Biomedical Microdevices

, Volume 12, Issue 2, pp 333–343

Glass-composite prototyping for flow PCR with in situ DNA analysis


  • Ilija Pješčić
    • Institute for MicromanufacturingLouisiana Tech University
  • Collin Tranter
    • Institute for MicromanufacturingLouisiana Tech University
  • Patrick L. Hindmarsh
    • School of Biological SciencesLouisiana Tech University
    • Institute for MicromanufacturingLouisiana Tech University

DOI: 10.1007/s10544-009-9389-2

Cite this article as:
Pješčić, I., Tranter, C., Hindmarsh, P.L. et al. Biomed Microdevices (2010) 12: 333. doi:10.1007/s10544-009-9389-2


In this article, low cost microfluidic devices have been used for simultaneous amplification and analysis of DNA. Temperature gradient flow PCR was performed, during which the unique fluorescence signature of the amplifying product was determined. The devices were fabricated using xurography, a fast and highly flexible prototype manufacturing method. Each complete iterative design cycle, from concept to prototype, was completed in less than 1 h. The resulting devices were of a 96% glass composition, thereby possessing a high thermal stability during continuous-flow PCR. Volumetric flow rates up to 4 µl/min induced no measurable change in the temperature distribution within the microchannel. By incorporating a preliminary channel passivation protocol, even the first microliters through the system exhibited a high amplification efficiency, thereby demonstrating the biocompatibility of this fabrication technique for DNA amplification microfluidics. The serpentine microchannel induced 23 temperature gradient cycles in 15 min at a 2 µl/min flow rate. Fluorescent images of the device were acquired while and/or after the PCR mixture filled the microchannel. Because of the relatively high initial concentration of the phage DNA template (ΦX174), images taken after 10 min (less than 15 PCR cycles) could be used to positively identify the PCR product. A single fluorescent image of a full device provided the amplification curve for the entire reaction as well as multiple high resolution melting curves of the amplifying sample. In addition, the signal-to-noise ratio associated with the spatial fluorescence was characterized as a function of spatial redundancy and acquisition time.


MicrofluidicsContinuous-flow PCRDNA melting analysisXurographyThermal gradientRapid prototyping

1 Introduction

Continuous-flow PCR (cf-PCR) is the microfluidic DNA amplification technique that reverts the polymerase chain reaction (PCR) into a steady-state process (Kopp et al. 1998). By maintaining the device at a single thermal and flow profile over time, simpler controls and faster analysis times result (Crews et al. 2008b). In addition, massively serial DNA testing can be achieved with no cross-contamination between successively amplified samples (Schaerli et al. 2008). Two significant limitations to the versatility of cf-PCR have recently been resolved in the literature. #1—Since the number of thermal cycles is designed into the microfluidic structure, it had not been possible to select the number of PCR cycles for a given sample prior to its subsequent analysis. Obeid et al. responded by designing a cf-PCR device with multiple outlets spaced at intervals along the microfluidic channel, thereby allowing for the PCR product to be removed for analysis after 20, 25, 30, 35, or 40 amplification cycles (Obeid et al. 2003). #2—Another challenge associated with cf-PCR, as well as other micro-PCR technologies, involves the integration of analysis subsystems for characterization of the PCR product. Nakayama et al. demonstrated an in situ fluorescence technique to confirm that amplification was occurring during cf-PCR (Nakayama et al. 2006), although product verification was not attempted. In the past, product verification could only be achieved through a post-PCR analysis. Recently, however, Crews et al. showed that the DNA target can be analyzed while it is amplifying, rather than afterwards (Crews et al. 2008c). This was achieved by performing cf-PCR within a stable thermal gradient, during which the florescence of the PCR mixture was imaged. The cyclic melting of the amplifying DNA was characterized, and used to identify the DNA product. This spatial melting analysis is analogous to a time-domain DNA melting analysis (Zhou et al. 2004), and has recently been shown capable of even single nucleotide polymorphism (SNP) genotyping (Crews et al. 2009).

