Anal. Chem. 2005, 77, 4474-4480
DNA Microarray Enhancement Using a Continuously and Discontinuously Rotating Microchamber Johan Vanderhoeven,† Kris Pappaert,† Binita Dutta,‡ Paul Van Hummelen,‡ and Gert Desmet†
Department of Chemical Engineering, Vrije Universiteit Brussel, Brussels, Belgium, MicroArrayFacility Lab, Flemish Institute for Biotechnology (VIB), Leuven, Belgium
It is demonstrated that the most efficient way to enhance DNA microarray analysis consists of a maximal reduction of the total device volume (to keep the concentration of the available DNA as high as possible), combined with the creation of a strong lateral convective transport of the sample. In the present study, DNA microarray hybridizations are performed in a set of rotating, circular microchambers covering exactly the spotted area of the microarray and with a depth varying between 70 and 1.6 µm. Rotating the microchamber substrate while keeping the microarray stationary, the rotating microchamber bottom wall literally drags the sample past the microarray spots with a velocity which is independent of the fluid layer thickness. Interestingly, it was found that transporting the sample in a discontinuous mode (with stop periods of several minutes) not only yields a more stable and reproducible operation, it also yields significantly larger hybridization intensities (typically a factor of 2-3 larger) than a continuous rotation. This seems to be due to the fact that the velocity field disturbs the binding process at the binding site level. Working under limiting DNA sample mass conditions, the system yielded in a short, 30-min experiment already a 5-fold increase of the hybridization intensity, as compared to a conventional microscope slide/coverslip system operated overnight under diffusiondriven conditions. Compared to a commercial pumparound hybridization system, the gain was even more impressive, precisely due to the fact that the pump-around system requires larger volumes, which with a fixed amount of available genetic material leads to the application of more diluted samples. In the past years, many research groups have focused on the enhancement of DNA microarray analysis, and a large number of different solutions have been proposed and commercialized. These methods proposed include surface acoustic wave (SAW)based microagitation,1 cavitation microstreaming,2 electronic assistance,3 pressure- and electrically driven flow generation,4-6 the * To whom correspondence should be addressed. Phone: (+)0.32.2.629.37.81. Fax: (+).32.2.629.32.48. E-mail:
[email protected]. † Vrije Universiteit Brussel. ‡ Flemish Institute for Biotechnology. (1) Toegl, A.; Kirchner, R.; Gauer, C.; Wixforth, A. J. Biomol. Technol. 2003, 14, 197-204.
4474 Analytical Chemistry, Vol. 77, No. 14, July 15, 2005
use of pump-around systems,7 chaotic mixing,8 sample oscillation,9 shear-driven flow systems,10 etc. Central to all these approaches is that they use some form of mechanical or electrical forces to increase the transport rate of the sample DNA strands beyond their normal (extremely slow) diffusion transport rate. Trying to compare the different proposed systems on the basis of the published hybridization experiment results is very difficult because the available data relate to widely differing experimental conditions (probe spot density, length and complexity of the sample target strands, size and concentration of the applied samples, etc.). In addition, the reported final spot intensity data are very diffuse and disperse. In some cases,1,8 gains of up to a factor of 5 are claimed, whereas in other cases, the diffusion control and the transport enhanced experiment clearly converge to the same limit.2,6,9 Sometimes, the gain with respect to the traditional diffusion-driven coverslip experiments is not even mentioned. Approaching the problem of enhancing DNA microarray analysis with a minimum of mathematical complexities, the sample concentration can be premised as the key factor determining both the initial hybridization rates and the final spot intensities. The dependency of the initial hybridization rates on the sample concentration follows directly from the well-established kinetics expression,11-12
dH/dt ) konCinitial(Hmax - H) - koffH
(1)
