1. Introduction

A current research trend in molecular detection is the development of transduction mechanisms that do not require fluorescent labeling of the sample. Even though fluorescence-based detection is highly sensitive (to the level of a single molecule Ref [1]), it has the disadvantages of requiring extra steps in sample preparation and the labeling can influence the binding characteristics of the target molecule, Label-free methods typically rely upon the shift in an electrical, mechanical, or optical resonance in proportion to the concentration of one or more species bound to a surface, which may result through the process of molecular capture. Traditional techniques include the quartz crystal micro balance (QCM) Refs [2, 3] and surface plasmon resonance (SPR) Refs [4, 5]. With SPR, for example, an optical resonance exists at a metal-dielectric interface which is sensitive to dielectric properties near the surface via the evanescent wave. These techniques (SPR in particular) are popular for protein binding studies, but are less effective for DNA due to the smaller molecular weight of sequence lengths in common use; sensitivities for short sequences (∼20-mers) are typically in the nM range Ref [6], but sub-nM sensitivities have been reported Ref [7]. The sensitivity of these techniques also scale poorly when the size of the capture spot is reduced. As a result, numerous label-free transducers have been developed based upon localized surface plasmon (LSP) resonances in metal nanostructures Ref [8, 9, 10], optical resonance in dielectric resonators and microcavities Ref [11, 12], and mechanical resonances in micro- and nano-cantilevers Ref [13], for example. These methods promise to provide improved sensitivity and/or greater transducer density than QCM or SPR, but it is not clear that they will achieve the routine performance levels of fluorescence detection (with readout sensitivities in the 1-10 molecules/μm² range using commercial microarray substrates and scanners Ref [14]). Further, work continues on new, more sensitive, methods of fluorescence-based detection.

We have developed a technique that takes advantage of fluorescence readout, but does not require the sample to be fluorescently labeled. This technique, termed Competitive Displacement Detection Method (CDDM), uses a known “competitor” molecular species that is fluorescently labeled and spiked into the sample. The binding kinetics of the competitor provides quantitative information about the unlabeled molecules of interest via the mechanism of competitive displacement Refs [15, 16, 17]. The only requirements of the technique are that: 1) the competitor molecule must produce a signal that is distinguishable from all others in readout (thereby making fluorescence highly convenient), and 2) signal acquisition must occur in real time. These requirements can be met either with solution or surface capture. In solution, real-time fluorescence readout can be obtained with fluorescence resonance energy transfer (FRET) based mechanisms Ref [18] to isolate binding events from background. At a surface, evanescent wave excitation can be used via a total-internal reflection fluorescence (TIRF) arrangement Ref [19], a dielectric waveguide Refs [20, 21], or surface-enhanced fluorescence in a metallic film or nanostructure Refs [22, 23, 24, 25], for example. Nanoparticle scattering labels Refs [26, 27] could also be used on the competitor as an alternative to fluorescence. In this paper, we demonstrate the technique for DNA detection using evanescent-wave excitation from a thick planar waveguide.

2. Competitive displacement mechanism

In order to understand the mechanism of competitive displacement, consider the chemical reaction of two different species (target and competitor) to a single probe species:

where B represents bound concentrations, k_a association rate constants, C target solution concentrations, R_t concentration of probe molecules, and k_d dissociation rate constants. In the case of surface capture, mass transport from solution to the surface can be described by Fick’s Law

where D is the diffusion coefficient, assumed the same for the two species, and convective transport is neglected. Under the equilibrium condition (i.e. all time derivatives are zero), the occupation fraction of target and competitor to the probes is

assuming that solution concentrations are not significantly depleted by binding to the probes. Because association rates are always connected with their respective solution concentrations, the dynamic range of selectivity (i.e. the ability to discriminate one species from the other) is determined by the dissociation rates Ref [16].

Of a less general, but highly illustrative, nature, Eqs. 1 and 2 can be solved, assuming C constant over all time (the so-called well-mixed case):

where the α, β, γ, and δ coefficients originate from the rate constants and concentrations of the target and competitor species, and α + β = Ψ and δ - γ= Ψ_comp. For the target, Equ. 6, the exponential terms add to produce a monotonically-increasing curve; however, for the competitor, Equ. 7, non-monotonic behavior is possible. At early times, the τ ₂ term controls a monotonic increase, and, during displacement (i.e. t ≫ τ ₂), the τ ₁ term controls with exponential decrease. Although not shown explicitly from Eqs. 6 and 7, the kinetics of the competitor strongly depend on the concentration of the target species, and hence forms the basis of CDDM.

