Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Microfluidics is a burgeoning area of analytical chemistry that will impact many fields, including clinical diagnostics. The ability to miniaturize and expedite chemistry with smaller volumes presents the possibility for testing with expedited turnaround times at lower cost and possibly in a portable and handheld format.
This chapter describes the basic concepts necessary to understand microfluidics at a fundamental level. This includes the methods and materials used for fabrication, both historically and currently, as well as aspects of microfluidic architecture necessary to carry out reactions, chemistry, labeling, and detection. The chapter highlights some of the basic developments associated with the microfluidic manipulation or analysis of diagnostically relevant analytes such as cells, nucleic acid (NA), proteins, and small molecules. It also presents some exemplary applications (e.g., circulating tumor cell capture and pathogen detection) that have paved the path for the development and adoption of microfluidics in clinical chemistry and molecular diagnostics.
The first miniaturized analytical device with some degree of functionality was reported in 1979, when the separation of a simple mixture of low-molecular-weight (MW) compounds was achieved in only a few seconds in a microdevice fabricated in silicon. The significance of this development can be appreciated only by comparison with the state of the art of separation techniques common to the laboratories of that time. The 1970s saw most chemical separations carried out by open-column, paper, and thin-layer chromatographic processes; gas chromatography (GC) was only beginning to mature into a bona fide technology for small molecule analysis. Pressure liquid chromatography began to evolve into a method that would bring faster analysis times with higher resolution, and the late 1970s brought new chemistries amenable to reverse-phase liquid chromatographic separations for improved resolution of more polar compounds, including biological analytes. Electrophoresis was the workhorse for large biomolecule separations, with Laemmli having established denaturing separations of proteins in acrylamide gels and O’Farrell adding the power of a second separation dimension in two-dimensional gel separation of proteins ; the advent of sequencing gels and the use of agarose gels for nucleic acid (NA) separations followed later that decade.
With that historic context, visualize the stark contrast of a paper that reported a GC analyzer fabricated in silicon ( Fig. 27.1 ) using the tools under development for a microelectronics industry that was still in its infancy. The work of Terry et al. provided the first evidence that microfabrication could be used to provide microminiaturized devices for analytical separations. Despite that evidence, the scientific community did not immediately recognize the significance of this achievement nor see this as a glimpse into a future that would be seeded, in no uncertain terms, by Widmer’s group at Central Analytical Research, Ciba-Geigy AG (Basel, Switzerland) approximately a decade later. Although several papers followed this initial work, it was not until the concept of a total analysis system (TAS), first proposed by Manz et al. in 1990, that the idea of miniaturization had a significant impact. The TAS concept was originally motivated by a lack of adequate sensors for detection of specific species from a complex mixture. It was hypothesized in the seminal Manz et al. paper that, by improving the sample treatment steps and thereby increasing the selectivity of the sensor, an ultrasensitive sensor would not be required if interfering chemical compounds were removed. A TAS was proposed that entailed initial sampling, transport of the sample, sample pretreatment steps, and the final detection of the analyte. Miniaturization of all of the processes required for total analysis, a micro-total analysis system (μTAS), would allow development of systems that perform these functions at the site of measurement (i.e., not in a central laboratory). Additional benefits are also derived from decreasing the size of a TAS, including decreased volumes of sample and reagent and the potential for drastically reduced analysis times.
Early μTAS devices were fabricated in glass (borosilicate or soda lime) for two reasons. First, the wet etching of features in glass could directly leverage the tools developed for etching silicon for microelectronics fabrication. Second, modification of the surface, for passivation or adduction of reactive groups could leverage the abundant capillary electrophoresis (CE) and chromatography literature. However, as the complexity of the architecture increased, glass became less attractive as a substrate and substantially less cost-effective. The field expanded to exploit polymeric substrates, primarily because the processes that evolved for injection molding and hot embossing provided a path to cost-effective mass production. Initial work involved pour-and-cure materials such as polydimethylsiloxane (PDMS) and materials that could easily be milled, such as poly(methylmethacrylate) (PMMA). Numerous reviews highlighting the traditional processes used for fabricating microdevices in glass and silicon, namely, masking, photolithography, and wet etching, can be found in the literature. As a result, in this chapter, only a cursory coverage of these processes is provided. It is more pertinent to discuss polymeric microchip fabrication broadly, with a focused coverage on the more popular plastic substrates.