wherein H and Hmax represent the molar surface concentration of (2) Liu, R. H.; Lenigk, R.; Druyor-Sanchez, R. L.; Yang, J.; Grodzinski, P. Anal. Chem. 2003, 75, 1911-1917. (3) Erdman, C. F.; Raymond, D. E.; Wu, D. J.; Tu, E.; Sosnowski, R. G.; Butler, W. F.; Nerenberg, M.; Heller, M. J. Nucleic Acids Res. 1997, 25, 49074914. (4) van Beuningen, R.; van Damme, H.; Boender, P.; Bastiaensen, N.; Chan, A.; Kievits, T. Clin. Chem. 2001, 47, 1931-1933. (5) Benoit, V.; Steel, A.; Torres, M.; Yu, Y.-Y.; Yang, H.; Cooper, J. Anal. Chem. 2001, 73, 2412-2420. (6) Lenigk, R.; Liu, R. H.; Athavale, M.; Chen, Z.; Ganser, D.; Yang, J.; Rauch, C.; Liu, Y.; Chan, B.; Yu, H.; Ray, M.; Marrero, R.; Grodzinski, P. Anal. Biochem. 2002, 311, 40-49. (7) http://www1.amershambiosciences.com/aptrix/upp01077.nsf/Contect/ Products?OpenDocument&ParentId)460778. (8) McQuain, M. K.; Seale, K.; Peek, J.; Fisher, T. S.; Levy, S.; Stremler, M. A.; Haselton, F. R. Anal. Biochem. 2004, 325, 215-226. (9) Liu, Y.; Rauch, C. B. Anal. Biochem. 2003, 317, 76-84. (10) Pappaert, K.; Vanderhoeven, J.; Dutta, B.; Van Hummelen, P.; Clicq, D.; Baron, G. V.; Desmet, G. J. Chromatogr., A 2003, 1014, 1-9. 10.1021/ac0502091 CCC: $30.25
© 2005 American Chemical Society Published on Web 05/13/2005
hybridized target strands at equilibrium and the maximal molar surface concentration of hybridized target strands,respectively; kon and koff represent the forward and backward hybridization reaction rate coefficient, respectively; and Cinitial represents the initial molar volumetric concentration of target strands present in the solution and showing that the larger the Cinitial, the larger the initial hybridization rates will be. From the consideration that the DNA hybridization is an equilibrium-dominated process with an equilibrium constant KA determined by the ratio of the kon and koff values
KA )
koff Hfinal ) kon (Hmax - Hfinal)Cfinal
(2)
and assuming probe excess conditions for the sake of simplicity (i.e., Hfinal , Hmax), eq 2 readily shows that if the equilibrium state is reached, which corresponds to the maximally achievable spot intensity, the number of hybridized strands is also always directly related to the final concentration of DNA in the sample. Knowing now that both the initial binding rates and the maximally achievable final spot intensities are proportional to the sample concentration, it can be inferred that a maximal sample concentration is beneficial during the entire course of the hybridization process. The following very simple design rule can, hence, be put forward: If a given mass of DNA sample is available, the fastest hybridization rates and final spot intensities will be obtained if the available amount of DNA dry mass is dissolved in the smallest possible buffer volume. To handle these small volumes, minimal volume hybridization chambers are needed, and the use of pumparound sample reservoirs and tubing should be avoided or at least minimized. If, at the same time, the total microarray surface has to remain constant and has to be fully covered by the entire sample, the minimization of the hybridization chamber volume can only be achieved by decreasing the thickness of the fluid layer covering the microarray spots. Sinc, in this case, the volume of fluid which is in direct contact with each individual target spot becomes very small, a strong convective transport enhancement is needed to continuously renew the sample at the spot surface. Similar conclusions have recently also been made from the study of a planetary centrifugal DNA hybridization enhancement system.13 A theoretical calculation underpinning the above argumentation has been presented in ref 14. As demonstrated in our earlier work on shear-driven flow,15-16 this type of flow is preeminently suited for the generation of largevelocity flows in thin fluid layers and could, hence, also be advantageous for microarray hybridization. Shear-driven flows simply exploit the dragging effect of a moving channel or chamber wall and display the unusual property that the established fluid velocities are independent of the thickness of the channel or fluid (11) Gadgil, C.; Yeckel, A.; Derby, J. J.; Hu, W.-S. J. Biotechnol. 2004, 114, 3145. (12) Sartor, M.; Schwanekamp, J.; Halbleib, D.; Mohamed, I.; Karyala, S.; Medvedovic, M.; Tomlinson, C. R. Biotechniques 2004, 36, 790-795. (13) Bynum, M. A.; Gordon, G. B. Anal. Chem. 2004, 76, 7039-7044. (14) Vanderhoeven, J.; Pappaert, K.; Dutta, B.; Vanhummelen, P.; Baron, G. V.; Desmet, G. Electrophoresis 2004, 25, 3677-3686. (15) Desmet, G.; Baron, G. V. Anal. Chem. 2000, 72, 2160-2165. (16) Clicq, D.; Vervoort, N.; Vounckx, R.; Ottevaere, H.; Gooijer, C.; Ariese, F.; Baron, G. V.; Desmet, G. J. Chromatogr., A 2002, 979, 33-42.