For the simulations and experiments, we have chosen a simple model system consisting of 20-mer sequences: target CGAGGGCAGCAATAGTACAC, competitor CGAGGGCAGCATTAGTACAC which differs from the target by a single base, and probe that is a perfect complement to the target. During simulations we assume association rates k_a = k_a,comp = 10⁶ M^-1s^-1, dissociation rates k_d = 3.4 × 10^-6 s^-1 and k_d,comp = 3.7 × 10^-3 s^-1, a diffusion coefficient D = 1.3 × 10^-10m²s^-1, and probe concentration R_t = 10^-11 M∙m. These parameters produce results which are in reasonable agreement with the experimental results in the following section. Further computational details can be found in Refs [16, 28].

Figure 1(a) shows results from simulations of the binding kinetics of the competitor (with 10 nM solution concentration C_comp) as a function of the solution concentration of the target species. One distinguishable feature from these curves is the maximum height of B_comp, which depends on C (which represents the unknown/unlabeled target concentration). Using the maximum height of B_comp, a calibration curve can be generated, as shown in Fig. 1(b). Note that at very large or very small target to competitor concentration ratios, the peak height (normalized to the equilibrium value in the absence of the target) asymptotically approaches its limiting values of 0 and 1, respectively. Combined with the SNR in detection (which determines the smallest changes that can be measured), this effect ultimately limits the dynamic range of C. The dynamic range can be extended towards lower C by measuring the exponential decay during the displacement regime, but this requires longer acquisition times approaching equilibrium. Further, dynamic range can be shifted in concentration space by either changing C_comp or by altering the ratio of k_d values by changing the competitor sequence or the temperature of the reaction. Another approach to lower the detection limit is to reduce the value of R_t, so that competition can occur with low target concentration Ref [28].

3. Label-free detection experiments

Experimental evidence of competitive displacement was provided in Ref [17]. Here, we demonstrate the change in competitor kinetics as a function of target concentration using the same sequences as we used during modeling. The experimental setup used to perform CDDM is illustrated in Fig. 2. A 532 nm laser was end-fire coupled into a quartz microscope slide which acted as an optical waveguide. Real-time detection of hybridization by a CCD camera (Santa Barbara Instruments ST-7XMEI) was facilitated by the evanescent field due to total internal reflection inside the waveguide, with the detected fluorescence signal proportional to B_comp(t). The imaged area encompassed reference and hybridization spots, as shown in Fig. 3, where the reference spot consisted of a fluorescently-labeled probe sequence that was non-interacting with the other sequences in the experiment. The purpose of the reference spot was twofold: it aided with initial alignment of the system, and the measured fluorescence intensity was used to normalize out fluctuations in laser intensity over long acquisition times.

 

Fig. 1. a) Simulation of the binding kinetics of 10 nM competitor concentration in the presence of varying target concentration. b) Calibration curve. The symbols represent peak values of competitor curves.

Download Full Size | PDF

 

Fig. 2. Two-color real-time microarray imaging setup. Only a single color (green) is used in these experiments.

Download Full Size | PDF

The camera was cooled to -25°C using internal Peltier devices and by water from a refrigeration unit to reduce dark signal. Each frame captured by the camera was exposed for 2.5 seconds and then saved for post processing. A TTL signal from the camera shutter was used to modulate the laser output in order to reduce photobleaching during the time interval between acquisitions. Further experimental details, such as surface modification chemistry, target and probe preparation, and probe immobilization can be found in Ref [17].

 

Fig. 3. Image frames illustrating the reference and hybridization spots used in the experiments; the images have been compressed horizontally for display purposes. The reference spot is approximately 500 μm in diameter. The dashed outlines indicate the areas over which intensity is integrated to obtain the total fluorescence signal; the areas are the same for the reference and hybridization spots to account for differences in morphology. In these frames, the competitor concentration is 10 nM and there is no target. Note that at the concentration used in this experiment, the surface occupancy of hybridized DNA is less than 100%.

Download Full Size | PDF

The first set of experimental results were obtained with a relatively high competitor concentration of 10 nM in order to reduce the time per hybridization run, as shown in Fig. 4(a). As in the computational results of the previous section, the hybridization curves were normalized to the maximum height of the competitor curve in the absence of the target. Target concentrations were varied from 0 nM to 100 nM; in each experiment, 250 μL of solution containing target and competitor was dispensed over the hybridization spots. Before each hybridization experiment, the duplexes formed during the previous experiment were melted by first raising the surface temperature above the melting temperature and then washing with buffer. Even though this technique improves the repeatability of the results, only a limited number of regeneration cycles can be employed before degradation of the probes occurs.

