1. Introduction

Nanostructures with a certain concentration of surface free electrons can generate an enhanced hot spot on their surface under the condition of surface plasmonic resonance, which is the result of collective surface free electron oscillation. As a classical physical effect, it has important applications in scientific research, such as surface enhanced Raman scattering (SERS), enhanced infrared absorption and enhanced optical nonlinear conversion efficiency [1–5]. Among them, the intensity of SERS signal is directly proportional to the strength of local surface electric field, and the enhancement of surface electric field is directly related to the substrate of SERS. Here, the surface plasmonic resonance generated by SERS substrate is the reason of the surface electric field enhancement. Therefore, improving the efficiency of SERS substrate coupling excitation light is important to signal enhancement and sensitivity improvement.

Usually, the interaction between two plasmonic particles gives a rise to interesting hybridization of the plasmonic modes, which can be utilized to tune the resonance frequency of the system. In particular, the interaction between two coupled oscillator-models leads to the electromagnetic induced transparency (EIT). EIT is a coherent quantum interference effect originally observed in three-level atomic gas systems [6,7], however, the occurrence of EIT-like effects is not necessarily limited to quantum systems. The design of metamaterials based on the two coupled oscillator- models has been reported, which provides a variety of options for the design of novel photonic devices and optical sensors [8–10]. Two coupled oscillators driven resonance model is used to describe the optical EIT (Section 1 of Supplement 1). Usually, the resonant mode with larger damping is manifested as a wide stop band in transmission. However, the transmission can be resonantly restored by the coupling to the undamped mode with the opposite phase, and it leads to a narrow band of optical transparent window [11–14]. Compared with the traditional method for generating EIT by using two beams to interact with atomic media, the metamaterial designed based on the theoretical model can be realized only by excitation with a single beam [15]. The large near-field enhancement generated by this method is beneficial to harmonic generation, SERS and other optical processes [16–18]. SERS is often used in molecular level detection of substances. In particular, the functionalized SERS chip has great potential in the application of cancer diagnosis due to specific detection of protein molecules.

Liver cancer is the fourth most common cause of death from cancer worldwide [19], and the incidence of liver cancer shows a trend of younger. Since the mortality of cancer is directly related to the pathological state at the time of diagnosis, early diagnosis of HCC is necessary. However, for early HCC, traditional clinical detection techniques are not only difficult to detect, but also costly [20,21]. In addition, alpha-fetoprotein (AFP) is often used as a biomarker for early diagnosis of liver cancer, and serum AFP level is an important basis for the diagnosis of liver cancer. But this method is still flawed, because other liver diseases can also lead to elevated serum AFP levels, and this is detrimental to the diagnosis of early HCC. Therefore, the development of a new biomarker with specific and sensitive signal expression is necessary for the diagnosis of early HCC.

Recently, Lens culinaris agglutinin (LCA)-reactive fraction of AFP (AFP-L3) is considered to be a specific marker for the diagnosis of HCC [22], and AFP-L3 is a biological protein only found in patients with liver cancer. Clinical studies indicated that the ratio of AFP-L3 to total AFP level (AFP-L3%) should be a new clinical diagnostic standard for accurate judgment of liver cancer [23]. In general, AFP concentration in healthy individuals is less than 10 ng/mL, and the diagnosis of HCC should be considered when AFP-L3%≥10% [24]. At present, several methods for detecting AFP-L3 have been reported, including fluorescence immunoassay [25], light-scattering immunoassay [26], electrochemistry [24]. However, these methods are limited by complicated detection processes, long time, high cost and poor sensitivity, and not applicable to low total AFP concentration. The detection of AFP-L3 using SERS method has been reported [27], and the detection limit reached 8 ng/mL, which could meet the diagnostic requirements for early HCC. Further improvement of detection sensitivity will be more helpful for the prevention of early HCC and postoperative monitoring.

Herein, we constructed a sandwiched immunoassay method based on SERS, in which two independent SERS probes, 4-mertobenzoic acid (4-MBA) and 5, 5-dithiobbis (succinyl-2-nitrobenzoic acid ester) (DSNB) were used. Immunoassay was based on both the frequency shift and intensity change of SERS probes. The peak located at 1075 cm⁻¹ (“black line 1” in Fig. 1(a)) from the SERS spectra could be ascribed to the benzene ring-S-silver (ph-S-Ag) bond after the conjugation of anti-AFP/4-MBA to Ag. Binding with AFP caused deformation of the phenyl ring-S-silver (ph-S-Ag) surface complex because of the high molecular weight of AFP. Thus, the fingerprint of MBA in this condition will only cause a change to the band resulting from a change in the polarizability of the ph-S-Ag complex. AFP was quantitatively determined by frequency shift method (“red line 2” in Fig. 1(a)).