layer. They also do not exhibit any pressure drop or double layer overlap flow velocity limitation, as in the traditionally employed pressure- and electrically driven transport enhancement methods.17-18 In previous work, it could be demonstrated that flows of several centimeters per second can be generated in channels with a thickness ranging from 8 µm19 down to only 100 nm (see ref 16). Shear-driven flows can be either linear or rotational, but for the present purpose of enhancing DNA microarray analysis, it has been preferred to use a rotational shear-driven flow system based on a rotating circular microchamber. With this layout, any unnecessary prechannel and postchannel reservoir volumes are avoided. It should be noted that the preface “micro” in the word “microchamber” refers to the depth of the chamber and not to the diameter, which was always 2 cm. All other main experimental parameters have been varied in the present study: microchamber depth, rotation rate, and contact mode. The rotating microchamber experiments were always compared with (i) purely diffusion-driven experiments conducted in the same microchamber (to assess the gain originating from the increased lateral transport rates during the rotation experiments); (ii) purely diffusion-driven experiments conducted under a coverslip (which is the traditional method and which typically uses larger volumes than the rotating microchamber volume); and (iii) a commercial pump-around system, creating a very strong convective mixing but requiring an increased sample volume. All experiments have been carried out using DNA sample amounts typical for microarray analysis. The total amount of genetic material (expressed in dry mass of cDNA) applied to each system was always the same. EXPERIMENTAL SECTION DNA Hybridization and Detection Procedures. All hybridization experiments were conducted using conventional microarray procedures. To simplify the analysis, only six different probe molecules (six different cDNA fragments: Nras (716 bp), PolA (330 bp), Rad52 (683 bp), Nia12E (914 bp), Nia12F (1000 bp), and Nia12G (1400 bp)) were used. All probes were obtained by PCR amplification. After amplification, the fragments that were diluted in DMSO solution to denature them were arrayed at a concentration of 200 ng/µL on aminosilane-coated slides (Takara Bio Inc., Otsu, Japan) using a commercial Lucidea Array Spotter (Amersham Biosciences, Buckinghamshire, UK). After the spotting, the probe molecules were cross-linked by UV radiation at 50 mJ. The spots (diameter 100 µm and spaced 65 µm apart) were grouped in five different blocks, each block consisting of six different columns, each column consisting of 12 spots of the same probe DNA, and spotted in such a mode that the spots and blocks were spaced in a regular way across the microchamber (Figure 1). The third block lies exactly in the middle of the microchamber, whereas the second and fourth blocks were positioned ∼4 mm from the center of the microchamber, and the first and fifth blocks were positioned ∼8 mm from the center. Investigation of false positive hybridization events was performed by applying a sample mixture that only contained matching (17) Desmet, G.; Baron G. V. J. Chromatogr., A 2002, 946, 51-58. (18) Desmet, G.; Vervoort, N.; Clicq, D.; Gzil, P.; Huau, A.; Baron, G. V. J. Chromatogr., A 2002, 948, 19-34. (19) Desmet, G.; Vervoort, N.; Clicq, D.; Baron, G. V. J. Chromatogr., A 2001, 924, 111-122.
Analytical Chemistry, Vol. 77, No. 14, July 15, 2005
4475
Figure 1. Bird’s eye view schematic of the rotating microchamber hybridization setup, showing the rotatable sample chamber (1, dark gray) and the DNA microarray (2) with five spotted regions (3) covering the sample chamber during the operation of the device. The outer etched ring (4) only serves to limit the contact area between the rotating bottom substrate and the microarray slide surface.
Figure 2. Microarray layout and position of the spots of half a block. Each block consists of 6 columns of each 12-probe DNA spots (Nras, PolA, Rad53, Nia12E, Nia12F, and Nia12G, respectively).
