Medical Device Link.

ASSAY DEVELOPMENT

The ongoing biological sciences revolution has generated high hopes for the advent of true personalized and preventive medicine. As the necessary biological tools are being developed at a fast pace, it has become clear that their cost, operation, and manufacture are equally challenging issues that must be addressed before they can be widely adopted in medicine. This is especially true in light of today's skyrocketing medical costs and shrinking insurance coverage.

Figure 1a. State-of-the-art commercial machines for immunoassay blood tests. The Elecsys 2010 by Roche Diagnostics Corp. (Indianapolis), a typical midsize system, is 1.4 m long, weighs 110 kg, costs approximately $100,000, and can generate 85 results per hour.

In diagnostics, decentralized near-patient or point-of-care testing has brought the promise of fast, quantitative results at the bedside or in the clinic, thereby allowing early detection and prevention, decreasing hospital stays, and eliminating transportation and administrative expenses. Although a few such systems have been developed (e.g., blood glucose meters), the enormous potential remains untapped. The vast majority of medical diagnostics are still conducted in clinical labs using large equipment.

This article examines current commercial systems and their emerging alternatives. One technology in particular—high-throughput multiantigen fluorescence immunoassays in polydimethylsiloxane (PDMS) microfluidic chips—could help bring finger-prick blood tests to every physician's office and hospital bedside, in the process revolutionizing protein-based in vitro diagnostics.¹

The criteria for judging these systems include cost, sample size, overall turnaround time, versatility, and portability. Specificity and sensitivity are also critical, although the former is primarily determined by the quality of antibodies while the latter needs only to best the clinically relevant abundances, which are already high for many analytes.

Benchtop Commercial Systems

Figure 1b. State-of-the-art commercial machines for immunoassay blood tests. Modular Analytics, by Roche, is one of the largest inmmunoassay systems. When configured to contain four E 170 immunoassay modules, it can produce 680 results per hour but it weighs 2850 kg and requires 12 kW of power.

Today, large desktop machines conduct the bulk of clinical blood tests and represent the gold standard for diagnostics (see Figures 1a and 1b).

For the Elecsys 1010 and 2010 systems—two common analyzers from Roche Diagnostics Corp. (Indianapolis)—30 µl and 20 µl of sample, respectively, are listed as sufficient for one test of one patient. However, in practice, the amount necessary is 0.5–1.0 ml per test per patient. This is because some volume is lost when serum is prepared from whole blood. In addition, each measurement uses about 250 µl, and each test must be performed in multiple copies to minimize the effect of potential errors. Thus, multiple tests on the same patient might require 1–10 ml of sample. Since such volumes cannot be obtained through a simple finger prick, blood must be drawn from a vein by a trained phlebotomist. Hence, the entire procedure generally requires an extra appointment and transportation, incurring additional costs and delays. In addition, the required volume makes multiple blood tests difficult for pediatric patients, since an infant has only about one liter of blood.

Large desktop machines cost approximately $100,000 and require skilled technicians to maintain and operate them. Consequently, the use of these instruments is relegated to a centralized facility such as a hospital reference laboratory, to which samples are transported after being collected. At these labs, heavy daily loads can generate bottlenecks, resulting in further expenses and delays, while on slower days, unused capacity is wasted. A single poorly prepared serum sample can clog and disable an entire machine. Because the cost of operating such a facility can run $10,000 per sq ft per year, both delays and unused capacity are expensive.

The Elecsys system can measure 50 analytes, pertaining to infectious disease, cardiac condition, tumors, thyroid function, bone, and more, and in an emergency, the processing time for a particular test can be as short as 9 minutes. However, a sample is typically not processed until enough have been collected to run a specific test. As a result, it may take 7–14 days for results to become available. In the meantime, the health of a patient may deteriorate, requiring costly emergency procedures. In waiting for favorable results that warrant release, a patient might also be kept hospitalized.

In addition, the Elecsys 1010 is hardly portable. Although it could be installed in a vehicle, doing so requires an extra power source supplying 110 V and 610 W of power. Also, the glass tubes and mechanics of the robotic sample retrieval make shock resistance low. The Elecsys 2010 has the same limitations and is about 30% larger and heavier.

Literature for the E 170 module by Roche claims that the instrument requires 10–50 µl of sample, can handle 25 tests at a time among the same selection of 50, and outputs 170 results per hour. However, the E 170 also occupies 4 m³, weighs 750 kg, and requires 4 kW to run. The E 170 can be used in Roche's Modular Analytics work area analyzer, which offers combined measurements of substrates, specific proteins, drugs, ions, and more, with throughput of several hundred to several thousand results per hour. If all four components used in the analyzer are E 170 modules, the throughput can reach 680 results per hour. However, in this configuration, the system weighs 2850 kg, requires four times the space, and consumes 12 kW of power.

