1. Introduction

Surface plasmon resonance (SPR) biosensors have become an important tool for exploring the kinetics of biomolecular interactions and have been widely adopted in detection of chemical and biological analytes [1–4]. Moreover, by combining the SPR technique with an imaging system, one can readily achieve high-throughput real-time label-free biosensing in two-dimensional microarrays and parallel monitoring of numerous biomolecular interactions [5–8].

Four practical SPR imaging approaches based on intensity, angle, phase and spectrum interrogations have been developed and widely reported [9–12]. Both spectrum and angle interrogations have a wide dynamic range, and can provide a homogeneous and optimal response on the entire surface analyzed [13]. More importantly, spectral surface plasmon resonance (λSPR) is much more flexible for optimization because a range of operation wavelengths can be freely selected for obtaining the best SPR excitation [14]. Typically, a λSPR scan is obtained by using either a spectrometer that analyzes the reflected beam or a monochromator that sequentially selects the incidence wavelength [15–19]. The spectrometer-based SPR sensor can provide fast detection of a single sensor site or multiple sensor sites in a 1D line array. However, it is not straightforward for measuring analytes in 2D arrays. So far 2D detection using spectrometer-based SPR sensors has been demonstrated only through intricate mechanical scanning of the sensor chip [20,21]. The measurement speed has been greatly compromised. On the other hand, wavelength-scanning of the incident light is an attractive alternative as the SPR response from each sensor site within the 2D array can be obtained directly through analyzing the intensity distribution of the collected images. The imaging device is effectively monitoring the SPR reflectance distribution of the entire sensing surface. This makes λSPR most suitable for imaging SPR applications. However, data throughput immediately becomes very demanding: Each pixel has to provide a time sequence of signal data points while a specified spectral range is being scanned to locate the SPR absorption dip with an acceptable level of accuracy (typically <0.01nm) [22]. Typically, a monochromator-based SPR system takes ~2s to produce a data point for each wavelength. This mechanical point-by-point wavelength-scanning approach is too time-consuming to be practical for 2D λSPR imaging sensors. Liu et al reported a one-dimensional optical line scan λSPR system for imaging of 2D arrays. For an area of 8mm × 8mm, it took 60s to measure the SPR dips of the 2D array [23]. Wong et al. developed a scanless 2D spectral SPR imaging sensor based on the combination of a polarization control scheme and a color CCD camera. It realized 2D spectral SPR imaging with high sensitivity of 2.7 × 10⁻⁶ in a linear response range from 1.3333 to 1.3365 RIU [24]. Sereda et al. recently developed a special fitting algorithm, which uses only five parameters, for spectrum interrogation of SPR sensing [25,26]. The measurement time has been reduced to 10s. Clearly, fast scanning 2D λSPR imaging technique with a wide dynamic detection range required in the real-time detection of rapid changes in sample refractive index still has a considerable room for improvement.

In this paper, we address the key issue of data throughput in imaging λSPR biosensors through the introduction of an adaptive spectral-dip tracking technique. The system does not have any movable parts. The overall measurement time is approximately 700ms. As far as we know this is the fastest 2D imaging λSPR sensor ever reported in the literature.