DNA strands for three of the six different probe types: Rad52, Nras, and PolA (Figure 2). These target strands were prepared by performing a nick translation with a labeled dCTP (Cy5) on a sample mixture containing the different DNA fragments. These were subsequently purified and mixed with Huntsman hybridization buffer (50% formamide, 5 × SSC, 0.1% SDS, 100 µg/mL salmon sperm DNA) to obtain the desired sample concentration, which was, together with the amount of incorporated Cy5 molecules, measured using a NanoDrop spectrophotometer (NanoDrop Technologies, Montchanin, DE) prior to each experiment. The target strands were then denatured by heating them to 96 °C for 5 min before the hybridization was carried out at 42 °C, either on the microchamber system, on the ASP system (Amersham Biosciences, Buckinghamshire, UK), or under coverslip. Before each hybridization experiment, the slides were prewashed in 2 × SSPE/0.2% SDS at 26 °C for 30 min, rinsed with MilliQ water, and dried by centrifugation. After hybridization using the rotating microchamber system and under coverslip, the microarray slides were washed two times for 10 min and one time for 4 min in different sodium dodecyl sulfate (SDS)/sodium citrate (SSC) solutions (respectively, 1 × SSC/0.2%SDS, 0.1 × SSC/ 0.2%SDS, and 0.1 × SSC/0.2%SDS) at 56 °C and one time for 1 min in 0.1 × SSC before rinsing it for a single second with MilliQ water and drying the slides by centrifugation. Washing after hybridization in the ASP system is a fully automated process performed at 56 °C: Similar to the manual washing procedure for diffusion-driven and shear-driven hybridizations, the microarray slides are washed two times for 10 min, one time for 4 min, and 4476
Analytical Chemistry, Vol. 77, No. 14, July 15, 2005
one more time for 10 min in 1 × SSC/0.2%SDS, 0.1 × SSC/ 0.2%SDS, 0.1 × SSC/0.2%SDS, and 0.1 × SSC, respectively, before rinsing them for a single second with MilliQ water. The slides were then scanned at 532 nm using a commercial Array Scanner Generation III (Amersham Biosciences, Buckinghamshire, UK). Image analysis was performed using ArrayVision 7.0 software (Imaging Research, St. Catharines, ON, Canada) and Spotfire 8 software (Spotfire Inc., Somerville, MA). Rotating Hybridization Chamber System. Figure 1 shows the main components of the employed system for microarray hybridization on a rotating microchamber. The circular microchamber (diameter ) 2 cm, depth varying between 1.6 and 70 µm) was etched in the center of a round, flatly polished (flatness ) λ/4 at λ ) 512 nm) borosilicate glass wafer (Radiometer Nederland, The Netherlands) with a thickness of 6 mm and a diameter of 5 cm. The etching was carried out in a laminar flow cabinet using a 50% HF solution (Fluka Chemie GmbH, Buchs, Switzerland) for different time intervals, depending on the desired microchamber depth. After the etching, the chamber depth was measured using a Talystep step profiling apparatus (Rank Taylor Hobson Ltd., Leicester, UK). In a second etching step, an ∼150µm-deep annular zone was etched in the outer region of the substrate (cf. Figure 1) to minimize the contact between the microarray slide and the chamber substrate surface. The microchamber is sealed off by simply putting a microarray slide upside down (spots facing the bottom of the microchamber) on the spacer regions of the rotating disk. The sample volume is applied in the inner etched ring so that exposure to air and the accompanying evaporation problems can be avoided. The chamber substrates were subsequently hydrofobically coated according to the coating procedure described in ref 14: 2 h in a 1 M KOH solution (E. Merck, Darmstadt, Germany), 1 h in a 0.03 M HCl solution (Fluka Chemie GmbH, Buchs, Switzerland), and overnight in a methanol solution (Fluka Chemie GmbH, Buchs, Switzerland) with [3-(heptafluoroisopropoxy)propyl]trichlorosilane, 97% (Aldrich Chem. Co, Steinheim, Germany). Adding the hydrophobic coating to the microchamber surface was found to prevent the deposited sample droplets from spreading out across the chamber bottom surface and over the microchamber borders. It was also observed that a compact droplet shape promoted homogeneous filling of the microchamber during the deposition of the microarray slides. During hybridization experiments, the microchamber substrate was clamped in a home-built, circular, stainless steel holder, with a stainless steel rotation shaft attached to its bottom wall, which in turn was connected to an automated M-060 rotation stage (Physik Instrumente, Karlsruhe, Germany) equipped with an ActiveDrive DC stepper motor (Physik Instrumente, Karlsruhe, Germany) and controlled with NetMove420 software (Physik Instrumente, Karlsruhe, Germany) running on a MicroSoft Windows ME-controlled PC. The rotation shaft was running through the bottom of a home-built, circular holding cup machined in a piece of PVC and having inlet and outlet ports to allow the circulation of a temperature controlled water flow. The outlet port was positioned such that only the top few millimeters of the microchamber substrate were above the heating water level. The temperature of the water flow was controlled using a Julabo F32MD heating circulator (JD Instruments Inc., Houston, TX) and
was set at a constant temperature of 42 °C during all the experiments. Prior to each experiment, a sample droplet is manually applied to the bottom of the microchamber center with a pipet (Gilson Inc., Middleton, WI). Subsequently, a microarray slide is manually deposited on the microchannel substrate to close off the microchamber. During the deposition of the slide, it has to be ensured that it is kept as parallel as possible with the microchamber substrate to prevent the formation of air bubbles. During the rotation of the substrate, the microarray slide was held in place by four small metal pins arranged around the circumference of the PVC cup, one near every corner of the microarray, to avoid movement of the microarray slide. A load of ∼500 g was put on top of the slide to guarantee intimate microchamber substrate/ microarray slide contact and, as a result, to guarantee the nominal microchamber depth. After the hybridization experiment, the slide can simply be picked up from the substrate, since both are completely detached from each other. Diffusion-Driven Hybridization