Medical Device Link.

IVD Technology Magazine
IVDT Article Index

Originally Published November/December 2000

Microfluidics-based lab-on-a-chip systems

A small-scale separation and detection technology that puts a diagnostic laboratory on a card holds promise for point-of-care applications.

Miniaturization and automation have revolutionized the world of microelectronics. Whereas computers once were room-sized, steady technological advance has led to laptops, palmtops, and even game consoles that are much more powerful than were those behemoths of the 1950s and 1960s. But until recently few of these cutting-edge engineering technologies have been applied to the needs of the medical diagnostics laboratory. Consequently, much laboratory work long remained—and some still is—inefficient, laborious, and time-consuming.

Over the past five years, however, research and development for clinical diagnostic systems based on lab-on-a-chip technologies have increased tremendously. To date, more than 2000 research papers on labs-on-a-chip, sometimes also called micro total-analysis systems (µTAS), have been published. Such systems hold great promise for clinical diagnostics. They consume sample material and reagents only in extremely low volumes. Individual small chips can be inexpensive and disposable. Time from sampling to result tends to be very short. The most advanced chip designs can perform all analytical functions—sampling; sample pretreatment; separation, dilution, and mixing steps; chemical reactions; and detection—in a single integrated microfluidic circuit. Lab-on-a-chip systems allow designers to create small, portable, rugged, low-cost, and easy-to-use diagnostic instruments that offer high levels of capability and versatility. Microfluidics—fluids flowing in microchannels—makes possible the design of analytical devices and assay formats that would not function on a larger scale.

Figure 1. Flow schematic of an H-filter used for the diffustion-bassied separation of smaller from larger particles.

Several lab-on-a-chip manufacturers, including Aclara (Mountain View, CA), Caliper (Newton, MA), Orchid Biosciences (Princeton, NJ), and Cepheid (Sunnyvale, CA), have developed technologies that work very well with the highly predictable and homogeneous samples that are common in genetic testing and drug discovery processes such as dilute solutions of genetic material in buffers or aqueous solutions of drug lead compounds. The first generation of instruments based on these technologies is now entering the non-FDA-regulated life sciences research market. However, these current lab-on-a-chip systems can be seriously challenged when asked to perform analyses of such complex and heterogeneous clinical samples as whole blood.

Lab-on-a-chip devices now being commercialized generally consist of a small microfluidic chip surrounded by a substantial, typically desktop-sized, analytical instrument.^1–5 Pursuing further integration of electronic and fluidic system components may lead to greater miniaturization. But even more-substantial miniaturization, as well as automation advances and analytical-cost reduction, could result from taking advantage of some of the inherent characteristics of microfluidics. This article presents one company's efforts to develop microfluidic diagnostic systems capable of handling difficult real-world clinical samples such as whole blood and other bodily fluids.

Diffusion-Based Chip Technologies

Micronics Inc. (Redmond, WA) has developed several lab-on-a-chip technologies that are implemented as low-cost plastic, disposable, integrated microfluidic circuits, typically in credit card–sized cartridges. These technologies have demonstrated potential diagnostic efficacy for a variety of analytes and samples, including enzymes, proteins, electrolytes, heavy metals, and drugs, and may have utility in cellular analysis.

Figure 2. A schematic representation of diffusion-based detection using the T-Sensor

A diffusion-based separation and detection technology used with the company's H-Filter is based on the parallel flow of two or more fluid streams in a single microfluidic channel. Under microfluidic conditions, fluids usually travel in a very predictable laminar fashion that allows miscible fluids such as whole blood and phosphate buffered saline (PBS) to flow next to each other without any turbulent mixing and without being physically separated by a membrane. However, diffusion does occur under these conditions. This is the physical process by which molecules and particles transported in the fluid incline to migrate from areas of higher to areas of lower concentration. Smaller particles, such as ions, small proteins, and many drug molecules, generally diffuse quickly across the boundary layer, whereas large molecules and particles such as cells tend to diffuse only minimally. This effect can be exploited not only to separate particles by size, but also to extract components from samples. For example, H-Filter technology can be used to generate a serum equivalent directly from whole blood without the use of a filter or centrifuge. It also has application as a desalting step in DNA sample preparation.

