1 Introduction

Quick, sensitive, and specific clinical diagnosis is critical for saving lives and curing diseases. Diagnostic devices (biosensors) designed for this purpose have to be disposable and cheap, i.e., fabricated from inexpensive materials via affordable technologies, yet need to be sensitive enough to be useful. Several disposable sensor designs have been investigated within the last few years [1,2,3].

A particular focus has been put on DNA biosensors based on the recognition of specific nucleic acid sequences as these targets are of significant importance in clinical analysis [4] for the detection of pathogenic microorganisms such as bacteria and viruses by their genetic materials and for the detection of infectious diseases [5]. Moreover, due to the huge amount of genetic information that needs to be analyzed in routine medical procedures, there is a high demand for efficient screening methods for DNA, e.g., for the identification of point mutations in a patient’s genome [6].

One of the methods developed to increase the sensitivity of such devices is by implementing DNA detection schemes on the surface of a sensor platform [7,8,9,10,11]. Most nucleic acid detection schemes exploit the specificity of base recognition between complementary DNA strands and the high binding constants of the resulting duplex. The sensor surface is usually modified with a probe DNA of known sequence and is exposed to an aqueous solution of the target sequence. Monitoring the hybridization between the target and the modified sensor surface indicates the level of interaction between the probe and the target, thus allowing for the identification of the target sequences [7, 12]. Particularly in the case of DNA chips, thousands of different probes are immobilized and analyzed simultaneously to discriminate between fully matched and partially mismatched oligonucleotide pairs.

One of the widely used methods for DNA detection is based on surface-plasmon resonance (SPR) and surface-plasmon fluorescence spectroscopy (SPFS) [13, 14]. The latter, in particular, has been shown to allow for the quantitative evaluation of hybridization reaction rates and binding affinity constants between surface-attached capture probes and their complementary strands binding from the analyte solution. The classical approach of surface-plasmon field-enhanced fluorescence spectroscopy in biosensor formats is based on the Kretschmann configuration, employing a prism for the resonant coupling of laser photons from the backside of the prism through the thin noble metal layer (mostly Au because of its chemical inertness) to surface-plasmon waves at the metal–solution interfaces [15]. These surface modes, propagating at the interface with their evanescent field decaying exponentially into the analyte solution within approximately 150 nm, can excite chromophores (Fig. 1a) like normal photons can do. Hence, analyte molecules in the solution that carry a fluorescent dye and that can bind to specific recognition sites at the surface of the correspondingly functionalized sensor substrate contribute to the observed fluorescence intensity (monitored typically from the base side of the prism) in at least two ways: one contribution is proportional to the analyte surface density after binding, thus providing a direct measure of the surface coverage Θ; in addition, another contribution originates from free analyte molecules that are in solution at a concentration c. This contribution is proportional to c; however, originates only from the thin slice of the evanescent plasmon field. The ratio of the two contributions is governed by the affinity of the surface reaction: according to a Langmuir model, the surface coverage depends on the bulk concentration of the analyte according to:

Fig. 1
figure 1

a Schematics of surface-plasmon fluorescence spectroscopy in the Kretschmann configuration where the incident light is coupled via a glass prism through the thin metal layer to surface plasmons propagating at the metal–dielectric (here the analyte solution) interface, which in turn excites chromophores positioned near the surface within the evanescent optical field; b schematics of surface-plasmon fluorescence spectroscopy employing a surface grating used to couple light to plasmons, c schematic illustration of the experimental setup as described in the text

Full size image
$$\Theta = \frac{{c \cdot K_{{\text{A}}} }}{{1 + c \cdot K_{{\text{A}}} }},$$
(1)

hence, depends only on the affinity constant KA, which is the inverse of the dissociation or half-saturation constant, Kd:

$$K_{{\text{A}}} = { 1}/ \, K_{{\text{d}}}$$
(2)

For most hybridization reactions that we studied in the Kretschmann configuration by global analysis for duplexes of high complementarities, i.e., with dissociation constants, Kd, between probe and target strands in the range of nano- to lower micromolar, the contribution of the bulk fluorescence could be ignored; the measured fluorescence signal was completely dominated by the surface-bound target strands. Only in the case of the binding of a 15 mer target with 2 mismatches to the capture probe strand, with Kd > 100 µM a small fluorescence “bulk jump” signal, following immediately the injection of the sample solution into the flow cell, could be observed [16].

