Microfluidic devices for fluidic circulation and mixing improve hybridization signal intensity on DNA arrays

Abstract

Reactions of biomolecules with surface mounted materials on microscope slides are often limited by slow diffusion kinetics, especially in low volumes where diffusion is the only means of mixing. This is a particular problem for reactions where only small amounts of analyte are available and the required reaction volume limits the analyte concentration. A low volume microfluidic device consisting of two interconnected 9 mm × 37.5 mm reaction chambers was developed to allow mixing and closed loop fluidic circulation over most of the surface of a microscope slide. Fluid samples are moved from one reaction chamber to the other by the rotation of a magnetic stirring bar that is driven by a standard magnetic stirrer. We demonstrate that circulation and mixing of different reagents can be efficiently accomplished by this closed loop device with solutions varying in viscosity from 1 to 16.2 centipoise. We also show by example of a microarray hybridization that the reaction efficiency can be enhanced 2–5 fold through fluid mixing under conditions where diffusion is rate limiting. For comparison, similar results were achieved with a disposable commercial device that covers only half of the reaction area of the closed loop device.

---

Introduction

In biological or chemical reactions, diffusion is typically the dominant mechanism for bringing different reactants together, and if limited by diffusion, the reaction may take days to complete. This is a particular problem for reactions where only small amounts of analyte are available and the required reaction volume limits the analyte concentration. In that situation fluid mixing is desirable in order to enhance reaction kinetics. Although efficient mixing can be achieved through diffusion alone by significantly reducing the dimensions of a system,¹ it is not always practical, and fluid circulation is the preferred mechanism.

Fluid pumping can be achieved by either mechanical or non-mechanical means in an open loop format, but is typically more difficult to implement in a closed loop system.² A toroidal thermal closed convection loop has been studied extensively to introduce fluid motion in either time-dependent, periodic or chaos flows as well as for developing various fluid motion control strategies.^3–5 In these cases, heating and cooling are required to generate fluid motion, which is not practical for temperature sensitive reactions, and is difficult to implement in miniaturized systems.

Recently, it was demonstrated that analyte binding could be enhanced by a factor of 60 or more on a polydimethylsiloxane (PDMS) device with very small reaction vessels in the micron range.^6,7 However, such a device is not suitable for applications that require a relatively large reaction area, such as a DNA microarray (typically >1 cm²). To this end, we have developed a low volume microfluidic device for closed loop fluidic circulation and mixing which consists of two interconnected reaction chambers of sufficient size to cover most of the usable area on a standard microscope slide. The fluid sample is circulated and mixed by the movement of one of the two magnetic stirring bars that are part of the device. The effect of fluid circulation on reaction performance was tested with a microarray hybridization assay under conditions where the reaction kinetics were limited by diffusion of the target across a relatively large reaction area.

Experimental

Design and fabrication

The closed loop microfluidic device consists of two interconnected 9 mm × 37.5 mm reaction chambers, that are designed to cover approximately two thirds of a standard 25.4 mm × 76.2 mm (1″ × 3″) glass slide. The chambers are connected at each end via a small circular reservoir, which also serve as inlet and outlet ports, respectively. The reservoirs each hold a magnetic stirring bar (Fig. 1A). Placing the device on a standard magnetic stirrer (Corning Stirrer/Hot Plate, Model PC-420) allowed one of the bars to be rotated, but resulted only in a vibration of the second bar. A more efficient set-up would allow both bars to spin simultaneously, but was not available at this time. For movement and mixing of solutions, a stirrer setting of 6 was typically used, resulting in a 20 min fluid cycle time with water-like viscosity, and was adjusted to accommodate higher viscosity solutions.

Fig. 1 (A): Design of the closed loop microfluidic device. (B): Schematic diagram of the plastic closed loop microfluidic device. One side of the double-sided adhesive membrane (3M Tape number 9731; 3M St. Paul, MN) (polycoated paper lined side) was attached to a glass substrate such as a standard 1″ × 3″ microscope glass slide while the other side (film lined slide) was attached to a 1″ × 3″ plastic substrate. Sample was injected into the chamber through one of the reservoirs in the plastic substrate. Stirring bars were placed into the reservoirs. The reservoirs were sealed with a single-sided adhesive membrane.

Two rapid prototyping techniques were used for fabrication. The first technique was soft lithography using PDMS which has been widely described in the literature.⁷ Briefly, a prepolymer of PDMS was cast onto a molding template and cured at 65 °C for one hour. The polymer replica of this template which contained a negative relief of the closed loop device (110 μm–130 μm in depth) was peeled away from the molding template and placed on a standard 1″ × 3″ microscope glass slide to form the PDMS closed loop device.

