Polymerase Chain Reaction in High Surface-to-Volume Ratio SiO2

Anal. Chem. 2004, 76, 6588-6593

Polymerase Chain Reaction in High Surface-to-Volume Ratio SiO2 Microstructures Madhavi Krishnan,†,‡ David T. Burke,§ and Mark A. Burns*,|

Department of Chemical Engineering, Department of Human Genetics, and Department of Biomedical Engineering, The University of Michigan, Ann Arbor, Michigan 48109

We have performed the Taqman β-actin PCR system in high-surface-to-volume ratio (0.02-0.13 µm-1) SiO2 microchannels and observed the reaction yield and uniformity. The concentrations of the enzyme, magnesium chloride, and reaction template were varied in the reaction mix, and PCR amplification was performed in devices of various surface-to-volume ratios. We found that microchannels with higher surface-to-volume ratios required higher enzyme concentrations to achieve the same amplification efficiency. We investigated the possibility that the observed reaction nonuniformity was related to the specific adsorption of magnesium ions to the negatively charged SiO2 surface. The effect of several modifications to the reaction chemistry, the addition of the cagedmagnesium dye DM-Nitrophen, the replacement of human DNA template with PCR product, and the coating of the microchannel surface with Teflon were all studied. These modifications resulted in improved reaction uniformity in the microchannels and present opportunities for further studies on enhancing the efficiency and uniformity of PCR in high surface-to-volume ratio SiO2 microchannels. Miniaturization of biochemical analysis systems has necessitated studies on the feasibility of adapting traditional benchtop reaction biochemistries to the microscale environment. One the one hand, there are several obvious advantages to scaling down reaction volumes: greatly reduced reagent consumption, lower power requirement, faster processing times, increased portability, and, more recently, the ability to detect single molecules.1-4 On the other hand, the biochemistry of many analytical techniques in molecular biology is highly sensitive to minor changes in the chemistry of the reaction and, therefore, pose a significant challenge in process scale-down to the microchip format. The * To whom correspondence should be addressed. Phone: +49-351-463-40355. Fax: +49-351-463-40342. E-mail: [email protected]. † Department of Chemical Engineering. ‡ Present address: Institut fu ¨ r Biophysik/BioTec, Technische Universita¨t Dresden, Tatzberg 47-51, D-01099, Dresden, Germany. § Department of Human Genetics. | Department of Biomedical Engineering. (1) Burns, M. A.; Johnson, B. N.; Brahmasandra, S. N.; Handique, K.; Webster, J. R.; Krishnan, M.; Sammarco, T. S.; Man, P. M.; Jones, D.; Heldsinger, D.; Mastrangelo, C. H.; Burke, D. T. Science 1998, 282, 484-487. (2) Woolley, A. T.; Hadley, D.; Landre; P.; deMello, A. J.; Mathies, R. A.; Northrup, M. A. Anal. Chem. 1996, 68, 4081-4086. (3) Belgrader, P.; Benett, W.; Hadley, D.; Richards, J.; Stratton, P.; Mariella, R.; Milanovich, F. Science 1999, 284, 449-450. (4) Lagally, E. T.; Medintz, I.; Mathies, R. A. Anal. Chem. 2001, 73, 565-570.