In standard photolithography practices, glass wafers are obtained with photoresist and metal layers already deposited on the wafer ( Fig. 27.2 ). These films function to allow for channel design and for glass protection during the chemical etching step. A chrome metal layer is first deposited onto the glass because of its tight bonding with both glass and silicon. This layer provides extra protection from hydrofluoric (HF) acid during wet chemical etching. On top of the metal is a layer of photoresist that allows for the channel patterning. The traditional mode of application is to spin coat the photoresist on top of the metal layer, although there are alternatives to this approach, such as the constant-volume injection method. As a result of the microscale precision attainable by photolithography, the application of the photoresist is performed under cleanroom conditions using specialized instrumentation.
The transfer of the channel pattern to the photoresist involves using a photomask containing the design. The photomask can either be a film mask or metal mask. For either mask, the microfluidic design can be created using standard design software (e.g., Adobe Illustrator, Adobe Systems, San Jose, CA; AutoCAD, Autodesk, San Rafael, CA). A relatively inexpensive film mask can be printed by a specialized high-resolution printer, allowing for microfluidic channels 40 μm or greater in diameter. Alternatively, metal masks can be used for higher-resolution structures (<20 μm). Then the photomask is placed on top of the photoresist or metalized wafer and exposed to a wavelength of ultraviolet (UV) light that destroys the exposed photoresist as determined by the design of the photomask. With the mask removed (and retained for further use), the wafer is then placed in a developer solution that removes the photoresist where the channels will exist. The exposed chrome is then removed, exposing the glass only in the places where channels are to be etched.
The glass channels are obtained by wet chemical etching with HF acid. Extreme caution when using HF acid is warranted. The etch rate in glass depends on the concentration of HF acid, with faster rates correlating to higher HF acid concentrations. Etch rates of 10 μm/min in borosilicate glass can be achieved with a 49% HF acid etch solution. The use of wet chemical etching results in an isotropically etched channel in which the glass is etched equally in all directions. This type of etching results in a channel architecture that is semihemispheric in shape (see Fig. 27.2 ). After the channels have been etched to the desired width and depth, access holes are drilled in a capping layer, and both glass wafers are thoroughly cleaned, aligned, and brought into intimate contact for thermal bonding. Although grounded in the scientific method, there is an “art” associated with the bonding process at temperatures between 620 and 690 °C for 6 to 8 hours. The use of a programmable furnace is critical to step ramping the temperature (up and down), and this must often be determined empirically for each system. Multiple bonding cycles may be required to get an effectively bonded microdevice, representing one significant drawback to thermal bonding. Other approaches have been developed, such as room temperature bonding techniques, UV-curable adhesives, , and silicate solutions.
If not purchased in ready-to-etch form, the application of a photoresist requires a spin coater and cleanroom conditions, which can add expense. Alternatives to photoresist have been developed, such as patterned PDMS affixed to a glass wafer that is etched by HF acid flowing through PDMS channels. Enzyme nanolithography can also achieve nanometer-sized depressions in a protein surface. With this method, scanning probe microscopy controls a nanopipette, which delivered proteases to a dried bovine serum albumin (BSA) film. , Although wet chemical etching does not require expensive instrumentation, the desire for high aspect ratio structures or for more control over the dimensions of the channel creates a need for other types of etching techniques. For example, high aspect ratio structures were obtained in glass and fused silica wafers by using powder blasting, fast atom beam etching, and deep reactive ion etching (DRIE). All of these methods involve bombarding the wafer surface with particles; however, whereas DRIE uses chemically reactive ions, typically fluoride, to collide and remove surface atoms, fast atom beam etching uses neutral particles.