 

Fig. 1. Schematic of the SERS-based immunoassay. (a) EIT-like metamaterials were modified different types of protein molecules on the surface, subsequential gold nanoparticle immune colloids were captured to form a sandwich structure. The curve shows that SERS spectra of antibody capture chip fabrication, and SERS spectra of the two Immuno-processes. Curve 1 represents the SERS spectrum of 4-MBA and antibody modified metamaterial (balck line). Curve 2 represents SERS spectrum of AFP combined with substrate. Due to the high molecular weight of AFP, binding with the antibody on the substrate surface results in the deformation of the phenyl ring-S-Silver (ph-S-Ag) surface complex, which manifested as a shift of the Raman peak at 1075cm⁻¹ (red line). Curve 3 represents the SERS spectrum of immune colloids combined with AFP-L3. After the addition of immunogold colloids, a new peak appeared at 1331cm⁻¹ because of the strong scattering cross-section of its symmetric NO₂ stretch of DSNB (blue line). (b) Synthesis principle of immune colloids. DSNB, as a coupling agent, was modified on the surface of gold nanoparticles by coordination bonds. Antibodies were modified on the surface of the gold nanoparticles via peptide bonds, which formed an immune gold colloid.

Download Full Size | PDF

AFP-L3 is a biological protein only found in patients with liver cancer. Clinical studies indicated that the ratio of AFP-L3 to total AFP level (AFP-L3%) should be a new clinical diagnostic standard for accurate judgment of liver cancer. Based on the above reasons, we needed to measure the specific concentration of AFP-L3 by SERS method. In the experiment, AuNPs were modified with DSNB and used as immunogold colloid, in which DSNB was used as a coupling agent between the antibody and AuNPs (Fig. 1(b)). The strong SERS band at 1331 cm⁻¹ was chosen as a signal of immunogold, because of the strong scattering cross-section of its symmetric NO₂ stretch of DSNB (“blue line 3” in Fig. 1(a)). The plot of the SERS intensity ratio (I₁₃₃₁/I₁₀₇₅) showed a response to AFP-L3 concentration change. The intensity of the band at 1331 cm⁻¹ gradually increased with an increase in the AFP-L3 concentration. Here, the Raman peak of 4-MBA at 1075cm⁻¹ is taken as the internal standard because its Raman peak strength is stable.

In this paper, we designed and prepared EIT-like Ag metamaterial based on Ref. [15]., and used it to construct SERS biosensor substrates through surface molecular modification [28–30]. Compared with the conventional surface plasmon resonance based on a metal surface, the coupling and interference between the dipole (“bright”) and electric quadrupole (“dark”) mode from EIT-like metamaterial structure could generate a stronger surface electric field, which would significantly increase the SERS enhancement factor. As a result, the signal intensity of SERS could be further enhanced. The substrate has obvious Raman scattering enhancement effect, and the detection limit of AFP-L3 is ∼0.01 ng/mL. The feasibility and accuracy of EIT-like metamaterials for biological protein molecular detection are verified by comparing to clinical detection results.

2. Results and discussion

2.1 Fabrication and characterization of the EIT-like metamaterial

The resonant frequency of Ag metamaterial can be adjusted by changing the size and period of the structure. The balance of energy loss and group velocity dispersion should also be considered in the structure design of the EIT-like plasmonic substrate. Figure 2(a) shows the SEM image of the fabricated metamaterial. The metamaterial consists of periodic unit structures. Each unit structure includes longitudinal single strip and horizontal parallel double strips, which are fabricated by electron beam etching of a 70 nm thick Ag film grown on a 0.5 mm thick sapphire substrate. The Ag material was chosen because of its low inherent loss and easy surface molecular modification. This structure has Fano-like resonance, leading to classical simulation of EIT peak. For each unit, a single Ag strip acts as a dipole oscillating antenna, and its polarization along the long axis is strongly coupled to the incident light, which can serve as the “bright” mode for an EIT-like plasma system, meanwhile, the “dark” mode consists of two separate parallel metal strips, which have an antisymmetric mode.