In principle, the fluorescence could be monitored with an additional enhancement mechanism by coupling the fluorescence emitted from the chromophores back to (Stokes-shifted) surface-plasmon modes that couple out again through the prism. However, this requires a highly demanding optical setup [14, 16]; hence, typically the fluorescence is monitored from the base side of the coupling prism. But even in this case, for prisms made of materials of a moderate refractive index (e.g., n ~ 1.5…1.6) the involved coupling angles for experiments in an aqueous phase are rather high and thus require a complex optical/mechanical setup. Moreover, the need for a precisely defined metal thickness at the base of the prism (typically 50 nm of Au) adds another challenge to the reproducible fabrication of sensor modules resulting in concerns for quality management in mass-produced sensors.

This led to considering the development of grating structures as surface-plasmon couplers [17,18,19,20]. Their use in Raman spectroscopy is well established and enhancement factors of ~ 104–105 have been documented [21, 22]. As in the case of prism coupler, photons of a given laser wavelength couple at a particular resonant angle θ which is defined by the energy–momentum matching condition between photons and surface plasmons. These surface modes again excite chromophores which are (either in the vicinity of or) at the sensor surface by a specific recognition and binding reaction or still freely in the bulk solution but within the slice of the evanescent field (schematically represented in Fig. 1a).

However, in the case of surface-plasmon excitation by a grating coupler from the front side of the sensor, an additional main drawback of the whole arrangement is that the exciting laser beam must pass first through the whole sample cell containing the fluorescently labeled analyte molecules before it reaches the grating to excite a surface-plasmon mode. Even for thin microfluidic flow cells, the thickness of this bulk layer is typically 1000 times thicker than the evanescent field of the surface-plasmon wave. This then leads to a scheme of grating coupling for sensor applications as shown in Fig. 1b.

Depending on the reflectivity at the chosen angle of incidence, the laser beam travels even a second time through the cell (though with reduced intensity, cf. Figure 1c). In both cases the chromophores of the free analyte molecules in the sample solution are excited and emit fluorescence light which generates a background signal, i.e., the detected signal consists of three components: the signal of the bulk solution, the contribution of chromophores in the evanescent plasmon field, and the actual sensor signal which originates from the analyte molecules bound to the active sites on the surface. While the first two appear immediately after the sample solution is injected into the flow cell, the third intensity, the sensor signal of interest, increases linearly with time because the surface reaction is mass transfer limited. The “bulk jump” of the first two contributions limit the sensitivity of grating coupled plasmon-optical sensors and asks for strategies to minimize this contribution to allow for a fast response of the sensor without rinsing the cell as an additional preparation step.

In this work, we present a surface-plasmon fluorescence spectroscopic protocol for oligonucleotide identification by hybridization from solution to complementary strands, immobilized on the surface of the transducer, which is a grating coupler. The advantages of this approach are (i) a much compacter instrumental configuration; (ii) ease of integration into microfluidic cells, e.g., at the bottom of the fluidic channels, or in the bottom of a 96 well plate; (iii) the free choice of the incident and emission angles for surface-plasmon excitation by the appropriate selection of the grating constant (the periodicity of the grating grooves); (iv) independence of the coupling efficiency of the Au coating thickness (a serious quality management issue in prism coupling); (v) ease in mass production of the coupling elements by hot embossing, the technique used by industry to produce millions of DVDs. However, to handle the disadvantage of this way, which is that exciting surface-plasmon modes at the transducer surface involves passage of the excitation laser beam through the analyte solution layer in the flow cell in front of the grating, which leads to a contribution to the fluorescence signal from the bulk, we developed a protocol based on the bulk fluorescence quenching of the analyte chromophore by hybridization in solution with a complementary (anti-target) strand with one mismatch but carrying a dye molecule that acts as an acceptor for the donor dye at the target strand.

2 Materials and methods

2.1 Grating fabrication

The main parameters that describe a grating structure, and that are relevant to analyze its performance as a plasmon coupler, are the grating constant, Λ, its amplitude H (depth), and the surface roughness, α. For this series of studies, master gratings with grating constants of L = 470…650 nm, and amplitudes of typically H = 15…50 nm, were fabricated by a holographic technique [23]. This generates a sin2 pattern which is generally well reproduced by the development of the exposed photoresist. The resulting surface corrugation of the photoresist layer is then transferred into the glass (or quartz) substrate by ion beam etching. Depending on the processing parameters, this preparation step results in surface corrugations that can vary from near rectangular [24] to nearly perfectly sinusoidal master gratings.