The specific fabrication techniques used for cf-PCR micro-devices greatly determine their expense and complexity. Glass (Crews et al. 2008b; Kopp et al. 1998; Obeid et al. 2003) and glass/silicon (Schneegass et al. 2001) microfluidic devices have been made using conventional micromanufacturing methods. Such devices are chemically robust and posses optimum thermal characteristics for flow PCR. However, even low production volumes of these devices usually require a significant investment of time and financial resources, as well as the support of a dedicated laboratory infrastructure. Cf-PCR devices have also been fabricated from polymers (Becker and Gartner 2008), such as polycarbonate(PC) (Hashimoto et al. 2004), polydimethylsiloxane(PDMS) (Nakayama et al. 2006), and polymethylmethacrylate (PMMA) (Schaerli et al. 2008). Although polymers are more susceptible to flow-induced thermal variations (Crews et al. 2008a; Hashimoto et al. 2004), the transition of cf-PCR microfluidics to these materials has resulted in a reduction in long-term fabrication costs. However, there is typically a significant initial investment to create the original master, mold, or cast required for polymer fabrication. Therefore, when considering iterative device design and development, the conventional glass, silicon, and polymer fabrication techniques are equally impractical.

A recently developed rapid prototyping technique, xurography, is characterized by a low startup cost in addition to a reduction in the long-term investment of time and resources (Bartholomeusz et al. 2005). Greer et al. estimated that Xurographic microfluidic devices can be fabricated for less than $0.40 USD for materials or for less than $7.00 USD for labor and materials (Greer et al. 2007). In addition, the initial investment for the xurography equipment is less than $2000 USD, and no part of the fabrication protocol requires exposure to hazardous chemicals (HF, HNO3, etc.) or cleanroom processing. This technique involves the patterning of double-sided tapes with a precision-guided cutting blade, after which the thin patterned film is sandwiched between blanks of a thicker substrate material (Sundberg et al. 2007). This rapid prototyping technique has been used to fabricate microfluidic devices capable of time-domain DNA melting analysis (Greer et al. 2007; Sundberg et al. 2007) and spatial DNA melting analysis (Crews et al. 2009). While short-term contact with the exposed adhesive layers does not interfere with DNA melting analysis, Erill et al. found that the adhesive in acrylic tape “induces an acute inhibition of PCR” (Erill et al. 2003).

This current article describes the manufacture and use of cf-PCR microdevices fabricated with the xurography technique. Thin films with a silicone adhesive were used to enhance PCR biocompatibility, and glass substrates were used to enhance the thermal performance of the prototype devices.

2 Experimental setup

The experimental system consisted of a microfluidic chip, a syringe and syringe pump (KDS100, KD Scientific, MA, USA), a heating subsystem, and an optical subsystem. A computer was used to store and analyze the images acquired by the optical subsystem. The microfluidic chip had a serpentine channel through which a PCR mixture was pumped. A stable temperature gradient was generated across the chip, such that the flowing fluid would experience the cyclic heating and cooling for PCR to occur. A syringe containing the PCR sample was attached through 1/16” tubing (Upchurch scientific, WA, USA) to the chip inlet. The chip outlet was connected to a waste reservoir. Periodically during the PCR, the fluorescence of the intercalating dye (LC Green Plus, Idaho Technology, UT, USA) in the PCR mixture was imaged. Both the amplification curve and characteristic melting curves of the DNA were obtained from each image.

2.1 Microfluidic device design and fabrication

Microfluidic devices were comprised of a patterned adhesive tape between two blank substrates (see Fig. 1(a)), along with the appropriate fluidic interfaces. The choice of substrate material was based primarily on performance and cost. Since the extent of undesired temperature shift within microfluidic flow channels is inversely proportional to the thermal conductivity (k) of the bulk material (Crews et al. 2008a), glass (k = 1.1 W/mK) microscope slides rather than polymer (k = 0.1−0.2 W/mK) substrates were used. Aminosilane-coated slides (S4615, Sigma-Aldrich, MO, USA) were used rather than plain glass slides, in order to immobilize bovine serum albumen (BSA) and thereby reduce adsorption of the PCR reagents onto the microchannel walls. A 100 µm thick, double-sided polyimide tape (PPTDE 1 1/2,, CA, USA) was patterned using a sign cutter (CE 5000-40-CRP, Graphtec, Yokohama, Japan), using a previously documented technique (Crews et al. 2009). This tape was selected in part for its high temperature rating and for its silicone adhesive (thereby avoiding acrylic). Prior to the assembly of the composite device, access holes were mechanically drilled (850-010C, NTI, Kahla, Germany) through the glass slides, over which improvised fluidic interfaces were attached. These cylindrical ports (diameter of ∼1 cm) were punched (6122A26, McMaster-Carr, IL, USA) out of a 8 mm thick sheet of PDMS (Sylgard 184, Dow Corning, MI, USA). A 1.5 mm hole was then punched (33-31A, Miltex, PA, USA) in the port (see Fig. 1) to create a compression fit interface for the tubing. The PDMS ports were then bonded to the glass over the pre-drilled hole positions. This was done by exposing both mating surfaces to ionized air for 2 min at a pressure of 250 mTorr (Harrick Plazma, NY, USA) and then positioning and pressing them together.1 To prevent the plasma treatment from damaging the aminosilane coating, polyimide tape was used to temporarily mask the opposite side of the glass, where the microchannels would be located. All cleaning of the glass pieces during the assembly process was done by rinsing with a 1% solution of detergent (Alconox, NY, USA), then distilled water, and then manually blowing the material dry with compressed nitrogen.
Fig. 1