under Coverslip. The standard overnight hybridization procedure20 of the MicroArray Facility Lab was followed for the diffusion-driven hybridization experiments under coverslip. This procedure simply consists of pipetting a 30-µL sample volume, which was sandwiched between a microarray slide and a thin glass coverslip (2 × 5 cm) and sealed off with rubber glue. The slides were subsequently kept overnight (16 h) in an incubator at 42 °C. Hybridization Using the Automated Slide Processor (ASP) System. Hybridizations using the ASP system are performed in independently temperature-controlled chambers, allowing customized process parameters to be used for each slide. The system consists of up to five modules, each containing six chambers that clamp to slide surfaces. Labeled sample is injected via a septum injection port, and wash solutions are circulated from wash storage bottles via tubing to each chamber. More than 200 µL can be injected into each chamber. Chambers are constructed with chemically resistant polyetheretherketone (PEEK) and have a patented O-ring design. Wash solutions are pumped in and out of each chamber using a 1000-µL syringe pump. The flow rates can be adjusted with the ability to toggle flow back and forth through the chamber (10-50 µL). Clean, dry air is used to dry slides as well as to actively pump wash solutions back and forth across the slide surface.21 Microchamber Experiments. All experiments were conducted with the same mass of DNA sample and were compared with the conventional, purely diffusion-driven assays carried out under a coverslip and run overnight, as well as with an overnight pump-around hybridization ASP system (Amersham Biosciences, Buckinghamshire, UK). From the sample volume (30 µL) applied in the purely diffusion-driven coverslip experiments, it can be calculated that the liquid layer height during these experiments was ∼30 µm (coverslip surface ) 2 × 5 cm). The rotating microchamber experiments were carried out in systems with four different depths: one depth (d ) 70 µm) corresponding to an increase (i.e., approximately a doubling) of the fluid layer height with respect to the overnight coverslip control experiment and (20) Puska´s, L. G.; Zvara, AÄ .; Hackler, L., Jr.; Van Hummelen, P. Biotechniques 2002, 32, 1330-1341. (21) http://www5.amershambiosciences.com/aptrix/upp01077.nsf/Content/ Products?OpenDocument&ParentId)460778.
three depths (d ) 10, 3.7, and 1.6 µm) corresponding to a decrease (3-, 8-, and 14-fold reductions, respectively) of the sample fluid layer height. All rotating microchamber experiments lasted for exactly 30 or 60 min. For the continuous rotation experiments, six different rotation velocities were compared. For the discontinuous rotation experiments, wherein a sequence of exactly timed stop periods is alternated with a short rapid rotation of the microchamber over a predefined angle (12°), five different contact modes were considered (30 × 1 min, 15 × 2 min, 10 × 3 min, 6 × 5 min, and 3 × 10 min). The displacement angle was the same in all discontinuous rotation experiments and was selected such that the microchamber elapsed exactly one full circle in the 30 × 1 min displacement experiment. Each angular displacement between two successive waiting periods lasted for only 5 s and was, hence, negligible as compared to the total stop period. To keep the total contact time always exactly equal to 30 min, the 1-min stop periods in fact lasted for only 55 s, the 2 min waiting period lasted for only 115 s, and so on. All reported values have been obtained by averaging the intensities of the 60 matching PolA spots lying in the microchamber area. The data for Nras and Rad52 (not represented) showed a huge similarity to the results obtained for PolA. Each experiment was conducted in duplicate. The reported values are, hence, also averaged over the two experiments conducted for each set of experimental conditions. As indicated by the experimental variance bars (indicating the minimal and maximal obtained value) added to the result plots shown below, the largest deviation between two experiments carried out under identical experimental conditions was never larger than 9%, a variability which in DNA microarray analysis is considered very small. The overnight diffusion control was conducted five times. RESULTS AND DISCUSSION Comparison of the Hybridization Uniformity. Figure 3 compares the hybridization intensity across the entire chamber of a 30-min discontinuous rotation hybridization experiment with stop periods of 175 s and a 30-min diffusion-driven hybridization experiment in three different microchambers. As can be noted, the rotation-driven hybridization experiments yield a uniformity that is comparable to (and even slightly better than) the purely diffusion-driven experiments. The fact that the hybridization intensity does not vary significantly with the radial position in the chamber can be explained from the fact that in the discontinuous rotating experiments, the chamber was always displaced over 18° between two successive stop periods. It can be calculated that for spots positioned at a distance of only 1 mm from the center, the fluid is already displaced over a distance of 300 µm, implying that the liquid above these spots is already completely replaced by fresh sample liquid, as is the case for the spots lying more toward the border of the microchamber. Comparison of the Hybridization Intensity: Continuous Rotation versus Diffusion-Driven Experiments. Figure 4 compares the hybridization intensities obtained after 30 min of continuous rotation for various values of the rotation rate with that obtained after 30 min of purely diffusion-driven hybridization in the same microchamber. As can be noted, the hybridization intensity increases slowly with the angular rotation rate until a small optimum is reached around ωopt ) 1 rpm. Beyond this Analytical Chemistry, Vol. 77, No. 14, July 15, 2005
4477
Figure 5. Comparison between the hybridization intensities obtained after 30 min each of pure molecular diffusion and continuous rotation (ω ) 1 rpm) for the four considered chamber depths: 1.6 (black), 3.7 (dark gray), 10.0 (light gray), and 70.0 µm (white). The spot intensities for the continuous rotation experiments are disappointingly small, as compared to the purely diffusion-driven control.
