1. Introduction

Cancer has become a world intractable disease because of its high diffusion and low curability. However, it is difficult to cure at terminal stages of cancer. Therefore, the early detection of cancer plays an essential role in cancer diagnosis. Cancer biomarkers are widely used for cancer sensing in some traditional methods such as enzyme-linked immunosorbent assay (ELISA) [1], electrochemical sensing [2], and fluorescence-based methods [3], which are complex and time-consuming. Compared with these traditional methods, label-free spectrum methods such as surface plasmon resonance (SPR), localized surface plasmon resonance (LSPR), and surface-enhanced Raman scattering (SERS) [4–6] have more benefits for rapid detection, easy operation, and high sensitivity. In particular, label-free biosensors based on the terahertz (THz) spectrum have achieved much outstanding progress due to the development of THz spectrum technology. THz wave energy resides in the vibrational and rotational modes of molecules. Therefore, the THz spectrum has been widely used in molecular fingerprint spectroscopy and protein detection [7,8]. However, the wavelength of the THz wave (30-3000 µm) is much larger than the size of the molecules, indicating that THz spectroscopy is insensitive to the molecule concentration.

Terahertz metamaterials composed of artificially designed microstructures are often used to improve the sensitivity of terahertz spectroscopy biosensing. The microstructures include split rings, squares, and circles [9–11]. These artificial microstructures can resonate with the terahertz wave, which leads to a local enhancement of the surface electromagnetic field and the formation of a resonance peak. The resonance peak is very sensitive to the surface electromagnetic field environment. As a result, slight changes in the surface-wrapped electromagnetic field environment can result in a shift of the resonant peak. Therefore, the terahertz metamaterial has been evaluated as a promising method for biosensing. In recent years, terahertz metamaterials have gradually been used to sense cells, proteins, ribonucleic acid (RNA), and other biological fields with good performance [12–14]. Terahertz metamaterials have also been utilized in the sensing applications of cancer biomarkers. For example, Cui et al., used an insert grate metamaterial biosensor to detect carcinoembryonic antigen (CEA) [15]. The simulation sensitivity was 387 GHz/RIU when the whole surface of the biosensor was covered with 5-µm-thick analyte. CEA protein (500 ng/mL) caused a resonance peak shift of 29 GHz. In the study of Geng [16], a split-ring metamaterial biosensor with a simulated sensitivity of 150 GHz/RIU was applied to detect the Alpha fetoprotein (AFP) and Glutamine transferase isozymes II (GGT-II) of liver cancer biomarker. Lin proposed four metal split-ring-resonators to detect the concentration of CEA. The sensitivity reached 76.5 GHz/RIU with 5-µm-thick analyte covered on the surface [17]. However, the shortcomings of the low modification efficiency and longtime consumption should be improved.

Traditional metamaterial biosensors use thiols to form gold-sulfur bonds on a gold surface to connect biomolecules and the gold, which changes the dielectric environment around the gold surface for biosensing. However, the electromagnetic field of terahertz metamaterials is enhanced on the whole surface of the biosensors, including non-metal areas without modification. For example, the strongest electric field of a split ring metamaterial is found at the split position instead of the metal arms [18]. The resonant peak shows different shifts when Penicillium chrysogenum is placed at different positions on the metamaterial surface. The peak shift caused by depositing Penicillium chrysogenum at the opening gap is the largest, which indicates that the opening gap is the most sensitive to the changes in the dielectric environment. In recent years, much research has also been carried out for making full use of surface electric field enhancement. For example, a microchannel at the opening gap of a metamaterial split resonance ring has been fabricated by etching the substrate. A liquid sample was injected into the microchannel to improve the sensitivity [19]. Kailing Shih placed the particles in the groove at the open gaps for biosensing [20]. Zhou et al. increased the sensitivity of a THz metamaterial biosensor by etching the substrate to form a cross-shaped absorber to excite hot spots covering the analyte [21]. Jeong filled the gap of the metamaterial with a liquid sample instead of flowing on the top of the metamaterial to improve the sensitivity [22]. However, how to make full use of the effective sensing area with simple modification methods remains unresolved. In addition, terahertz metamaterial biosensors usually utilize rigid substrates such as high-resistance silicon and quartz, which could not meet the needs of clinical detection in the future.

