Sequential DNA hybridisation assays by fast micromixing

The prospects of performing DNA hybridisation assays in a novel sequential scheme are explored in this article.

It is based on recording the kinetics of hybridisation on a microfluidic device measuring only 10 by 5 mm.

It contains a split channel system for fast mixing and a subsequent meandering channel to observe the evolution of the mixture by optical means.

The problems of diffusion limitations in the laminar flow regime are overcome by reducing the average diffusion distance to a few micrometers only.

DNA oligomers (20-mers) of different sequences were injected on the chip for mixing.

The detection of hybridisation was based on the fluorescence of DNA-intercalating dyes.

Two modes of operation were investigated.

First, the samples were injected into the micromixing device at a high flow rate of 40 µl min−1.

When the sample passed through the actual micromixing unit, the flow rate was reduced to allow for measurement of fluorescence levels at various steady-state reaction times in the range of 2–15 s, as defined by the channel geometry.

Using this continuous flow approach, photobleaching of fluorophores could be avoided.

In a buffer containing 0.2 M NaCl, 2 base-pair mismatches could routinely be detected within 5–20 s.

Single base-pair mismatches were successfully identified under low salt conditions.

In the second mode, the flow was completely stopped and the evolution of the total fluorescence signal influenced by the hybridisation of oligomers and photobleaching was observed.

Whereas the sequence-dependent effects remained unchanged, the assay times between the mixing of two oligomers and clear identification of their hybridisation properties could be reduced down to a maximum of 5–7 s, in some cases even below 1 s.


DNA hybridisation reactions are at the core of biological and medical analysis.

Performed in vast numbers, one of the most valuable platforms are DNA microarrays where probe oligomer sequences representing genes of interest are immobilised and incubated with a mixture of unknown target DNA.

Fluorescent labelling techniques are predominantly used to identify the sites where probe and target sequences match.

The performance of DNA microarrays is impressive as integration with microsystem technology progresses to increase the number of probe oligomer per unit area.1,2

Thereby, the information density is increased considerably and enables large-scale genomic analysis such as gene expression profiling of whole organisms.3–5

Although it seems that many application needs in industrial research and high-throughput screening can be satisfied with microarrays, there is an on-going interest in alternative DNA hybridisation technology, e.g. hybridisation-based sensors that depend on other sensing devices like fiber optics,6 surface plasmon resonance sensors7 or acoustic wave sensors.8

Especially attractive solutions were presented particle-based probe oligomer immobilisation schemes.9,10

Yamashita et al. proposed a microfluidic system based totally on laminar flow, where the DNA hybridisation occurs at the interface between the streams.

By tailoring the geometry of channel bends on their chip, only the newly formed and fluorescence-labelled DNA duplexes migrate into the bulk region of the second laminar stream for detection.11

Another interesting contribution from Tamiya's group suggested the use of silicon micromachining to create platelet-like particles for immobilisation whose shape characteristics alone could code for different functionalisation and still be easily identifiable in mixtures.12,13

However, there is a distinct lack of scale-up potential towards high-throughput applications in most of these approaches.

In some cases, the beads coated with probes could be sorted into a silicon frame and became individually addressable.14

We wish to present a different approach with potential for high-throughput application by combining purely solution-based DNA hybridisation reactions with the use of a microfluidic system.15,16

This device overcomes the mixing problem that is inherent to most microfluidic systems which work in the laminar flow regime and is based on the fine distribution of two laminar inlet flows (Fig. 1).

DNA hybridisation assays are performed in series by injecting the sample solutions one after the other and mix them rapidly on the chip.

Single nucleotide polymorphisms (SNP) are detectable.

If such a scheme is to be used for a large number of analyses in sequence, the single assay has to be performed within a short time to become comparable to the enormous multiplexing performance of microarray technology.

The prospect of decreasing the time of one assay to the order of 1 s is presented.

Potentially, this technology could allow for many thousand assays within just a few hours.

By using solution-based DNA chemistry,17,18 the need for immobilisation schemes is eliminated.