Because of the expense involved with the fabrication of glass microdevices, there has been a major shift in the field toward the use of plastic materials as microfluidic substrates. Some of the key factors to consider when choosing a polymeric substrate for microchips include surface chemistry, optical clarity, background fluorescence, chemical compatibility, and temperature stability; these decisions are largely driven by the desired application. Glass microfluidic devices make an excellent platform for electrophoretic separations that require a range of electro-osmotically driven flows; the presence of an abundance of silanol groups on the surface facilitates control of the electro-osmotic flow. However, this same surface can be problematic with applications such as polymerase chain reaction (PCR) in which the reactivity of the glass surface can inactivate crucial components (e.g., proteins), thereby inhibiting the desired chemistry. Polyimide is a thermoplastic polymer with a neutral surface charge providing an excellent vessel for PCR, yet it has less than optimal visible light transmittance, making optical detection difficult. Finding easier fabrication methods is a driving force for finding alternatives to glass for microfluidics.
As the list of possible substrates for microfluidic devices grew, so did the innovative ways for simplifying and streamlining fabrication. Polyester-toner microdevices have been made on overhead transparencies using laser printers. The channel geometry is simply printed onto the transparency (where the absence of toner dictates the channels) and placed together with another piece of blank transparency containing only access holes. The sandwich is bonded by lamination, and the toner printing results in channels that are approximately 6 μm deep. These microdevices have been used for electrophoretic separations using a number of different detection methods. Another material is SU-8, a negative photoresist that can also be used as a substrate for microfluidic devices. Fabrication takes advantage of the photoresist properties of SU-8 and allows for multiple layers. The first layer is homogeneously spun onto silicon, followed by a complete crosslinking with UV exposure. A second layer is spun onto the first layer, covered with a photomask, and exposed to UV for channel pattern transfer. The third layer is separately spun onto a piece of glass, completely crosslinked through UV exposure, and then placed on the second layer. A low-temperature baking step, UV exposure of the whole device, and removal of the glass and silicon pieces completes the device. Thermoset polyester (TPE) is an unsaturated polyester that can be used as a substrate. Crosslinking occurs upon addition of a catalyst, and fabrication methods are similar to those PDMS (see next section). Poly(cyclolefin) copolymer, the material used for making compact discs, has also been used for microfluidic devices. Hot embossing is used to make these chips, and a silicon master dictates channel design. A cover plate is then bonded to enclose the device, using heat and pressure to generate a complete seal.
PDMA is one of the most popular polymeric materials used as a microdevice substrate. Fabrication is simple, involving the combination of elastomeric base and curing agent. PDMS has methyl groups on the surface, imparting a hydrophobic nature, and is flexible and optically transparent. The flexibility of PDMS is useful for valving, and it can be both reversibly and irreversibly sealed to a variety of surfaces. A number of different PDMS fabrication methods have been established (see Sun and Kwok for a review), with replica molding and rapid prototyping perhaps the most popular. Rapid prototyping involves the use of a master, which once made allows for fast and easy replication of a particular design. The master is created by using a negative photoresist, such as SU-8, spun onto a silicon or glass wafer. The microfluidic channel design is imparted onto the photoresist using photolithography, and the wafer is subsequently developed. The negative photoresist retains the structures exposed to light, and the masked areas are removed by a developer solution. This leaves elevated structures where channels are desired in the final PDMS device. The height of the elevation over the glass dictates the depth of the PDMS channels. Although photolithography is needed for the fabrication of the PDMS master, after it has been created, multiple devices can be easily made.
PDMA as a solid substrate is produced when an elastomeric base and curing agent are mixed in the appropriate ratios. The mixture is poured over the master and allowed to cure; this can be accomplished overnight at room temperature or expedited using a heating process. After curing, the PDMS can then be peeled from the master and bonded with the substrate of choice. The bonding method will depend on the substrate used as the base or capping layers; PDMS-PDMS devices can be either reversibly or irreversibly bonded via a number of different methods, one being the activation of the surface by plasma oxidation. , Another method alters the elastomer base to curing agent ratios, using 20:1 for the fluidic layer and 5:1 for the cover plate. The device is assembled and placed in an oven, at which point the two layers “fuse” because of the distribution of the excess reagents between the two layers.