A second approach for fabricating such a device was implemented to solve the sealing problem of the PDMS device that occurred during overnight incubation at 42 °C. A plastic device was formed by cutting a template from a 140 μm thick double-sided adhesive membrane (3M Tape number 9731; 3M St. Paul, MN) and placing it between a 1″ × 3″ custom-made plastic substrate and a standard 1″ × 3″ microscope glass slide (Fig. 1B). Two standard 1″ × 3″ plastic cell culture slides (160005 Permanox^® Cell Culture Slide, Nalge Nunc International, Naperville, IL) were taped together to form the 1″ × 3″ plastic substrate for increased stability. Two 5 mm diameter holes were machined through the plastic substrate for the inlet and outlet of the device and to serve as reservoirs for the magnetic stirring bars (Cat. No. 37119-0005; Bel-Art Products, Pequannock, NJ). The bars were reshaped from their original size (2 mm diameter × 5 mm) to fit the reservoirs. After loading the fluid sample and placing the stirring bars, the inlet and outlet were sealed by the adhesive membrane and the whole device was placed on top of the stirrer.

Fluid circulation and mixing

A series of experiments was performed at room temperature to characterize fluid flow in either of the closed loop devices as a function of the stirrer setting. The PDMS based devices tested contained either 120 μm or 60 μm deep reaction chambers with 5 mm or 8 mm diameter reservoirs, respectively. Conventional photolithography was used to fabricate the PDMS molding template for the 60 μm deep reaction chamber device (total volume of 120 μL). Two 2 mm diameter × 7 mm stirring bars (Cat. No. 58948-976; VWR International, West Chester, PA) were used in the device with the 8 mm diameter reservoirs (a total volume of 360 μL), and 2 mm diameter × 4 mm stirring bars (reshaped from their original size) were used for the 5 mm diameter reservoirs.

Deionized water and hybridization buffers of two different compositions (viscosity) were tested as fluid samples. For visual inspection, 10 μL of food dye was added to each of the reservoirs after loading the fluid sample (Assorted Food Colors & Egg Dye; McCormick & Co., Inc., Hunt Valley, MD). Different stirrer settings were used for fluid movement; in control experiments the stirrer was turned off.

Microarray hybridization

PCR amplified human and bacterial gene sequences were used to fabricate test arrays.⁸ The same sequences were then amplified in the presence of primers that contained the T7 promoter for synthesis of mRNA using T7 RNA polymerase (Ambion, Inc., Austin, TX). These mRNAs were copied into cDNA and simultaneously labeled with Cy3 by reverse transcription (PerkinElmer Life Sciences, Inc., Boston, MA). Details of the microarray fabrication and hybridization assay protocols have been described elsewhere.⁹ Hybridization buffer B (used for all hybridization assays in this study unless otherwise stated) contains dextran sulfate as a volume displacer and has a viscosity of 16.2 centipoise. Buffer A (1.5 centipoise) does not contain any dextran sulfate but is otherwise identical to buffer B. The hybridization solution was introduced into the device through the ports followed by the magnetic stirring bars. The ports were sealed and the device was placed on top of the stirrer inside an oven (Isotemp Premium Lab Oven, Fisher Scientific International Inc., Pittsburgh, PA) and the fluid was circulated at 42 °C with a stirrer setting of 6. After 42 °C hybridization, the glass slide was separated from the device, washed to remove unhybridized target, dried and scanned (GenePix 4000A Microarray Scanner with GenePix Pro 3.0 data analysis software; Axon Instruments, Inc., Foster City, CA) as described before.^10,11 In this study, the PDMS device was used for 2 hour hybridization while all the overnight hybridizations were performed with the plastic device.

Results and discussion

Fluid circulation and mixing

Time-lapse photography demonstrated fluid circulation and mixing of water inside the PDMS closed loop device after addition of red and blue food dyes (Fig. 2A).

Fig. 2 (A): Time-lapse photography (from left to right, t = 0, t = 6 min and t = 16 min) of fluid circulation inside the PDMS closed loop microfluidic device with 120 μm deep reaction chambers (stirrer setting: 6). (B): Fluid cycle times at room temperature in differently configured PDMS closed loop devices as a function of stirrer setting. A 120 μm deep chamber was tested with deionized water (1 centipoise) and either 2 mm diameter × 4 mm (■) or 2 mm diameter × 7 mm long stirring bars (●). The same test was done with a 60 μm deep chamber and 2 mm diameter × 4 mm long stirring bars (▲). Viscosity effects were tested in a 120 μm deep chamber with 2 mm diameter × 4 mm long stirring bars. Hybridization buffer A (◆), hybridization buffer B (◇).