2. Experimental setup

Figure 1 shows a schematic of our experimental setup based on the use of an adaptive feedback loop. Light from a 100 W halogen tungsten lamp is coupled to the inverted prism through a multimode optical fiber for SPR excitation. A bandpass filter selects an input wavelength window with a center wavelength of 610 nm and full width half maximum (FWHM) of 220 nm. The LCTF rapidly scans the output wavelength in a stepwise manner with different wavelength steps. For our LCTF unit (VariSpec, VIS-10-20-STD), the FWHM, spectral tuning resolution and response time are respectively 10 nm, 0.5 nm and 50 ms. While the Kretschmann configuration has been adopted, the SPR cell consists of an equilateral prism made from BK7 glass, a microscope glass slide coated with typically 48 nm thick gold film, and a flow chamber for sample injection. The incident angle can be manually set to an optimized value as calculated using Fresnel equations. The output light is captured by a 12-bit CCD camera. The choice of camera exposure time depends on the intensity of light reaching the CCD camera and its sensitivity. An exposure time of 50 ms per wavelength is matched with the wavelength stepping time of the LCTF [8]. The A-D card (NI DAQ 6259) also generates a triggering signal that synchronizes the operation of the LCTF and CCD camera. A timing diagram of control signals is shown in Fig. 2. In our experiments, it takes 50ms for the LCTF to switch the wavelength before the CCD starts its exposure cycle. Consequently, the CCD camera records a long series of images taken at different wavelengths within a prescribed spectral range. Each pixel within the CCD image frame provides one wavelength data point within the λSPR profile of a small region in the sensor surface. Theoretically this one-to-one correspondence between the sensor surface and the 2D CCD imaging pixel arrays may offer massively parallel biosensing capability. Upon analyzing a series of images, one can build a set of spectral plots that cover all the pixels within the sensor surface. Here we use a second order polynomial-fitting algorithm to locate the spectral dip, which corresponds to the resonant wavelength at each pixel. Data analysis is performed in an automatic mode using a custom-made software based on a numerical comparison algorithm.

 

Fig. 1 Schematic of our feedback loop-based imaging λSPR system in the Kreschmann configuration. L_1-7, lens; MF, multimode fiber; DA, diaphragm aperture; LCTF, liquid crystal tunable filter; P₁ and P₂, polarizer.

Download Full Size | PDF

 

Fig. 2 A timing diagram of the control signals.

Download Full Size | PDF

3. Feedback loop

The feedback loop shown in Fig. 1 is designed to produce automatic tracking of the shift in the SPR dip caused by binding of target biomolecules. Most importantly, incorporation of this feedback loop will drastically reduce the spectral width one needs to sweep before hitting an SPR dip. Consequently, the time required to arrive at an SPR resonance data point is significantly reduced. For example, it takes 60s to obtain the resonance dip with a traditional stepping monochromator based on a spectral range of 30nm and a step-size of 1nm. Whereas with the help of a feedback loop, the time required by the present system to find the SPR dip is 700ms. Typically, a conventional λSPR system continuously sweeps through a fixed wavelength range to ensure that the SPR dip never moves beyond the range of interest. Large changes of refractive index may occur because of a long interaction time, high analyte concentration levels or in situations where the size of the target molecules is large. For example, if a fixed wavelength range is $\Delta \text{λ}$ = 30nm, the maximum change of refractive index is ~$\Delta \text{n}$ = 0.69 × 10⁻² RIU, which is calculated using the expression:

Figure 3 schematically shows how the feedback loop operates. First we obtain an initial SPR spectral profile by scanning the incident wavelength across a large spectral range in order to extract the baseline resonant wavelength λ₀. We then choose a spectral range ${\lambda }_0-p\cdot x_S-q\cdot x_L$ to ${\lambda }_0+p\cdot x_S+q\cdot x_L$ with which we can obtain the desired SPR dip with sufficient resolution (<0.01nm). Here, x_S is the small step, x_L is the large step, and p and q are the number of steps on either side of the SPR dip modified by x_S and x_L, respectively. The introduction of x_S and x_L is aimed at speeding up optimization of the feedback loop with variable step-sizes. The corresponding partial SPR plot near the SPR dip (black dotted curve) will be repeated if the refractive index n₀ of the baseline solution does not change. When a change of refractive index has taken place, e.g., from n₀ to n₁, a new partial SPR plot (solid red curve) will be generated by scanning the incident wavelength from ${\lambda }_0-p\cdot x_S-q\cdot x_L$ to ${\lambda }_0+p\cdot x_S+q\cdot x_L$. Thus a new resonant wavelength λ₁ will be extracted, and then the new scanning spectral range will be modified to ${\lambda }_1-p\cdot x_S-q\cdot x_L$ to ${\lambda }_1+p\cdot x_S+q\cdot x_L$. If the refractive index continues to change from n₁ to n_2, the corresponding partial SPR plot (solid green curve) is then updated by scanning the incident wavelength from ${\lambda }_1-p\cdot x_S-q\cdot x_L$ to ${\lambda }_1+p\cdot x_S+q\cdot x_L$. This process will continue until the system has identified the next new resonant wavelength λ₂, which will trigger further modification of the scanning range, i.e., from ${\lambda }_2-p\cdot x_S-q\cdot x_L$ to ${\lambda }_2+p\cdot x_S+q\cdot x_L$. In a more general representation, the k + 1th spectral scanning range, which covers ${\lambda }_k-p\cdot x_S-q\cdot x_L$ to ${\lambda }_k+p\cdot x_S+q\cdot x_L$, is generated from the resonant wavelength λ_k as extracted from the partial SPR plot near the SPR dip that is obtained in the k^th scan covering the spectral range of ${\lambda }_{k-1}-p\cdot x_S-q\cdot x_L$ to ${\lambda }_{k-1}+p\cdot x_S+q\cdot x_L$. For a 2D sensor array, one will obtain a number of different resonant wavelengths from the sensor elements as each of them is designated to perform a specific sensing task. In such a situation, our system will try to cover the entire range of all the elements by registering the lower and upper limits, i.e., λ_kmin and λ_kmax, of the spectral dip, and the range will become ${\lambda }_{k\min }-p\cdot x_S-q\cdot x_L$ to ${\lambda }_{k\max }+p\cdot x_S+q\cdot x_L$. In this case, the first and last wavelength data points of the k + 1th spectral scanning range near the SPR dip, λ_begin and λ_end, can be calculated using the following expressions:

 

Fig. 3 Schematic showing the tracking of an SPR dip based on the feedback loop technique. n₀, n₁…, n_k and n_{k + 1} represent a series of refractive index values at the Au/sample interface when molecular binding is taking place at the sensor surface. The curves represent the partial SPR plots near the SPR dips on the different refractive index values.

Download Full Size | PDF

In particular, if the index spread on the sample is small enough, i.e., the corresponding shift in resonance dip is less than the system resolution of 0.01nm, the change may become smaller than a single step, and λ_kmin and λ_kmax will be the same. In our experiments, to achieve the fastest speed with the best sensitivity, the choice of parameters is p = 1, q = 2, x_S = 1nm, and x_L = 7nm, so that the width of the spectral scanning range is 30nm, and the measurement time for a SPR dip becomes the number of steps multiplied by the measurement time for each SPR dip, which is 7 × 100ms = 700ms.

4. Results and discussions

Fast spectrum interrogation capability based on our adaptive spectral-dip tracking approach is experimentally demonstrated by measuring the shift of refractive index of different salt-water mixtures from an imaging λSPR sensor. Salt solutions were prepared with concentration level ranging 0% to 30% in increments of 5% by volume, which corresponds to a refractive index ranging from 1.3330 to 1.3885 RIU [12]. In the experiment, an incident angle of 72.5° was chosen based on a calculated dip location with water as the sample. The wavelength scan range was about 30nm, and the time for finding the minima of the spectral dip was about 700ms. The 3 × 3 array sensor sites, which were uniformly distributed on the chip, were monitored simultaneously in terms of their spectral SPR characteristics to assess the performance of our system. The step changes of resonant wavelength obtained from different salt concentration samples are shown in Fig. 4. Based on measurement data obtained in the wavelength range from 600nm to 700nm, the calculated sensitivity limit is 4.69 × 10⁻⁶ RIU, and the measured dynamic detection range of the system is 5.55 × 10⁻² RIU. Here, the conversion from minimum measurable wavelength shift to the corresponding change in refraction index value is performed with the formula [12]

 

Fig. 4 Shift in resonance dip versus salt concentration in water as detected by sensor sites within the 3 × 3 array on the SPR chip. Each data point was obtained by averaging the intensity values of 30 × 20 pixels, corresponding to the measurement region of 139 µm × 179 µm on the sensing chip, to reduce the effect of noise with no compromise in spatial resolution.