Table I. (click to enlarge) A comparison of selected commercial immunoassay systems for blood testing.

In this discussion of benchtop and larger systems, there are recurring motifs. Immunoassays can measure about 50 blood protein analytes. Even so, a linear increase in throughput and simultaneously loaded tests requires a roughly linear increase in size, weight, power consumption, and cost (see Table I). The desire for comprehensive integrated test panels is offset by a linear modularity, resulting in very large systems. Even the smallest of these is not portable.

Improvements Offered by Miniaturization

Figure 2. (click to enlarge) The microfluidic miniaturization of blood tests would produce fundamental technical improvements that would lead to organizational changes and savings in modern healthcare.

Reduced instrument costs would free diagnostic testing from the centralized laboratory. Full blood tests could be conducted at the hospital bedside or in every physician's office, eliminating transportation and handling requirements and their concomitant costs and delays. If the needed sample volume is greatly reduced, then a simple finger prick could suffice, eliminating phlebotomist appointments and the danger of hema-tomas, as well as making blood tests far more accessible to pediatric patients. In addition, a small, robust, and shock-resistant system could enable rapid blood testing in field conditions, such as those encountered by law enforcement and military personnel (see Figure 2).

Easier, faster, less-expensive testing would also mean more-frequent testing and, as a result, earlier detection and diagnosis. The importance of early treatment becomes apparent when it is considered that 20% of insured patients— in particular, those with chronic conditions—use 80% of all healthcare costs.

Figure 3. (click to enlarge) The i-State by Abbott Laboratories (Abbott Park, IL) is a handheld system that can measure gasses, ions, glucose, pH, urea, hematocrit, and a cardiac marker. Droplets of blood are inserted into a cartridge, which is plugged into the device. Results are generated within two minutes. However, i-Stat does not yet offer the standard protein panels available for many benchtop machines.

A major step toward miniaturization has been the development of the i-Stat system by Abbott Laboratories (Abbott Park, IL). A cartridge hardwired to a set of tests is filled with a blood sample as small as 20 µl and inserted into the i-Stat handheld device, which produces a readout in as little as 2 minutes (see Figure 3). Measurements are based on ion-selective electrode potentiometry (for ions, pH, and urea), electrical conduction (for hematocrit), or generated current (for glucose). The i-Stat is already widely used in hospitals, especially for quick testing before surgery. However, with the exception of a cardiac marker (cTnl), the device, unlike the large systems discussed earlier, cannot process panels of blood proteins.

The Reflotron, a desktop system by Roche, is also a compact, inexpensive analyzer that generates multiple, fast measurements using small amounts of sample. It uses reflection to detect color changes in test strips and can measure 15 analytes (including lipids, glucose, metabolites, and enzymes). Results are delivered within 3 minutes and 18–30 can be produced each hour. The system occupies about 1 cu ft without its computer and weighs 5.3 kg. However, like the i-Stat, the Reflotron also does not offer the standard broad blood panels of its large robotic counterparts.

A major challenge in high-throughput multiantigen immunoassays is the simultaneous handling of multiple reagents whose function and detection method are identical. This obstacle may explain why the i-Stat uses different methods on the same chip while the rest of the separation is caused by chemical orthogonality among the analytes (e.g., proteins versus salts). This also explains why diffusion and cross-contamination are not issues.

By contrast, systems similar to the Elecsys, which measures blood proteins with immunoassays, rely on the same method (antibodies and optical detection) for all samples and tests. Since the available optical bandwidth in standard fluorescence and chemiluminescence is too small for large-scale multiplexing, the only orthogonality is provided by antigen-antibody recognition. As a result, tests must be isolated to prevent diffusion and to ensure that the signals are correctly attributed. Such isolation, however, entails robotic fluid handling and the use of large, heavy, complex, expensive machines that cannot be straightforwardly and economically miniaturized.

The Microfluidic Approach

If, indeed, fluid handling is the primary reason for the size of current machines, it stands to reason that fluidic miniaturization should allow overall miniaturization. Extensive literature is available on microchannel devices in silicon and glass that use etching and patterning techniques borrowed from the semiconductor industry. However, multiplexed immunoassays also require compartmentalization through active nonleaky microvalves. This technology has remained elusive, although significant steps toward active devices have been made by researchers fabricating microelectromechanical systems.

Another solution has been presented by the advent of PDMS microfluidics. PDMS is soft, transparent, elastomeric, and inexpensive. Since PDMS is also impermeable to water, it is widely used as a sealant (e.g., for caulking bathtubs). In addition, its biological inertness and nontoxicity have led to its use in cosmetic implants and contact lenses.