Figure 3. A disposable detector card with an hydrostatically driven integrated T-Sensor design. Dimensions for height (H), width (W), and length (L) of system components are given in millimeters. Tanks are 1.78 mm thick, and channels are 0.1014 mm thick. The photograph at right shows such a T-Sensor in operation determining the pH of a buffer solution.

The H-Filter is diagrammed in Figure 1. A sample (for example, blood) is put into a reservoir at one end of one post of the H, and an acceptor reagent such as water or saline is placed in the reservoir at the other end. Two parallel laminar streams will flow along the crossbar of the H as a result of hydrostatic pressure. Smaller components of the sample stream will diffuse across the interface into the acceptor stream in accordance with natural principles. The parallel flows are then sent by design into two separate reservoirs in the other post of the H. One reservoir holds the sample solution with a reduced concentration of the extracted component, and the other holds the acceptor reagent containing extracted sample material. Both reservoirs can then be either harvested from the cartridge for further use or processed through additional integrated microfluidic structures.

The technology behind the T-Sensor is based on a principle similar to that involved in the H-Filter, but with the difference that the sample molecules diffuse into a parallel stream that contains a reagent or an indicator, thus allowing for both qualitative and quantitative detection.

Figure 4. A passive chemical separation device with a hydrostatic-pressure-driven integrated H-Filter design. Depicted is an H-Filter experiment in progress, separating flouroscein from labeled dextrane molecules.

A T-Sensor is operated by flowing a sample, a reagent, and a control solution in a microfluidic channel in parallel, again through the agency of hydrostatic pressure (see Figure 2). Adjacent flowing solutions form diffusion interaction zones between them. Typically, a fluorescent or absorption indicator is added to the reagent. The sample and control analyte molecules then diffuse into the reagent stream, reacting with the indicator and forming visible diffusion interaction zones. The width and intensity (absorption, fluorescence intensity, chemiluminescence, etc.) of these diffusion interfaces at a particular location in the channel and for a particular flow speed are proportional to the concentration of the analyte.

These zones can be visually interpreted, by comparing width and color to a chart similar to a test strip for qualitative or semiquantitative assays, or they can be monitored with optical systems such as CCD cameras, linear diode arrays, or scanning lasers. Through determination of the ratio of an optical property exhibited in these two zones (chemiluminescence, fluorescence, or absorption), an essentially calibration-free concentration measurement can be derived.

With T-Sensors, using standard chemistries developed for serum analysis, quantitative analysis can be performed in whole blood, making the time-consuming centrifugation step unnecessary. T-Sensors have been tried successfully in many applications, including kinetic assays, protein and enzyme assays, immunoassays, and electrolyte assays.

Figure 5. An absorption-driven integrated T-Sensor design. The fluid is initially aspirated via capillary action, and then flow is controlled by the absorption pad.

Another key aspect of the T-Sensor is that both the sample material and one or more reference solutions can flow in parallel in real time through the same detection channel and past the same detector. This enables designers to devise self-calibrating, continuously controlled detection systems that would be particularly useful in handheld point-of-care devices for which quality control and calibration, on a per-test basis, are typically more difficult and expensive than with central-laboratory instruments.

One of the main applications of lab-on-a-chip devices is expected to be point-of-care diagnostics. In this arena, instruments often have to be miniaturized to handheld size and their operation simplified so that they can be used by nurses and doctors at the bedside and in the field. One avenue leading to further miniaturization and integration of µTAS elements in existing analytical instruments is to eliminate as many power-consuming and otherwise complex elements as possible, and to replace them with passive components that operate via such natural power sources as gravity, air pressure, absorption, capillary forces, or simple manual attention. Fruits of the investigations reported in this article include such microfluidic functional elements as mixers, separators, and detectors, and also complete microfluidic devices that function without any moving parts or external power sources. These devices lend themselves to applications such as ultra-low-cost disposable qualitative and semiquantitative clinical and environmental assays for home, office, and field use, and as sample- or reagent-preparation tools that produce processed liquids for downstream analysis or synthesis.

Microfluidic Devices

One such device is a hydrostatic-pressure-driven disposable microfluidic detector card that combines the ease of use of a paper test strip with the versatility of a microfluidic system (see Figure 3). The card is based on the T-Sensor diffusion-based detection method.^2,5,6–12 Here, a sample is put into the reservoir at the top left of the figure, a reagent (for example, an indicator dye) is put into the top middle reservoir, and a reference solution with a known concentration of analyte is put into the third reservoir. Comparison with a reference chart of the intensity and position of the two diffusion interaction zones that form provides a semiquantitative analyte determination. The detector card can also be monitored using a simple digital-camera-based readout device.