 

Fig. 2. Characterization and stimulation of the fabricated metamaterial. (a) The SEM image of the Ag metasurface on a sapphire substrate. A single unit cell consists of a dipolar Ag bar antenna and a double parallel Ag bars resonator to its left. Geometric parameters of the metamaterial (above the sapphire) are: Ag thickness = 80 nm; length of the Ag bar antenna = 200 nm; width of the Ag bar antenna = 180 nm; length of the parallel Ag bars = 150 nm; width of the parallel Ag bars = 70 nm; the interval of the parallel Ag bars = 70 nm; distance from the centre of the Ag bar antenna to the centre of left parallel Ag bars = 235 nm; unit cell period = 700 nm. (b) Energy level diagram of mode coupling of three level EIT resonant system and represents the coupling parameter. (c) Transmission spectra of metamaterials obtained from simulation (blue line) and experiment (red line). (d) The electric field amplitude profiles of the “bright” mode (left) and “dark” mode (right).

Download Full Size | PDF

The antisymmetric mode has a reverse propagation current and thus cannot be excited directly by normal incident light, but can be coupled to the “bright” mode through the near field. The resonant frequencies of the corresponding structures of the two modes coincide with each other, and the interference of the broad band “bright” mode and the narrow band “dark” mode forms a typical three-level EIT system, as shown in Fig. 2(b), which generates a sharp EIT peak in the transmission spectrum. Compared with atomic EIT system, the coupling strength of two energy levels in plasma EIT system is determined by their space distance. We simulated the linear optical properties of the structure based on FDTD. Figure 2(c) shows the transmission spectra of simulation and measurement. In the experiment, an EIT peak was observed near 788 nm, which is in good agreement with the simulation results. The wide band absorption corresponds to the bright mode of the metamaterial, and the sharp EIT peak corresponds to the excitation of the double dark mode. Figure 2(d) corresponds to the electric field distribution of the two modes.

2.2 SERS performance of the EIT-like metamaterial

We used R6G molecule to evaluate the SERS performance of EIT-like metamaterial and employed the single resonance mode metamaterial as the comparison, and the results showed that the SERS enhancement of EIT-like metamaterial was one order of magnitude higher than that of the single resonance mode metamaterial (Fig. S1 of Supplement 1).

2.3 SERS-based determination of AFP/AFP-L3

EIT-like metasurface was used as the SERS substrate to determine the concentration of AFP. First, the 4-MBA was anchored onto the EIT-like substrate through the Ag-S bond. The conjugation of anti-AFP antibody to the substrate was achieved through the coupling between the amino group in anti-AFP and the carboxyl group of MBA using EDC and NHS as activators. Subsequently, the anti-AFP antibody binded to the substrate via a derived coupling agent to form an antibody capture substrate. We used SERS method to detect the effect of surface modification. Specifically, the peak at 1370 cm⁻¹ of MBA's COO- stretching pattern disappeared when it was combined with biological macromolecules such as antibodies (Fig. 3). At the same time, 1573 cm⁻¹ peak corresponding to NH-CO mode appeared, and the changes of the two peak positions indicated that MBA-based antibody was successfully modified onto the substrate.

 

Fig. 3. (a) Schematic of antibody modification on substrate. (b) SERS spectra before (1) and after (2) antibody addition.

Download Full Size | PDF

We detected AFP using a modified substrate. As shown in Fig. 1, the detection of AFP was reflected in the deviation of 1075 cm⁻¹ peak, which was a vibrational frequency shift owing to antibody conjugation [27,31,32]. In the experiment, the substrate blocked by BSA were immersed in AFP (in PBS) solutions with different concentrations, and then the Raman test was conducted, as shown in Fig. 4. It could be seen that, compared with the blank sample, the AFP concentration increased from 0.1 ng/mL to 100 ng/mL, and the peak at 1075 cm⁻¹ had obvious deviation. The change of peak offset was a function of AFP concentration (Fig. 4(c)). The frequency shift met a linear relationship with the logarithm of AFP concentration,, where was the logarithm concentration of AFP, and was the change of frequency shift. In the experiment, the minimum detection concentration of the substrate was 0.1 ng/mL. Currently, the most common diagnostic method for HCC was chemiluminescence which detected an increase of only AFP levels in the blood. However, in addition to HCC, other factors could also lead to an elevated serum AFP concentration, such as hepatitis and cirrhosis. Compared with the concentration of AFP alone, it was more accurate to analyze the ratio of AFP-L3 to AFP as a diagnostic index of HCC, which might be a more effective analysis method to avoid misdiagnosis due to false positive or false negative results. Therefore, the determination of AFP-L3 is necessary.