In PDMS microfluidics, microchannels are made by pouring the mixture of monomer and cross-linker over photoresist mounds on a plain silicon wafer.² After curing, the PDMS retains the reverse features. This molding process is simple, inexpensive, and fast.

Separate layers of PDMS can be aligned on top of one another and bound together to produce a variety of devices. For example, a microfluidic valve can be constructed by crossing two channels traveling in separate, but bound, layers.³ Pressure in one of the channels causes a membrane to pinch the other channel. If the pressure is released in the former, the membrane snaps back and the latter is opened once again. In this respect, the valve is similar to a foot stepping on and off of a garden hose.

The typical width of a channel is 100 µm, while channel height and valve membrane thickness range from 10 to 30 µm. Different valve dimensions produce different reproducible closing pressures.³ Three valves acting on the same channel and activated in the proper sequence produce a peristaltic pump. Channels, valves, and pumps can be arrayed to build complex large-scale integration (LSI) devices, which, when applied to biotechnology, are called BioLSI.⁴ Most recently, the development of microfluidic vials has allowed vertical connectivity between channels in different layers, offering further architectural flexibility and allowing novel autoregulatory devices.

Within its first decade, PDMS microfluidics has evolved from the plain channel to a plethora of specialized components organized by the thousands in LSI devices. The now-mature technology has been successfully used in a number of important applications, including protein crystallization, DNA sequencing, nanoliter polymerase chain reaction, cell sorting and cytometry, nucleic acid extraction and purification, and cell studies.^5–10

Multiplexed Microfluidic Fluorescence Immunoassays

Figure 4. (click to enlarge) Multiplexed microfluidic fluorescence immunoassays. (a) A 100-chamber polydimethylsiloxane chip bound to an epoxide slide allows 5 tests of each 10 samples, with 2 chambers per sample-test combination. The scale pictured is in inches. (b) A functional diagram of an individual coliseum (one chip contains 10 such paths). Control channels and valves are drawn in red, and flow channels in blue. Vertical and horizontal comb-like valve arrays enclose a pair of immunoassay chambers for each of five tests. Valve arrays 1, 2, and 3 pump the sample in a circle along the coliseum (e.g., clockwise for actuation order) with a lap time of 20 X 50, in which case a desktop machine could reach a throughput of 15,000 results per hour while still using less than 1 μl per sample.5,6

Using PDMS microfluidic technology, a high-throughput multiantigen fluorescence immunoassay system quantifying blood analytes at their clinically relevant levels was developed (see Figure 4). This chip multiplexes sandwich immunoassays to allow five simultaneous tests for each of 10 samples. Microfluidic valves direct the pressure-driven reagent flow along a network of channels measuring 10 µm tall.³ The network contains five rows and 20 columns. The columns are connected in pairs to form 10 circular paths, or coliseums. Four-way valving at each intersection of rows and columns forms a capture microchamber, within which the immunoassay stack is built for a particular sample-test combination.

Monoclonal antibody for each of five analytes is fed along its own row and bound to the glass substrate by epoxide chemistry. The unbound excess is flushed with Tris buffer in all rows in parallel while unreacted epoxide moieties are passivated by binding Tris molecules. Similar passivation is conducted along columns to prevent the direct binding of sample proteins. Along the columns, 10 samples are fed to fill their corresponding coliseums. Each sample is trapped in its coliseum and cycled over the capture sites by peristaltic pumping. Unbound material is then flushed out along the columns, and biotinylated polyclonal antibodies are fed in parallel in their corresponding rows. Unbound excess is removed again by parallel buffer feeds. Next, fluorescently tagged streptavidin is fed along all rows, and unbound excess is flushed once more with buffer. Fluorescence detection at each chamber quantifies the captured antigen. This quantity is divided by the processed sample volume to produce the respective analyte concentration for each sample-test combination.

The chip uses 10–100 nl of sample for each test, thus enabling portable devices to be developed that can accommodate common finger-prick blood tests. Simultaneously, the chip uses 300 nl (as low as 0.8 ng) of antibody per sample-test combination. In contrast, the state-of-the-art Elecsys SPSA kit from Roche uses 200 ng of sample per test, or approximately 250 times more.

In principle, nonspecific attachment (e.g., to PDMS) could produce a false-positive signal and thus degrade specificity, but such problems have not been encountered.1 This may be due to the bovine serum albumin treatment, or to the microfluidic architecture, which precludes a polyclonal antibody from coming in contact with any capture sites other than the ones patterned with its corresponding monoclonal antibody.

