Chloroauric acid tetrahydrate (HAuCl₄-4H₂O), silver nitrate (AgNO₃), hydrogen peroxide, sulfuric acid and ascorbic acid (AA) were obtained from YangZhou LanTian Chemicals Co. Ltd. (China). 4-aminothiophenol (4-ATP), sodium chloride and ethanol were acquired from Jiangsu Younuo Chemicals Co. Ltd. (China). All of the materials were applied directly without further processing. Meanwhile, ELISA kits, superoxide dismutase and four necked round bottom flasks were all purchased from Sangon Biotech (Shanghai, China). Deionized water (resistivity >18.2 Ω) was used for the preparation of the specimens and throughout all the experiments. All glassware was dipped in aqua regia [HNO₃/HCl = 1:3 (v/v)] for over 24 h and washed with deionized water.

Peripheral blood samples from clinical medical college of Yangzhou university in 50 cases of healthy subjects and 50 patients with low grade squamous intraepithelial lesion, high-grade squamous intraepithelial lesion in 50 cases of patients and 50 cases of cervical cancer patients were centrifuged at 3,000 rpm for 12 min at 4°C. Then serum samples were collected and based on its classification to store in −80°C before analysis. Consent documents were obtained from all donors. Table 1 summarizes the details of age and histopathological stage.

Au-AgNBs modified by nanodots was synthesized by one-step method (Li et al., 2018). The experiment was carried in 100 mL conical tubes. First, 90 μL of 1% HAuCl₄ solution was added into 10 mL ultrapure water by continuously stirring for 1 min. Then, 170 μL of AgNO₃ (6 mM) was dropped into the mixture and the mixture became turbid pearl white. After dropping the 125 μL of AA (0.1 M), the solution turned to distinct blueviolet color indicating the Au-AgNBs were prepared. After 10 min of continuously stirring, the Au-AgNBs were concentrated by centrifugation (4,000 rpm, 4 min). Collecting the bottom sediment, disperse the particles in 5 mL deionized water and store at 4°C.

The silicon wafer was divided with the size of 0.8 × 0.8 cm², then they were placed in a beaker and used after ultrasonic cleaning with ultra-pure water and ethanol in turn. Then, the silicon wafer of appropriate size was immersed in piranha solution (hydrogen peroxide (30%) was added to concentrated sulfuric acid in a ratio of 3:7 by volume) for 30 min to make the silicon wafer hydrophilic. Then ultrapure water and ethanol were used again to clean the silicon wafer for three times. Au-AgNBs prepared the array by using the method of oil-water interface self-assembly. In brief, by mixing 8 mL of the prepared Au-AgNBs solution sequentially with 4 mL of hexane in a beaker and then adding 4 mL of ethanol drop by drop, it was found that Au-AgNBs formed neat arrays at the oil-water interface. Next, the Au-AgNBs array was picked up using the prepared hydrophilic silicon wafer and placed in a ventilated place to dry. Then, the Au-AgNBs array is obtained. These prepared Au-AgNBs array were uniformly stored in a sealed glass cover at 4°C. Every time they were used, Au-AgNBs arrays made in the same batch were selected to reduce the error caused by the detection platform.

Before the test, 20 μL of 2.5 × 10⁻⁵ M 4-ATP was dropped onto the prepared Au-AgNBs array surface. Then an appropriate amount of buffer solution was added to the array surface to adjust PH = 9 and left it at room temperature for 2 min to mix evenly. Next, 20 μL of sample solution was dropped and stayed for 2 min at room temperature to make it evenly covered. The ordered array was irradiated by 785 nm laser. The laser power at the sample location was 2.3 mW. Under the continuous irradiation of the sample, the SERS spectrum of 1 s was continuously measured in 1 min steps. All SERS spectrum reported in this study were collected in a continuous mode within the range of 400–1800 cm⁻¹. The average SERS spectra measured at 10 different points at random in one platform were used to quantify SOD in the sample solution, which ensured the authenticity and rationality of the data. The characteristic bands of 4-ATP and DMAB were listed in the table S1 (supporting information).

Uv-vis-near-infrared (UV-VIS-NIR) spectrometer (UV-3000PC, Mapuda, China) was used to detect the absorption spectrum of UV-VIS-NIR. Transmission electron microscope (TEM) images were taken with transmission electron microscope (Tecnai 12, Philips, Netherlands). Scanning electron microscopy (SEM) images were studied using an S-4800II ⅐ laser emission scanning electron microscope (Gemini SEM 300, Carl Zeiss, Germany). High-resolution TEM (HRTEM) images, selective region electron diffraction (SAED) and element mapping images were obtained using a field emission transmission electron microscope (Tecnai G2F30 S-Twin, FEI, United States). Raman spectrometer (Renishaw inVia, United Kingdom) was used to record SERS mapping with a mapping step of 1 μm and a pinhole of 25 μm. All experiments were performed at room temperature.