Herein, we propose a flexible high-sensitivity terahertz metamaterial biosensor based on parylene-C substrate used for the label-free detection of cancer biomarkers. Parylene-C is a polymer material with good biocompatibility and a capability for moisture isolation and is often applied in electronic packaging. A study shows that parylene-C is extremely hydrophobic, creating a good adsorption capacity for proteins [23]. It has a more robust adsorption capacity than quartz and polystyrene (the substrate materials commonly utilized in ELISA biosensing) [24]. Therefore, studies using parylene-C as an adsorption layer to detect biomolecules have shown good performance for SPR, quartz crystal microbalance (QCM), and capacitive sensors [25–27]. In this paper, a protein was modified on the parylene-C substrate through a simple modification method, making full use of the sensing area on the surface of the terahertz metamaterial. Simulation analysis shows that the sensing sensitivity of modification on the substrate is higher than that on the gold surface. The fluorescence experiment results proved that the protein is effectively modified on parylene-C and that the unmodified surface is completely blocked by bovine serum albumin (BSA). The specific adsorption result indicates that the biosensors have a good selective absorption capability. Furthermore, standard samples and patient serum were examined showing that our sensor has potential for clinical application, and had reference significance for future flexible, lightweight, and wearable sensors.

2. Materials and methods

2.1 Materials and reagents

CEA was purchased from Abcam Company (USA). BSA was purchased from Solarbio Company (China). Anti-CEA was purchased from Creative Biomart Company (USA). Alexa Fluor 488 antibody was purchased from Thermo Fisher Company (USA). Alpha fetoprotein (AFP) antigen was purchased from Creative Biomart Company (USA). Anti-AFP was purchased from Abcam Company (USA). Deionized water was purchased from Solarbio Company (China). Parylene-C dimers were purchased from Specialty Coating Systems Company (USA).

2.2 Preparation of the terahertz metamaterial biosensor

The structural configuration of the metamaterial biosensor is shown in Fig. 1(a). The simulation results showed that the double U-shaped metamaterial structure is more sensitive than the traditional split-ring resonator (SRR) and single U-shaped structure (see Supplement 1 Fig. S1 and Fig. S2). Significantly, when the thickness of the parylene-C film is over 15 µm, the frequency of the resonance peak does not change which indicated high stability. The refractive index sensitivity is decreased as the thickness is increased over 15 µm. The quality factor, which is the ratio of wavelength to the full width at half maximum ($Q = \frac{{{f_0}}}{{FWHM}}$), is increased as the thickness is increased (see Supplement 1 Fig. S3). However, increasing the thickness of parylene-C requires more time for growth. By comprehensively considering the growth time, stability, sensitivity and quality factor, the thickness of the parylene-C film was set to 15 µm.

 

Fig. 1. The designed metamaterial structure, fabrication and biofunctionalization processes. (a) Geometric configuration of the metamaterial with structural parameters of t₁= 4 µm, t₂= 15 µm, w = 36 µm, l₁= 20 mm, l₂= 44 µm, and l₃= 36 µm; (b) The fabrication processes for the THz metamaterial biosensor: (1) growth of the parylene-C film; (2) lithography; (3) metal layer deposition and lift-off; (4) peeling off parylene-C; (c) illustration of biofunctionalization: (1) cleaning metamaterial biosensor; (2) anti-CEA incubation; (3) BSA blocking; (4) CEA incubation.

Download Full Size | PDF

The THz metamaterial biosensor fabrication processes are shown in Fig. 1(b). The operations consisted of the following steps: (1) A parylene-C film (thickness:15 µm) was deposited on a 4-inch silicon wafer by chemical vapor deposition (CVD) (PDS2010, Specialty Coating Systems company); (2) Standard photolithography was used to form a double U-shaped pattern; (3) The following methods were used sequentially to obtain the metamaterial sensing unit, electron beam evaporation (Cr 10 nm/Au 100 nm) and lift-off; (4) Parylene-C with the U-shaped metamaterial was peeled off from the silicon wafer and was clipped as a 20 mm×20 mm sensing unit.