Very good qualitative agreement with simulations is obtained. Further, since the concentrations used were relatively high, Eqs. 6 and 7 can be used to fit the experimental data, as indicated by the curves in Fig. 4(a). Using these fitted curves, a calibration curve can be generated which relates the maximum height of the competitor curve to the concentration of the target, as shown in Fig. 4(b). One important aspect to note is that the calibration used here does not require equilibrium be reached; all that is necessary is that the displacement regime be reached so that the peak height can be determined.

As the concentration of target decreases, the time needed to reach the displacement regime increases. Further, at low concentrations with respect to the competitor, the ability to discriminate via the height of the competitor curve diminishes (see Fig. 1(b)); in these cases, the exponential decay can be used, but this requires longer acquisition times, as stated previously. Even though microarray hybridizations are typically performed for 16 to 18 hours to try to reach equilibrium with lower concentrations, CDDM can be performed at these lower concentrations over shorter times by reducing the competitor concentration, thus shifting the dynamic range. Figure 5 shows experimental curves with 1 nM competitor concentration and demonstrates sub-nM detection of the target. These reactions are more strongly diffusion controlled, so Eqs. 6 and 7 are less applicable in fitting. Nevertheless, it is clear from the experimental curves that concentrations below 0.1 nM should be detectable with this system.

 

Fig. 4. a) Experimental demonstration of CDDM with 10 nM competitor concentration. The symbols represent experimental data points while the solid curves are model fits (with R ² values greater than 0.93). Target concentrations are as indicated. b) Calibration curve. The symbols represent peak values of curve fits.

Download Full Size | PDF

 

Fig. 5. CDDM with 1 nM competitor concentration, demonstrating detection of sub-nM target concentrations.

Download Full Size | PDF

4. Discussion and conclusion

We have demonstrated a new label-free detection method, CDDM, in which competitive displacement is used to indirectly detect a primary target. The preliminary experimental results show good correlation with simulation results and theory. Using CDDM it was shown experimentally that a sensitivity of one-tenth the competitor concentration with a dynamic range of detection greater than two orders of magnitude is achievable.

Another aspect of sensitivity lies with the transduction mechanisms. In our experiments, we used a thick-film (∼1 mm) planar waveguide to excite fluorescence. Utilizing a thin-film waveguide Refs [20, 21] (1 μm, for example) reduces the effective mode size by about a factor of 1000, thereby increasing the intensity at the surface by roughly the same factor. In the shot-noise limit of detection, this would result in an improvement in SNR by the square-root of the intensity enhancement factor Ref [29], or by about 30 in this case. More advanced approaches promise to improve the detection limits even further Refs [23, 29, 24, 30]. It is conceivable then that sufficient sensitivity in fluorescence-based detection can be obtained to render molecular amplification steps (such as cell culture or PCR) unnecessary in some circumstances; the additional advantage offered by CDDM is that sample preparation can be further simplified by not requiring a labeling step. In the case of DNA, the necessary sample preparation steps would then consist of cellular disruption, DNA isolation, and shearing to a desired average length.

Aside from sensitivity limitations, the quantitative capability of CDDM depends upon a method of calibration of the fluorescence signals. A simple calibration method would be to use calibration spots, where probe and target sequences associated with those spots are designed to be non-interacting with other sequences in the system, but with similar equilibrium binding constants as the actual competitor molecules. This has the advantage of allowing calibration and experiment to be performed simultaneously. Another method would be to simply run an experiment with the competitors only so that their native kinetics can be determined, then reuse the array (by melting of surface bound duplexes, for example) with both competitor and sample.

A number of improvements on the method can be made. One such improvement is the use of multiple competitor molecules used for the detection of multiple targets (e.g. multiple SNPs). Depending upon the relative cross-reactivity between targets and each probe spot, a single competitor may even be used for more than one target.

This research was performed within the Center for Microarray Technology, which is supported in part by the Utah Centers of Excellence Program, National Science Foundation Career ECS-0134548, and University of Utah Technology Commercialization Project. Financial support for J. Bishop has been provided by a National Science Foundation fellowship (No. NSF IGERT:DGE 9987616) and by the Centers of Excellence.