For the series of experiments presented here, a master grating (the stamps) made from quartz glass were used to emboss the structure on a large copy number of polymer substrates. Typically, poly-(methyl methacrylate) (PMMA) substrates of 1.5 mm thickness and a glass transition temperature of 110 °C, were used for this purpose. The embossing procedure was performed by first applying an anti-adhesive layer to the surface of the stamp to lower its surface energy and prevent sticking to the substrate [25, 26]. The anti-adhesive layer was deposited by immersing the stamp in dimethyldichlorosilane for 30 min and then annealing the functionalized master grating at 100 °C for 10 min. The stamp was then heated up to 140 °C, and pressed into the polymer substrate for a few seconds, before both the master and the substrate were cooled down before separation. Figure 2a shows the tailor-made embossing setup. Embossing time and temperature was optimized to achieve the best results (defined by the plasmon resonance properties of the embossed gratings and their reproducibility during multiple use of the same master stamp). The measured surface-plasmon resonance of the embossed gratings, coated with a 100-nm-thick gold layer, displayed an excellent and reproducible resonance behavior; Fig. 2b shows a series of surface-plasmon resonance curves of embossed gratings (the 1st, 11th, and 21st replicas) using the same stamp indicating great reproducibility. The embossed substrates were also characterized by AFM (Fig. 2c, d).

Fig. 2
figure 2

a Setup used to fabricate the embossed gratings on polymer substrates, where the master grating is placed on a hot plate and the substrate fixed to a holder is pressed onto the stamp; b SPR reflectivity scans of a series of embossed gratings showing reproducibility up to 21 embossing steps; c AFM image of a grating embossed into a PMMA substrate; d AFM profile of the same embossed grating showing a grating constant and grating amplitude similar to those of the master grating

Full size image

The embossing process could be repeated many times using the same stamp; only after 30 replicas, the stamp started to lose its anti-adhesive layer. However, cleaning and remodifying with the anti-adhesive dimethyldichlorosilane made the stamp again ready for embossing more samples. Hence, even at this laboratory level of preparation protocol, this method offers a fast, reliable, and inexpensive way to produce disposable chips. All measurements in this study were done—after proper coating with Au—with these stamped gratings.

2.2 Optical setup

The grating coupling scheme requires the independent control and variation of the incident angle, θin, and the angle of emission of the fluorescence light, θout (Fig. 1a and b, respectively). To this end, the coupling grating with the attached flow cell (cf. Figure 1c, inset) was mounted onto one of the two-circle rotary stages of a goniometer. Relative to the fixed direction of the incident laser (λ = 632.8 nm, 5 mW, PL-750P polarized helium–neon laser, Polytec), the angle of incidence could then be continuously varied. The reflectivity was monitored with a photodiode mounted to the second stage of the goniometer rotating at 2θ. To block the scattered light, two interference filters that allow for the fluorescence emission wavelength to pass, and a sufficiently narrow slit aperture to define the angle of observation, θout, was used. All other components including the Photomultiplier (PMT) and photon-counting unit were identical to the normal SPFS setup in a Kretschmann configuration [14, 16]. The actual grating coupler is attached to a flow cell sealed by an O-ring and a cover glass with a low level of intrinsic fluorescence, with inlet and outlet for continuous sample solution or buffer supply (cf. inset of Fig. 1c).

2.3 Surface architecture

A typical surface architecture for DNA hybridization studies is given schematically in Fig. 3a. The metal surface was chemically modified by a self-assembled monolayer of a mixture of two kinds of thiols: a biotin-terminated thiol (cf. Figure 3b), and an ethylene glycol (EG)-terminated thiol, mixed in solution at a molar ratio of 1:9 before being chemisorbed onto the gold surfaces. To this mixed layer, streptavidin can be bound via the biotin moieties. This way, a generic binding matrix is generated through the two binding sites for biotinylated oligonucleotide catcher strands (probes), to which target analyte strands with their covalently attached fluorophores can hybridize to. The DNA sequences employed in these studies are represented in Fig. 3c.

Fig. 3
figure 3

Biomodification of the sensor surface. a The figure sketches the individual layers of the functionalization architecture with some artistic freedom. It is schematic and not to scale; e.g., the gratings used are of very small aspect ratios (grating constants between 475 and 515 nm versus a typical amplitude of 20 nm) and the subsequent layers follow the slight underlying surface undulation; b the chemical structure of the biotinylated thiol [12-mercapto(8-biotinamide-3,6-dioxaoctyl] dodecanamide; c the structure of the probe P2, the target T2, and the Anti-T2 which acts as a quencher through its chromophore Cy 5.5

Full size image

As shown in the Fig. 3a and c, the probe is coupled to the Au sensor surface via a binding matrix comprised of a thiol SAM/biotin/streptavidin multilayer, and contains a thymine spacer of 15 bases; with the target carrying the chromophore at the 5′-end this leads to a total chromophore-Au surface separation of more than 15 nm, corresponding to nearly 2 Förster radii which allows the labeled target to be far enough from the surface to avoid nearly completely quenching by the metal layer. The chromophore pair used in the experimental realization of a bulk-quenching scheme was Cy5 and Cy5.5. The target strand was labeled at its 5´-end with Cy5, the anti-target carries a Cy5.5 labeled at its 3´-end (Fig. 3c). Both dyes show a significant spectral overlap to ensure efficient fluorescence resonance energy transfer (FRET) as shown in Figure S1. The biotinylated thiol derivatives were synthesized in our laboratories, the streptavidin was purchased from Sigma-Aldrich, and the DNA sequences labeled with Cy5 and Cy5.5, respectively, were purchased from Biomers GmbH.