Download Full Size | PDF

 

Fig. 5 Resonance wavelength shift obtained from salt-water solutions with weight concentrations from 0% to 1% in the uniform and variable step-size modes.

Download Full Size | PDF

 

Fig. 6 Shift in resonance dip of pure water as a function of time in the uniform and variable step-size modes.

Download Full Size | PDF

We also analyzed the effects of chromatic aberrations on the incident angle at the interface of Au/prism when the incident angle at the prism’s bevel is fixed. Based on the relationship between the refractive index of BK7 glass and wavelength, for our system it can be calculated that the incident angle at the interface of Au/prism changes 2.7 × 10⁻⁴ degrees when the incident wavelength changes 1nm. In the scanning range ${\lambda }_0-15nm$to${\lambda }_0+15nm$, the change of the incident angle could be limited in range −4.05 × 10⁻³ to 4.05 × 10⁻³ degrees, much less than the angle resolution (0.01°) in angular SPR systems [27]. Moreover, when the incident wavelength changes from ${\lambda }_0-15nm$to${\lambda }_0+15nm$the movement distance of the position of excitation light on the sensing chip along the tangential wave vector of the incident light is limited in the range −0.5 to 0.5μm. The position of the excitation light on the sensing surface does not move along the direction perpendicular to the tangential wave vector of the incident light. The 27.8 µm × 35.8 µm spatial resolutions on our sensing chip (corresponding to 6 and 4 pixels on the CCD imaging chip) are measured using the resolution plate. The distance of 0.5μm corresponds to 0.06 pixels of the CCD imaging chip. Therefore, the effects on the performance of our system of the chromatic aberrations created by scanning the incident wavelength can be ignored.

The biosensing capability of our imaging λSPR sensor is demonstrated by detecting interactions between protein molecules. In our experiments, two specific binding interactions involving goat anti-rabbit IgG binding with rabbit IgG are monitored. The protein probe, rabbit IgG, was first immobilized on 3 × 4 array sites on the sensor chip. We manually tune the incidence angle to achieve minimum intensity at the SPR dip for a PBS sample. We have also developed software to analyze the signal captured by the CCD to convert the SPR shift in the 12 sensing sites. When PBS sample was injected on the sensor chip, it was possible to obtain a stable baseline within reasonable time. When goat anti-rabbit IgG (10μg/ml) were injected, the subsequent antigen–antibody binding reactions taking place on the spotted sensing regions were monitored continuously. A set of 12 spectral SPR plots representing the binding reactions is shown in Fig. 7(a). The image of SPR dip shift obtained by comparing the resonance wavelengths before and after binding interactions is shown in Fig. 7(b).

 

Fig. 7 Measurement of antigen-antibody interaction between goat anti-rabbit IgG and rabbit IgG. (a) Real-time wavelength response of antigen–antibody binding reactions. (b) Image of SPR dip shift.

Download Full Size | PDF

5. Conclusion

We have demonstrated a novel adaptive spectral-dip tracking approach for high-throughput imaging λSPR biosensors with no mechanical moving parts. The combination of scanning the incident light wavelength using LCTF and a feedback loop that also incorporates variable wavelength step-size, the reported feedback loop-based λSPR sensor offers the possibility of real-time detection of rapid changes in sample refractive index. As compared to conventional imaging-based λSPR sensors, the experimental setup shows similar values in detection sensitivity [24] and dynamic detection range [23]. The detection speed is among the fastest reported in literature [22,28]. The detection sensitivity is slight larger than that expected in theory, which is primarily related to the passband width of 10 nm imposed by the LCTF device, there is much scope for further improvement if one uses a tunable filter of higher performance. Secondly, since the software fits the SPR profiles based on the sparse sampling data, uncertainties of fitting error may occur between the present measured dip location and the next measured dip location. Our fitting algorithm should also be further investigated in the future. The reported scheme may lead to truly real-time parallel sampling of a number of molecular interactions. This will enable real-time affinity comparison between many biomolecular species as required by many applications such as drug screening and proteomics.