PMMA has become a popular substrate for microfluidic devices. PMMA is considered to be a thermoset polymer, which softens into a viscous liquid when heated above its glass transition temperature. Multiple methods of fabrication are available for PMMA, and the channel surface is easily modified to adapt to a particular application (see reviews of polymer fabrication techniques). , Hot embossing , consists of heating the polymer above the glass transition temperature and stamping the substrate with the desired microfluidic pattern. To make PMMA devices using this technique, a metal or silicon stamp is brought into contact with the PMMA, and constant force and elevated temperatures partially melt the substrate. The embossing process can be completed in less than 10 minutes, and after the apparatus has cooled, the stamp and substrate can be carefully separated. This same process can also be done at room temperature using PMMA as a substrate, yielding channel dimensions that vary by less than 2%. Injection molding, another method that can be used for fabrication of PMMA devices, also incorporates the use of a metal or a silicon master. This involves heating the PMMA pellets to temperatures that result in the formation of a viscous liquid. The PMMA is then pushed into the mold by applying pressure and allowed to cool, and it is removed by “lift-off” from the mold. Both of these methods result in devices with micron-sized channels and can be used for mass production of devices with the same microfluidic architecture. For applications that require submicron features, x-ray lithography is capable of making these intricate structures. PMMA poststructures can function as a solid phase for DNA chromatography ( Fig. 27.3 ). Solid phases for DNA extraction traditionally are created with silica beads, sol-gels, or silica bead–sol-gel hybrids. DNA is loaded in a solution that forces DNA binding to the solid phase. It is washed, and then the purified DNA is eluted off the solid phase and captured. Although traditional forms of extraction involve packing the solid phase into a capillary or tube that solution can pass through, phases can also be microfabricated during the microchip fabrication process. This is done using either a direct write process or the use of a mask, typically fabricated in a thick gold layer that functions to absorb x-rays. Where the x-rays are incident upon the PMMA, the polymer is degraded, and it can be removed with a developer solution to solubilize the degraded products. The length of exposure, along with the energy used, dictates the depth of the channels. Laser ablation is also used to make submicron features in PMMA devices. This involves bombarding the surface with a pulsed UV source to ablate (vaporize) the polymer structure. With these last two techniques, high aspect ratio features can be obtained; however, neither of these methods is ideal for mass production.
Miniaturized electrophoresis was the first and most dominant embodiment of a μTAS, with the driving force for the miniaturization related to an enhancement in its analytical performance rather than a simple reduction in size. Critical in the development of microfluidic analytical technology was the use of microchips as an analytical platform for electrophoretic separations. The components required for microchip-based electrophoresis (ME) include a detection system, a power supply, and a computer with programmable software for controlling application of voltage, data collection, and the conditioning of the raw data. Although a number of different detection methods can be used with microchips, fluorescence is most widely used. Power supplies with multiple outputs are needed for electrokinetic injection of sample and application of the desired electric field. For chip-based electro-driven separations, the cross-t design originally proposed by Verheggen et al. for electrokinetic injection of sample is the most popular. For this design, a minimum of four voltage outputs is required to control the sample injection and separation. As the complexity of the chip design increases, more outputs may be necessary, particularly if electrokinetic pumping of reagents and buffers is involved. The effective electrokinetic mobilization of solutions through the microchip architecture is limited to reagents that are low in conductivity and organic solvent concentration; this limits the use of many clinical samples and thus may necessitate the implementation of pressure-driven fluidic control.