By sandwiching a patterned tape between microscope slides, disposable glass-composite cf-PCR devices were fabricated. (a) shows an exploded diagram of the chip assembly; (b) shows the dimensions of the repeating feature of the serpentine microfluidic channel in the final design; (c) shows the average temperature profile experienced by the PCR mixture when assuming a flow rate of 2 µl/min and a linear temperature gradient from 60–90°C. Under these conditions, each thermal cycle is achieved in 26 s. The solid section of the curve in (c) corresponds to the repeating feature shown in (b), and the dotted section shows the continuation of the cyclic temperature ramping

The thermal heating and cooling cycles were performed as the molecules in the fluid moved between the hottest periphery of the microfluidic channel and its coolest periphery. The product extension by DNA polymerase, which occurred at low to intermediate heating temperatures, was considered the only time-dependent aspect of PCR kinetics (Wittwer and Hermann 1999). While this rate of reaction restricted the maximum heating rate of the PCR mixture, only physical limitations dictated the maximum cooling ramp rate from the denaturing to the annealing temperature. For this reason, wide heating sections (decreasing mixture velocity) and narrow cooling sections (increasing mixture velocity) were characteristic of every tested design. The channel geometries for the iterative design and fabrication were created in Adobe Illustrator (Adobe Systems, CA, USA). At least one microfluidic device of each design was fabricated and tested. Several criteria were used to optimize the design, such as the propensity to form or trap air bubbles, ease of fabrication, the maximum number of channels that would fit on a device, the ramping ratio between heating and cooling sections, and the quality as well as quantity (i.e. spatial redundancy) of fluorescent data.

2.2 Temperature control subsystem

The temperature control subsystem was designed to generate a stable, controllable temperature gradient within the device that would be virtually linear in the vicinity of the microchannel. The uniformity of the thermal gradient would ensure a reliable thermocycling protocol for the PCR as well as a precise DNA melting analysis of the product during its amplification. The thermal subsystem was comprised of a heating platform and a closed-loop temperature controller. A 3D model of the heating platform, along with further assembly and circuitry details about the temperature control subsystem are provided as “Supplementary material”. An infrared (IR) camera (A320, FLIR, OR, USA) with a 320 × 240 pixel array was used during the development of the temperature control subsystem to characterize the spatial temperature distribution across the glass substrate. Prior to using the IR camera, manufacturer’s instructions were followed to calculate the emissivity of the glass device.

2.3 Optical subsystem

The optical subsystem consisted of a diffuse LED light source and a CCD camera, designed to excite and detect the spatial emission of the fluorescent dye that would bind to the DNA as it was amplifying. LC Green Plus absorbs light between 440–470 nm and, when double-stranded DNA is present, strongly emits light between 470–520 nm. Therefore, the LED source (HPLS-Dragon, LightSpeed Technologies, CA, USA) was filtered to a 50 nm wide band between 425–475 nm (HQ450/50x, Chroma, VT, USA).12-bit images of the microfluidic device were acquired with the 0.3 megapixel monochrome camera (LU075, Lumenera, ON, Canada) fitted with a 50 mm macro lens (Canon EF 50 mm f/2.5, Canon, Tokyo, Japan) that was filtered to block out all wavelengths below 485 nm (HQ485LP, Chroma, VT, USA). The field of view for this camera covered approximately ten PCR cycles, with a resolution on the order of 35 µm per pixel. When examination of more cycles was desired, the heating platform and microfluidic device were shifted beneath the stationary optics, so that multiple regions of the chip could be imaged in quick succession. These images were merged together using an automated function in Adobe Photoshop (Adobe Systems, CA, USA), resulting in a single image of the entire channel length.