Figure 6. The Hematology Cartridge (Micronics Inc.), is designed to determine red-cell and platelet counts, hemoglobin concentration, a white-cell differential count, and various derived parameters.

Another passive device is able to separate chemical compounds by their diffusion coefficients and produce in 1 minute several microliters of cleaned-up sample for further processing (see Figure 4). It is based on the H-Filter method.^5,7,9 Such a device, requiring no power or moving parts, can be used as a simple and cheap replacement for a centrifuge.

The two devices just discussed are driven by hydrostatic pressure. Devices using other passive drive methods, including absorption and capillary force, have also been developed.

Figure 5 shows a T-Sensor with a built-in absorbing pad that acts as both a fluid driver and a containment cell for waste. The fluid is aspirated at first by means of capillary action. The flow is then controlled by the absorption pad, whose shape, geometry, and directional preference can be configured to achieve the desired flow speed for the fluid being absorbed. Plasma treatment of the channel surfaces makes the cartridge self-wetting.

Showing how these microfluidics-based devices can interface with a machine-controlled analytical system is a hematology cartridge that was developed to determine red-cell and platelet counts, hemoglobin concentration, a white-cell differential count, and various derived parameters (see Figure 6). The cartridge contains a variety of sample preprocessing structures, including mixers, diluters, and chemical reactors. It is connected with a toaster-sized external system containing a laser-optical interrogation system, a flow-control and fluid-interface system, and a reagent cartridge. Cell counting and characterization are accomplished with an array of laser-light-scattering detectors. The amount of scattered light at various angles carries information about the size, shape, and internal structure of the cells.

Figure 7. Schematic of a T-Sensor-based device for determining reaction rate via observation of the diffusion interaction zone along the sensor channel.

A few drops of whole blood are introduced into the card. The sample is then split up into three largely separate circuits. In the hemoglobin circuit, red blood cells undergo lysis. The extracted hemoglobin is converted to cyanomethemoglobin, which is then quantified in an optical-window portion of the circuit. In the second circuit, whole blood is diluted but not lysed. The cells are then focused into a microcytometer structure where red cells and platelets are counted and identified by means of light scattering. In the third circuit, whole blood again is diluted; but this time it is also lysed with "soft lyse," which renders red cells invisible to the detector while leaving white cells largely unaffected. The cells are then focused into the microcytometer, and white cells are classified into subpopulations with the use of multiangle light-scatter detectors.

T-Sensors in their basic form can be used to detect analytes in complex sample solutions such as whole blood. A number of variations on this technology have been developed. For example, a T-Sensor can be used to perform kinetic measurements (see Figure 7). With this device, reaction rate is determined by observation of the diffusion interaction zone along a T-Sensor channel. Spatial intensity distribution replaces time measurement for kinetic measurements.

Figure 8. (a) Schematic of the initial conditions in a diffusion immunoassay. The graph in (b) represents the concentration of unlabeled conjugate of the antigen to be monitored (LA) across the diffusion dimension at an early stage of diffusion (free LA, antibody-bound LA, and the total of the two). The graph in (c) represents the case when the concentration of binding antibody is much less than that of the sample and labeled tracer antigens combined. The concentration profiles shown in (b) and (c) were generated via an analytical model.

Another variation of T-Sensor technology, the diffusion immunoassay (DIA), is based on the difference in the diffusion coefficients of unbound antigens and antigen-antibody complexes (see Figure 8). It works particularly well with smaller analyte molecules such as drugs and haptens, and generally with most assays now used with fluorescence polarization immunoassay analysis systems.

A sample containing analyte antigen molecules is mixed, on- or off-chip, with a known concentration of labeled tracer antigen and flows (on the chip) next to a reagent stream containing high-molecular-weight antibodies for the analyte. (The former fluid may also contain diffusing and nondiffusing interferent compounds.) The antigens, both labeled and unlabeled, diffuse into the reagent stream, where they interact with the slower-diffusing antibodies. The higher the concentration of unlabeled antigens, the further the labeled tracer antigens have to diffuse in order to bind to antigens. Bound antigen molecules diffuse much more slowly, resulting in an accumulation of signal near the fluid interface. This effect is visualized and measured by the position of a fluorescent line in the microfluidic channel that indicates the position of bound tracer molecules flowing down the channel.