For injection, both pressure and electric field-driven injection modes are used as they are in CE. The pressure-driven injection mode commonly used in CE, which avoids bias associated with analytes that have a wide range of electrophoretic mobilities, is less popular with chip-based injection. Integrated diaphragm pumps on a hybrid PDMS–glass microchip can perform pressure sample injection for electrophoretic separations. The same type of electrokinetic bias associated with CE can be avoided on chip with pressure injection using injected sample volumes as low as 500 nL while maintaining sample composition. This approach has been used for integrated microchip detection of infectious pathogens in blood. However, the injection mode that dominates for most applications in ME is electrokinetic injection (driven by electrophoretic mobility, or electro-osmotic effects, or both).
The more popular electrokinetic sample injection is dominated by two modes—pinched and gated—and both are possible using the standard cross-t design chips. With pinched injection, a small sample plug is electrokinetically mobilized into the injection cross-t via a voltage applied between the sample and sample waste reservoir and orthogonal to separation channel. After the cross-t is adequately populated with sample, injection is accomplished by reconfiguring the primary electric field between the inlet buffer and outlet buffer reservoirs. Injection “bleed” of sample into the separation channel is minimized by applying a small voltage to the sample and sample waste reservoirs. This contrasts with gated injection, which allows for larger sample plugs to be injected into the separation channel.
When performing chip-based separations, the separation channel in a cross-t design provides the counterpart of the capillary in CE but with a substantially truncated separation length ( Fig. 27.4 ). Many of the parameters governing electrophoretic separation in capillaries apply to chip-based separations. For example, similar to capillaries, one of the benefits of glass-based devices is their inherent ability to dissipate the Joule heat (because glass is an effective conductor of heat) generated by the high electric fields needed to facilitate rapid separation. However, the same type of Ohms law plot (applied voltage vs. current) used to test buffer compatibility with high electric field applications should be carried out here as well. If performing separations requiring electro-osmotic flow, the rigorous cleaning and hydration methods suggested for fused silica capillaries is required to ensure proper surface regeneration. The entire cross-t channel architecture of the device is filled, including the reservoirs, to ensure good electrical contact between the platinum electrodes and the solution. The holes drilled into the microfluidic device to function as reservoirs (i.e., buffer and sample microvials) may not be large enough to hold a sufficient volume of buffer to avoid buffer depletion. The user must remain cognizant of this and, even with expanded reservoirs generated by commercial “nanoports” or homemade trimmed pipette tips, the replenishment of buffer may be needed.
For sample injection, separation, and detection in a chip of simple design (see Fig. 27.4 ), sample is placed in the reservoir, and a voltage is applied between sample inlet (SI) to sample outlet (SO) to inject the sample into the separation channel in a manner consistent with the injection mode chosen. Separation is initiated by application of voltage to buffer outlet (BO) and buffer inlet (BI), where electrophoresis of the sample analytes down the separation channel ensues. A point along the separation channel, most commonly at a location close to BO, is used for detection. The most frequent detection mode is optical, with laser-induced fluorescence (LIF) popular because of its sensitivity. The detector system (e.g., a laser and photomultiplier tube [PMT]) is positioned at a point in the separation channel commensurate with providing the necessary effective separation length (L eff ) for the separation; the sample analytes passing that detection point are excited and the resultant fluorescence collected. Similar electric fields ( E ; volts/cm) are applied in both CE and ME; however, reduction of the L eff is more easily achieved in microchips. The reservoirs and electrodes required to apply voltage to capillaries limits the minimal L eff to tens of centimeters; microchips, in contrast, can easily complete separations in a few centimeters. Interestingly, the U-shaped channel in microchips has little effect on the resolution of separations under high electric fields. Similarly, the plug profile typically associated with capillaries and known to play a role in high-efficiency separations is not adversely affected by hemisphere-shaped channels.