2.4 Image analysis

A software program was used to obtain DNA amplification and melting data from the fluorescent images. These in-house graphical user interfaces (GUIs) were developed in MATLAB (The MathWorks, MA, USA), based partially on algorithms that have been previously described (Crews et al. 2008c). The program allowed for the user to interactively orient each image and indicate the regions of interest for the analysis.2 Fluorescence was measured as a function of cycle number and of temperature in order to generate an amplification curve and one or more DNA melting curves for each image.

In the amplification curve GUI, it was important to minimize the effect of any spatial variations in the excitation light. Therefore, the background fluorescence in each cycle was subtracted from the measured signal before the amplification curve was obtained. The fluorescence intensity was averaged within a 10 × 10 pixel region of interest within each section to obtain these values. The melting analysis GUI allowed for the user to select any PCR cycle or multiple cycles in which to examine the amplifying DNA. The analysis algorithm took advantage of the spatial redundancy which was inherent in the 2-D fluorescence distribution. In channel sections that were oriented parallel to the 1-D temperature gradient, it was assumed that adjacent pixel lines would exhibit virtually identical fluorescence, differing only by the effects of the optical and electrical noise in the system. By grouping these parallel data lines together for combined analysis, it was hypothesized that the signal-to-noise ratio (SNR) of the data would increase. Therefore, the GUI allowed the user to select the pixel width of the data subset for the spatial DNA melting curve analysis.

2.5 PCR reagents

An identical PCR mixture was used for all of the experiments conducted as part of this work. Unless otherwise noted, all reagents were purchased from Sigma-Aldrich (MO, USA). The mixture contained 108 copies/µl of a viral phage DNA template (ΦX174, New England Biolabs, MA, USA), 0.5 µM of each of the forward and reverse primers (Integrated DNA Technologies, IA, USA), 200 µM of each deoxynucleotide triphosphate (dNTP), 0.04 U/µl of KlenTaq1 polymerase (AB Peptides, MO, USA), 2 mM MgCl2, 2.5 mg/ml BSA, and 1X LCGreen Plus in a 30 mM Tris (pH 8.3) buffer. One of two primer sets3 was used at a time to target either a 110-bp or a 181-bp segment of the template DNA. The concentration of the DNA template used in these tests was approximately 1,000 times higher than is used in clinical applications with human genomic DNA (Crews et al. 2008c). This was done to reduce the number of cycles by ten that would result in a measurable fluorescence signal (10 cycles∼210 increase). Positive and negative controls for each batch of PCR mixture were amplified and analyzed on a commercial system (LS-32, Idaho Technology, UT, USA). The amplification protocol for the LS-32 consisted of a 1 min initial denaturation at 95°C, followed by 30 cycles of 95°C for 1 s, 60°C for 1 s, and 75°C for 3 s. All temperature ramping on the LS-32 was at a rate of 5°C/s. At the conclusion of the PCR, a high resolution melting analysis of each amplified sample was performed serially, by monitoring the fluorescence during a steady ramp of 0.3°C/s from 60°C to 90°C.

Previous work indicated that reactants in the mixture can adhere to the glass walls of the cf-PCR channels, thereby inhibiting initial amplification (Crews et al. 2008c). To reduce this effect in the current system, the microchannel was passivated prior to its first use with a solution identical to the PCR reaction mixture, minus the nucleic acids and Taq polymerase. Forty µl of this passivation mixture was passed through the device at a flow rate of 2 µl/min, and PCR testing was initiated shortly after the microchannel was emptied.

2.6 Microfluidic PCR and analysis

To prepare each device for testing, a microfluidic chip was placed in the heating platform and the heating protocol was initiated. While the gradient was equilibrating, the passivation mixture was passed through the channel and the PCR mixture was loaded into a syringe. At the conclusion of the pre-treatment, the sample was injected into the channel at a constant flow rate of 2 µl/min. The device was periodically imaged while the microchannel was filling. Upon filling, all thermal and flow conditions were maintained. Fluorescence images continued to be taken at intervals, in order to observe any change in the behavior of the system over time. This experiment was repeated multiple times, with multiple chips, and with each of the target-specific primer sets.