Figure 3. Hybridization intensity obtained after 30 min each of diffusion-driven (A) and discontinuous rotation (10 × 3 min) hybridization (B) in 1.6-([), 3.7- (0), and 10-µm (2)-thick microchambers, showing that the shear-driven hybridization yielded a slightly better uniformity than the purely diffusion-driven hybridization experiments.
Figure 4. Hybridization intensity obtained in a 3.7-µm thin microchamber after 30 min of continuous rotation as a function of the chamber rotation rate. A small optimum is reached around ωopt ) 1 rpm, followed by a sharp decrease with increasing rotation rate. The dashed horizontal line represents the hybridization intensity level obtained for a 30-min purely diffusion-driven hybridization. Reported values are obtained by averaging the intensities of all matching spots lying in the chamber area.
optimum, a sharp decrease of the hybridization intensity is noted. It is assumed that above this angular velocity, the forces originating from the convective flow field are so large that they either disturb the binding process or increase the koff value. Our current experimental setup does not allow us to distinguish between both hypothesis. The reason the hybridization intensities depend only very weakly on ω in the range of small rotation rates is also not entirely obvious at present. As can be noted, the hybridization intensities in this range in fact do not differ much from those 4478 Analytical Chemistry, Vol. 77, No. 14, July 15, 2005
obtained under purely diffusion-driven conditions. This was a consistent trend in all conducted experiments. Repeating the rotation-versus-diffusion-driven conditions comparison for the other available chamber depths, it was always found that both conditions yielded comparable hybridization intensities. We were unable to conclude that one of both conditions yields a significant advantage over the other (Figure 5): in some cases, the intensities obtained by the purely diffusion-driven experiments were slightly larger; in other cases, the continuous rotation experiments yielded larger intensities. The fact that the rotation-driven experiments yield no significant advantage over the purely diffusion-driven experiments should, nevertheless, be considered as a rather disappointing result. Part of the explanation certainly comes from the fact that in a 30-min experiment, where for an estimated diffusion coefficient of Dmol ) 10-11 m2/s, the probe strands only need to come from a circular region with a radius of ∼200 µm (value estimated from Einstein’s law of diffusion (l 2 ) 2Dmolt). The diffusion-driven experiment is not so diffusion-limited as during a typical overnight experiment, in which the lateral transport diffusion limitation is much more pronounced because there, the diffusion limitation affects a circular region with radius of ∼1 mm. However, given that the continuous rotation prevents the formation of a depletion zone by continuously supplying fresh, maximally concentrated sample to the probe spots, this should be somewhere noticeable in the obtained hybridization intensities. As can be noted from Figure 5, however, it clearly is not. At the present time, the most plausible explanation for the poor rotation experiment results is that the rotational movement also includes some kind of a counteracting effect. Surveying all possible effects, it seems most likely that at least one of the successive steps involved in the hybridization process (collision between probe and target strand; formation of a binding nucleus involving three consecutive matching base pairs and the subsequent zippering reaction) is hindered by the velocity field. Comparison of the Hybridization Intensity: Discontinuous versus Continuous Rotation Experiments. To verify the hypothesis of hindered hybridization during the continuous rotation experiments, a series of discontinuous rotation experiments has been conducted. Comparison of 30-min hybridization
Figure 6. Comparison between the hybridization intensities obtained after 30 min of discontinuous rotation hybridization for the four considered chamber depths, 1.6 (black), 3.7 (dark gray), 10.0 (light gray), and 70.0 µm (white), with stop periods of 1, 2, 3, 5, and 10 min. As can be noticed, the optimal stop period in the 1.6-µm-deep microchamber is 1 min, but a 3-min stop period should be selected for the other microchamber depths.