Au-AgNBs is a new type of nanomaterial, which has the advantages of high hotspot density and good stability. TEM and SEM were used to characterize the structure and size of Au-AgNBs. The SEM image of Au-AgNBs could be seen in Figure 1A, which indicated that a large amount of Au-AgNBs could be obtained by one-step process with uniform size and good dispersion. It could be seen that in the TEM image, the hollow inner wall and outer wall of Au-AgNBs were obviously different (Figure 1B). The mean side length of Au-AgNBs was 70 nm and the thickness of the wall was 5 nm. As shown in Figure 1C, the four bright rings {111}, {200}, {220}, and {311} indicated that Au-AgNBs were polycrystalline (Gomez-Grana et al., 2013). Usually, nanocages were composed of bimetals and almost all hollow nanostructures needed to use templates to form nanocages (Zhang et al., 2010). The template was often silver nanocubes. In particular, Cl⁻ promoted the formation of silver cube templates through the cap effect, which was consistent with the halide selective stabilization of the {100} surface of Au-AgNBs, while the silver nanocrystals were oxidized to form Au-AgNBs in the inner pore wall (Huang et al., 2009). Figure 1D indicated the plane distances between the tip crystal faces of the inner and outer walls of Au-AgNBs were 0.210 nm and 0.225 nm respectively. The basic diagram of energy dispersive X-ray energy spectrum (EDX) of Au-AgNBs could be seen from Figure 1E, it could be seen that the outer wall composition of Au-AgNBs were dominated by silver and gold. Figure 1F showed the UV-visible spectrum of Au-AgNBs with a broadband maximum of 692 nm, which indicated that a large amount of Au-AgNBs had been prepared. The physical drawing showed the picture of the Au-AgNBs solution in visible blue color. Figure 1G was the energy dispersive X-ray spectrum (EDX) of Au-AgNBs. It showed that Au-AgNBs is mainly composed of gold and silver and the peak of copper in the electron spectrum was mainly due to the use of copper mesh as the test substrate. The Raman spectra of 4-ATP and 4-ATP-labeled Au-AgNBs were shown in Figure 1H. 4-ATP and Au-AgNBs were connected to each other mainly through Au-S bonds (Jang and Keng, 2008). As could be seen from the figure, the Raman spectrum of 4-ATP showed that the SERS signal was very weak. In contrast, significantly enhanced SERS signal could be observed for 4-ATP-labeled Au-AgNBs indicating that Au-AgNBs had a strong SERS effect. The analytical enhancement factor (EF) of Au-AgNBs was calculated as EF=(I_SERS/C_SERS)/(I_RS/C_RS) (Mahmoud and El-Sayed, 2010). SERS and RS represent SERS condition and non-SERS condition, while C and I represent the concentration and intensity, respectively. When C_SERS = 1 × 10⁻⁶ M, C_RS = 10⁻¹ M and the intensity measured at 1,081 cm^-1, the EF calculated was 7.531×10⁵. This effect was due to the coupling of its inner and outer surfaces, leading to strong optical absorption and high SERS enhancement. Therefore, compared with the ordinary noble metal substrate, the Au-AgNBs array had significant SERS signal enhancement ability.

Excellent SERS substrate performance is the key to practical application. As shown in Figure 2A, the SEM image showed the side view of Au-AgNBs array. It could be seen that Au-AgNBs was highly uniform and orderly arranged, with an average height of about 70 nm. 40 × 40 μm² area was randomly selected on the array marked with 4-ATP for SERS intensity mapping measurement. Each pixel in the spatial position of the mapping image represented the signal strength at 1,081 cm^-1. These signals were related to the distribution of Au-AgNBs on the array surface. Although there were a few blue and yellow areas in the image, most areas show relatively stable green, indicating that the SERS substrate had a high uniformity as shown in Figure 2B. In order to more intuitively verify the uniformity of the detection substrate, 10 random points were selected on the substrate surface for SERS spectrum measurement. Figure 2C showed the SERS spectrum of the random points. It could be seen that the signal strengths of the selected points were relatively consistent. The histogram in Figure 2D intuitively showed the slight fluctuation of the spectrum and its relative standard deviation (RSD) was 7.832%, indicating that the Au-AgNBs array had good signal uniformity.