6588 Analytical Chemistry, Vol. 76, No. 22, November 15, 2004

behavior of many benchtop reaction chemistries, particularly polymerase chain recation (PCR), has been found to be quite different in the presence of inorganic reaction vessel materials and higher surface-to-volume ratios encountered in microstructures.5-7 In scaling a reaction down from a microfuge tube to a microchannel, the reaction mix is subject to at least 1 order of magnitude increase in the surface-to-volume ratio. The walls begin to dominate, and the chemistry of the wall surface can become a significant component in the reaction. The situation can be particularly complicated in a multicomponent reaction such as PCR where the concentrations of several components need to be maintained in a fairly tight range and are also sensitively balanced with respect to one another (e.g., dNTPS and DNA bind a certain amount of free magnesium, making it unavailable for the reaction). In addition, traditional semiconductor microfabrication techniques used in the fabrication of microchips rely on silicon-based materials that are not traditionally used in bench-scale biochemical reactions. While the effects of various silicon-based materials, such as silicon oxide and nitride, on the performance of microscale PCR have been studied and reported in the literature, it has generally been concluded that certain materials such as SiO2 do not inhibit PCR performance, while others like silicon nitride do.5 PCR reactions in microfabricated devices can be easily studied using the biochemistry developed for real-time systems. The first demonstrated real-time PCR assay used the intercalation dye ethidium bromide.8 This intercalation dye as well as others such as SYBR Green I demonstrates a significant increase in fluorescence when bound to doublestranded DNA. While this method of detection is relatively simple and straightforward, DNA binding dyes indiscriminately bind all double-stranded DNA products. Therefore, nonspecific amplifications also result in an increased fluorescent signal that makes it difficult or impossible to disitinguish between target amplifications and misamplifications. Fluorescent oligonucleotide probes on the other hand, enable real-time monitoring of the PCR assay and ensure that increases in fluorescence result only from the accumulation of the desired product. One popular probe strategy is the TaqMan assay (Applied Biosystems) which capitalizes on the 5′-exonuclease activity of (5) Cheng, J.; Shoffner, M. A.; Hvichia, G. E.; Kricka, L. J.; Wilding, P. Nucleic Acids Res. 1996, 24, 380-385. (6) Shoffner, M. A.; Cheng, J.; Hvichia, G. E.; Kricka, L. J.; Wilding, P. Nucleic Acids Res. 1996, 24, 375-379. (7) Taylor, T. B.; WinnDeen, E. S.; Picozza, E.; Woudenberg, T. M.; Albin, M. Nucleic Acids Res. 1997, 25, 3164-3168. (8) Higuchi, R.; Fockler, C.; Dollinger, G.; Watson, R. Bio/Technology. 1993, 11, 1026-1030. 10.1021/ac0488356 CCC: $27.50

© 2004 American Chemical Society Published on Web 10/13/2004

Figure 1. Process flow for fabrication of glass-silicon hybrid devices with high surface-to-volume ratio microchannels (a) Channels fabricated in glass. (b) Channels fabricated in silicon.

Taq polymerase to cleave a labeled hybridization probe during the extension phase of PCR.9 Although the experiments in this study focus on end point detection of PCR amplification and do not involve real-time observation of amplification kinetics, use of the TaqMan biochemistry confers on the system the advantage of in situ detection of target-specific amplification using fluorescence microscopy. Here we focus on the performance of PCR in microchannels with SiO2 surfaces and study the effect of surface-to-volume ratio on the efficiency and uniformity of PCR amplification throughout the length of a microchannel. We report interesting anomalies in PCR amplification uniformity with increasing channel surface-tovolume ratio using the β-actin TaqMan PCR system. The β-actin Taqman PCR system is very robust and has been widely studied in various microsystem applications.7,10 This study adds to previous studies that deal with PCR in SiO2 microstructures11,12 and those that specifically addressed the dependence of PCR amplification efficiency on surface-to-volume ratio of silicon-based microstructures5-7 by exploring a range of microchannel surfaceto-volume ratios up to 1 order of magnitude higher than those described previously. MATERIALS AND METHODS Device Fabrication. Traditional photolithography techniques were used to pattern glass and silicon wafers. Two distinct types of glass-silicon hybrid devices were fabricated and assembled. Channels of lower surface-to-volume ratio (0.02-0.08 µm-1) were fabricated by wet-etching the structures in the glass side, while the high surface-to-volume ratio (0.13 µm-1) structures were fabricated by etching the channels into the silicon side using an anisotropic plasma etch. In either case, depending on whether the channels had been fabricated in silicon or glass, the die (9) Holland, P. M.; Abramson, R. D.; Watson, R.; Gelfand, D. H. Proc. Natl. Acad. Sci. U.S.A. 1991, 88, 7276-7280. (10) Kalinina, O.; Lebedeva, I.; Brown, J.; Silver, J. Nucleic Acids Res. 1997, 25, 1999-2004. (11) Waters, L. C.; Jacobson, S. C.; Kroutchinina, N.; Khandurina, J.; Foote, R. S.; Ramsey, J. M. Anal. Chem. 1998, 70(1), 158-162. (12) Kopp, M. U.; de Mello, A. J.; Manz, A. Science 1998, 280 (5366), 10461048.