Figure 7. Comparison between the hybridization intensities obtained after 60 and 30 min of discontinuous rotation hybridization (optimal stop periods; cf. Figure 6) for the four considered chamber depths, 1.6 (black, optimal stop period: 1 min), 3.7 (dark gray, optimal stop period: 3 min), 10.0 (light gray, optimal stop period: 3 min,) and 70.0 µm (white, optimal stop period: 3 min), diffusion-driven overnight (16 h) hybridization under coverslip and 12-h overnight hybridization in the ASP system.
experiments with stop periods of 1, 2, 3, 5, and 10 min is given in Figure 6. An optimal stop period can be noticed for each microchamber, and this optimal stop period becomes shorter as the microchambers become smaller. These findings are in agreement with one’s physical expectation, since local target depletion will occur sooner in smaller microchambers, and therefore, the required frequency of sample refreshment will be higher than in larger microchambers. Figure 7 shows that whereas the continuous rotation experiments yielded intensities which were never significantly larger than the purely diffusion driven conditions, the hybridization intensities obtained during the discontinuous rotation experiments are, without any exception, at least 2-fold larger. The introduction of stop periods, during which the disturbing effect of the velocity field is switched off, hence, clearly has a beneficial effect. From a pure diffusion and reaction kinetics point of view, it would, however, have been expected that the continuous rotation experiment would always yield the largest hybridization rates, because a continuous refreshment of the sample liquid always results in a larger average target concentration above the probe spots than with a discontinuous supply. The explanation for the larger hybridization intensities of the discon-
tinuous rotation experiments should, hence, be related to the disturbing action of the velocity field. To verify this hypothesis, we performed an experiment with a similar discontinuous displacement pattern, but instead of keeping the microchamber perfectly stationary during the stopping period, the chamber was continuously oscillated back and forth, with a rotation angle amplitude of 1.2 degrees and with a rotation rate of ω ) 1 rpm, that is, the same rate as the one used in the continuous rotation experiments. By our doing so, the hybridization intensities were decreased by ∼54% (data not presented), that is, significantly larger than the largest deviation between any two experiments conducted under the same conditions in the present study. The significant reduction of the hybridization intensity in the oscillating microchamber experiment, hence, clearly supports the hypothesis of the disturbing velocity field. Further research covering a broader range of oscillation angles and oscillation velocities is needed to fully understand and describe the clearly velocity-field-related disturbance effect in the hybridization experiments performed in the present study. The experimental variability bars also indicate that the continuous rotation experiments are far less reproducible than the discontinuous rotation experiments. Although we have at present no clear-cut explanation for this observation, it certainly constitues another important advantage of the discontinuous operation mode. The reader should note that the final equilibrium is typically reached in a period of a few hours, as can be noticed from the time series shown in ref 14 and as can be noticed from the relatively small difference between the 30 and 60 min discontinuous rotation for the 3.7-, 10-, and 70-µm chambers. It is only for the 1.6-µm channel that the final equilibrium is not yet achieved in the first hour. The differences between the results obtained in the different microchamber depth cases shown in Figure 7 are also very interesting. First of all, the advantage of using the thinnest possible chamber is obvious, as was already theoretically predicted.14 The 1.6-µm-deep microchamber, for example, yields a 5-7 times larger final hybridization intensity than the 10-µm-deep microchamber. This roughly compares with the concentration factor obtained when putting the same amount of DNA into a 10/1.6 = 6 times thinner fluid layer. A similar conclusion can be made by comparing the 1.6-µm microchamber with the 3.7-µm microchamber, where the concentration factor amounts up to ∼2.5. Comparison with the 70-µm microchamber is more difficult, because in this case, the obtained intensities flirt with the background signal. Nevertheless, plotting the largest intensities for each considered microchamber depth versus the inverse of the microchamber depth, which is a direct measure for the concentration factor, shows a strongly increasing relation between the hybridization intensity and the concentration factor (Figure 8). Obviously, the gains stemming from this concentration factor (spanning a range of nearly a factor of 40) are much larger than the gains originating from the rotation itself, as can be noted from the fact that the difference between the purely diffusion-driven experiments and the best possible discontinuous rotation experiments is only maximally a factor of 4. Figure 8 also shows that the optimal contact modes depend on the microchamber depth. The thinner the microchamber, the shorter the optimal stopping periods. This is in agreement with Analytical Chemistry, Vol. 77, No. 14, July 15, 2005
4479
and the relatively thick hybridization chamber. This volume is obviously much larger than the volumes applied in the microchamber system and in the coverslip experiments (requiring 30 µL of buffer volume). Although the pump-around sample transport in the ASP system circumvents the diffusion limitation governing the hybridization rates in the coverslip system, this obviously is insufficient to compensate for the required sample dilution of a factor of ∼7.