A detection system is required to measure the number and intensity of analytes separated in a channel as they pass the detection point. With microchip analysis originally rooted in separations, a variety of different detection options evolved and included optical detection methods that are, for the most part, similar to those used in CE (primarily fluorescence). However, electrochemical, conductivity, and mass spectrometric approaches have also been developed. For optical detection, the most prevalent method in CE and high-performance liquid chromatography (HPLC), UV-visible spectroscopy (UV-VIS) detection, is not as common with chip-based electrophoresis. This is primarily because of challenges with the short optical path length (channel depth) in miniaturized systems and the difficulties in coupling the light into and out of these channels. UV-VIS detection in analytical microsystems is making progress ; however, no universally successful method akin to that of CE has been reported. A number of other optical detection schemes for microchips have evolved with varied success and applicability including infrared, nuclear magnetic resonance (NMR) spectroscopy, surface plasmon resonance, thermal lens microscopy, and chemiluminescence and electrochemiluminescence.
The most prevalent detection system used for chip-based separations is LIF. It is both extremely sensitive and well-suited to detection in small volumes, hence its widespread popularity in CE separations. A number of reviews detail the components required for LIF, alignment of the optical system, and key elements (analyte velocity, interrogation volume, and optimized signal-to-noise ratio as a function of excitation intensity) necessary to achieve sensitive detection. , The components used in an LIF system are a light detection module, such as a PMT or charge-coupled device (CCD) camera, a laser source, optics for alignment of both excitation and emission, an excitation source, and a computer for data collection. LabVIEW software (National Instruments, Austin, TX) is commonly used for application-specific programming to provide data collection and control of the system. , A typical method for performing LIF detection on a microchip is to bring the excitation light from a laser or light-emitting diode (LED) into the microchannel orthogonal to the device plane, which is relatively simple to align, and to then focus the beam into the channel using a microscope objective or some combination of achromatic lenses. LIF emission is then collected with a separate objective or lens combination, spatially and spectrally filtered, and recorded on a photodetector. Epifluorescent detection systems were popular in the early CE literature. In an epifluorescent system, focusing and collection optics are one and the same, and the excitation and emission wavelengths are separated with a dichroic beam splitter. However, variations in LIF setups exist. For example, in one system, the excitation is brought into the microchannel from an angle. Two reviews provide detailed descriptions of the fundamentals and practice of LIF detection in microanalytical systems. ,
Chip-based detection systems can involve either multipoint detection in a single channel or multichannel detection; multichannel detection is key to high-throughput analysis. For multipoint detection, waveguides can be incorporated into the microchip design for detection at multiple points along the same channel and applied to microchip isoelectric focusing (IEF). Whole-channel imaging has been demonstrated during the IEF separation of proteins using a CCD array using organic LEDs as the excitation source. Alternatively, a scanning fluorescence detector and a single PMT have been used to scan the entire separation channel. In terms of multichannel scanning, an acousto-optic deflector (AOD) can change the diffraction angle of an incident laser beam to address an eight-channel device designed for parallel DNA separations. However, the multichannel microdevices designed by Mathies and coworkers are even more significant for high-throughput analysis. Ninety-six channels can be multiplexed on a radial CE microchannel plate, demonstrating that a microchip platform can compete with 96 capillary arrays that evolved for CE. Radial microdevices of this type for DNA analysis containing as many as 384 channels have been demonstrated. More recently, this group has reviewed the types of applications possible with these high-throughput chip-based systems.
Multicolor LIF detection is not essential for all applications but has been very useful for some key genetic analysis techniques. Although expanding the number of photodetectors needed to sample at multiple emission wavelengths is a simple approach used for multicolor DNA sequencing and genotyping on microchips, diffraction gratings can also disperse emissions onto a CCD array for four-color sequencing. Furthermore, a transmission imaging spectrograph with a wide imaging area has been used to disperse LIF emissions from a two-color genotyping experiment run in 48 parallel channels onto a CCD array. More recently, a system was developed to complement the microchip platform by featuring a single acousto-optic tunable filter (AOTF) and a single PMT. The filter behaves similarly to the previously described AOD, except that when an acoustic wave traveling through the crystal at a specific frequency interacts with the light, only select wavelength bands are diffracted by the crystal and collected by the PMT.
Become a Clinical Tree membership for Full access and enjoy Unlimited articles
If you are a member. Log in here