Experiments were also conducted to evaluate the thermal stability of the temperature distribution within the glass-polymer-glass microfluidic devices. The melting of the PCR product was used as the metric, such that a shifting or a distortion of the observable fluorescence transition would be indicative of flow-induced temperature drift. Volumetric flow rates up to 4 µl/min were examined. For these tests, a microfluidic device was filled with pre-amplified PCR product of 110-bp in size. When the chip reached thermal equilibrium, the flow rate was set to a value, and 60 s (± 5 s) elapsed before the device was imaged and the flow rate was changed. This was repeated for all of the flow rates of interest.

2.7 Spatial data quality

The effect of spatial redundancy and of camera exposure time on the data quality obtained from the CCD images was investigated by measuring the SNR of the spatial DNA melting curves. These calculations were performed by importing the GUI-generated melting curves into an analysis software (MeltingWizard, University of Utah, UT, USA). A representative amplification of the 110-bp plasmid target was examined for this evaluation. With a full chip, under steady-state flow and thermal conditions, three images of the fluorescence were taken with exposure times of 125 ms, 250 ms, and 1,000 ms. The MATLAB GUI was used to perform spatial melting analyses on each of these images at five arbitrary positions within the 22nd PCR cycle. Line clusters from one to 13 pixels wide were evaluated, and the corresponding SNRs of these curves were examined as a function of exposure time and pixel line width.

3 Results and discussion

3.1 Iterative device design and fabrication

Microfluidic chips were fabricated according to the aforementioned protocol. The final microchannel design, as shown in Fig. 1, contained 23 PCR cycles. During the design development process, device variations were quickly applied and tested. The fabrication of each iterative design was achieved in less than one person-hour, beginning with the initial concept idea and ending with an experiment-ready prototype. The fabrication yield with this technique was approximately 80% when all geometric features were at least 200 µm in size. The occasional unusable prototypes were characterized by wrinkles in the tape leading to leaking, shifting of the serpentine pattern in the tape resulting in channel blockage, and/or leaking around the PDMS ports. Mishandling of the components and inadequate cleaning during fabrication have been identified as the leading causes of these failures.

Figure 1(b) shows the periodic geometry that resulted from the iterative design and testing process. The channel sections where sample cooling would occur were designed to be as narrow as possible. Channel widths below 150 µm, however, proved challenging to fabricate consistently, due to the difficulty associated with manually removing the material on the interior of the cut channel outline. In the final design, the widths for the cooling channels were 200 µm to increase the fabrication yield. In addition, abrupt width changes resulted in more frequent formation of bubbles, and were avoided in the later design iterations. To maximize the extension time for a given volumetric flow rate, gradually increasing widths, up to a maximum of 1.0 mm in the middle of the chip, were selected for the final device design. The thermocycling protocol resulting from this optimized design is shown in Fig. 1(c), which represents the analytical solution derived from the local average fluid velocity. A volumetric flow rate of 2 µl/min was assumed, as well as a linear temperature gradient within the channel from 60°C to 90°C. The total time for each thermal cycle was calculated to be 26 s. Within this cycle, the PCR constituents would be within an acceptable temperature range for polymerase extension (70–80°C) for roughly 9 s, which is more than adequate for full extension of small amplicon. In the optimized design, the two final heating channel sections were a particularly wide 1.5 mm. During the analyses of the fluorescence images, these two channels were the ones most commonly used for the spatial melting analysis, and therefore designed for maximum spatial data redundancy.

The resulting microfluidic devices had a footprint of approximately 75 mm × 25 mm, and a total thickness of 2.3 mm. Excluding the PDMS ports, whose thermal properties did not affect the temperature distribution in the proximity of the microfluidic channel (see Fig. 2), the chip was approximately 96% glass when considering the bulk materials. The walls of the microfluidic channel itself were between 80–90% glass, depending on the specific geometry used.
Fig. 2

By placing a microfluidic device in the heating platform, a controlled 1-D temperature gradient develops. (a) shows an IR image of a heated device, taken from above. The dotted rectangle indicates the location of the cf-PCR microchannel within the device, which corresponds to a 120 × 24 pixel array in the thermal image. (b) shows the average temperatures at the 24 vertical positions. The error bars in the graph indicate the maximum variation of the 120 horizontal temperature values for each vertical position

Use of the aminosilane coated microscope slides increased the material cost per device by approximately $1.30 USD over the $ 0.40 USD estimated by Greer (Greer et al. 2007). The disposable PDMS interfacing used here adds an additional $0.20 USD material cost per device, although the reusable Nanoport interfaces (∼$20 USD, Upchurch Scientific, WA, USA) suggested by Greer were not included in their estimates. Since the labor involved in mounting the PDMS interfaces versus the Nanoports is similar, the labor costs given by Greer would also apply to this present device manufacture. Therefore, the estimated cost to fabricate the prototype devices described in this article is less than $2 USD for materials, and $10 USD with labor.