Due to restrictions of molecular diffusion near surfaces, the kinetics of heterogeneous hybridisation assays are slower, as characterised in literature.19–22

The detection of DNA hybridisation is based on the fluorescent dye PicoGreen that selectively enhances its fluorescence upon binding to double-stranded DNA (dsDNA).

Typically, such intercalating reagents are used for DNA quantification23–27.

Experimental procedures

Instrument setup

The setup consisted of a Leica DM IL inverted microscope with fluorescence detection equipment as depicted in Fig. 1c.

The glass-silicon-glass sandwiched micromixing device was secured in a custom-made aluminium holder by pressing PEEK capillaries and silicone ferrules (Upchurch Scientific, Eugene, Oregon, USA) into the access holes of the microchip.

There were two light sources installed, one a halogen lamp for bright-field imaging, the other lamp a high-performance Hg-lamp (50 W) for fluorescence excitation.

Various neutral density filters could be inserted before the fluorescence filter set in order to attenuate the excitation light.

Connected to the microscope was a photomultiplier tube (PMT) Cairn Integra (Cairn Research Ltd., Faversham, UK).

Its preamplifier analog voltage output was connected to a National Instruments PCI DAQ card for direct digitalization of the PMT data.

Liquid pumping was done using a Harvard PhD 2000 syringe pump equipped to hold up to two Hamilton syringes.

One input line was connected to a Rheodyne EV750-100 6-port injection valve (Rheodyne LLC, Rohnert Park, California, USA) which was equipped with a 20 µl sample loop.

A video camera for capturing live and still images was installed on the top port of the PMT mirror assembly.

By the use of dichroic mirrors which split the light at 600 nm, the camera remained connected in parallel to the PMT so that detection window and positions could be checked shortly before and during measurement.

The detection window was defined by an adjustable rectangular aperture of 160 by 200 µm.

The fluorescence light was collected at highest magnification through a 32× objective (NA = 0.4).

The whole setup was computer-controlled using LabView and was placed in a cloth-clad cubicle in order to avoid stray light.

Micromixer flow visualisation

A solution of 0.385 mM fluorescein isothiocyanate (FITC) dissolved in MeOH was used in a mixing experiment with water.

The components were mixed at a total flow rate of 120 µl min−1.

DNA hybridisation assays

Gel-filtered DNA oligomers were synthesized by Eurogentech, Romsey, UK and used as received.

The lyophilised DNA oligomers were reconstituted using sterile and RNAse free water.

The following sequences were used for the present work, a base sequence A: 5′-GTTTATTAATGCGGCCCGCG-3′, its perfect match A′: 5′-CGCGGGCCGCATTAATAAAC-3′, as well as sequences containing increasing numbers of base pair mismatches, A-1: 5′-CGCGGGCTGCATTAATAAAC-3′, A-2: 5′-CGCGAGCCGCATTGATAAAC-3′ and A-5: 5′-CGTGAGACGCACTGATATAC-3′.

Location of mismatches are indicated in bold-oblique letters.

The hybridisation properties of various oligomer mixtures were also confirmed by DNA sizing on an Agilent 2100 bioanalyser.

To ensure temperature stability and constant lighting conditions etc., each experiment consisted of different injections done in the same run within a few hours.

PicoGreen was purchased from Molecular Probes Inc., Leiden, Netherlands.

For a typical experiment, 3 ml of buffer solution containing 34.5 mg NaCl (0.2 M) and 8.9 mM tris-borate/0.2 mM EDTA at pH 8.3 were prepared and put under vacuum for degassing.

To this buffer, the appropriate amount of oligomer and PicoGreen dye (typically 5 µl of PicoGreen diluted in water 1:10) was added shortly before the experiment.

The first syringe connected directly to one inlet usually contained 450 µl of a 0.8–8 µM (5–50 µg ml−1) target oligomer solution, the syringe supplying the other inlet contained 450 µl pure buffer and was used as carrier stream through the Rheodyne injection valve.

Probe oligomer solutions for injection were at the same concentration unless indicated otherwise.