 

Fig. 4. (a) SERS spectra of the substrate modified with different concentrations of AFP. (b) SERS spectra is offset at 1075cm⁻¹ (The labeled region in Fig. 5(a)). (c) Function plot of the AFP concentration logarithm and the 1075 cm⁻¹ peak frequency shift. (d) SERS spectra of different AFP-L3 concentrations (0.01−100 ng/mL) are detected by substrate. (e) Function curve of the intensity ratio (I₁₃₃₁/I₁₀₇₅) and the AFP-L3 concentration.

Download Full Size | PDF

In the experiment, we used the classical “sandwich” structure to detect AFP-L3. Different from AFP detection, we synthesized DSNB and modified Au nanoparticles with DSNB (Fig. S2 of Supplement 1). DSNB was used to link antibodies to capture AFP-L3 on the substrate. Figure 4(d) showed the concentration dependence SERS spectrum of the immunogold modified substrate, and the peak 1331 cm⁻¹ generated by the symmetric NO₂ stretching of DSNB was used as the characteristic immunogold signal. The X-ray photoelectron spectroscopy (XPS) data (Table S1 of Supplement 1) also prove that the AFP-L3 antigen on the SERS-active chip is successfully immunoreacted with the AFP-L3 antibody on the immunogold.

Here, the peak value near 1075cm⁻¹ of MBA was selected as the benchmark to correct the influence of signals caused by non-AFP-L3 concentration changes on the results. The relationship of the intensity ratio of SERS (I₁₃₃₁/I₁₀₇₅) to the concentration of AFP-L3 was shown in Fig. 4(e). In the range of 0.01∼100 ng/mL, the linear relationship fitting was obseved, where the x-axis was the logarithmic concentration of AFP-L3, the y-axis was the ratio of I₁₃₃₁/I₁₀₇₅. The above results showed that SERS substrate based on EIT-like metamaterials could be used to evaluate the proportion of AFP-L3.

2.4 Clinical sample analysis

To verify the feasibility and accuracy of EIT-like metamaterials for detecting biological protein molecules, the substrate was used to detect the percentage of AFP-L3 in a clinical serum sample from a 29-year-old female patient with HCC. Human serum sample were collected from patient in the People’s Hospital of Xili Shenzhen with approval by the Ethics Committee with informed written consent. According to the obtaining time of the sample, the serum AFP concentration was determined to be 611.3 ng/mL by chemiluminescence method (See section 4 of Supplement 1), and the results were directly compared with the clinical results by SERS substrate. In the experiment, after the serum was diluted 100 times, AFP was determined using SERS with substrate. For the determination of AFP-L3 level, the standard curve of AFP-L3 (Fig. S3 of Supplement 1) was constructed by ELISA (0.3∼20 ng/mL), and then the serum was determined.

We performed SERS test on serum from HCC patients. As shown in Fig. 5(b), the peak offset of MBA at 1075 cm⁻¹ was 4.51 ± 0.3 cm⁻¹. According to the standard curve in the experimental results, the corresponding AFP concentration was 607.17 ng/mL. The intensity ratio of SERS (I₁₃₃₁/I₁₀₇₅) was 0.4937. According to the standard curve, the corresponding concentration of AFP-L3 was 601.1 ng/mL. The optical density value of ELISA standard curve was 0.415, and the corresponding AFP-L3 concentration was 605.2 ng/mL. The proportion of AFP-L3 determined by both methods was 99%, and the patient was diagnosed with HCC, which was consistent with the clinical diagnosis. The results showed that compared with the traditional method, we proved that SERS substrate based on EIT-like metamaterial was feasible and accurate for the detection of HCC. Moreover, 4 clinical samples from the serum of patients with cancer were tested with the results are shown in Fig. S4 of Supplement 1. As a control experiment, we also tested the serum of cirrhosis patients and healthy people (Figs. S5 and S6 of Supplement 1), which also support the feasibility and sensitivity of the proposed SERS-based assay for the early diagnosis of HCC.