When the concentration of available binding antibodies is much less than that of unbound sample and tracer antigens combined, only a small number of antigen molecules are able to bind owing to the saturation of binding sites (Figure 8(c)). This results in a diffusion profile similar to that of free diffusion. Less labeled antigen accumulates near the fluid interface.

Figure 9. Predicted extraction efficiency for extracting salts from a DNA solution. The blue-shaded area represents the extracted salts.

The DIA has been demonstrated with an assay for phenytoin (Dilantin). It represents one of the fastest immunoassays published to date, with a complete response over the entire clinical range available in less than 20 seconds.

Theory and Modeling

All circuits described in this article were first designed using microfluidic flow modeling. A chief feature of microfluidics is that flow in structures built on the microfluidic principle is usually very predictable. Thus, flow models yield results that usually are very close to experimental data obtained with prototype devices.

For example, it is possible to predict with high accuracy the extraction efficiency of H-Filter devices for a given set of molecules and particles. Figure 9 shows the predicted efficiency of extraction of salt ions from a solution containing DNA molecules. With the particular H-Filter design used for that model, the concentration of salts in the DNA solution can be reduced by about 80% in less than 20 seconds.

Modeling is also extremely useful for optimizing quantitative detection of analytes using T-Sensors. The design of the T-Sensor shown in Figure 3, for example, required that the volume flow rates of the fluid streams be predicted through modeling.¹²

Microfluidic circuits can be represented as analogous to electrical circuits for ease of modeling the behavior of fluids in complex integrated fluidic circuits (see Figure 10). Resistances in a fluid circuit are like electrical resistors: a smaller-diameter, longer channel has a higher resistance than a relatively short, wide channel. A fluid capacitor, such as a fluid chamber covered by a membrane, is like an electrical capacitor; a pressure applied to the chamber "charges" the capacitor, and the charge is released over time into the fluidic circuit. Fluid pressure can be represented as if it were voltage, and fluid flow rate as current.

Figure 10. A hydraulic-circuit diagram for the T-Sensor shown in Figure 3. Values in the key are based on the circuit element dimensions indicated in Figure 3.

The values for the hydraulic-circuit elements in Figure 10 were based on the resistance and inertance of channels and the capacitance of fluid reservoirs having the physical dimensions of those system components in the Figure 3 device, and on the initial height of the reservoirs. The resistance, inductance, and capacitance of each fluidic channel and tank were determined from their dimensions and from the properties of their fluids (the density and viscosity of water were used). The initial head in the tanks of the reference, indicator, and sample streams was 196 Pa, and the initial waste tank head was –586 Pa, all calculated from the initial height of the liquid in each tank. The predicted flow rates were 99 nl/sec for the reference and sample streams, and, owing to the greater resistance of its inlet channel, 97 nl/sec for the indicator stream. Knowledge of these flow-rate parameters is vital to the device designer, as the diffusion extraction efficiencies are directly proportional to the time the different fluids spend in contact with each other.

The diffusion and chemical reaction of the chemical species in the T-Sensor main channel were predicted by a mathematical model based on finite element analysis.¹² The diffusion constants were derived from the Einstein-Stokes relation and the effective molecular diameters. The chemical-reaction model was diffusion-controlled, and the extent of reaction was based on the equilibrium constants of the chemical species. The high Peclet number (the ratio of convective flux to diffusion flux) of the flow in the main channel guarantees that axial convection is dominant over axial diffusion. This allows the algorithm to march downstream sequentially, solving the grid points at each streamwise location, since the field value at each grid point does not depend on its downstream neighbors.

Figure 11. Flow and diffusion modeling results for the T-Sensor in Figure 3 are shown at right. The concentration of the base form of the indicator in the main channel of the T-Sensor indicates the optically observable change as a function of pH. The concentration color chart is shown at left.

Figure 11 displays the results of this fluid modeling. Experimental data obtained with the system verified the modeling results.