Between experiments, the chip was filled with liquid in order to avoid air bubbles forming and occasionally flushed with buffer and EtOH/MeOH or 2-propanol to keep the microchannels clean.

Calibration experiments to assess the detection limit of dsDNA within a large background of ssDNA were performed.

All solutions were prepared with excess oligomer to mimic the initial stages of hybridisation with the corresponding amount of unreacted oligomer.

A 5 µM stock solution of dsDNA (A:A′) was diluted to 1, 2, 20, 40, 80, 160 × 10−8 M concentrations to which another equivalent of A was added to bring the total oligomer concentration in the system to a total of 8 µM (comprising dsDNA and ssDNA).

This time, the solutions were injected on the microchannels through the outlet since the micromixing function was not necessary.

Fluorescence levels were recorded for both stopped flow and continuous 40 µl min−1.

As the results of the calibration run indicate, dsDNA concentrations of 200–400 nM could readily be detected (not shown).

Where applicable in terms of comparable experimental conditions, a linear fit to this calibration was used to convert the PMT voltage scale to concentration units.

The PMT data was first normalised to the measured basic fluorescence level of one oligomer/PicoGreen solution mixed 1 : 1 with buffer.

Results and discussion

The high efficiency of mixing in the laminar flow regime is brought about by splitting each of the two inlet streams into 16 smaller substreams which then are joined and recombined (Fig. 1a).

Details concerning the design and microfabrication procedures of the device are given elsewhere.28

Briefly, channels and through-holes were micromachined into a silicon wafer from both sides by deep reactive ion etching (DRIE) using three photolithography masks.

The silicon wafer then was sandwiched by anodic bonding of a plain glass wafer at the bottom and a glass wafer featuring sand-blasted access holes on top and finally diced to yield single chips.

The channel surfaces had not been functionalised in any way.

Hence, a layer of adsorbed oligomers on the open SiO2 surfaces had to be expected.

Due to the excitation light source containing significant intensity in the UV range, an irreversibly bound monomolecular layer is likely to have formed.

However, due to the dynamic operation mode featuring only short interruptions of a continuous flow, the immensely larger detection volume, the DNA concentrations used and frequent flushings, no evidence of adverse effects on the fluorescence signals was found.

In addition, the intense excitation light source was constantly illuminating the channel system (see positioning of aperture in Fig. 1c), thus quenching any remaining fluorescence from possible adsorbates by photobleaching.

With the present system geometry, Reynold numbers of Re = 1.6 to 4.8 have been calculated for the flow rates of 40 to 120 µl min−1 of aqueous oligomer solutions used within this project.

In the absence of turbulence at Re < 2,000, the only transport mechanism for mixing is diffusion.

The splitting of flows therefore reduces the average diffusion distance to approx.

10 µm, see also Fig. 1b where micrographs of the stratified laminar flow pattern after mixing fluorescein isothiocyanate (FITC) with water are shown.

A meandering channel allows for downstream observation of the flow, e.g. after 30 ms the laminar pattern has been washed out by diffusion.

The splitting of flows leads to a significant decrease in mixing times down to the order of ms, since the diffusion distance x follows a root law as function of diffusion coefficient D and time t

A literature overview of some diffusion properties in water for various molecules of interest are summarised in Table 1.

Smaller molecules like FITC easily cover 7–8 µm within 100 ms, whereas DNA oligomers and double-stranded pieces of 20 bp length take up to approximately 1 s for the same distance.

This is due to their diffusion coefficient being approx.

10× smaller.

It is important to note that these values may vary depending on the actual experimental conditions like buffers, pH, temperature, DNA sequence and functionalisation etc.

Therefore, these literature values were used as a guideline only.

All the characteristics addressed above led to the decision to vary the flow rate in the present work by mixing probe and target oligomer quickly using a high flow rate and subsequently slowing down or stopping the flow to allow for enough time to observe the evolution of the DNA hybridisation reaction.

The downstream channel volume of 480 nl would only allow for a time window of 600 ms at typical flow rate of 40 µl min−1.

