1. Introduction

Identification of severe diseases and advanced medical diagnoses are of increasing importance in aging societies [1], and therefore daily examinations of diagnostic biomarkers in blood and humor will have long demands. Although some easily-detectable biomarkers are already being serviced in health clinics, low-cost and high-throughput detection of very small amounts of specific biomarkers suitable for earlier and more reliable diagnoses is still a challenge [2]. Common immunological methods such as enzyme-linked immunosorbent assay allow a detection limit (DL) of pM order [3], which barely meets the requirement for these biomarkers, and yet has problems that they need time-consuming functionalization procedures of fluorescent labels where the labels themselves are suspected to change the original properties of the biomarkers. As label-free methods, photonic sensors [4–10] and those based on other physical parameters [1,11,12] have been developed. However, none of them simultaneously satisfy all the requirements of practical medical sensors: low cost, disposable use, simple procedure, high sensitivity, and high selectivity against contaminants.

In this paper, we present remarkably high-performance biosensing using photonic crystal (PC) nanolasers. The PC consists of GaInAsP semiconductor thin membrane with periodic airholes (PC slab) and some non-periodic airholes as a nanocavity [13]. Since its total area is no larger than 20 × 20 μm², high-throughput low-cost fabrication and production will be feasible even using e-beam lithography. It is easily operated by room-temperature photopumping using a standard micro-photoluminescence (μ-PL) setup [13–15]. In this device, the evanescent field of the laser mode penetrates outwards from the PC slab, and the laser wavelength is changed by the environmental index, which is the principle of the sensing. In addition, incorporating a nanoslot (NS) [13,16,17] into the nanolaser enhances the localization of the laser mode in water and improves the sensitivity. The localization also improves the thermal stability because the positive thermo-optic effect in semiconductors is canceled by the negative effect in water. This suppresses the spectral chirping due to heating and greatly narrows the spectrum, resulting in higher resolution of the sensing. Using the NS nanolaser, we have detected bovine serum albumin (BSA) of 255 fM concentration [18], which means 3 − 4 orders higher sensitivity than that of common surface plasmon resonance (SPR) sensors [6]. In this paper, we demonstrate an even higher sensitivity for biotin-streptavidin (SA) specific binding and detection with an ultra-high selectivity against BSA as the contaminant. Consequently, we show that this device satisfies potentially all the above requirements for practical medical sensors.

2. Device and method

PC nanolasers are fabricated on a 180 nm thick GaInAsP quantum-well active layer epitaxially grown on (100) InP substrate, as shown in Fig. 1. We employ the simple process of e-beam lithography and HI inductively coupled plasma etching for making holes into the GaInAsP layer and HCl etching of InP beneath the holes to form the air-bridge structure. In this experiment, we prepared devices with and without a NS (NS width w_NS = 20 − 80 nm) at the center of the cavity. The basic cavity structure is H0-type consisting of four shifted airholes [13]. The lattice constant a = 500 nm, airhole diameter 2r = 220 nm, and airhole shifts s_x = 120 nm and s_y = 80 nm. The periodic modification (airhole diameter 2r’ = 190 − 210 nm inside dotted lines) is added to enhance the vertical emission and enable easy detection in our setup [19]. Such devices are integrated in a 4 × 4 arrayed configuration within a footprint of less than 100 × 100 μm². In one array of NS devices, w_NS is varied so that we can evaluate sensor signals from various designs.

 

Fig. 1 Optical micrograph (left) and scanning electron micrograph (right) of GaInAsP NS PC nanolasers. Dark and bright areas of the left picture are those with lower InP removed and remained. Diameter of airholes inside dotted lines in the right picture is reduced from 2r to 2r’.