Figure 8. Maximal spot intensity versus the inverse of the microchamber depth for 30-min discontinuous rotation ([) and 30-min diffusion-driven (0) experiments, showing that in the investigated 30min time frame, the shear-driven convective flow transport approximately yields a 4-fold gain in spot intensity, as compared to the purely diffusion-driven control.
one’s intuitive expectations based on the shorter radial diffusion times that are needed if the microchamber depth decreases. Comparison of the Hybridization Intensity: Influence of Hybridization Time and Comparison with Other Hybridization Systems: Coverslip and ASP. In the comparisons discussed above, all experiments have been performed in the microchamber and in a 30-min time period. In the present section, these experiments will be compared with experiments performed in existing hybridization systems, for which typically overnight waiting periods are required to obtain sufficiently high hybridization intensities. As can be noted from Figure 7, the gain in hybridization intensity between the 30-min discontinuous rotation in the thinnest microchamber and the overnight diffusion-driven hybridization under coverslip is already a factor of ∼4. Waiting for 30 more min in the discontinuously rotating microchamber system to obtain a total hybridization time of 1 h, the gain with respect to the diffusion-driven overnight coverslip experiment amounts to almost a factor of 6. This is quite impressive when one takes into account that the hybridization time in the discontinuous rotating system is 16 times shorter. From the findings in Figure 5, we can attribute this gain largely to the increased sample concentration obtained when switching from the 30-µm liquid layer height in the coverslip experiment to the 10-, 3.7-, and 1.6-µm microchamber depths. A clear indication of the fact that an enhancement of the transport rates is not capable of overcoming a reduction of the sample concentration can be found in the fact that experiments in the 70-µm-deep microchamber obviously yield a much lower hybridization intensity than the coverslip experiment with its 30 µm liquid layer height, despite of the applied convection enhancement and despite the fact that equal amounts (in terms of dry mass) of DNA have been used. The poor performance of the 70-µm-deep microchamber also explains the poorer results obtained in the pump-around ASP system. Here, the available amount of 120 fmol of DNA had to be dissolved in 209 µL of buffer to fill the entire recirculation loop
4480
Analytical Chemistry, Vol. 77, No. 14, July 15, 2005
CONCLUSIONS The present study has demonstrated that if a given amount (expressed in mass or number of moles) of DNA is available, each attempt to increase the DNA microarray hybridization intensity or the hybriddization speed should preferentially involve a reduction of the total volume of the hybridization system, because this allows an increase in the initial sample concentration, which is the key factor controlling both the initial hybridization rate and the equilibrium spot intensity. Due to the induced dilution, hybridization systems requiring an enlarged sample volume to generate an enhanced convective transport cannot be expected to yield a large final equilibrium spot intensity, despite the enhanced transport rates. The dilution also acts counterproductively in the initial phase of the hybridization. The currently proposed rotating microchamber system allows combination of a miniaturization of the sample chamber with the creation of a strong lateral convective transport. In only 30 min, the system can yield hybridization intensities that are ∼4 times larger than a traditional overnight experiment under a coverslip and more than 5 times larger than in an experiment using the same amount of DNA conducted in a commercial pump-around hybridization station. One of the essential keys to this gain seems to be that the convective displacement is performed in a discontinuous mode, because the continuous rotation mode only yielded intensities that were of the same order as those obtained in the purely diffusion-driven mode. Another indication of the existence of a velocity field-related disturbance of the hybridization process is the steep decrease of the hybridization intensity noted if the rotation velocity exceeds a certain limit in the continuous rotation experiments. It was found that the optimal condition for hybridization in the microchamber system is to perform the hybridization in the shallowest microchamber in a discontinuous contact mode with a stop period of 1 min. The hybridization time can be varied between 30 min and a few hours. Obviously, there is a limitation to approaches pursuing a maximal concentration of the DNA sample, since there is a dissolution limit. With the typical surface waviness of commercial microarray slides, it also seems unfeasible to think of fluid layers that are thinner than 1 or 2 µm and still have a sufficiently uniform thickness. Received for review February 2, 2005. Accepted April 20, 2005. AC0502091
