zymark’s Twister plate handler.
5. Functional testing of the pipettes: the function of the pipettes filled with sample can be tested before spotting. If, for instance, air is trapped, no droplets are delivered. This can be checked manually by observing the droplets in a stroboscope (an automatic check via image processing is in preparation) or automatically in a microelectronic droplet sensor,⁸ which is an array of interdigitated electrodes produced by microstructuring of platinum (Fig. 2B). It is possible that, after a run, missing spots can be repaired by spotting with another pipet, adding a high degree of quality control to the production of microarrays.
6. Use of piezoelectric pipettes for biochemical assays: non-contact pipettes in an XY stage can also be ‘abused’ for mixing. Since the pipetted volume can be varied by varying the number of droplets dispensed (Fig. 2C), not only can microarrays be spotted, but also assays can be performed, e.g., for diagnostics or environmental testing. Miniaturisation of the volume will be an issue where costs are to be reduced by minimizing the consumption of reagents. For this purpose, GeSiM is offering nanotitre plates (see Fig. 1C) produced by anodic bonding of a sieve plate from silicon and a transparent glass bottom, allowing the recording of optical data. It has been shown that mixing occurs upon injection of droplets into samples; this is because a droplet leaves the pipette with a velocity of 1 to 5 m s⁻¹. Thus, as long as the total volume remains small, the Nano-Plotter is sufficient to perform biochemical assays; no further agitation by, e.g., surface acoustic waves,⁹ is necessary. A humidifier to minimise evaporation of samples during pipetting is available.

Disadvantages of the piezo-actuated pipetting are the more complicated setup of the instrument, the need to exactly adjust pipette parameters (for washing, drying, sample uptake, and dispensing), the inability to handle highly viscous media, and the need to regularly check for dirty or broken pipettes and air bubbles or particles in the fluid system.

One might think that it is more difficult to avoid carry-over with pipettes than with pins. However, experiments carried out in our laboratory with DNA probes and proteins have shown that cross-contamination is not an issue if the pipettes are rinsed for a few seconds from the inside and outside. In fact, this process is faster than the cleaning of pin tools, which generally contain a small uptake channel which is not easily accessible and thus requires several cycles of washing while applying ultrasound. Moreover, the pipettes, having an SiO₂ surface, are chemically inert and can be treated with acid, alkali, and detergents.

Microfluidics and engineering service

As a microsystems manufacturer, GeSiM are constantly cooperating with customers to develop new lab-on-a-chip systems and to eventually become an original equipment manufacturer (OEM) for these devices, as was the case with the dielectrophoretic cells (see below). Aside from rapid construction (including finite element treatment), prototyping, and quality assurance, GeSiM can offer a series of microsystems techniques, such as the manufacturing of very cost-effective glass photomasks for MEMS using LCD technology (for structures ≥ 20 μm) and of more accurate chrome masks, physical and chemical vapour deposition, dry and wet bulk micromachining, double-sided lithography, anodic bonding, die and wire bonding, micro-adhesive bindings, and micromachining of LTCC (low-temperature co-fired ceramics; allowing the manufacturing of stacks containing more than 50 layers). Custom electrode patterns made of platinum, gold, aluminium, titanium, tantalum, titanium dioxide or indium tin oxide, which can be effectively isolated by PECVD (Plasma Enhanced Chemical Vapour Deposition) with Si₃N₄, SiO₂, or plasma-polymerised Teflon), can be integrated into the microsystems. Moreover, manufacturing of moulds for the structuring of poly(dimethyl siloxane) (PDMS) is possible. An overview of the microsystems technology available at GeSiM is given in Fig. 3.

Fig. 3 Micromachining techniques available at GeSiM. (A) Silicon–glass technology (with anodic bonding), allowing channel heights of 10–800 μm (aspect ratio, i.e. height/width >10) and layer thicknesses of 20–1000 μm. (B) Glass–silicon–glass structures. The glass layers can vary between 200 and 1000 μm whereas the silicon spacer and thus the channel height might be as thin as 10 μm. (C) Silicon–glass–silicon structures. The glass layer can be 150–1000 μm allowing aspect ratios of the channels of <1, the silicon layers can vary between 50 and 1000 μm. (D) Silicon–silicon–silicon technology, allowing channel depths of 10–500 μm at a wafer thickness of 200–500 μm. A possible application is the production of very deep cavities, as shown. (E) Glass–polymer–glass structures. SU-8 resist (grey) is structured as channel wall and the second glass layer is glued onto it. This technique is compatible with light microscopy, i.e., removable cover slips can be integrated into the manifolds. (F) LTCC, here shown as a four-layer structure; up to 50 layers are possible. Structures in (E) and (F) allow the integration of metal electrode structures.