All of the disposable cartridges discussed in this article were manufactured by means of a proprietary rapid prototyping process based on the laser cutting of single-layer plastic sheets that are subsequently laminated to form complex three-dimensional flow structures such as the ones displayed in Figures 3–6 and 12. Once the design modeling was complete, a computer-aided design drawing produced from the modeling results was prepared to provide the basis for the microfabrication process.

Figure 12. The Microfluidics Evaluation Kit from Micronics Inc.

This process is now performed, partly manually, at prototype rates of up to 100 cards a day. A scaled-up system planned for mass manufacturing would use reel-to-reel processes to produce microfluidic disposables economically.

Conclusion

Microfluidic lab-on-a-chip systems hold great potential for many laboratory applications, including clinical diagnostics and life sciences research. The research on microfluidic systems described here has yielded a number of devices that appear to be well suited to handle the tough requirements of real clinical samples such as whole blood and other bodily fluids. The devices can be used in stand-alone applications as self-contained passive disposables for qualitative and semiquantitative assays and separation applications, or they can serve as components of sophisticated instrument-based systems. Products based on the technologies outlined above are currently being developed for several applications, including hematology.

References

1. A Kopf-Sill, "Commercializing Lab-on-a-Chip Technology," in Proceedings of Micro Total Analysis Systems 2000, ed. A van den Berg et al. (Dordrecht, Netherlands: Kluwer Academic Publishers, 2000), 233.

2. BH Weigl and P Yager, "Microfluidic Diffusion-Based Separation and Detection," Science 283 (1998): 346–347.

3. LJ Kricka, "Miniaturization of Analytical Systems," Clinical Chemistry 44 (1998): 2008–2014.

4. NH Chiem and DJ Harrison, "Microchip Systems for Immunoassay: An Integrated Immunoreactor with Electrophoretic Separation for Serum Theophylline Determination," Clinical Chemistry 44 (1998): 591–598.

5. P Yager et al., "Applying Microfluidic Chemical Analytical Systems to Imperfect Samples," in Proceedings of Micro Total Analysis Systems '98, ed. DJ Harrison and A van den Berg (Dordrecht, Netherlands: Kluwer Academic Publishers, 1998), 207–212.

6. AE Kamholz et al., "Quantitative Analysis of Molecular Interaction in a Microfluidic Channel: The T-Sensor," Analytical Chemistry 71, part 23 (1999).

7. JP Brody, AE Kamholz, and P Yager, "Prominent Microscopic Effects in Microfabricated Fluidic Analysis Systems," in Proceedings of Micro- and Nanofabricated Electro-Optical Mechanical Systems for Biomedical and Environmental Applications (Seattle: University of Washington, 1997), 103–110.

8. RB Darling et al., "Integration of Microelectrodes with Etched Microchannels for In-Stream Electrochemical Analysis," in Proceedings of Micro Total Analysis Systems '98, ed. DJ Harrison and A van den Berg (Dordrecht, Netherlands: Kluwer Academic Publishers, 1998), 105–108.

9. MR Holl et al., "Optimal Design of a Microfabricated Diffusion-Based Extraction Device," in Proceedings of the 1996 ASME Meeting (Atlanta: ASME, 1996), 189–195.

10. B Weigl et al., "Simultaneous Self-Referencing Analyte Determination in Complex Sample Solutions Using Microfabricated Flow Structures (T-Sensors)," in Proceedings of Micro Total Analysis Systems '98, ed. DJ Harrison and A van den Berg (Dordrecht, Netherlands: Kluwer Academic Publishers, 1998), 81–84.

11. PJA Kenis, RF Ismagilov, and GM Whitesides, "Microfabrication inside Capillaries Using Multiphase Laminar Flow Patterning," Science 285, no. 5424 (1999): 83–85.

12. BH Weigl et al., "Passive Microfluidics—Ultra-Low-Cost Disposable Lab-on-a-Chips," in Proceedings of Micro Total Analysis Systems 2000, ed. A van den Berg et al. (Dordrecht, Netherlands: Kluwer Academic Publishers, 2000), 299.

Bernhard H. Weigl, PhD, is manager of business development and senior scientist at Micronics Inc. (Redmond, WA). Researchers contributing to the work reported in this article include Tom Schulte, Ron Bardell, Anson Hatch, and Andy Kamholz at Micronics, and Paul Yager at the University of Washington (Seattle).