3.2 Thermal control and uniformity

By heating a microfluidic device from room temperature with the heating platform and controller, a stable temperature gradient would develop within 5 min, as determined by the IR images of the system. Figure 2(a) shows an IR image after thermal equilibrium was achieved, with a thermal resolution of approximately 400 µm per pixel. In this overhead view, the gradient is oriented vertically, with the cooler edge along the bottom of the chip. The dotted rectangle overlaid on the image in Fig. 2(a) indicates the position within the microfluidic device of the serpentine channel. As is shown in Fig. 2(b), the gradient within this rectangular region encompassed the entire temperature range within which PCR can be achieved. The error bars in the graph show the maximum temperature variation across the rectangular region indicated in Fig. 2(a), with a maximum error of less than 1°C at the lower temperatures, and of approximately 1.4°C at the higher temperatures. With the cross-chip gradient being on the order of 3°C/mm, precise placement of the microfluidic device within the heating platform was critical to reducing thermal variations between experiments. Therefore, the alignment features designed into the heating platform (see “Supplementary material”) proved particularly useful.

3.3 Simultaneous amplification and analysis

The microfluidic PCR was performed and the device was imaged, always with the reaction proceeding at full speed. Figure 3(a) shows a representative fluorescence image taken shortly after the PCR mixture filled the channel, such that there was PCR mixture simultaneously in every stage of the amplification process. The orientation of the gradient in Fig. 3(a) is the same as that shown in Fig. 2. The flow rate of the PCR mixture was 2 µl/min from left to right, such that the fluid observed in the left-most section in the image was experiencing its 2nd thermal cycle and the fluid observed in the right-most section was experiencing its 23rd thermal cycle. The fluorescence in the cooler regions of the channel increased with cycle number, first observable around cycle 12 and reaching a plateau near cycle 20. This indicated that that amplification was occurring, and agreed qualitatively with the real-time data obtained on the LS-32. Within the higher cycle channels, the strong fluorescence that was observed at lower temperatures abruptly disappeared at a specific and consistent vertical position. This transition signaled the denaturation of the specific PCR product. An amplification curve and multiple DNA melting curves were extracted from the acquired images.
Fig. 3

By analyzing one fluorescence image, the entire amplification process can be analyzed. (a) shows a representative image of the microfluidic device after PCR mixture has filled the serpentine channel. The temperature gradient, approximated with the labels to the right of the image, is uniform in the horizontal direction. Fluid flow is from left to right in this image, such that the fluid seen on the far right of the image was in the process of experiencing its 23rd thermal cycle. The increasing fluorescence is shown as a function of cycle number in (b). The decreasing fluorescence as a function of temperature is shown in (c). These spatial melting analyses were performed at multiple cycle locations within the image, as indicated. These curves show the raw fluorescence within each region of interest

The amplification curve, as shown in Fig. 3(b), included one data point for each cycle. Each data point was obtained by examining the intensity of a pair of 10 × 10 pixel squares, one at a vertical position where the product was double-stranded (labeled “signal” in the figure), and the other at a vertical position where the product was single-stranded (labeled “background” in the figure). Since the analysis is of a 12-bit image rather than raw photodiode data, it was not feasible to divide the signal value by the background value. Instead, the difference between these two values (signal minus background) was used as the data point in the amplification curve shown in Fig. 3(b). A first order Savitzky-Golay filter was also used to obtain a best-fit line through the data points. The resulting curve is also shown in the figure. Spatial melting curves, shown in Fig. 3(c), were obtained from cycles 10, 14, 16, 18, and 22 of the image shown in Fig. 3(a). Using the GUI, a 13 pixel wide vertical line of fluorescence intensity was used to produce the DNA melting curves shown. Although a single vertical pixel line could be used to characterize the melting within a channel section, this 13-fold spatial redundancy resulted in a significant increase in the SNR, which approached 200 for higher cycle melts. The curves shown in Fig. 3(c) constitute the raw fluorescence data as described. No additional smoothing was performed.