2.4 Background measurements

To study the impact of bulk fluorescence on the kinetics and affinity measurements on the sensor surface, a DNA titration experiment was performed with a fully mismatched 15mer target strand (MM15), expecting no surface hybridization at all. The concentration of the analyte MM15 was changed three times by a factor of ten to map the effect on the detected fluorescence signal over a wide concentration range from 3 pM to 3 nM. A grating with Λ = 474.7 nm was illuminated with P-polarized light near its resonance angle of θin = 10.6°, at the minimum of the reflectivity curve. Given the chosen grating constant, the back-coupling emission lobes [27] are in close angular proximity to each other. By opening the aperture next to the lens (diameter 22.4 mm), halfway between the sample and the PMT (each distance 2f = 100 mm), emission angles within Δθout = θout ± 6.08° of the grating normal were covered, integrating both fluorescence lobes at the same time, thus maximizing the signal. This arrangement resulted in very high intensity levels exceeding the input limit for the PMT, and the laser intensity had to be further reduced by a second neutral density filter of T = 10% to allow for safe operation. To prevent bleaching during the anticipated extended length of the experiment, illumination by the laser beam was controlled and blocked by an automatic shutter (open for 5 s, closed for 300 s).

3 Results and discussion

3.1 Background measurements with a fully mismatched target strand (MM15)

Before injecting the fully mismatched complementary strand, background and bulk signal intensities were measured while pumping PBS through the flow cell. The first column in Fig. 4a was recorded before the injection of the fluorescently labeled inert oligomer MM15. Its height represents the background signal level for an empty flow cell (in terms of fluorophores) at an average of 2169 counts/s of the PMT. It is composed of mainly two contributions: electronic noise and illumination crosstalk. The electronic noise is always present, even when the shutter is closed (visible in the three gaps between the last four columns) and amounts to a mean value of 266 counts/s. It is mainly caused by the dark current of the PMT. Assuming the electronic noise to be independent of PMT illumination, the remaining 1903 counts that appear upon opening the aperture are caused by laser photons scattered into the detector, despite the two interference filters. The background value corresponding to an empty flow cell (indicated by the dashed line in Fig. 4a) was deducted from the average height of all other columns in order to quantify the pure bulk fluorescence contribution.

Fig. 4
figure 4

Bulk fluorescence emission is one of several mechanisms contributing to the signal background. a The graph shows background measurements for different concentrations of the fluorescently labeled inert sequence MM15. The intensity emitted from the bulk is concentration dependent and rides on a constant level of electronic noise and illumination crosstalk (dashed line). b The pure bulk contribution scales linearly with the dye concentration (the error bars are determined by the RMS value of the PMT signal)

Full size image

Following the injection of a MM15 solution at a concentration of 3 pM, the second column was recorded. Its height is only slightly higher than that of the first one. It appears noisier since the data acquisition rate of the PMT had been increased at this point. The following columns correspond to MM15 concentrations of 30 pM, 300 pM, and 3 nM, respectively. After deduction of the “empty” background level, the remaining counts correspond to the pure bulk fluorescence emission. These intensities are plotted in Fig. 4b as a function of the bulk concentration. The intensity is expected to scale linearly with the concentration. For that reason, the data are fit with a line through the origin. More emphasis is placed on the intersection of the line fit with the origin (i. e. no fluorescence in the absence of fluorophores) than on the optimal coverage of all data points.

3.2 Affinity measurements with duplexes of high affinities (MM0)

Before starting the hybridization experiments with target strands of different degrees of (mis-)matches, the flow cell was thoroughly rinsed, and the regeneration protocol applied to ensure full removal of all MM15 oligomers. After that, the fully complementary strand (MM0) titration experiment started with the injection of a 3 pM solution. The hybridization was monitored for more than 11 h before the solution was exchanged against pure buffer and the flow cell continuously rinsed for about 5 h. Similarly, the titration was continued with the adsorption measurements from solutions of 30 pM, 300 pM, and 3000 pM concentrations, respectively, and subsequent rinsing steps (Fig. 5). The adsorption was monitored until the curve shape started to approximate the equilibrium region. Assuming the validity of the Langmuir model for this immobilization reaction, the four adsorptions and the four desorption curves were fit one by one to determine the association rate constant, kon, and the dissociation rate constant, koff.