2.3 Fluorescence experiments

Alexa Fluor 488 antibodies (goat anti-mouse), which are a kind of antibody directly conjugated with Alexa Fluor 488 fluorescence, were used to investigate the adsorption capacity of parylene-C. The fluorescent antibody was physically adsorbed onto the surface of parylene-C because of its hydrophobicity. In this process, the prepared metamaterial biosensor was washed in acetone, ethanol, and deionized water for ten minutes, respectively, and then dried by nitrogen. After washing with phosphate buffer saline (PBS, 100 µL), the metamaterial biosensor was treated with 100 µL of fluorescent antibody. The concentration of fluorescent antibodies ranged from 1 µg/mL to 10 µg/mL. The metamaterial biosensor was placed on a shaker and incubated for 1 h at 100 rad/min and 25 °C. After incubation, the metamaterial biosensor was washed with PBS three times for 10 min. Finally, after nitrogen drying, the fluorescence intensity of the metamaterial biosensor was measured by fluorescence microscopy. In addition, to explore the specific adsorption capacity of the parylene-C membrane, a mouse anti-human anti-CEA was used to capture the goat anti-mouse fluorescent antibody. Anti-CEA at a concentration of 10 µg/mL was first incubated onto the surface of parylene-C for 1 h. Then, the biosensor was washed with PBS three times to remove the unabsorbed anti-CEA and treated in 1% BSA (100 µL) solution to block the unoccupied surface of parylene-C. The metamaterial biosensor was then washed three times with PBS. Different concentrations of Alexa Fluor 488 fluorescence (100 µL) which can be captured by anti-CEA, were placed onto the biosensors. Finally, the metamaterial biosensor was washed with PBS three times to remove the fluorescent antibody that was not captured by the anti-CEA. The fluorescence intensity of the metamaterial biosensor was examined with an exposure time of 200 ms.

2.4 CEA incubation

The biosensor was used to quantify the concentration of the cancer biomarker CEA. The biological modification processes are shown in Fig. 1(c). First, the metamaterial biosensor was washed with acetone, ethanol, and deionized water, as shown in Fig. 1(c-1). Second, the metamaterial biosensor was treated with a mouse anti-human anti-CEA (100 µL) concentration of 10 µg/mL at room temperature (25 °C). Then, the biosensor was placed onto a shaker and incubated for 1 h. The shaker was rotated at 100 rad/min. After that, the metamaterial biosensor was washed with PBS 3 times for 10 min each time to remove the unadsorbed anti-CEA. This step was used to modify the anti-CEA on the parylene-C surface. The schematic diagram is shown in Fig. 1(c-2). Third, the metamaterial biosensor was treated with 1% bovine serum albumin (BSA, 100 µL) solution for 1 h under the same operations to block the unoccupied parylene-C surface, as shown in Fig. 1(c-3). Then, the metamaterial biosensor was washed with PBS 3 times, for 10 minutes each time. Finally, CEA (100 µL) was placed on the surface of the metamaterial biosensor, which was incubated for 1 h (Fig. 1(c-4)). Next, the metamaterial biosensor was washed 3 times with PBS to remove unmodified protein. To obtain the concentration curve for CEA, 6 different concentrations of CEA were utilized for detection. Moreover, the AFP antigen was used as a reference experiment instead of the CEA, which can verify the specific biosensing capacity of the metamaterial biosensor.

2.5 Patient serum modification

All patient serum was collected and provided by the School of Basic Medical Sciences, Shandong University. Samples were stored at -80 °C. For each measurement, 100 µL of each patient serum was dripped into the metamaterial biosensor, which was incubated with anti-CEA, blocked with BSA, and cleaned with PBS. Then, the metamaterial biosensor was incubated for 1 h on a shaker at room temperature (25 °C). The shaker was rotated at 100 rad/min, and the shaker was then washed with PBS solution 3 times, for 10 minutes each time.

2.6 THz time-domain transmission spectrum test

The THz spectrum of the terahertz metamaterials was measured by a commercial terahertz time-domain spectroscopy test system (CIP-TDS, Daheng Optics) with a resolution less than 5 GHz. The resolution will be further reduced to 1 GHz by increasing the delay line and Fourier transforms fitting. The frequency region ranges from 0.1 to 4.0THz. The test environment temperature was 23 ± 0.5 °C, and the relative humidity was less than 5%. The simulation indicates that the resonance peak of the biosensor is related to the polarization of the incidence. A designed fixture was used to prevent the biosensor rotation (see Supplement 1 Fig. S4). First, the terahertz time-domain transmission spectrum was recorded, which was used as the reference spectrum. Second, the metamaterial biosensors after different modifications were blown and dried with nitrogen gas. Then, the metamaterial biosensor was placed into the system to record the terahertz time-domain transmission spectrum data in the same environment. The terahertz transmission spectra obtained for the samples after different modifications were the signal spectra. The frequency-domain signal spectrum and the reference spectrum were obtained by Fourier transformation from the time-domain spectra. The frequency-domain signal and reference spectra were defined as ${E_s}\; $ and ${E_{r\; }}$. The transmittance T is expressed as:

The resonance peak position of the transmission spectrum of the metamaterial biosensor without modification is ${f_b}$. The resonance peak position of the transmission spectrum of the biosensor with the sample modification is ${f_s}$. The shift in the resonance peak due to the change in surface refractive index $\Delta f$ is:

2.7 Simulation

The resonance absorption peak of the metamaterial biosensor and the surface electric field distribution simulation were explored using COMSOL Multiphysics software.

3. Results and discussion

3.1 Theory of metamaterial biosensing

Figure 2(a) shows the main experimental program used for the parylene-C film metamaterial biosensor. Figures 2(b) and (c) show the fabrication results obtained for the metamaterial biosensor on a flexible parylene-C film and a microphotograph of the double U-shaped structure, respectively. When the incident terahertz wave passes through the metamaterial biosensor, an induced charge is generated and concentrated at the split position, eventually forming an electric field. Meanwhile, the movement of the charge can also generate a circular current that forms an inductance. Therefore, the resonance can be approximately regarded as an LC resonance generated under the combined action of inductance and capacitance. The resonance frequency can be expressed as follows [28]:

 

Fig. 2. The experimental program and fabrication results for the flexible metamaterial biosensor. (a) Schematic illustration of the THz metamaterial biosensor used for CEA detection; (b) Photograph of the THz metamaterial biosensor. (c) Micrograph of the double U-shape metamaterial.

Download Full Size | PDF

3.2 Simulation

Full-field electromagnetic simulations of the double U-shaped metamaterial structure were completed using COMSOL Multiphysics. The dielectric constant of the parylene-C substrate is ${\varepsilon _p} = 2.7$, and the extinction coefficient K is negligible. The conductivity of Au is $\sigma = 4.56 \times {e^7}S/m$, in accordance with the Debye distribution. The sensitivity of the metamaterial with different gold thicknesses is close (see Supplement 1 Fig. S5). Therefore, the thickness of Au was 100 nm to reduce the cost. The periodic boundary conditions were used. The electric field of the incident electromagnetic wave pointed along the X direction, and the magnetic field pointed along the Y direction. Figure 3(a) illustrates a comparison diagram for the transmittance curve between the simulation result (red) and the experimental result (blue), which indicates that the metamaterial shows two absorption peaks in the frequency range of 0-2.5 THz. The first peak position is observed at 1.12 THz whose half-width is relatively wide. The second peak position is located at 1.706 THz with a narrow half-width. Therefore, the second peak was more suitable for biosensing. The simulation results obtained for the sensing sensitivity for the two peaks also show that the sensitivity for the second peak is higher than that for the first peak (see Supplement 1 Fig. S6), which is in good agreement with the result obtained using the terahertz time-domain spectroscopy system. Therefore, the following experiments only focused on the second resonance peak. To explore the generation of the absorption peak at 1.706 THz, the surface current at the peak frequency was explored, as shown in Fig. 3(b). The results indicate that the internal U-ring generates a ring-shaped current at 1.706 THz, which is a typical LC resonance. Therefore, the resonance absorption peak at 1.706 THz was formed by the LC oscillation of the inner U-ring. To investigate the effect of electric field distribution, the electric field distribution on the X-Y plane and X-Z plane was simulated. The frequency was set at 1.706 THz. The electric field of the planes Z= 0.1 µm and Y = 40 µm were observed respectively. Figures 3(c) and (d) show the surface electric field distribution on the X-Y plane and X-Z plane at the resonance frequency, respectively. Figure 3(c) demonstrates that the electric field is not only distributed around the metal metamaterial structure but is also present in the non-metal areas with a stronger intensity. A change in the dielectric constant of the surrounding environment of the non-metal area can also have a significant impact on the resonance. A strong electric field distribution is observed at a distance of a few microns above the entire device surface (Fig. 3(d)), which illustrates that the biosensing area can reach a few microns above the surface of the metamaterial biosensor.