Table 1 channel cross section (h, µm × w, µm)

total channel vol (µL)

surface-to-volume ratio (µm-1)

100 × 650 70 × 150 35 × 80 30 × 30

7.15 1.15 0.30 0.1

0.02 0.04 0.08 0.13

bearing the channels was anodically bonded either to a surfaceoxidized silicon die (channels in the glass die) or to an unpatterened glass die (channels in the silicon die), to create a glasssilicon hybrid device. Channels were fabricated in glass as follows: glass substrates were coated with Cr/Au (600/4000 Å) and were then patterned using photolithography. Cr and Au were etched in the exposed locations, and channels were etched into the glass in these locations using HF/HNO3 (7:3) (Figure 1a). Channels were patterned on the silicon substrate using photolithography and etched using anisotropic etching in an inductive coupled plasma. The 200 nm of silicon oxide was then grown on the wafer surface in a furnace oxidation step (Figure 1b). Holes were drilled in the glass dies, both patterned and nonpatterned, to provide access ports to the microchannel. The silicon wafers, both patterned and nonpatterned, were diced after furnace oxidation and anodically bonded to the glass dies to produce the microchannel devices for PCR. Surface-to-Volume Ratio Estimation. Surface-to-volume ratios of the microchannels were estimated by measuring the height and width of the microchannels using surface profilometry and applying the formula, 2(h +w)/hw, where h represents the height and w the width of the microchannel, respectively. Device Dimensions. Cross sections of the channels fabricated and the corresponding surface-to-volume ratios are shown in Table 1. The length of channels were 110 mm, except in the case of devices with a surface-to-volume ratio of 0.13 µm-1, where channel lengths of 110 and 220 mm were tested. The parameter “dimensionless distance” was calculated by dividing the actual distance of the point along the channel length, measured from the loading Analytical Chemistry, Vol. 76, No. 22, November 15, 2004

6589

vent, at which fluorescence intensity data were acquired, by the total length of the microchannel. Surface Coating Channels with Teflon. Experiments to investigate the effect of a Teflon microchannel surface coating on PCR amplification were performed using devices of surfaceto-volume ratio 0.08 µm-1. The channels were loaded with a 0.6% dispersion of Teflon PFA 335 fluoropolymer resin (DuPont) in water, and the water was allowed to evaporate. The procedure was performed twice, rinsing the channel between coatings and after the coating procedure with PCR grade water. Device Operation. Rubber gaskets (i.d. 2 mm) were attached to the device surface over the access holes using epoxy, and these served as reservoirs over the inlet and outlet ports to the microchannel. The PCR reagents were mixed in a microfuge tube, and a drop was placed at one end of the microchannel and allowed to wick into the channel by capillary action. When the advancing meniscus of the drop reached the end of the microchannel, the excess PCR mixture at the loading vent was aspirated, and the channel was sealed by loading the inlet and outlet reservoirs with PCR grade mineral oil. The devices we tested did not possess on-chip heaters, and thermocycling was performed by simply placing the device in a thermocycler (Biometra). Thermocycling conditions were as follows: precycle 50 °C for 2 min, 95 °C for 10 min; cycling 92 °C for 30 s, 54 °C for 30 s and 72 °C for 1 min, 35 cycles. At the end of the reaction, the device was removed from the thermocycler and fluorescence intensity data were gathered along the length of the microchannel using a fluorescence microscope. Fluorescence Measurements. The fluorescence intensities were recorded along the length of the microchannel using an inverted fluorescence microscope (Nikon TE 200). Images were acquired using a 12-bit high-resolution cooled Digital Interline CCD (Micromax: 1300 YHS, Princeton Instruments) and analyzed using MetaView software (Universal Imaging Corp.). A 100-W mercury lamp (Nikon) was used as the excitation source. The microscope was equipped with a motorized mechanical stage that was moved in 1-mm increments along the length of the microchannel, enabling images to be acquired along the length of the microchannel. Fluorescence excitation was performed using a band-pass filter in the range 450-490 nm. Fluorescence emission intensity data were acquired using two different emission filters. A band-pass filter in the green range (505-560 nm) was used to obtain green fluorescence emission, and a long-pass filter (515 nm LP) was used to obtain the total (red + green) emission separately. The difference between the total emission intensity and the green emission intensity was used to infer the red emission intensity. The ratio of green/red fluorescence intensity (G/R), which is a measure of FRET probe conversion, was used to assess the extent of target-specific amplification. Additionally, visual observation of a successful amplification under the microscope using the long-pass emission filter showed a marked color change from the initial red (“unamplified”) state to the green (“amplified”) state and thus served as an easy method to visually confirm the success of a reaction. PCR Reaction Conditions. PCR amplification was demonstrated by amplifying a 295-bp segment of the human β-actin gene (Applied Biosystems). Forward and reverse primer sequences 6590 Analytical Chemistry, Vol. 76, No. 22, November 15, 2004