Fig. 5
figure 5

A global analysis experiment with the fully matched target strand MM0. a The target concentration was increased from 3 to 30 pM, 300 pM, and 3000 pM. The flow cell was rinsed with pure buffer before each increase in target concentration. The resulting adsorption and desorption kinetics can be fit well with the Langmuir model, using the rate constants given in Table 1. Each injection or rinsing step defines the beginning of a new adsorption or desorption phase, respectively. b A closer look at the adsorption and desorption curves for the two lowest concentrations of the series

Full size image

Both rate constants and the fluorescence intensities for the different concentrations are given in Table 1. Imax gives the maximum fluorescence signal detected for a given concentration at the end of the adsorption phase. Ibulk indicates the pure bulk fluorescence emission previously measured (cf. Figure 5). It can be clearly seen that the maximum intensities measured for each MM0 concentration always exceeded the corresponding MM15 signal levels by several orders of magnitude. The bulk emission can be considered irrelevant in this concentration range and was, therefore, neglected in the quantitative analysis. Values for kon can be derived by analyzing the four adsorption phases 1, 3, 5, and 7, respectively (cf. Figure 5), and are given in Table 1. The mean value amounts to kon = 1.86 × 105 M−1 s−1. Moreover, all eight phases can be used to find koff. As typically found in such measurements, there is some variation in the values, but generally the numbers are in the range of a few 10−5 s−1; the mean value of all fits amounts to koff = 5.9 × 10−5 M−1. From these (mean) values of the two rate constants, one obtains a value for the affinity constant KA = kon/koff = 3.1 × 109 M−1 which compares favorably to the mean of the affinity constants calculated from the four kinetic adsorption measurements (cf. Table 1) which amounts to KA = 2.8 × 109 M−1.

Table 1 Hybridization and dissociation rate constants, kon and koff, the corresponding affinity constant, KA = kon/koff, and the fluorescence intensities, Imax and Ibulk, respectively, measured during the hybridization and dissociation process of the target strand MM0 to the surface-immobilized capture probes
Full size table

A different approach for the analysis of KA by this global analysis assay is given by plotting Imax as a function of the bulk concentration as done in Figure S2 (a); the fit to a Langmuir model (red fit curve in Figure S2 yields KA = 4.6 × 109 M−1, which confirms the validity of the model in that the affinity constant derived by the kinetic experiments is in very good agreement with the titration data. Moreover, the hybridization reaction of the very same pair of probe and target oligomer has been studied before by prism coupling SPFS [7] with a reported affinity constant of KA = 5.3 × 109 M−1 which is in excellent agreement with the data presented here.

The fluorescence emission from the bulk and from the surface at equilibrium coverage are both functions of the analyte solution concentration. The former can be modeled by a linear relationship, the latter is predicted by the Langmuir isotherm. The intensity ratio of the equilibrium fluorescence originating from the surface-bound species divided by the full background (electronic noise + illumination crosstalk + bulk emission) is plotted in Figure S2 (b). The curve exhibits a pronounced maximum at around 250 pM. The range from 10 pM to 30 nM is best suited for high SNR operations with a ratio of approximately 100. This is the case for hybridization measurements with the MM0 target with a Kd = 200 pM; hence, any background correction can be ignored.

Distinction between the background and the equilibrium signal from binding is still possible if both are equally strong and their ratio equals one. This criterion yields a theoretical lower limit of detection LOD = 40 fM for this DNA assay. (One must keep in mind, however, that it takes about 35 h to reach 95% of the equilibrium value at this concentration, which may not be practical for most applications. At the same time, there is an upper LOD of 1.5 μM, since increasing the concentration even further increases practically only the background when operating in the plateau region of the Langmuir isotherm. For the hybridization experiment with the MM0 high affinity duplex described so far, this limit was of no relevance.

3.3 Affinity measurements with duplexes of lower affinities (MM1)

This situation changes if one deals with hybridization reactions (or other ligand-receptor binding studies) with lower affinities; then this effect can cause significant complications. Monitoring lower affinity binding reactions requires, by definition, higher bulk concentrations to reach the same surface coverage and hence fluorescence contribution form the bound species. This then shifts the ratio of the fluorescence contributions from the surface relative to the bulk to smaller values. In the following, we demonstrate this for hybridization reactions between the identical surface probe layer and complementary target strands binding from solution exhibiting, however, a single mismatch, i.e., MM1, resulting in an affinity constant KA which is nearly 3 orders of magnitude lower than for the MM0 case [12].