 

Fig. 3. The simulated results for the metamaterial biosensor. (a) Numerically calculated (red) and experimentally probed (blue) transmission spectra for the metamaterial biosensor; (b) Surface current distribution in the X-Y plane; (c) Electric field distribution in the X-Y plane; (d) Electric field distribution in the X-Z plane.

Download Full Size | PDF

As shown by the electric field distribution in Fig. 3, the sensing effects between modification on the metal surface and the non-metal surface are different. To compare the different modification methods, the sensitivity of the refractive index for the three modification methods was determined. The thickness of the biological sample layer was 10 µm. The refractive index changes of the sample layer range from 1.3 to 1.4 with a scanning step length of 0.02. Figures 4(a) and (b) provide the transmittance curves for the sample with different refractive index values for the sample covering the metal surface and the non-metal surface. The peak shift $\Delta {f_2}$ (32.5 GHz) is larger than $\Delta {f_1}$ (15 GHz) which indicates that the resonant peak frequency shift is larger for the sample covering the non-metal surface than that covering the metal surface when the refractive index of the same sample changed. The resonant peak frequency under different refractive indices was extracted and calculated by linear fitting (see Supplement 1 Fig. S7). The formula for the metal function can be expressed as $y = 1.85 - 0.147x$, and $y = 2.044 - 0.325x$ for the non-metal function. The refractive index sensitivity for the sample covering on the metallic surface was determined to be 147 GHz/RIU. However, when the sample was covered on the non-metal surface, the refractive index shows a sensitivity of 325 GHz/RIU. The sensitivity of the latter is larger than that of the former, which showed that the sample covering the non-metal surface performs better. The refractive index sensitivity of different sample thicknesses covering the metal, non-metal, and whole surface was also explored. As shown in Fig. 4(c), the refractive index sensitivity of the three modification methods increases with increasing sample layer thickness. Finally, it reaches saturation at approximately 10 µm, which indicates that the effective sensing range of the sensor above the surface is approximately 10 µm. Figure 3(d) shows that the electric field was distributed approximately 10 µm above the sensor surface, indicating that the sensitivity simulation results are consistent with the electric field simulation results. The effective sensing area was related to the surface electric field distribution. The whole surface covering method fully utilizes the electric fields above the surface. Consequently, the largest refractive index sensitivity is obtained. The metal surface occupies a small part of the electric field. As a result, the refractive index sensitivity here is the smallest. The refractive index sensitivity of the sample covering the non-metal surface is better than that of the sample covering the metal surface. Finally, the influence of different substrate dielectric constants on sensitivity was also explored. Figure 4(d) shows that the sensitivity of the sensor is increased as the substrate dielectric constants decreased. Therefore, parylene-C was chosen as the substrate because it is better than polyimide (PI) and quartz due to its lower dielectric constants.

 

Fig. 4. The simulated transmission spectra and sensitivity of the metamaterial biosensor. (a) The simulated transmission spectra for the metamaterial with analyte on metal areas. The inset graph shows the details for the peak; (b) Analyte on non-metal areas. The inset graph shows the detail of the peak; (c) Refractive index sensitivity of the high-frequency peak with analyte on the whole chip (blue), on the non-Au areas (red), and on Au areas(black); (d) Refractive index sensitivity of the high-frequency peak under different substrate dielectric constants.