Figure 2. Amplification signal uniformity in microchannels of various surface-to-volume ratios. (a) PCR amplification was performed in microchannels with surface-to-volume ratios ranging from 0.02 to 0.13 µm-1, and the value of the fluorescence ratio, G/R, was measured along the length of the channel. The fluorescence data for each channel were normalized against the maximum G/R value observed along the length of that channel, typically a value between 0.9 and 1.2, and plotted with respect to dimensionless length, beginning at the loading end of the channel. Channel lengths were typically 110 mm, but were 220 mm in some experiments. We found that for such large lengths of channel, for a given cross section, the amplification signal profile remained the same with respect to dimensionless length. (b) Comparison of amplification profiles in two devices of surface-tovolume ratio 0.13 µm-1, loaded with reaction mix from opposite vents but analyzed in the same direction. High amplification signal is consistently observed in the portion of the device close to the loading vent.

were 5′-TCACCCACAATGTGCCCATCTACGA-3′ and 5′-CAGCGGAACCGCTCATTGCCAATGG-3′. Reactions contained 200 nM TaqMan Probe, 10 mM Tris-HCl, pH 8.3, 50 mM KCl, and 0.2 mM each dATP, dCTP, dGTP, and dUTP, 4 mM MgCl2, 10 ng/ µL human DNA, and varying amounts of AmpliTaq polymerase (PE Applied Biosystems) depending on the surface-to-volume ratio of the device used. RESULTS AND DISCUSSION An important effect we observed in these high surface-tovolume ratio, glass-silicon hybrid devices was a sharp decay in amplification signal intensity with respect to microchannel length measured from the loading port. The relative location of this decay with respect to the length of the microchannel varied depending on the surface-to-volume ratio of the microchannel but was constant in devices of a given surface-to-volume ratio (Figure 2a). The surface-to-volume ratios we studied ranged from 0.02 to 0.13 µm-1. Figure 2b shows a plot of amplification signal intensity with length measured in devices that were loaded from opposite ends of the channel but read in the same direction. This result confirms that the loss of signal occurs in the region of the device where the reaction fluid has traversed a greater length of channel and has been exposed to a larger effective surface area than the fluid closer to the loading vent.

The fluorescence signal decay profile was generated using measurements at distance increments of 1 mm along the length of the channel. Reaction components such Mg and dNTPs could be assumed to be nondiffusing, as they are present in high concentrations throughtout the channel and remain relatively constant over the course of the reaction. The only new species that are actually generated and diffuse along the channel, thereby determining the fluorescence profile measured along the channel length, are the generated PCR products and the degraded FRET probe. The green fluorophore resulting from FRET probe degradation has a higher diffusion coefficient than the 300-bp-long PCR product and determines the precision with which the distribution of PCR products in the reaction channel can be measured. An estimate of the length of diffusional spreading of the fluorescence signal resulting from target amplification was of the order 1 mm (assuming a diffusion coefficient of the fluorophore, resulting from FRET probe degradation, of 10-6 cm2/s and an amplification time of 2 h). As this distance is of the same order as the resolution of the profiling process, it would not be possible to know the distribution of PCR products within this experimental framework to a precision better than 1 mm. The variation of the fluorescence intensity profiles with surface-to-volume ratio were reproducible within margins of experimental error and day-to-day variability. Modifications that Enhanced Signal Uniformity. To investigate the possibility that the oxide surface might be stripping the magnesium ions out of solution and making them unavailable for reaction, we added 1.5 mM of the caged magnesium compound DM-Nitrophen (DMNP) to the PCR reaction mixture in addition to 4 mM MgCl2. The use of DMNP to supply magnesium ions in reactions was previously demonstrated in a study on light-directed restriction digestion of single DNA molecules.13 The addition of 1.5 mM DMNP to the PCR mix was found to have no adverse effect on amplification in polypropylene tubes. The reaction mixture was loaded into a microchannel of surface-to-volume ratio 0.08 µm-1, and then the entire channel was exposed to UV light in order to cleave the caged compound and release the magnesium ions. The chip was thermocycled and then observed under a fluorescence microscope. The results indicate that target amplification took place in regions of the device that usually showed no amplification when free magnesium alone was used in the reaction (Figure 3a). However, the magnitude of signal intensity from amplification was not as high in the reactions with DMNP as in the case of reactions with free magnesium alone (the reaction with DMNP showed an average G/R of 0.42 over the entire device, while the control without DMNP showed an average G/R of 0.55). Nevertheless, the important outcome of this investigation was that amplification was observed in regions of the channel that usually showed little or no amplification without the caged compound. Another modification to the reaction that resulted in better signal uniformity was the use of the 295-bp amplified target sequence itself in place of human DNA as the reaction template. The concentration of PCR product in the reaction was in the picogram per microliter range while that of the human DNA template was in the nanoggram per microliter range. A low level (13) Namasivayam, V.; Larson, R. G.; Burke, D. T.; Burns, M. A. Anal. Chem. 2003, 75, 4188-4194.