The concentration of 4-ATP and SOD had a strong influence on the detection sensitivity. By optimizing these parameters, the sensitivity of SERS detection platform could be further improved. When 4-ATP was adsorbed on the surface of Au-AgNBs, under the irradiation of laser, 4-ATP would be oxidized into DMAB. In the presence of SOD, SOD could combine with OH⁻ in solution to remove oxide, thus inhibiting the formation of DMAB and then SERS signal showed the characteristic peak of 4-ATP. The value of I₁₁₇₀/I₁₁₈₅ was selected as the index parameter of the optimization result, the detection sensitivity was positively correlated with I₁₁₇₀/I₁₁₈₅. In the process of preparing SERS substrate, 4-ATP with different concentrations was added and then SOD solution with the same solubility of 150 U mL⁻¹ was dripped respectively. As shown in Figure 3A, the peak intensity of I₁₁₇₀ gradually increased with the increase of 4-ATP concentration. It was shown in Figure 3B that when the 4-ATP concentration was 2.5 × 10⁻⁵ M, the ratio of I₁₁₇₀/I₁₁₈₅ was close to 1, and I₁₁₇₀/I₁₁₈₅ ratio remained unchanged with the increase of 4-ATP concentration. Therefore, when the SOD concentration was 150 U mL^-1, the optimal 4-ATP concentration was 2.5 × 10⁻⁵ M. Similarly, by controlling the concentration of 4-ATP to 2.5 × 10⁻⁵ M and changing the concentration of SOD from 0 U mL⁻¹–150 U mL⁻¹, as shown in Figures 3C, D, the ratio of I₁₁₇₀/I₁₁₈₅ gradually decreased with the increase of SOD concentration. When SOD concentration was 150 U mL^-1, the ratio of I₁₁₇₀/I₁₁₈₅ was close to 1 and when SOD concentration was increased again, the ratio of I₁₁₇₀/I₁₁₈₅ hardly changed, indicating that when SOD concentration was 150 U mL⁻¹, the conversion of 4-ATP could be completely inhibited. Therefore, SOD with the concentration of 150 U mL⁻¹ and 4-ATP with the concentration of 2.5 × 10⁻⁵ M were selected as the best concentrations.

The stability of SERS substrate was evaluated. The prepared SERS substrate was stored in a sealed container at 20°C and SERS detection was taken on the SERS substrate at the 0, 5, 10, 15, 20, and 25 days respectively. As can be seen in Figure 4A, there was no significant difference in the SERS spectrum peak and spectrum shape. Figure 4B showed the corresponding scatter plot, with the peak intensity of 1,081 cm⁻¹ as the characteristic peak and the peak intensity of the 15th day was 8.803% lower than that of the 0 day. Among them, the peak intensity on the 25th day was still maintained at 80% of the initial intensity compared with the peak intensity on the 0 day, indicating that the SERS array base had stable SERS enhanced effect and storage stability. As we all know, the selectivity of SERS platform was of great significance in the actual analysis of biological samples. In order to evaluate the selectivity of SERS immune substrate. The substances were selected that may exist in the detection environment as interfering substances and detected specific markers SOD and non-specific biomarkers or proteins (SCCA, CA125, IL-6, survivin, glucose) of the same concentration (100 U mL⁻¹) in PBS buffer. I₁₁₇₀/I₁₁₈₅ was selected as the tracer of SOD. As shown in the spectrum in Figure 4C, the peak intensity of I₁₁₇₀ of SOD was similar to that of I₁₁₈₅, while the peak intensity of I₁₁₇₀ of other substances was obviously higher than that of I₁₁₈₅. The histogram of Figure 4D indicated the results more clearly. When SOD existed, the I₁₁₇₀/I₁₁₈₅ ratio was obviously lower than that of the solution without SOD. Under the above optimal conditions, the reproducibility of another important parameter of the SERS platform was studied. According to Figure 4E, the SERS spectra was studied by selecting of ten independent experiments conducted at different times. There was almost no difference between these SERS spectra. The broken line graph of the SERS spectrum was shown in Figure 4F. With 1,081 cm⁻¹ as the characteristic peak, the peak intensity deviation was 7.625%. This small change showed that the SERS platform had good reproducibility. In order to study the differences between different batches of Au-AgNBs arrays, the Au-AgNBs arrays made at different batches were compared. Figure S1A showed the differences of SERS spectrum. The Au-AgNBs array marked with 4-ATP prepared at different batches were detected by SERS. As shown in Figure S1B, with 1,081 cm^-1 as the reference peak, the intensity deviation of the four peaks was small (2.407%), indicating that there was almost no difference between Au-AgNBs array prepared at different times.