 

Fig. 5. (a) Raman spectra of the (1) SERS-active substrate alone, and after (2) exposure to the serum of an HCC patient, and (3) anti-AFP-L3 modified on immune gold colloids. (b) Detection of AFP in clinical samples based on SERS active substrate.

Download Full Size | PDF

3. Conclusion

In this paper, we constructed a metamaterial EIT-like structure according to the simulation results of FDTD, which proved its superiority in SERS detection. Based on this result, SERS substrates for both AFP and AFP-L3 detection were constructed, and we accomplished quantitative determination of AFP and AFP-L3 with a good linear relationship to concentration. Meanwhile the detection limit of AFP-L3 reached 0.01 ng/mL. The AFP-L3% could be calculated for the diagnosis of HCC, which was consistent with the clinical results. The proposed EIT-like SERS substrate can be used as a highly sensitive rapid detection tool, providing valuable reference for the early diagnosis of cancer and other diseases.

Experimental section

Materials. 5,5′-Dithiobis(2-nitrobenzoic acid) (DTNB), dicyclohexylcarbodiimide (DCCD), Tetrahydrofuran solution, HAuCl₄, AFP-L3 ELISA kits, bovine serum albumin (BSA), phosphate-buffered saline solution (PBS; 0.01 M, pH 7.4), Borate buffer, 1-(3-(dimethylamino) propyl)-3-ethylcarbodiimide hydrochloride (EDC), and N-hydroxysuccinimide (NHS) were bought from Beijing Dingguo Changsheng Biotechnology Co., Ltd. (China). All chemicals were analytical-grade reagents and used without further purification. The ultra-pure water used in the experiment is self-made.

SERS characterization. A Raman microscope system (WITech) with a 785 nm laser source was used for Raman investigation with an operational range of 400 − 2000 cm⁻¹. All the Raman measurements were performed under the same experimental conditions (i.e., ×20 microscope objective lens, 66 mW laser power, 600 g/mm diffraction grating, and 5 s acquisition time).

Preparation of antibody capture substrates. The substrate was immersed in 4-MBA ethanol solution (10⁻⁴ M) for 4 h. The substrates were immersed in a mixed solution of NHS/EDC (in MES solution) for 3 h, and washed three times by ultrapure water, and then dried softly with Ar gas. Anti-AFP antibodies (50 µL) were then immobilized on the substrates by incubation at 37 °C for 4 h. Then, the substrates were washed three times with PBS solution (0.1% BSA, pH 7.4) and dried gently under Ar gas. Finally, nonspecific binding was minimized by incubation of the substrate with blocking buffer (0.1 mg/L BSA in PBS, pH 7.4) for 2 h. The substrates were washed three times with PBS and stored under Ar gas. PBS and NaCl (150 mM) were simulated the serum, and then the mixed solution was utilized to dilute 1000 ng/mL AFP antigen to prepare samples with different concentrations. Diluted AFP antigens were added into the centrifuge tubes containing SERS-active substrates. After 4 h, the substrates were washed three times with PBS solution and dried with Ar gas.

Immunoassay protocol. The substrates were immersed in PBS containing 0.1% BSA (1 mL). The chips were then incubated in the presence of AFP diluted with PBS at 4 °C overnight. After washing the substrate three times with PBS and drying under Ar gas, SERS spectrum measurements were performed. After the first SERS characterization, the substrates were put back in the plastic centrifuge tube, and 500 µL of immunogold colloids in PBS were pipetted into the tube and allowed to react for 4 h at room temperature. Similarly, the substrate was washed three times with PBS and dried under nitrogen before the second SERS spectrum measurement.

FDTD simulations. For the FDTD simulations, the simulated area is indicated by the blue box and the light green slices show the locations of the frequency-domain field monitoring, which are in the x−y planes, respectively. The wavelength of the source was set as 785 nm. The light source is set as a y-polarized plane wave source with a phase of 90°. The simulation results of Continuous wavelength in the corresponding frequency domain are accurately obtained by removing the Fourier transform of the time signal from the monitor. Different boundary conditions were selected for different directions according to the periodicity of the ideal morphologies. To obtain relatively high precise field enhancement resolution, the grid size was set to 0.05 nm with a conformal variant 0.5 mesh refinement. The overall simulation time was 1000 s, and the autoshutoff minimum was set to 1×10⁻⁶. The parameter values for the Ag permittivity and permeability were obtained from the software material database.