Figure 3. Representative plots of improvement of amplification signal uniformity under various conditions and comparison with untreated controls. (a) Addition of DM-Nitrophen to the reaction in a microchannel of surface-to-volume ratio 0.08 µm-1. (Device efficiency: DMNP 100%; no DMNP 80%.) (b) Replacement of human DNA with PCR product as template in reaction in a microchannel of surface-to-volume ratio 0.13 µm-1 (Device efficiency: PCR product template 95%; human DNA template 65%.) (c) Amplification in a microchannel of surface-to-volume ratio 0.08 µm-1 coated with Teflon. (Device efficiency: Teflon coated 78%; no Teflon coating 80%.) The average G/R values in the usually inactive portion of the device (G/R ) 0.1) for these cases: DMNP 0.77, PCR product 0.43, Teflon coating 0.79.

of amplification signal (G/R ) 0.4) was observed in regions of the device that usually showed no target-specific amplification when human DNA was used as template (Figure 3b). We also investigated the effect of coating the microchannel surface with Teflon on amplification signal uniformity. This modification to the microchannel surface was found to extend the length of the amplification region in the device (Figure 3c). High signal intensity was observed in regions of the device that showed no amplification in the absence of the coating treatment, although some signal loss was observed close to the loading end of the channel. The average G/R value over the length of a Teflon-coated microchannel was found to be 0.76 while that in an uncoated device was ∼0.65. Analytical Chemistry, Vol. 76, No. 22, November 15, 2004

6591

Figure 5. Variation of “device efficiency” with MgCl2 concentration. The concentration of MgCl2 in the PCR reaction was varied between 1.5 and 5 mM, and PCR amplification was performed in microchannels of surface-to-volume ratio 0.13 µm-1(the concentration of human DNA template in these reactions was 10 ng/µL). The percentage of microchannel length with a G/R value greater than 0.3 was estimated. The signal uniformity profiles showed little variation around the optimum MgCl2 concentration of 3.5 mM. However, amplification was suppressed at lower and higher concentrations, which seemed to have no effect on the amplification efficiency in microfuge tubes (G/R values were consistently around 1.0 over this range of MgCl2 concentrations in PCR reactions in microfuge tubes).

Figure 4. Effect of enzyme concentration on amplification efficiency in SiO2 microchannels. (a) Variation of enzyme concentration with surface-to-volume ratio. PCR amplification was performed in devices of surface-to-volume ratios ranging from 0.02 to 0.13 µm-1. The amount of enzyme required to achieve a high G/R value of ∼1.0 in the active portion of the channel was found to increase with increasing surface-to-volume ratio. The point at 0.001 µm-1 represents data for a polypropylene microfuge tube. (b) Variation of amplification efficiency with enzyme concentration in a microchannel of surface-tovolume ratio 0.13 µm-1. The maximum G/R value observed in the active portion of the device was taken as a measure of the efficiency of specific amplification of the target sequence under specific reaction conditions. At an enzyme concentration below the optimal value for a give surface-to-volume ratio, no amplification was observed. Enzyme concentrations higher than the optimum resulted in a reduction in the FRET probe conversion efficiency, reflected in a lower G/R ratio. This effect could be attributed to a reduction in specific amplification of the target sequence in the presence of large amounts of enzyme.