Next, the fluid cycle time of differently configured PDMS devices was estimated at room temperature. Fluid cycle time was defined as the time required for the front edge of the colored fluid to travel through both reservoirs once and return to the starting point. Since this was estimated based solely on visual inspection and given that the limit of detection of the human eye is rather poor, the cycle times may actually be faster than reported. As expected, the fluid cycle time depends on the reaction chamber depth, the size of the stirring bars and the stirrer setting (Fig. 2B), and can be significantly reduced either by increasing the size of the stirring bars, resulting in more pressure, or by increasing the reaction chamber depth, resulting in less friction. Since larger stirring bars require larger reservoirs, both of these approaches result in an increase in the minimum volume of fluid sample required. In our experiments, the fluid cycle time decreased concomitantly with an increased stirrer setting, but only up to a point after which it increased again. This result is an artifact of our current imperfect experimental set-up. At higher stirrer settings, the speed of one of the stirring bars could be increased but the vibrations of the second bar turned into a random jumping movement. As a result, the colored fluid was simultaneously being pushed into the two reaction chambers by the fast rotating bar while the vibrating bar counteracted the fluid flow, resulting in an increase rather than a decrease of the fluid cycle time.

The effect of viscosity was tested by comparing deionized water with two different buffer solutions. Surprisingly, the cycle time for the hybridization buffer A was found to be shorter than that of deionized water. We expect this is because the hybridization buffers contain sodium dodecyl sulfate (SDS) and buffer A spreads more easily on a glass surface than deionized water, due to reduced surface tension. For buffer B this effect was compensated by its increased viscosity due to the dextran sulfate.

Enhancement of microarray hybridization efficiency through fluid movement

Hybridization assays were performed on a small test array in the PDMS device with 2 mm diameter × 7 mm long stirring bars at 42 °C for 2 h with or without fluid movement (stirrer setting 6). As shown in Fig. 3, a 3 fold increase in hybridization efficiency was observed as a result of moving the hybridization mixture (buffer B) across the slide. However, extending the incubation times beyond 2 h at 42 °C resulted in the formation of bubbles around the edges of the two reaction chambers, and eventually the two reaction chambers would completely dried out, due to a loss of the seal between the PDMS and the glass surface.

Fig. 3 Effect of fluid circulation on 2 h 42 °C microarray hybridization assays in the PDMS device. Arrays consisting of two different oligos (positive and negative control, each repeated 27 times, 54 spots total) were hybridized with 1.25 pg μL⁻¹ of target complementary to the positive control oligo in 360 μL of buffer B. The average signal intensity ratio of positive control/negative control is shown for two experiments, one performed with fluid circulation and the other without.

To overcome this shortcoming, a plastic device was fabricated using a plastic substrate and a double-sided adhesive membrane as described in the Experimental section, and tested in overnight 42 °C hybridization assays (2 mm diameter × 4 mm stirring bars). Again, a 2–5 fold increase in hybridization efficiency was observed with fluid circulation (Figs. 4A and B).

Fig. 4 (A and B): Effect of fluid circulation on overnight 42 °C hybridization efficiency in the plastic device. A test array containing 8 different oligo probes (gene names indicated) with 12 repeat spots each was hybridized with 150 μL of hybridization buffer B containing different amounts of Cy3 labeled target for each gene as indicated on the x axis. (A): Image of the hybridized array with fluid circulation (right) and without (left). (B): Fold increase of the average net signal/background intensity ratio for eight probe genes as a result of fluid circulation. (C and D): Effect of viscosity and fluid circulation on 42 °C hybridization efficiency in the plastic device. Hybridization was for overnight in either buffer A or buffer B, but otherwise identical to the conditions described for A and B. (C): The average net signal/background intensity ratio is shown for two experiments with buffer B, one performed with fluid circulation (black) and the other without (white). (D): Fold increase of the average net signal/background intensity ratio as a result of fluid circulation for buffers A (white) and B (black).

The effect of fluid viscosity was investigated by comparing hybridization buffer A (1.5 centipoise) with buffer B (16.2 centipoise; identical to buffer A but with increased viscosity due to the presence of dextran sulfate) under otherwise identical hybridization conditions using the plastic device. Figs. 4C and D show a better hybridization efficiency in the presence of dextran sulfate in spite of the increased viscosity. As reported before,¹² dextran sulfate acts as a volume displacer, effectively increasing the concentration of the target. As expected, fluid circulation further improved the hybridization efficiency for buffer B. However, for hybridizations in low viscosity buffer fluid movement did not enhance the kinetics (Fig. 4D), suggesting that diffusion was not the rate limiting step under these conditions. It is important to note in this context that even though hybridization kinetics on arrays are pseudo-first order, the overall results (signal intensity) are actually the results of several reactions, whereby only the first step (target making first contact with the probe) is limited by target diffusion.