Download Full Size | PDF

3.3 Fluorescence experiment

To explore the ability of the parylene-C film to adsorb protein and the feasibility of the antigen-antibody sensing experiment, a series of fluorescence verification experiments were conducted. Alexa Fluor 488 antibodies were used as the target protein to be adsorbed. Figures 5(a) and (b) present the fluorescence pictures of the unmodified fluorescent protein and that modified by Alexa Fluor 488 antibodies at a 10 µg/mL concentration, respectively. Figures 5(a) and (b) show that the parylene-C film has a strong adsorption capacity for the fluorescent protein, and the modification is uniform. The modification results obtained for other concentrations of fluorescent protein are shown in Fig. S8 (see Supplement 1), which indicates that the fluorescence intensity is significantly changed with the protein concentration. In addition, 10 µg/mL Alexa Fluor 488 antibodies were modified on the silicon, quartz, and PI (see Fig. S9 in Supplement 1). The fluorescence images showed that parylene-C was the best substrate to adsorb the protein. Due to the requirements of specific modification, it is necessary to verify the blocking effect of BSA on the parylene-C film. The surface of the parylene-C film was modified with a concentration of 1% BSA. Then the fluorescent antibody at a concentration of 10 µg/mL was modified, as shown in Fig. 5(c). Compared with the strong fluorescence intensity observed for the unincubated BSA chip in Fig. 5(b), the biosensor modified with fluorescent antibody after blocking with 1% BSA shows almost no fluorescence. The parylene-C surface is sufficiently blocked by BSA, suggesting that BSA has a good blocking effect on the unoccupied surface. To investigate the specific adsorption capacity of the biosensor, the anti-AFP, which cannot capture the fluorescent antibody, was first incubated on the parylene surface instead of the anti-CEA. Then fluorescent antibodies were placed onto the biosensor. Figure 5(d) shows that there no fluorescence is observed on the surface of the biosensor after the modification with 10 µg/mL fluorescent antibody, which indicates that there was no fluorescent antibody is captured on the surface. The fluorescent antibody captured by the anti-AFP and physically adsorbed on the surface is negligible. These results illustrate that the parylene-C substrate metamaterial biosensor exhibits a strong specific sensing ability.

 

Fig. 5. The fluorescence experiment results. (a) Fluorescence photo of blank parylene-C membrane; (b) Fluorescence photo of parylene-C membrane directly incubated by 10 µg/mL Alexa Fluor 488 antibody; (c) Fluorescence photo of parylene-C membrane incubated by 10 µg/ mL Alexa Fluor 488 antibody after BSA blocking; (d) Fluorescence photo of parylene-C membrane when anti-CEA replaced by anti-AFP; (e)-(h) Fluorescence photos of biosensor modified by a three-step modified method with a fluorescence antibody the concentration of 10 µg/ mL, 4 µg/ mL, 2 µg/ mL, 1 µg/ mL, respectively; (i) THz spectrum of metamaterial before (blue) and after (red) incubation by the 10 µg/mL Alexa Fluor 488 antibody (shift direction shown by arrow); (j) The frequency shift for the modified processes: anti-CEA, BSA, and Fluorescence antibody. The error bars indicate the standard deviation (SD).

Download Full Size | PDF

The three-step modification method was verified through the following processes. (1) Mouse anti-human anti-CEA was incubated on the parylene-C surface as the primary antibody. (2) BSA (1%) was incubated on the surface to block the unmodified surface. (3) The goat anti-mouse Alexa Fluor 488 fluorescent antibody, which is connected with the mouse anti-human anti-CEA, was incubated on the surface as the target antibody. Figures 5(e), (f), (g), and (h) show fluorescence photographs of the metamaterial biosensor, which was modified by the fluorescent antibody through the three-step modification method at concentrations of 10 µg/ mL, 4 µg/mL, 2 µg/mL, and 1 µg/mL, respectively. The metamaterial biosensor modified through the three-step modification method shows obvious fluorescence in contrast to the metamaterial biosensor shown in Fig. 5(c), which was not modified by the anti-CEA before BSA blocking. Furthermore, the fluorescence intensity is increased with increasing antibody concentration. Therefore, the fluorescence experiment results indicate that the three-step modification method can be applied for the specific modification of antigens and antibodies, which can be used in biosensing experiments. A terahertz time-domain spectroscopy system was used to examine the metamaterial biosensor before and after modification. Figure 5(i) shows the transmission curves measured for the blank film (blue line) and the metamaterial biosensor modified by the 10 µg/mL fluorescent antibody through the three-step modification method (red line). The results show that the resonance peak frequency is shifted by 12 GHz to the low frequency after the whole modification. Figure 5(j) shows the shift for each process. The resonance peak position gradually shifted to a low frequency as the modification proceeds. The results indicate that every modification process can be probed by the THz spectrum, which is consistent with the results obtained from the fluorescence experiments.

The above fluorescence experiments prove that the parylene-C substrate has a strong adsorption capacity for protein, and the surface adsorption capacity of different concentrations of protein is different. After being modified with the antibody, the surface can be well blocked by the BSA. Therefore, impurity proteins will not be physically adsorbed by the metamaterial. As a result, only the antigen that can be specifically connected to the primary antibody will be well captured by the metamaterial biosensor. As a result, the parylene-C substrate terahertz metamaterial biosensor based on the three-step modification method is very suitable for molecules biosensing. Compared with traditional modified sensing methods, the three-step method has the advantages of simple operation, short modification time, fewer chemicals, and high modification efficiency.