Figure 6a shows a typical kinetic/titration experiment of a chromophore-labeled target strand hybridizing from solution to the surface-attached 15mer probe. The sequence of the 15mer strand was complementary to the probe except for 1 base in the middle of the sequence, hence constituting a mismatch 1 (MM1) situation. Upon stepwise increasing the concentration of the target solution two effects could be observed: for the lower bulk solution concentrations (10–50 nM) only the gradual occupation and final saturation of the probe binding sites by target molecules could be observed (cf. Figure 6a). For the higher bulk concentrations (100 nM and 250 nM), one first observes a jump of the fluorescence intensity right upon exchanging the bulk solutions. This is then followed by a further gradual increase of the fluorescence, originating from the surface hybridization reaction with the expected saturation at longer times. The final rinsing step again shows two steps in the decrease of the fluorescence: the first is again a stepwise decrease, following immediately the exchange of the bulk from the 250 nM target solution to pure buffer solution, followed by the time-dependent decrease of the fluorescence as a consequence of the gradual dissociation of the target strands from the capture probes. The latter again demonstrates the reversibility of the binding, a prerequisite for the establishment of an equilibrium between all the plateau values reached for the different concentrations and the bulk, justifying the application of the Langmuir model for the analysis the data [28] (cf. Figure 6a).

Fig. 6
figure 6

a A typical global analysis experiment of a targets labeled with Cy5, hybridizing to the surface-attached probes; the target is complementary to the probe except for one base, the affinity constant calculated from this experiment was 2.9 × 107 M−1, b Langmuir isotherm of the intensity data from (a), corrected by the bulk contribution (for details see text). The red curve yields an affinity constant KA = 2.9 × 107 M−1

Full size image

This observation is in agreement with the consideration presented in Figure S2 (b): the dashed green lines indicate the range of bulk solution concentrations (10–250 nM) used in this experiment for the quantification of the MM1 hybridization. The lower bulk concentrations rinsed through the flow cell correspond to an “equilibrium fluorescence/full background” ratio of > 50; hence the “bulk jump” by the bulk solution injection amounts to only a few % of the total signal and, hence, is not observed. For the two higher concentrations (100 nM and 250 nM, respectively), the situation is different: the mere solution injection results in a bulk jump of the fluorescence signal that constitutes a significant increase (cf. Figure 6a) that needs attention in the quantitative treatment of the hybridization data by a titration analysis.

For the kinetic runs, this is still simple: each one of them is recorded after the solution concentration has been completely changed, i.e., is monitored at constant bulk target concentrations; hence, the analysis of the time-dependent increase (during hybridization) or decrease (during dissociation) of the fluorescence intensity for the determination of the corresponding rate constants, kon and koff, respectively, is unaffected by the fluorescence contribution form the bulk solution. Operating with a flow cell guarantees that the latter is constant during the change of the surface coverage.

From the kinetic runs with increasing bulk solution concentrations and after rinsing with pure buffer presented in Fig. 6a one obtains a mean association rate constant kon = 1.5 × 10–4 M−1 s−1 and a dissociation rate constant koff = 2.1 × 10–4 s−1, and from this an affinity constant KA = 6.6 × 107 M−1.

The situation is different if one analyses the intensity data at different bulk solution concentrations via the Langmuir adsorption isotherm. Here, one needs to take into account (at least for the higher bulk concentrations) the jump of the intensity upon changing the analyte solutions from one level to a higher or lower concentration. As mentioned before, this is particularly pronounced for the introduction of the 100 nM and 250 nM solutions, respectively, and upon going back to zero concentration, i.e., upon rinsing pure buffer through the flow cell. In the latter case, one observes a corresponding jump of the fluorescence intensity to a lower level by about 10,000 cps (cf. Figure 6a), i.e., the bulk fluorescence contribution of the flow cell originating from a 250 nM solution amounts to that fluorescence intensity. The fluorescence contribution from the surface-bound targets amounts to 39,000 cps; this value is the one given for c = 250 M in the Langmuir adsorption isotherm, presented in Fig. 6b. Since the bulk intensity scales linearly with concentration (cf. Sect. 3.1), each of the recorded final intensities for the various lower concentrations must be reduced correspondingly to derive at the true contributions from the surface hybridized target strands at that bulk concentration, e.g., the surface contribution from the target coverage in equilibrium with the 100 nM bulk solution amounts about 32,000 cps.

All thus correspondingly corrected surface intensities are plotted in Fig. 6b as a function of the bulk concentration. The red curve is a fit thought the data using the Langmuir isotherm analysis. The obtained affinity constant KA = 2.9 × 107 M−1 compares well with the data from the kinetic analysis and is in excellent agreement with measurements performed with the same system in the Kretschmann configuration [7].