For comparison, we have tested a commercially available low cost microarray hybridization device, the GeneFlow™ microarray chamber from Gene Tec Corporation (www.genetecproducts.com). The unique chamber design of this device uses a fluid volume of 50 μL to cover a 4 cm² area and achieves passive fluid circulation by mechanically rotating the chamber through 360° (Fig. 5A). Using essentially identical hybridization conditions (same target concentration) as those described for Fig. 4, fluid circulation resulted in 3–4 fold increased hybridization efficiency, very similar to the results obtained with our devices (Fig. 5B). However, the GeneFlow™ chamber covers only slightly more than half the reaction area (20 mm × 20 mm) of the closed loop device (two 9 mm × 37.5 mm reaction chambers).

Fig. 5 Overnight 42 °C hybridization enhancement by the GeneFlow™ microarray chamber. (A): The dimensions of the slide area covered by this hybridization chamber are 2 cm × 2 cm, with a total fluid volume of 50 μL. (B): Arrays were hybridized overnight with target solution in buffer containing 4.5% dextran sulfate (13.6 centipoise) but otherwise similar to that described in the legend to Fig. 4. Fluid movement was achieved by mounting the device on a rotor inside an incubator (Model no. 308, Lab-Line Instruments, Inc., Melrose Park, IL) with rotor setting of 6 requiring 90 min time for complete mixing. (■): with fluid movement. (□): without fluid movement.

Conclusions

It was demonstrated that fluid circulation in closed loop microfluidic devices can improve the hybridization of DNA targets to surface mounted probes under conditions where the buffer viscosity is significantly higher than that of water. Depending on the exact hybridization conditions and chamber dimensions we have observed a 2–5 fold increase in reaction efficiency when comparing the signal intensities obtained with fluid movement to those without fluid movement. However, what we have not addressed herein is the effect of target concentration. Since hybridization of DNA targets to surface mounted DNA probes resembles first order reaction kinetics, the fluorescent signal obtained on each probe spot should scale linearly with target concentration over a wide range of target concentrations. Thus, decreasing the hybridization volume without reducing the total amount of target should also have a positive effect on hybridization kinetics, until the dimensions are such that fluid movement is difficult to achieve and diffusion becomes rate limiting. Thus, the optimum device design would have minimal reaction chamber dimensions and still achieve fluid movement.

Acknowledgements

We would like to thank Gene Tec Corporation (Durham, NC) for providing the GeneFlow™ microarray chambers.

References

1. G. H. W. Sanders and A. Manz, Trends Anal. Chem, 2000, 19, 364–378 . [*][*] A review of chip-based microsystems for genomic and proteomic analysis.
2. P. Schuck, D. B. Millar and A. A. Kortt, Anal. Biochem., 1998, 265, 79–91.
3. P. K. Yuen and H. H. Bau, J. Fluid Mech., 1996, 317, 91–109.
4. P. K. Yuen and H. H. Bau, Neural Networks, 1998, 11, 557–569.
5. P. K. Yuen and H. H. Bau, Phys. Fluids, 1999, 11, 1435–1448.
6. H.-P. Chou, M. A. Unger and S. R. Quake, Biomed. Microdev., 2001, 3, 323–330 . [*] This paper described a microfabricated rotary pump fabricated by multilayer soft lithography.
7. S. R. Quake and A. Scherer, Science, 2000, 290, 1536–1540 . [*][*] A review of soft materials for applications in areas ranging from microfluidic device technology to nanofabrication.
8. A. E. Murray, D. Lies, G. Li, K. Nealson, J. Zhou and J. M. Tiedje, Proc. Natl. Acad. Sci. USA, 2001, 98, 9853–9858.
9. P. Bao, A. G. Frutos, C. Greef, J. Lahiri, U. Müller, T. C. Peterson, L. Warden and Y. Xie, Anal. Chem., 2002, 74, 1792–1797.
10. P. Hegde, R. Qi, K. Abernathy, C. Gay, S. Dharap, R. Gaspard, J. Earle-Hughes, E. Snesrud, N. Lee and J. Quackenbush, Biotechniques, 2000, 29, 548–556.
11. U. R. Müller, Y. Bao, G. Li, Y. Xie and N. Raghavachari, in FISH Technology: Springer Laboratory Manual, eds. B. Rautenstrauss and H. Liehr, Springer-Verlag Publisher, New York, 2002, pp. 473–484.
12. G. M. Wahl, M. Stern and G. R. Stark, Proc. Natl. Acad. Sci. USA, 1979, 76, 3683–3687.

---

Footnote

† Current address: Nanosphere, Inc., 1818 Skokie Blvd., Northbrook, IL 60067, USA. Tel: (847)-656-0252.

---