3.4 CEA detection

The double U-shaped terahertz metamaterial biosensor chip was used to measure the standard curve for CEA biomarkers according to the three-step modification scheme. First, the anti-CEA was incubated on the non-metal surface. Second, BSA (1%) was used to block the surface without the anti-CEA. Finally, the anti-CEA specifically captured the CEA. Figure 6(a) shows the transmission curves obtained for the blank biosensor and biosensors with different concentrations of CEA. The inset figure shows the normalized result. Compared with the blank film, as the antigen concentration is increased, the resonance peak of the modified metamaterial biosensor shows a gradual shift to low frequency. Moreover, when the antigen concentration is increased, the surface refractive index is also gradually increased, and the resonance peak frequency is gradually decreased, consistent with the simulation results. Figure 6(b) shows the curve for the resonance peak shift for different antigen concentrations (two experiments for each data point). The concentrations of CEA ranged from 0 ng/mL to 1000 ng/mL, and the fitting function can be expressed as $\triangle f = 7.82log(C )+ 4.74$. Figure 6(c) shows the measurement data at various concentrations ranging from 0 ng/mL to 10 ng/mL. The limit of detection (LOD) of the biosensor was calculated from the 3-fold standard deviation of the blank control signal to be 2.97 ng/mL. This result is lower than the usual clinical detection limit of CEA cancer biomarkers (<5 ng/mL). The biosensor optimizes the modification processes and shortens the sensing time compared to other traditional detection methods (see Table S1 in Supplement 1). The better performance of the metamaterial biosensor was attributed to the biomolecules being modified on the non-metal surface to make full use of the electric field enhancement. At the same time, the double U-shaped metamaterial structure contains a larger gap area, which leads to a wider enhanced electromagnetic field in the non-metal area. In addition, the flexible parylene-C film has a low refractive index and low absorption and reflection for THz waves, which determines the high sensitivity of the metamaterial biosensor. Therefore, the double U-shaped metamaterial biosensor provides the advantages of low cost, simple method, and high sensitivity compared with other sensors.

 

Fig. 6. The THz detection results for different CEA concentrations. (a) THz transmission spectra for the bare metamaterial and different concentrations of CEA. The inset graph shows the normalized THz transmission spectra; (b) Peak frequency shift of the THz metamaterial biosensor for CEA detection. Error bars indicate the SD; (c) The linear relationship between the peak shift and CEA concentration ranging from 0 ng/mL to 10 ng/mL. Error bars indicate the SD; (d) The peak frequency shift obtained for the blank control, serum 1, serum 2, serum 3, specificity test using 1 µg/mL AFP antigen, and 1 µg/mL CEA.

Download Full Size | PDF

3.5 Patient serum and specific test

Figure 6(c) shows the sensing property with CEA concentrations ranging from 0 ng/mL to 10 ng/mL. The asterisk is shown in the curve in Fig. 6(c) denotes the patient serum data. The patient serum data tested by the THz metamaterial sensor expressed good consistency with the ELISA data provided by the hospital. However, there remain some deviations due to the impurity proteins in the patient serum. The 1000 ng/mL concentration of AFP antigen was used as a nonspecific protein to conduct the selective measurement, which proves the selective sensing capacity of the biosensor. As shown in Fig. 6(d), the resonance peak shift of the AFP antigen is small despite the AFP antigen concentration being sufficiently high. The peak shift for CEA is obvious at the same concentration, indicating that the AFP antigen cannot be captured by the modified anti-CEA. All the experimental results show that the metamaterial biosensor has very good specific sensing performance.

4. Conclusion

A novel flexible terahertz metamaterial biosensor based on non-metal modification is proposed for label-free biomarker CEA detection with high sensitivity and selectivity. The biomolecule was modified on the non-metal surface of the metamaterial biosensor by the three-step modification method. This strategy increases the sensitivity and reduces the time required for detection, which was verified by the results of simulation and fluorescence experiments. The metamaterial biosensor presents good selectivity and sensitivity for CEA detection with a LOD of 2.97 ng/mL. Moreover, the results obtained for patient serum detection illustrate that the metamaterial biosensor is suitable for clinical application and has promising potential for mass fabrication and early detection of cancer biomarkers.