Therefore, these reductions of the flow-rate or stop-flow schemes were required.

Since the detection of hybridisation is based on intercalating dyes which increase the quantum yield of fluorescence upon interacting with dsDNA, the following contributions to the overall fluorescence level can be expected: (i) the fluorescence of the intercalating dye in solution without DNA interaction, (ii) interaction of the dye with ssDNA, (iii) intercalation into oligomer-dimers, (iv) intercalation into dsDNA.

Of these contributions, (i) should be negligible29 (ii) is expected to remain constant throughout the experiments, or, in the case of matching oligomers, is expected to become negligible as ssDNA is consumed by hydridisation, (iv) desirably the strongest and (iii) will have to be controlled carefully, since it may vary depending on the probe sequences of the oligomers in use.

Newer intercalating dyes like PicoGreen seem to fulfil these requirements, especially since the brightness increase upon binding to dsDNA is reported to be 2 orders of magnitude.

Continuous-flow experiments, detecting SNPs

A typical experiment mixing oligomer A with A′ is depicted in Fig. 2a.

At 40 µl min−1 the injection valve is switched to its inject position for 10 s, injecting approximately 3.3 µl of probe oligomer solution.

Concurrently with switching the valve back to the load state, the flow is abruptly reduced to a flow rate of 2 µl min−1.

After 10–20 s another steady state of the fluorescence signal is reached.

For demonstration of the repeatability of this flow rate change, the flow rate is temporarily brought back to 40 µl min−1 for just a short moment (at 45 s).

The signal recovers to the same steady-state level after that.

The experiment is ended at 80 s by flushing the system at 40 µl min−1.

Both oligomer concentrations were 2.4 µM.

The observed difference in fluorescence intensity before and after the flow rate reduction step, ΔPMT, was evaluated to detect oligomers that match or, respectively, mismatch in varying degrees.

The ΔPMT values of sequence-dependent experiments are shown in Fig. 2b.

The same experiment as described above was performed with various mismatch combinations.

Matching oligomers A/A′ and 1-mismatch A/A-1 could not be distinctly resolved.

This is attributed to the high NaCl concentration of 0.2 M which was initially chosen to maximise the reaction rate at fixed oligomer concentrations.

The error of multiple injections was dominated by the fluctuations of the mercury-arc lamp that was used.

However, if the NaCl content is reduced to 40 mM, the stringency of hybridisation is increased.

Using these conditions, single base-pair mismatches A/A-1 were successfully distinguished from A/A′ at 4.8 µM oligomer concentrations (Fig. 2c).

As both syringes were driven by the same pump, the mixing ratio was fixed to 1:1.

However, due to individual pressure variations upon flow rate change, deviations from this 1:1 ratio had to be taken into account.

Therefore, the initial response always carries some unwanted spike which was characterised by observing the effect of flow-rate change without injecting a sample.

In those cases, the corresponding fluorescence response of only the change in flow rate without sample injection is also displayed.

The difference between the two steady-state levels is indicative of the hybridisation reaction rate.

As Fig. 2c also shows, the higher concentrations used in the single base-pair mismatch experiments yields a much shorter settling time of only 5 s to reach flow equilibrium.

Stop-flow experiments

A sharp increase in fluorescence brightness was obtained when injecting the complementary 20-mer A′ to the basic sequence A and stopping the flow.

The injection time of 10 s was again chosen in order to fill the channel downstream of the mixing unit long enough to perform several stop-flow measurements.

A series of three injections is shown in Fig. 3a for [A] = constant = 8 × 10−6 M and [A′] = 8 × 10−6 M, resp.

4 × 10−6 M and 1.6 × 10−6 M. The curves each show the behaviour after stopping the flow 3 times and give an impression of the reproducibility.

From top to bottom, the dilution of [A′] is shown.

Clearly, the steady-state fluorescence level follows the dilution series quantitatively.

The hybridisation kinetics of DNA oligomers can be described as a reaction following second order kinetics with a rate constant k1 of approximately 5.7 × 105 M−1s−1.18,30

The time-law of dsDNA formation for equal oligomer concentrations is given by where c0 is the initial concentration of both oligomers and k1 the rate constant.