Download Full Size | PDF

In the sensing experiment, the device chip is first thermally pre-oxidized at 180°C for one hour. If we do not this, GaInAsP surfaces are thermally oxidized during the laser operation in water. This reduces the effective index of the laser mode and results in the wavelength blueshift, which obscures the redshift due to the adsorption of proteins. The pre-oxidization is effective for suppressing additional oxidation in brief measurement. In addition, we recently found that the oxidation can be suppressed completely by coating the device with a transparent film. The detail on this matter is presented elsewhere [20]. Therefore, it will not be a problem even in future experiments. Secondly, the device surface is treated to be hydrophilic in O₂ plasma and functionalized with amino group in dehydrate toluene mixed with 0.05% 3-aminopropyl triethoxysilane (APTES) [21] for 5 min, and also with biotinylation in water mixed with 50 mM water-soluble carbodiimide (WSC) [22] and 50 μg/ml biotin for one hour. Then, the chip is soaked in a sample liquid for 5 min, where the sample is a mixture of pure water, SA (concentrations of 0 – 100 pM), BSA (0 − 10 μM), and 0.1% surfactant Tween 20 [23]. Both the specific adsorption of biotin-SA and non-specific adsorption of BSA can occur, as shown in Fig. 2, and the surfactant is expected to suppress the non-specific one. Finally the chip is rinsed by pure water, soaked in fresh pure water, and each device is photopumped at λ = 0.98 μm one by one through a μ-PL setup with a computer-controlled stage. The laser emission is coupled into optical fiber and its wavelength is measured using optical spectrum analyzer (OSA), Advantest Q8383 (resolution limit: 100 pm) or Q8384 (10 pm). Pulsed pumping (duty ratio = 200) is employed to avoid excess heating and further oxidation. In general, the laser wavelength redshifts with the adsorption. Wavelength shifts Δλ of the 16 devices (sometimes less than 16 when some devices do not emit enough power for the spectral analysis) are used for statistical evaluations such as averaging, confidence interval and t-test.

 

Fig. 2 Schematic of biotin-SA binding in impure sample.

Download Full Size | PDF

3. Results and discussion

First, we detected the biotin-SA binding in the absence of BSA and Tween 20. Here, samples of various SA concentrations were prepared by repeatedly pipetting and diluting the SA solution. To avoid any contaminations, the tip of the pipetter was replaced by a new one every time. We confirmed from the absorbance of dye-functionalized proteins that the protein concentration was diluted reasonably in this procedure at least in the μM regime (nM and lower concentrations are difficult to quantify in this method). The detection measurement was done in water, as aforementioned, after soaking the chip in each sample from low-concentration side and rinsing in pure water. When the above procedure of the soaking, rinsing, and measurement using Q8384 was repeated only with pure water not including SA, Δλ was less than 60 pm. We regard it as a noise level affected by the fluctuation in the measurement condition. Incidentally, we did not use buffer solutions including salt ions in this experiment to avoid the etching of the device surface. Similarly to the oxidation, such etching causes the wavelength blueshift and obscures the sensing. Since the aforementioned film coating also suppresses the etching completely, it will not be a problem even in buffer solutions in future experiments.

Figure 3(a) shows example lasing spectra, in which the spectral width of 150 pm is mainly limited by Q8383 used in this measurement. Clear wavelength shifts Δλ larger than the spectral width are observed even at extremely low concentrations of SA. Figure 3(b) shows the Δλ versus SA concentration characteristics for six devices with different w_NS. All the devices with w_NS > 0 exhibit such meaningful Δλ at concentrations from 10 − 100 zM. Fundamentally Δλ increases with increasing the concentration, but in some devices decreases in the fM − pM regime, which might be due to the slight oxidation of the device after repeated measurements. In addition, Δλ increases with decreasing w_NS, which is attributed to the enhanced interaction of the laser mode and adsorbed SA in the NS [17]. Figure 3(c) compares the average Δλ with and without NS. Corresponding to (b), Δλ starts appearing from the SA concentration of 10 − 100 zM. When the measurement was done without biotin, Δλ did not increase until higher than 1 nM. Therefore, Δλ at lower and higher than 1 nM are dominated by the specific binding of biotin-SA and the physisorption of SA, respectively. Solid lines are drawn from the Langmuir fitting assuming the linear combination of three different adsorption terms: two chemisorptions and one physisorption [18]. Table 1 summarizes the parameters and DLs obtained from the fitting. In Fig. 3(c) and Table 1, the results for SA are also compared with our previous results for the non-specific adsorption of BSA [18]. Affinity constants K_A1 and K_A2 associated with chemisorptions of SA are 6 − 9 orders higher than those of BSA. Here, K_A2 almost correspond to original affinity constants for SA and BSA seen in literatures [24,25]. Therefore, the very low DL for SA arises primarily from the large affinity constant for SA. Furthermore, K_A1 associated with Δλ at lower concentrations is enhanced and the DL is reduced by 3 − 4 orders with the NS. Such a huge K_A1 cannot be expected only from thenature of the chemisorption. We have discussed previously the possibility that the optical gradient force captures the protein and enhances the affinity and sensitivity effectively [18]. In this experiment, however, the adsorption must take place during the soaking in the samples before the devices are operated in pure water. Therefore, the optical gradient force during the laser operation cannot contribute to the enhancement of the adsorption but only to the suppression of the desorption. Although the results shown here is reproducible in repeated experiments, its mechanism is still an open question.