Future Microfluidic Benchtop Systems

Engineering Considerations. Pneu-matic control of the fluorescence immunoassay device is provided by valve arrays from Lee Co. (Westbrook, CT) and by electronic boxes, which are available from a number of companies. This subsystem can fit in less than one-quarter of a cubic foot and can be operated through a USB port by an external computer. Since capture sites are separated spatially on the chip, the use of one fluorophore type is sufficient. As a result, one bright light-emitting diode can replace the standard mercury lamp in other systems. In addition, the microscope can be discarded for a mechanized ministage that rasters a high-numerical-aperture objective under the chip.

Another possibility is an array of optical fibers aligned in a matrix to match the microfluidic chambers at one end while the other end is fed to an optomechanical or acousto-optic transducer that includes an emission filter. The detector itself could be an inexpensive cooled charge-coupled device camera, a photomultiplier, or an avalanche photodiode.¹ Thus, the engineering of the benchtop system would be relatively straightforward, and its cost should fall well below $10,000.

Test-Matrix Size. The desired size of the test matrix for such a system would depend on the user. For example, an individual physician's office might prefer to conduct a dozen tests with one sample per chip for the convenient screening of individual patients. A larger practice, on the other hand, might prefer a larger chip for the thorough study of multiple patients against a panel, to search for developing conditions that would otherwise be missed by the current symptom-based methodology and limited doctor-patient interaction. A major pharmaceutical company conducting drug tests might want as large a matrix as technically possible, to minimize reagent cost and make maximal use of the available experimental samples. As the cost of experimental drugs approaches $1 million per gram of protein, the competitive advantages of microfluidic immunoassays become clear. In principle, the parallelism of the fundamental technology should allow the building of 50 X 50 chips.⁴

Technical Considerations. If the coliseum geometry is retained, fluidic resistance increases and pumping across larger distances becomes more difficult to control. Therefore, higher pressures would be required to produce the same fluid velocities (an average chip cannot sustain pressures in excess of 20 psi). Finally, if quality control is gated to disallow a single failure, it might be difficult to produce superchips economically. Auspiciously, Fluidigm Corp. (South San Francisco, CA) already offers relatively large fluidic circuits.

Expected Throughput. With the PDMS prototype, taking pictures of each of the 100 chambers manually with a fluorescence microscope is time-consuming. However, in a commercial benchtop system, this task can be improved by the hardware solutions discussed earlier. Software would then integrate the signals in the respective spots, compare these to calibrations, and calculate the results. The system could complete its measurements within 10 minutes of sample loading, which would mean a throughput of up to 15,000 results per hour in a 50 X 50 matrix.

Since most of its fluid-handling hardware could be replaced by small, disposable, mass-produced elastomeric chips, the complete system would be significantly smaller, lighter, and less expensive than most current commercial benchtop machines. In addition, the use of disposable chips would offer a major advantage over current benchtops. Although one badly prepared serum sample might shut down a large machine, a clog in a disposable chip could be remedied by discarding the chip and quickly moving to another.

The versatility of the system would be limited by the availability and specificity of antibodies, but this is also the case with immunoassays in general. Moreover, as new analytes, such as cancer markers, are identified, incorporating them into the test panel would be straightforward. An orthogonal increase in versatility could also be achieved by validating the system for the interro-gation of different types of samples, including cerebrospinal fluid, synovial fluid, vaginal secretions, urine, saliva, and suspended biopsy samples.

Production Costs. Current chip pricing reflects the high costs and potential profits of pharmaceutical research. In the face of the enormous expense of producing experimental proteins, it is still economical to manually produce chips in small numbers. However, to become feasible for the personalized healthcare of the average insured patient, the price of chip production must be reduced by some two orders of magnitude. Fortunately, this goal can—and should—be met by developing processes and building machines that mass-produce microfluidic chips, in much the same fashion as the semiconductor industry.

Future Microfluidic Handheld Systems

Emil P. Kartalove, PhD, is a postdoctoral research associate in the department of biochemistry and molecular biology at the University of Southern California, Keck School of Medicine. He can be reached at kartalov@usc.edu.

There is little question that microfluidic immunoassay systems would be a great improvement over their benchtop counterparts, but is it possible to produce a handheld device based on the same technology? While the chips themselves may be quite small and light, it would be challenging to shrink the electrical macrovalves and optical system that control and read them. A better idea would be to discard optical detection in favor of electrical detection (like that of the i-Stat) while retaining the microfluidic separation that is critical for multiplexed immunoassays. Miniature pneumatic-valve actuation would also have to be developed. However, because such a system would contain no optical elements, and all moving components would be elastomeric membranes, it would be highly shock resistant.

The dream of personalized preventive medicine cannot be realized without ubiquitous and affordable diagnostics. Decentralization and miniaturization are especially critical in lowering the cost of testing. High-throughput multi-antigen microfluidic fluorescence immunoassay chips represent an important step toward fast, inexpensive, portable, versatile systems for protein-based diagnostics.