The amplification of both the 110-bp and the 181-bp targets were examined with this spatial analysis technique. Figure 4 compares the DNA melting curves and derivative plots for these two PCR targets on both the experimental system and the LS-32. The data sets from the experimental system consist of a seven pixel wide line cluster, minimally smoothed using a first order Savitzky-Golay filter. The variation between the melting temperatures and melting profiles of the two products agrees qualitatively with the control amplification and analysis experiments performed on the LS-32. Melt curve variations between the experimental system and the LS-32 were caused by the approximations used to obtain the temperature distribution in the microfluidic device (Crews et al. 2008c). While comparison between spatial melt curves has been used to genotype SNPs (Crews et al. 2009), the extent of the product analysis that was performed with the homozygous samples from this current study concluded that the process did in fact amplify the desired sequence, and that only the desired target was amplified. For all experiments, images such as that shown in Fig. 3(a) were analyzed to confirm the specificity of the spatial PCR, differentiate the products, and to obtain qualitative efficiencies for the respective experiments.
Fig. 4

(a) The melting behavior of the PCR products was imaged while the DNA was undergoing a spatial PCR process. (b) The melting temperatures of both targets were verified on the LS-32. As expected, the 110-bp target was characterized by a single melting regime, while the 181-bp target displayed two melting regimes. Such variations in the melting profile—not just the melting temperature—are sequence-specific characteristics that serve as “fingerprints” of the different PCR targets. Identical processing of both data sets in (a) included a 7-fold spatial redundancy and minimal smoothing

3.4 Robustness

The combination of aminosilane-coated glass and the surface passivation step were expected to reduce the reagent adsorption that can prevent initial amplification of the PCR mixture (Crews et al. 2008c). By monitoring the fluorescence of the PCR mixture as it filled the microchannel, the initial effect of this surface passivation protocol was examined. The air/liquid interface at the leading front of the PCR mixture can be seen in Fig. 5. The melting transitions are clearly observable in these leading microliters, thereby confirming the initial PCR biocompatibility of the microchannel. The extended biocompatibility of the system was also evaluated by identifying any change in the amplicon melting within a selected heating region over time. DNA melting curves from cycle 22 were measured after 10 µl, 40 µl, and 80 µl of the PCR mixture had been amplified. No statistical change in the melting curves was observed (data not shown).
Fig. 5

By pre-treating the microfluidic channel, amplification of the PCR samples occurred immediately. This image shows the air/fluid interface at the front of the flowing PCR mixture. The fluorescence of the sample is indicative of amplification, as is the observable melting transition of the amplifying DNA in the previous cycles

The thermal robustness of the glass-composite device was examined by quantifying changes in the spatial melting as a function of volumetric flow rate. Figure 6 shows the fluorescence, normalized in intensity only, for flow rates between 0.5 µl/min and 4.0 µl/min. The shapes of the six melt curves are virtually identical. The minute shifting in horizontal position of the melting transitions are random, rather than proportional to the flow rate. This can be seen in the figure inset. The maximum variation between the melting positions for these six curves is approximately five pixels, or 150 µm of channel length. Since very slight shifting of the optics or of the microfluidic device is anticipated, the differences between the melt curves at the six flow rates are within the uncertainty range of the optical and thermal subsystems. From this, it can be concluded that no significant temperature drifting is occurring within this range of flows. In contrast, numerical and experimental work with continuous-flow polymer microfluidics records significant thermal perturbations with flow rates even as low as 0.45 µl/min (Crews et al. 2008a; Hashimoto et al. 2004). The superior thermal robustness of these glass-composite cf-PCR devices is attributed to the nearly 10-fold higher thermal conductivity of glass over polymer substrates.
Fig. 6

The spatial melting curves at six flow rates are shown in this graph. The maximum melting temperature variation between these six curves is about five pixels, or 0.15 mm. To accurately compare the shape of the melting transitions, the fluorescence intensities of these six curves were normalized with respect to each other (vertically, not horizontally). The data shown here has not been smoothed beyond the clustering of adjacent pixel lines