Although treatment with Teflon and the use of DMNP in the reaction resulted in high signal in portions of the microchannel where no signal was usually observed, it is important to note that the overall device efficiency was only slightly higher than the “untreated” case, as the signal was not consistently high throughout the device. We define “device efficiency” as the percentage length of the channel that showed a G/R value greater than 0.3. Negative controls with no DNA amplification had a G/R value of 0.1, while a successful amplification reaction usually showed a G/R value of 0.9-1.2. A G/R value of 0.3 was therefore assumed to indicate weak amplification. [Device efficiency estimates for the cases of DMNP addition, reaction template replacement and Teflon-coating of the microchannel (Figure 3) are as follows: DMNP 100%, no DMNP 80%; PCR product template 95%, human DNA template 65%; Teflon coating 78%, no Teflon coating 80%. Investigation of Varying Reagent Concentrations. We found that glass-silicon hybrid devices of higher surface-to-volume ratio required higher enzyme concentrations in the reaction to give 6592 Analytical Chemistry, Vol. 76, No. 22, November 15, 2004

comparable probe conversion signals in the active portion of the device (Figure 4a). This observation is consistent with the general expectation of loss of enzyme from the aqueous phase to the walls through nonspecific adsorption to the surfaces. We also observed an optimum enzyme concentration for a given surface-to-volume ratio, an effect that was observed and reported for the β-actin system in another study.7 Addition of enzyme in excess of the optimum concentration resulted in a decrease in the maximum observed probe conversion signal, probably due to nonspecific amplification of nontarget sequences (Figure 4b). To investigate the possible adsorption of target DNA to the channel walls during loading, we altered the target concentration in the PCR mix. We performed PCR using four different human DNA template concentration: 0.3, 7.5, 10, and 18 ng/µL in devices of surface-to-volume ratio 0.13 µm-1. There was no detectable change in the location of the signal decay region in the microchannel over the concentration range 7.5-18 ng/µL. However the device with 0.3 ng/µL template human DNA showed no amplification at all, possibly due to statistical limitation of available template molecules in the 100-nL reaction volume (∼20 molecules). The magnesium ion concentration in the reaction was also varied to determine its effect on the amplification signal uniformity in the microchannel (Figure 5). The signal from the amplification of the target sequence appeared to be very sensitive to the magnesium ion concentration. We found that changing the magnesium concentration close to the optimal value for microfuge tubes (3.5 mM) did not alter the observed signal profiles in the device very much: the amplification intensity profiles for 3 and 4 mM magnesium were almost identical. However, increasing the magnesium concentration to 5 mM in the reaction mix resulted in complete inhibition of amplification throughout the device, except for a zone of weak signal close to the loading vent. This is behavior radically different from that observed in polypropylene microfuge tubes where the amplification efficiency and fluorescence signal resulting from FRET probe degradation are relatively insensitive to magnesium ion concentration in the range of 3-5 mM. However, the sensitivity of PCR to changes in the magnesium