The list would not be complete without mentioning a specialty of GeSiM, namely glass–silicon–glass multilayer systems (three layers, 150–500 μm thick, anodically bonded, fluid connections to both sides possible) that can be used as optical microcuvettes. A further microfluidic component available is an electrocalorimetric flow sensor (structured from silicon, glass, silicon nitride, and glue), which can detect flows from less than 1 μl min⁻¹ to 70 μl s⁻¹ at pressures of up to 40 bar (4 × 10⁶ Pa) by heating up the liquid by 5–10 K and measuring the profile surrounding a diaphragm containing temperature sensors. Moreover, the integration of heaters, heat sensors, and thermally driven hydrogel microvalves, plus the manufacturing of microflow cells for chemical sensors (e.g., ISFETs) in a configuration as described in the following section can be integrated into microsystems. In addition to micromachining, GeSiM can offer precision engineering and is thus able to manufacture complete instruments (e.g., the Nano-Plotter, Fig. 2) around microfluidic chips.

Application example: micro total analysis system (μTAS) to monitor water pollution

One of the first implementations of the piezo technology was a chemical sensor (Fig. 4) in which sample is introduced on the left end, while the main flow of the device runs in the horizontal direction, driven by a high-rate piezoelectric pipette on the right end (‘pump’). The omission of any standard pumps with moving parts leads to a high degree of integration. The sample is transported rightwards to an electrochemical sensor (an ISFET in this case), which is glued onto the chip. Since the baseline drift of the ISFET requires regular calibration, small volumes of calibration solution are injected by a second, vertically mounted pipette. This pipette fires into the main channel through a sieve plate etched in silicon through which fluid can enter the main flow from the outside without fluid leaking from the inside⁴ (fluid diode). The sieve plate also leads to a more even distribution of the calibration fluid in the main flow, i.e. the laminar flows mix faster than in a T junction, as shown by finite element simulation⁴ (Fig. 4, bottom). But although a fully functioning sensor for nitrate was finally obtained, microfluidic techniques were not popular in environmental testing at that time.

Fig. 4 Chemical sensor built into a channel system containing a microfluidic injector. The main channel contains buffer (or analyte) which is aspirated by the piezoelectric pipette at the right end. The injector pipette (built similar to the one in Fig. 1) is used to introduce small amounts of calibration fluid or analyte into the main flow via a sieve plate (fluid diode⁴). A scanning electron micrographic picture of a KOH-etched sieve plate is seen at the top of the figure. At the bottom, a FEM (finite element method) simulation of the mixing of the two fluids is given.

Application example: integration of electrodes in microfluidic manifolds

Integration of electrodes dramatically enhances the value of microfluidic systems. A popular lab-on-a-chip application of this kind is capillary electrophoresis (CE).^10,11 GeSiM has built custom-made prototypes for CE chips. But the greatest achievement so far is the manipulation of particles in microfluidic channels by the action of radio-frequency (MHz) electromagnetic fields. This technique, invented by Fuhr and coworkers,¹² was optimised to allow cell sorting and single cell processing^13,14 in modern inverse microscopes by Evotec Technologies GmbH with microsystems engineered at GeSiM.¹⁵ The flowthrough chips are multilayered structures of 150 μm thin optical glass. Electrode structures from platinum, indium tin oxide (ITO), or aluminium are used to generate the dielectric fields inside the fluidic channels. Flowthrough channels are formed by structuring of positive or negative photoresists (Hoechst) or of the negative photoexpoxy resist SU-8 and gluing of the second glass plate onto this setup (Fig. 3E). Channel heights can be 10, 20, or 30 μm, stacking of channels up to 120 μm is possible. Tube connectors are added, electrodes are bonded to printed circuit boards, and the entire chip is packaged into a plastic housing. The dielectrophoretic flowthrough cells are OEM products and are exclusively sold by Evotec Technologies GmbH.

The controlled application of alternating electromagnetic signals by specially designed electrodes allows the movement of cells or particles at will. Since cells are largely insensitive to electromagnetic radiation, microscopic analysis of cells without touching them has become possible, reducing any artifacts produced by the adherence of cells to surfaces. The microstructures contain different types of electrodes, such as funnel electrodes for queuing of cells, field cages for exact positioning of cells, deflectors and switches for highly specific and efficient sorting of single cells, electrode systems for cell permeation, and much more (Fig. 5).

Fig. 5 Dielectrophoretic cell manipulation chips. (A) Sorter chip. The fluidic channel running from left to right has an inlet on the left side and a waste outlet on the lower right; its walls from photoresist are dark. Selected cells are passed through the dielectric switch and leave the chip through the upper right channel. Several dielectrophoretic microelectrodes are seen (from left to right): a funnel for cell queuing, a deflector, a hook (parking element for cells), the cage in the centre (with four of the eight electrodes visible), and the switch at the right channel bifurcation. The fluidic cell can be combined with all common microscopic techniques, e.g., fluorescence imaging, phase contrast, and confocal laser scanning microscopy. (B) Loader chip with a four-way channel system for the stopped flow analysis of single cells and the isolation of cells after recording of reaction kinetics. The cells enter the system on the left and and leave the chip on the right side where they can be sorted by a switch as in (A). Two different chemicals can be applied through the upper and lower channels of the junction. Dielectrophoretic electrodes are parking elements (zigzag) in the left and the right channels, deflectors, a funnel, the cage at the channel junction, and a switch on the right. Again all usual microscopy techniques can be applied. (C) Microscopic picture of a human promonocytic leukemic U937 cell held in place in a dielectrophoretic field cage, taken at 40× magnification.