If we compare the concentration range of this titration experiment (between 10 and 250 nM, cf. the green dashed lines in Figure S2 (b) and the obtained half-saturation concentration Kd = 34 nM, also indicated in the figure), we see that with the MM1 experiment we are operating at the edge of the accessible range, near the LOD at the high concentration end where the fluorescence from the bulk gradually dominates the measured signal and a true Langmuir titration analysis requires a significant correction. Hence, the question comes up how one could deal with this situation or how to reduce or completely avoid the bulk contribution to the measured fluorescence signal.

3.4 Reducing the bulk contribution

A trivial step towards the reduction of the bulk contribution would be simply running the experiment in a thinner flow cell. This is possible but would give at most an improvement of the “equilibrium fluorescence/full background” ratio [cf. Figure S2 (b)] of about an order of magnitude; i.e., shifts the black curve by a factor of ten to higher concentrations.

Among the various other conceivable protocols for the reduction of the “bulk jump”, we present here the concept based on the principle of Fluorescence Resonance Energy Transfer (FRET) applied to quench the fluorescence from the target molecules in the bulk solution. The principle is schematically sketched in Fig. 7a: the labeled analyte strand to be detected by hybridization to a surface-immobilized capture probe strand, T2, is mixed in the bulk with oligonucleotide sequences that are complementary, Anti-T2, hence hybridize with the target strand in solution, though with an affinity that is weaker than the one to the capture probe sequence. In the experiments described in the following, this is realized by introducing a mismatch 1 between T2 and Anti-T2. Anti-T2 is also labeled with a fluorescence molecule that is able to interact optically with the chromophore of the target strand T2 via FRET; hence, can quench the fluorescence of the target molecules in solution. In the ideal case, this occurs without affecting the surface hybridization reaction between the target and its surface-immobilized probe strand.

Fig. 7
figure 7

Schematics of surface hybridization reactions: a target strands, T2, with a fluorescent chromophore attached (blue triangles) are injected into the flow cell and bind to the surface-immobilized probe sequence, P2, exhibiting a mismatch zero (MM0) situation; b schematics of the concept for the reduction of the fluorescence contribution form the bulk: the targets are pre-mixed in the bulk with a complementary strand, exhibiting a MM1 situation and being functionalized with another chromophore that quenches the donor fluorescence emission by FRET

Full size image

FRET is a process that concerns donor and acceptor molecules in their excited state, coupled by dipole–dipole interactions [29, 30], provided the emission spectrum of the donor fluorophore overlaps with the absorption spectrum of the acceptor chromophore. Moreover, the extent of energy transfer is determined—among other factors—by the distance between the donor and the acceptor. Since the Förster distance between the two fluorophores, i.e., the distance at which the fluorescence intensity of the donor chromophore is reduced to 50% by energy transfer to the acceptor dye, is comparable in size to biological macromolecules, FRET has been widely used as a spectroscopic ruler for measuring the distance between sites in proteins [31,32,33].

As explained above, in conventional sensing protocols, the probe-functionalized sensor surface is exposed to the solution of the chromophore-labeled DNA analyte strands; the detected signal then corresponds to the amount of hybridized DNA on the sensor surface yet contains also contributions from the bulk (cf. Figure 7a). In the modified protocol presented here for a MM0 situation between probe and target oligonucleotide strands, the analyte is a mixture of donor-labeled DNA and acceptor-labeled Anti-DNA. The anti-DNA is designed to exhibit a one mismatch with the target DNA and labeled with a chromophore that would allow for FRET to take place. Mixing the target, T2 in the example presented here, and the anti-target, Anti-T2, at the appropriate ratio in solution before applying the mixture to the surface allows for the free molecules to be quenched, ideally with minimal effect on the surface hybridization reaction (cf. Figure 7b).

In preparation for the demonstration of this concept, the quenching effect between these targets and anti-targets in bulk solutions was examined by measuring the change of the fluorescence intensity of the target at increasing quencher concentrations, varying the mixing ratios from 0:1 to 1:1. The measurements were performed using a Perkin Elmer fluorescence spectrometer at an excitation wavelength of λ = 630 nm, matching the wavelength used in the surface-plasmon fluorescence experiments. The results are presented in Fig. 8a: the first spectrum given in black was recorded from a 500 nM target solution in the absence of any quencher. The following 10 spectra were recorded sequentially, each after the addition of a 50 ml aliquot of a 500 nM Anti-T2 solution. This resulted in a gradual increase of the Anti-T2 quencher concentration, until after the 10th addition, a 250 nM concentration of the Anti-T2 quencher was reached. At the same time, the target T2 was successively diluted until after 10 additions of the quencher solution an overall target concentration of also 250 nM was reached; i.e., this spectrum was recorded at a 1:1 ratio between target T2 and Anti-T2.