Under the above optimization conditions, the capability of SERS platform for rapid analysis of SOD was evaluated. In order to combine the immunosensor with practical application, it was used to detect the concentration of SOD in serum. Au-AgNBs ordered arrays with the size of 0.8 × 0.8 cm² which adsorbed 4-ATP on their surfaces according to the detection principle were adjusted to the PH = 9. The SOD solution was diluted in the purpose-made serum (without SOD) to the concentration of 10 U mL⁻¹, 40 U mL⁻¹, 70 U mL⁻¹, 100 U mL⁻¹, 130 U mL⁻¹ and 160 U mL⁻¹ respectively. 20 μL of the above solutions was dropped on different SERS detection platforms, then the samples were continuously irradiated under Raman microscope. The SERS spectrum were continuously measured for 1 s in a step of 1 min until the spectral shape did not change. 10 points were selected randomly of the detection platform for measurement and the average SERS spectrum were calculated for quantitative detection of SOD in the sample solution. Figure 5A presented the SERS spectra of SOD solutions with different concentrations. It was obvious that with the increase of SOD concentration, the process of 4-ATP converting to DMAB was gradually inhibited, which was reflected in the gradual decrease of the ratio of I₁₁₇₀/I₁₁₈₅. By using the ratio of I₁₁₇₀/I₁₁₈₅ as parameter, a linear calibration chart for quantitative assessment of SOD concentration was constructed as shown in Figure 5B. When the SOD concentration was from 10 U mL⁻¹–160 U mL⁻¹, the ratio of I₁₁₇₀/I₁₁₈₅ was almost linearly related to the concentration of SOD. Its linear regression equation was y = −0.00332x+1.54406 and the relative coefficient (R ²) was 0.970. The limit of quantitation (LOQ) of the SERS platform for SOD was 10 U mL⁻¹. It showed that the SERS platform had a good linear relationship at the concentration range from 10 U mL⁻¹–160 U mL⁻¹, which could realize SERS detection of SOD activity in serum.

SERS platform was used to analyse SOD quantitatively to prove the accuracy, reliability and clinical practicability of the analysis. Clinical serum samples of healthy people, LSIL, HSIL and the cervical cancer were studied. By calculating the ratio of I₁₁₇₀/I₁₁₈₅ in SERS spectrum, the concentration difference of SOD in serum of different populations could be calculated. The used clinical serum samples needed not to be diluted and they were directly used as sample solution. Each sample was measured three times. 10 random test points were measured on the surface of the detection platform each time. Every spectra was the average result of 30 different serum samples. These results are analyzed and the standard deviation is calculated. Figure 6A showed the mean SERS spectrum of clinical samples. It could be seen that as the disease progresses, the characteristic peak of SERS spectrum gradually changed from 4-ATP to DMAB, which was the ratio of I₁₁₇₀/I₁₁₈₅ gradually increasing. The concentration of SOD in each actual sample was determined by fitting I₁₁₇₀/I₁₁₈₅ into the linear regression equation of the calibration curve. The concentration of SOD in the serum of patients with cervical cancer was significantly lower than that of HSIL, LSIL and normal people. Figure 6B directly showed the ratio difference of I₁₁₇₀/I₁₁₈₅ in SERS images. As a detection method for biomarkers, ELISA was used to detect the actual samples and calculated the average concentration. Figure 6C allowed a more intuitive comparison of the SOD concentration in the actual samples detected by ELISA kit and SERS platform. As shown in Table 2, the average concentration of SOD in the serum of healthy people, LSIL, HSIL, and the cervical cancer detected by SERS platform were 129.1 U mL⁻¹, 79.95 U mL⁻¹, 57.24 U mL⁻¹, and 28.96 U mL⁻¹, respectively. The concentration of SOD detected by ELISA were 121.7 U mL⁻¹, 82.19 U mL⁻¹, 59.72 U mL⁻¹ and 27.88 U mL⁻¹ respectively and the relative errors of the two methods were −3.87%, 4.1%, 2.73% and −6.08% respectively. The results showed that there was no significant difference between SERS platform and ELISA detection results, which confirmed that the SERS platform could be used for early clinical screening of cervical cancer.