Aside from cell sorting, dielectrophoretic caging of cells has allowed non-contact cell analysis. In a spatial array of eight electrodes at the corners of a cube, cells can be fixed (caged) at a certain position, allowing three-dimensional confocal imaging. Another application is a label-free viability assay: cells are turned by the electric field and cells can be irradiated by a laser beam or drugs can be introduced into the microchannel. Cell damage can be seen by increased rotation speed of the cells.¹⁴ A list of further possible applications, without being exhaustive, would contain the analysis of receptor/ligand interactions on cell surfaces (including assays using ligands bound on microspheres), fluorescent cell viability assays, drug screening, and, last but not least, cell sorting and archiving.

Dielectrophoretic processing thus represents a new micro-toolbox for the different tasks in biotechnology, such as sorting of individual cells, characterisation of single cells, cloning, and other manipulations. We expect it to become an indispensable addition to modern microscopes, maybe in combination with laser-based single molecule detection, and will find its applications in medicine, diagnostics, pharmacology, biotechnology, biophysics, and even physics.

Outlook

GeSiM's short- to middle-term focus will be the optimisation of the non-contact arrayer. A successor of the already successful Nano-Plotter 1.2 is almost ready to be shipped. A major advancement will be the integration of 16 independent channels (two rows of eight pipettes) in its pipetting head. The platform of the instrument is a new type of highly precise, but still cost-effective XYZ robotics system (repetition inaccuracy around 5 μm). Since it is now scalable, users can obtain a larger platform upon request. Droplet dispensing will be possible without stopping the head during disposal (spotting on the fly). The instrument will include the option for automatic target recognition in which positions highlighted by a different colour (e.g. gold pads) are automatically detected and used for high-precision spotting. In addition to the arraying functionality, the image processing capabilities in combination with additional tools (e.g., for picking) will allow this instrument to assemble, glue, and to electronically test the entire range of microelectromechanical components, including biosensors, bio-MEMS and optical MEMS.

Other activities will be the expansion of our program of flowthrough cells for cell analysis. In cooperation with Max-Planck institutes, new cell-based manifolds, with and without microscopic observation, are in preparation. These devices can be combined with other devices of our product line, i.e. a fluidic system including either conventional micro-syringe pumps¹⁶ or micropumps^3,4 (as in Fig. 4), hydrogel valves, fluidic sensors, and integrated electrochemical detection.¹ They will allow kinetic and thermodynamic measurements of the behaviour of cells against certain drugs or the recording of the movement of single fluorescently labeled biomolecules under the microscope. With these new developments and further acitivities in micro-PCR and the expansion of our pipetting technology (biochemical assays, protein crystallisation, integration into μTAS activities), GeSiM might enter a position to be a significant supplier of microdevices for pharmaceutical screening, diagnosis and environmental testing.

Acknowledgements

This article would not have been possible without the numerous contributors at GeSiM, especially Dr. S. Howitz who also proofread the manuscript. We are grateful to the German ministry of science and education (BMBF) to fund our past and ongoing projects, and to Professor G. Fuhr (Fraunhofer-IBMT, Humboldt-Universität zu Berlin) and Evotec Technologies GmbH for their cooperation with the dielectrophoretic cell processors.

References

1. http://www.gesim.de/.
2. http://www.fz-rossendorf.de/.
3. S. Fischer-Frühholz, Am. Lab., 1998, 2, 46–51.
4. S. Howitz, BioMethods, 1999, 10, 31–73.
5. M. Schena, D. Shalon, R. W. Davis and P. O. Brown, Science, 1995, 270, 467–470.
6. http://www.gesim.de/indexnp.htm.
7. http://www.cartesiantech.com/index.htm.
8. German Patent ‘Sensormessfeld’, PCT/DE00/00455, priority 1999/01/19.
9. http://www.advalytix.de/.
10. C. S. Effenhauser, G. J. Bruin and A. Paulus, Electrophoresis, 1997, 18, 2203–2213.
11. http://www.calipertech.com/products/index.html.
12. T. Müller, W. M. Arnold, T. Schnelle, R. Hagedorn, G. Fuhr and U. Zimmermann, Electrophoresis, 1993, 14, 764–772.
13. S. Fiedler, S. G. Shirley, T. Schnelle and G. Fuhr, Anal. Chem., 1998, 70, 1909–1915.
14. http://www.evotec-technologies.com/.
15. S. Howitz and G. Fuhr, Int. Pat. Appl. ‘Verfahren zum Herstellen einer 3-D-Mikrodurchflusszelle und 3-D-Mikrodurchflusszelle’, PCT/DE01/03324, priority 2000/09/07 (for GeSiM mbH).
16. http://www.cavro.com/.

---