Calculation of the 90% conversion value yields a value of t90 = 1.98 s.

Therefore, considering the specific conditions used in the experiments, the hybridisation should be faster than the mixing process by diffusion.

One expects to measure the response of the mixing system rather than the hybridisation kinetics themselves.

As Fig. 3a confirms, the times after the stopping of flow required to reach a steady fluorescence level do not change by diluting the concentration of one oligomer.

Linearisation to the root law of diffusion (1) of the time-dependent response shows that the hybridisation is indeed in the diffusion-limited regime for 20 s after which the increase in fluorescence deviates from the root law (Fig. 3b).

Possible explanations for this deviation are the completion of the mixing process or the onset of photobleaching as the intense excitation light is constantly focussed on the same volume element of the liquid after halting the flow.

The t90 values scale linearly with concentrations.

Hence, it has to be expected that using lower concentrations will eventually lead to a kinetic limitation of the response.

If this micromixing/stopped flow scheme is going to be used for genomic analysis, the dependence of the response to variations in sequence will be crucial.

Examples of sequence series are given in Fig. 4.

The A sequence was provided to the first inlet and the other oligomers were injected via the Rheodyne valve to the second inlet with injection times of 10 s.

A′ and A-1 containing one mismatched base pair coincide, whereas the first differences become evident for the two-mismatch oligomer, A-2.

In the case of A-5, the fluorescence increase is distinctly lower reflecting the thermodynamic equilibrium being shifted to the oligomer side.

True single-nucleotide-polymorphism (SNP) detection could again not be achieved due to the NaCl content (0.2 M).

However, the same way as control probe sequences containing an additional mismatch close to the target mismatch are included in microarray applications, such an approach could equally be applied for the proposed micromixing approach.

Further variations in buffer composition also lead to less ‘mismatch-tolerant’ and more stringent hybridisation conditions.

High-throughput application of this scheme demands for a quick succession of assays, e.g. another assay every 1–5 s.

It is not necessary to wait for the hybridisation to complete for obtaining the information whether two particular probe to target sequences match.

This will constitute an optimisation problem between fast detection and the resulting sequence-related levels of confidence.

Parameters to consider are the diffusion-controlled mixing or the kinetics of hybridisation and the sensitivity of detection, especially the detection of small dsDNA amounts in a large background of oligomers.

A true high-speed approach is shown in Fig. 5, where multiple stops on A mixed with A′, and A mixed with A-5 are depicted.

Particular to this experiment was the removal of all ND filters to have the highest excitation light intensity that the setup permitted.

Photobleaching effects now become very significant and the hybridisation/intercalation process produces just enough additional fluorescence to temporarily outweigh the photobleaching.29,31

Therefore, the signal obtained only increases in the case of matching oligomer sequences.

In the case of mismatching sequences, A-5 to A, only a decay of fluorescence is recorded.

First, such a scheme could provide for a signal with less ambiguity than the previous ones shown before by exhibiting a clear increase/decrease of fluorescence brightness.

Second, the magnified extracts that show the onset of the signal upon halting the flow are only 1 s, clearly indicating the possibility to discriminate matching/nonmatching situations with a signal change in the range of 10% within the first second of observation.

We have demonstrated an alternative approach for performing DNA hybridisation assays in sequence.

The enabling tool is a microsystem of passive mixing channels with subsequent observation channel.

DNA hybridisation at high concentrations was found to be diffusion-limited.

Sequence dependent signals down to SNP-level were successfully recorded.

A mode of operation suitable for high-throughput applications was found, where the signal shows within 1 s whether two oligomers match.

Whereas technological difficulties such as the periphery providing a steady stream of samples still need to be addressed, we believe this method could provide for a very flexible and versatile platform for genomic analysis.


M. Heule thanks for funding, by a postdoctoral fellowship award, the Swiss National Science Foundation (SNF).

Upchurch Scientific is acknowledged for providing a reference chip holder.