3.5 Signal-to-noise ratio

A representative image was used to qualitatively evaluate the effect of spatial redundancy and camera exposure time on data quality. Figure 7 displays this dependence, as calculated from cycle 22 within a representative image acquired during PCR. The error bars in Fig. 7(a) indicate the maximum variation of the SNR for the five values that were collected for each data point. The SNR increased with exposure time as well as line width. The positive effect of increased spatial redundancy is most pronounced for low values, approximately doubling the SNR when a single line is coupled with both its immediate neighbors. The increasing SNR with exposure time is shown graphically in Fig. 7(b). The three curves shown correspond to a pixel line width of 1. The noise itself can be seen to increase little, while the fluorescence change associated with the DNA melting increases significantly. The SNR as a function of line width is shown graphically in Fig. 7(c). These curves all correspond to an exposure time of 125 ms. The curves were exactly overlapping, as shown in the inset. They have been intentionally moved apart in Fig. 7(c) to clearly show the relative magnitude of the noise in each.
Fig. 7

Images were taken with exposure times of 125 ms, 250 ms, and 1,000 ms. Spatial melting analyses were performed at five arbitrary horizontal positions within cycle 22 of these images, and for seven different pixel line widths. The SNR associated with the spatial melting analysis of a given fluorescence image is a function of both the exposure time and the degree of the spatial redundancy in the image. (a) shows the SNR as a function of both parameters; (b) shows representative melt curves for the three exposure times with a line width of 1; (c) shows representative melt curves for several line widths with an exposure time of 125 ms. Note: These curves in (c) were artificially separated from each other for clarity. The original overlap of the three curves is shown in the inset of (c)

It is valuable to note that this inclusion of adjacent pixel lines in the spatial melting analysis—in contrast to averaging in the direction of the gradient—only strengthens the melting signal, and has a clarifying effect, rather than a smudging effect on the data (see Fig. 7(c)). This benefit is a product of the spatial redundancy naturally inherent in such systems, and is independent of the time of experiment.The relatively high SNR obtained with this system is very promising, given the grade of the CCD camera used for this work. By upgrading the optical system with a larger CCD pixel array, or by cooling the CCD chip, it is anticipated that the SNR would increase substantially. Even without these improvements, however, the current version of this system can achieve a SNR with a 1 s exposure time that exceeds the SNR obtained by some commercial systems after an hour long DNA melting analysis (Herrmann et al. 2006).

4 Conclusion

A low cost microfluidic device for simultaneous DNA amplification and identification was fabricated using a rapid prototyping technique. An iterative design, testing, and re-design process was achieved with minimal time and expense. The microfluidic device was comprised of a patterned polymer film adhered between two un-patterned microscope slides. The final product was a composite structure of approximately 96% glass, thereby maintaining essentially the same thermal and biocompatibility properties of glass. By inducing a steady-state thermal gradient in the substrate through which the PCR mixture was steadily pumped, temperature cycling was performed and efficient, specific amplification was achieved. A 23-cycle PCR was achieved on the device within 15 min, during which the amplifying plasmid target was detected and positively identified. This was done by quantifying the spatial fluorescence distribution produced by an intercalating dye in the mixture. Signal-to-noise ratios on the order of 200 were achieved with an entry-level CCD camera by evaluating areas of fluorescence rather than lines of fluorescence.

With the fabrication technique used in this work, the necessary expense of cf-PCR development and testing is significantly reduced. The virtually unmatched ease of iterative design is coupled with lower cost materials and equipment, while preserving much of the functional advantages of glass microfluidics. The experimental technique presented in this article incorporates the analytical features of real-time PCR into the cf-PCR platform. In addition, the PCR can become the final process of a genetic test, with the DNA being characterized while still within the cf-PCR channels themselves. The in situ melting analysis of the PCR product can be achieved by analyzing data obtained before the reaction is even completed, without interrupting the amplification process. By incorporating such optical analysis, cf-PCR now achieves a significant versatility, which is a welcome companion to its already proven speed and simplicity.


A short video demonstrating the fabrication and attachment of the PDMS ports is located at∼ncrews/Presentations.


A recent version of this MATLAB program and an operator’s tutorial is available for download at∼ncrews.





Authors credit funding by Louisiana Tech University and the Louisiana Space Consortium (LaSPACE). Authors thank Jimmy Cook for machine shop support, Debbie Wood, Dee Tatum, and the rest of the technical staff at the Institute for Micromanufacturing for their support on this work. Special thanks is also given to Dr. Rastko Selmic for his role in the formation of this research team, and Dr. Eric Gilbeau for his mentorship and assistance with the buildup of the research laboratory.

Supplementary material

10544_2009_9389_MOESM1_ESM.pdf (174 kb)
Supplementary Material(PDF 173 kb)

Copyright information

© Springer Science+Business Media, LLC 2009