ion concentration has been reported and may still be the major cause.14 In general, we had limited success with various other operational modifications directly related to the magnesium concentration. The system was very sensitive to alterations such as prerinsing the device with 1.5-2 mM MgCl2, and subsequently performing the reaction with 0, 1.5, and 2 mM MgCl2 in the reaction mix; these modifications consistently inhibited amplification throughout the device. Directly altering the magnesium ion concentration through MgCl2 in the reaction protocol therefore did not seem to have an effect on alleviating the signal nonuniformity issues in these devices. CONCLUSIONS High surface-to-volume ratio microstructures present a challenge to the successful adaptation of traditional biochemistries designed and optimized for the scale of benchtop operations to microscale systems. While the walls are usually assumed to play only a minor role in more macroscopic reaction systems, they play a significant role as a reaction component in the reaction chemistry at smaller size scales. Our studies show interesting effects with regard to reaction uniformity in high surface-to-volume ratio microchannels with SiO2 surfaces. These effects were not observed in previous studies on lower surface-to-volume ratio structures of the same material.5-7 One possible explanation for the location-specific decay in amplification signal intensity is the adsorption of one or more reaction components to the walls of the microchannel. We observed that the location of the amplification signal decay front shifted closer and closer to the loading vent with increasing surface-to-volume ratio. The packets of fluid that enter the microchannel first and traverse the entire length of the microchannel are exposed to a much larger effective surface area than the packets of fluid that enter the microchannel later and occupy the regions closer to the loading vent. It is likely that packets of fluids traveling longer distances through the microchannel are stripped of one or more reaction components by adsorption to the walls and, therefore, show no signal due to specific amplification of the target sequence. At some point along the microchannel length, it is possible that the arriving packets of fluid still retain a threshold concentration of components required for the reaction, and a high amplification signal is observed along the length of the microchannel preceding this point. The shift in the location of the decay front closer to the loading port, with increasing surface-to-volume ratio, further suggests the possibility that such an adsorption mechanism could be the underlying mechanism for the observed signal nonuniformity. Another possibility is that the actual solution concentration of magnesium in these SiO2 microchannels is affected not only by adsorption to the walls during the loading process but also by desorption off the walls at the elevated temperatures during the (14) Williams, J. F. Biotechniques 1989, 7, 762-768. (15) Misra, A.; Schmidt, B. L.; Hall, L.; Sees, J.; Hurd, T. Q. Semicond. Fabtech. 1999, 9, 173-178. (16) Yokota, H.; Sunwoo, J.; Sarikaya, M.; van den Engh, G.; Aebersold, R. Anal. Chem. 1999, 71, 4418-4422. (17) Doherty, E. A. S.; Meagher, R. J.; Albarghouthi, M. N.; Barron, A. E. Electrophoresis 2003, 24, 34-54. (18) Caelen, I.; Bernard, A.; Juncker, D.; Michel, B.; Heinzelmann, H.; Delamarche, E. Langmuir 2000, 16, 9125-9130.

reaction. Silicon oxide surfaces are negatively charged and have been known to cause surface adsorption of divalent cations.15 This effect has both beneficial and detrimental effects on device performance depending on the application. For example, the adsorption of magnesium ions to silicon oxide or glass surfaces has been used to tether and stretch DNA on surfaces.16 On the other hand the adsorption to an oxide wafer surface of divalent cations, such as the Ca2+ present in trace concentrations in deionized water used in semiconductor processing, has been known to produce deleterious effects on the performance of various semiconductor components.15 The number of surface oxide groups in a microchannel of surface-to-volume ratio 0.13 µm-1 was calculated to be more than 5 times in excess of the total number of magnesium ions in an amplification reaction occupying that volume. In our experiments on varying the concentrations of several components in the reaction, we found the signal uniformity issue was most affected in experiments related with magnesium. For example, a possible explanation for the disappearance of the sharp signal decay front upon addition of DMNP to the reaction mix is that the caged magnesium compound, an organic molecule, remained in solution during the loading of the microchannel and was not adsorbed to the oxide channel surfaces through ionic interactions. When the entire chip was then flood-exposed to the UV light, the magnesium was released uniformly into free solution along the entire chip, and the amplification was allowed to proceed in parts of the device where the reaction mix was possibly magnesium-depleted under normal circumstances. Coating the device with Teflon also resulted in a similar disappearance of the decay front, suggesting that the ionic nature of the untreated SiO2 surface might have a part to play in the abrupt loss of amplification. Further studies on these effects may be critical in furthering our fundamental understanding of reaction biochemistries themselves in addition to providing completely new information on the behavior of well-characterized biochemical phenomena at small length scales. An improved understanding of such phenomena would also aid in the development of techniques to alleviate observed anomalies, so that the desirable properties of materials such as silicon and glass may continue to be exploited in the fabrication of microstructures. While recognizing the importance of choosing biochemically inert or passivating materials to serve as walls in microstructures,17 understanding the effects of various wall chemistries on biomolecules or biochemical reactions could provide insight that could be exploited to a practical end. Microfabrication techniques greatly facilitate the construction of systems with surface-to-volume ratios ranging over orders of magnitude. A high surface-to-volume ratio microchannel provides an ideal platform for the exploitation of adsorption effects in various analytical operations such as purification, separation, or quantification of analytes on a chip.18 ACKNOWLEDGMENT This work was supported by the National Institutes of Health under Grant P01-HG001984. Received for review August 6, 2004. Accepted August 31, 2004. AC0488356 Analytical Chemistry, Vol. 76, No. 22, November 15, 2004

6593