Fig. 8
figure 8

a Fluorescence spectrum of a 500 nM DNA target solution (T2 labeled with donor dye Cy5, black curve); each of the following 10 spectra were recorded sequentially after the addition of a 50 ml aliquot of a 500 nM Anti-T2 solution (labeled with the acceptor dye Cy5.5), thus decreasing the target concentration and increasing the quencher concentration until at the end a 1:1 mixture was obtained; b surface-plasmon fluorescence spectroscopic recording of the binding kinetics of DNA target, T2, hybridizing to the surface-immobilized probe sequence, P2; the black curve is the binding from a 200 nM solution of the target to the probe showing a signal which corresponds to both, contributions from the surface and from the bulk. The red line is the signal measured after the injection of a mixture of 200 nM target and 40 nM of the quencher (Anti-T2)

Full size image

A few observations are noteworthy: the fluorescence intensity at λ = 670 nm, the peak emission wavelength of the donor, indeed, decreases monotonically with increasing quencher (Anti-T2) concentration. The shape of the emission spectrum remains nearly identical, just reduced in intensity as expected from a model in which the increasing Anti-T2 hybridizes to the target T2, quenches the emission of this duplex (Anti-T2 bound to T2), and only the chromophores of the remaining free T2 single strands emit fluorescence light with an unperturbed spectrum. However, this model is too simple to explain the complete observation: the intensity at λ = 670 nm does not go to zero for a 1:1 concentration ratio; instead, after the addition of 300 ml of the quencher solution, which results in a total quencher concentration of cAnti-T2 = 188 nM, and a total target concentration of cT2 = 312 nM, a second emission band with a peak intensity at λ = 690 nm appears. This wavelength corresponds to that of the spectrum of the pure acceptor dye (cf. the spectra presented in Fig. 5c). Further increasing the quencher concentration shifts intensity from the λ = 670 nm band to the λ = 690 nm band, with a well-pronounced isosbestic point at λ = 676 nm. Obviously, a more complex interaction model, both in terms of the oligonucleotide–oligonucleotide interaction, as well as with respect to the donor acceptor chromophore FRET coupling is needed to fully understand the quenching behavior in solution. However, for a qualitative demonstration of the concept of quenching the fluorescence of target strands in the bulk solution, these details are not needed.

This is shown in Fig. 8b by the following set of experiments: the black curve represents the fluorescence signal of a surface hybridization reaction upon the injection of a 200 nM T2 target solution (in the absence of any anti-targets) that is then interacting with the surface-bound capture strands, P2. The rapid increase in the measured fluorescence signal is composed of the contributions of the bulk jump and the relatively quick surface binding reaction. Upon rinsing, the bulk contribution is washed away, seen as the instantaneous fluorescence drop to the intensity level representing the contribution by the surface-bound species.

If the same experiment is repeated with a mixture of the target and 20% of the quencher (40 nM solution) the bulk jump upon injection of the solution is largely reduced, and so is the jump back upon rinsing the flow cell with pure buffer (red curve in Fig. 8b). Increasing the amount of pre-mixed quencher molecules certainly would allow for a complete reduction of the bulk jump; however, this can be achieved obviously only at the expense of a largely reduced hybridization reaction rate. This is understandable in view of the more complex reaction scheme: after the injection of the T2/Anti-T2 mixture, the competition with the hybridization of the target T2 to the surface-immobilized P2 probe strands starts; eventually, the higher affinity of the P2/T2 hybrid with mismatch 0 wins by its lower free energy compared to the T2/Anti-T2 which is by design a MM1 hybrid. However, the price is the reduced reaction rates. For practical application purposes, this is not attractive.

4 Conclusions

In this work, we used grating coupler as a transducer to excite surface-plasmon resonance, combined with fluorescence spectroscopy to detect the hybridization between oligonucleotides. Despite the advantages of this system, it comes along with a major disadvantage, which is the contribution of the bulk signal due to the passage of the laser light through the analyte solution. Hereby, we developed a protocol to reduce (or eliminate) the bulk jump. This protocol is based on using a mix of the target oligonucleotide to be detected, labeled with a dye molecule, and a complementary strand with one mismatch, labeled with another due molecule that exhibits FRET with the first one, to bind to the probe immobilized on the transduce surface, to suppress the fluorescence from the bulk solution. Upon injection of this quenched duplex into the flow cell, the competition to the surface probe strands leads to the dissociation of target and anti-target duplex by binding of the target to the surface probe, resulting in the appearance of fluorescence light from the sensor surface. This way, the bulk jump monitored upon the injection of a high concentration of the chromophore-labeled analyte solution could be completely suppressed; however, only at the expense of largely slowed down binding kinetics of the target hybridizing to the surface-immobilized probe layer.

This protocol can help in the direct detection of the oligonucleotide hybridization on the grating coupler surface, which might help in using it as a sensitive and quick sensor for point-of-care applications.