 

Fig. 3 Wavelength shifts due to adsorption of protein in pure samples. (a) Example laser spectra at different SA concentrations. (b) Δλ with SA concentration for six devices with different w_NS. (c) Comparison of average Δλ for devices with NS (closed circles) and without NS (open circles). Data for SA were obtained in this study, while those for BSA are from Ref [18].

Download Full Size | PDF

Table 1. Fitting parameters and DL evaluated from each solid line.

View Table

Next, we detected the biotin-SA binding in the presence of BSA and Tween 20. Here, surfactant Tween 20 hydrophilizes all the surfaces of proteins and devices. So it suppresses the non-specific adsorption of BSA, while does not disturb the biotin-SA binding as it desorbs from SA on this condition. Figure 4 summarizes the average Δλ with different SA concentrations in the presence of three different concentrations of BSA as the contaminant. Results in the absence of SA do not exhibit significant difference depending on the presence of biotin. When the BSA concentration is 100 nM, Δλ increases with increasing SA concentration only with the biotin; it was negligible without biotin. This indicates that biotin-SA specific binding occurs. When the BSA concentration is 1 μM, almost similar results are observed, although Δλ overall increased for both with and without biotin. This increase shouldbe due to the increase in the non-specific physisorption of BSA even with Tween 20 because of the very high concentration of BSA. Here, BSA might be adsorbed on the device surface as multi-layers and prohibits SA adsorption. This background noise increased more and obscured the main signal from the biotin-SA binding when the BSA concentration was 10 μM. Here we can see the gradual decrease in Δλ with increasing the SA concentration. This might also be due to the aforementioned oxidation of the device after repeated measurements. Thus we confirmed that the biotin-SA adsorption was detected up to a SA concentration of 100 zM against 1 μM BSA. The corresponding selectivity is as high as (SA/BSA) = 10¹³. It is known that such albumin is included by ~1mM in human blood. Such high concentration completely hampers the detection of low concentration target biomarkers even using our nanolasers. Therefore, our strategy is such that we first dilute the blood plasma to ~1μM, a thousand-fold lower concentration, and then detect the target. Antibodies of target biomarkers typically have affinity constants of 10⁹ – 10¹¹ M⁻¹ order, which are three orders lower than biotin-SA’s. Considering the DL of our nanolaser sensor for biotin-SA, the DL of target biomarkers will be in the sub-fM regime, which is equivalent to sub-pM regime before the dilution. This is a very promising number that enables the detection of various biomarkers of interest for medical diagnoses.

 

Fig. 4 Wavelength shifts due to adsorption of SA in impure samples. (a) Laser spectra of 16 devices under three different conditions. Some spectra for #6 and #10 were missing in the automatic measurement, which might be due to position errors of pump light and weak emission intensity from nanolasers. (b) Average Δλ due to the adsorption of SA in impure samples with three different concentration BSA as impurities. Red and blue plots show those with and without biotin, respectively. Error bars indicate the range for 95% confidence interval.

Download Full Size | PDF

4. Conclusion

Using GaInAsP photonic crystal nanoslot nanolasers modified by biotin, we observed the biotin-SA specific binding for which the detection limit concentration was evaluated from the Langmuir fitting to be 16 zM. We also succeeded in detecting 100 zM SA in the impure sample with 1 μM BSA as the contaminant. This means that the ultra-high selectivity (SA/BSA) = 10¹³ was achieved. Even when considering different affinities of the biotin-SA binding and actual antibody-antigen binding of biomarkers, this high sensitivity and selectivity will allow simple and fast examination of biomarkers. In addition, the simple fabrication process and measurement of the nanolaser will enable low-cost mass-production, disposable use, easy handling, simple procedure and testing. Thus, our nanolaser will be the promising candidate as a practical biomarker sensor in medical diagnoses.