Chloroauric acid tetrahydrate (HAuCl₄·4H₂O), ascorbic acid (AA), bovine albumin (BSA), ethanol (CH₃CH₂OH), N-hydroxysuccinimide (NHS), 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC), 4-mercaptobenzoic acid (4-MBA), phosphate-buffered saline (PBS), trisodium citrate dihydrate (C₆H₅Na₃O₇·2H₂O), silver nitrate (AgNO₃) and hexadecyl trimethyl ammonium bromide (CTAB) were purchased from Sigma-Aldrich. All reagents were provided at analytical grade and were used without further purification. CYFRA21-1, rabbit monoclonal anti-CYFRA21-1, CEA、AFP, SCC-Ag were purchased Sangon Biotech Co., Ltd. Oligonucleotides (Table 1) used were obtained from Shanghai Gene Pharma Co., Ltd.

The serum samples used in the experiment were obtained from the Clinical Hospital of Yangzhou University. Thirty blood samples were collected from 30 LC patients and 30 healthy subjects according to routine clinical laboratory procedures (Table 2). To demonstrate the excellent performance of the proposed sensor, only the LC patients at the Ⅰ stage were selected for the measurement. The blood samples were centrifuged (15 min, 3,000 rpm) and the serum was stored frozen and set aside. All experiments were performed according to the relevant laws and obtained approval from the Ethics Committee of the affiliated hospital of Yangzhou University. Consent documents were also obtained from all human subjects.

The Au seed solution was prepared by a one-step reduction method with sodium citrate (Wei et al., 2020). 80 mL of CTAB (100 mmol/L) solution and 0.8 mL of AgNO₃ (10 mmol/L) solution were added sequentially to the beaker containing 4 mL of HAuCl₄ (10 mmol/L) solution to obtain the growth solution. Add 1.6 mL of hydrochloric acid (1 mol/L) solution and 0.2 mL of AA (100 mmol/L) solution to the growth solution under continuous stirring. After 5 min, add 560 μL of the prepared Au seed solution to the mixed solution and placed in a water bath (28°C) for 24 h. After the color of the solution changes to dark purple, purification with ultrapure water was performed and dispersed in 20 mL of ultrapure water.

4 mL of GNBPs solution was mixed with 200 μL of 4-MBA solution (2 mmol/L) under continuous stirring for 1 h followed by centrifugation (15 min, 7,000 rpm). Subsequently, 60 μL of hpDNA1 solution (0.1 mmol/L) was added to 80 μL of freshly prepared TECP (1 mmol/L) buffer. The activated hpDNA1 solution was mixed with 2 mL of 4-MBA modified GNBPs solution for 12 h, then mixed with 40 μL of BSA (1 wt%) solution and incubated for 120 min. Then, the solution was dispersed in PBS buffer after three times of purifications (15 min, 5,000 rpm); thus, the SERS probes were prepared. Finally, GNBP probes were washed and incubated in 10 mL of 1 wt% BSA solution for 60 min.

At room temperature, 500 μL of MBs (0.5 mg/mL) solution was mixed with 5 μL of EDC (0.1 mol/L) solution and 5 μL of NHS (0.1 mol/L) solution and incubated with shaking for 30 min (500 rpm) to activate the carboxyl groups on the surface of MBs. After adding 60 μL of activated hpDNA2 solution (0.1 mmol/L) for 12 h, 10 μL of BSA (1 wt%) was added and incubated for 120 min. Then, the MBs were separated with a magnet, the supernatant was aspirated and washed repeatedly with PBS solution.

Serum sample was mixed with antibody-DNA conjugate, and incubated in the tube to ensure sufficient hybridization of the antibody tail DNA, followed by the addition of SERS probe solution, capture probe solution. Subsequently, a magnet was used to induce complex aggregation outside the tube wall, and the solid complex was removed and placed under Renishaw Raman microscope for testing. Test parameters: laser excitation wavelength of 785 nm, scanning range of 600–1800 cm⁻¹, objective magnification of ×50, laser power of 60 mW, exposure time of 10 s. To ensure the representativeness and validity of the SERS results, the SERS spectra of 15 different spot regions were selected randomly to gained the average result. UV-Vis absorption spectra were recorded with a Mapada UV-3000PC. Transmission electron microscopy (TEM) images were taken by FEI Tecnai G2 F30 S-twin field-emission transmission electron microscopy. Scanning electron microscope (SEM) images were taken by S-4800II field emission scanning electron microscope.

The morphology and structure of GNBPs were characterized by SEM and TEM. As shown in Figures 2A, B, GNBPs have two sharp tips with uniform size and good dispersion. The long diameter of GNBP is 41 nm and the short diameter is 16 nm. HRTEM image of the GNBPs (Figure 2C) shows that the lattice spacing is 0.235 nm, which corresponds to the (111) lattice plane located in the face-centered cubic crystal structure of Au. EDS spectrum of the GNBPs shown in Figure 2D reveals that the main elemental composition of GNBPs is Au and a small amount of Ag can be attributed to the underpotential deposition of Ag. Figure 2E shows a strong UV-Vis-NIR absorption peak at 693 nm due to the LSPR effect of the sharp tip on the GNBPs, while the narrow half-peak width further indicates the uniform size of the GNBPs. Figure 2F shows the SERS spectra of the Raman signal molecules 4-MBA (1 × 10⁻² mol/L) and 4-MBA (1 × 10⁻⁶ mol/L)-labeled GNBPs. The characteristic peak at 1,080 cm⁻¹ is generated by the in-plane cyclic respiration with υ (C-S) of 4-MBA molecular (Dziecielewski et al., 2020). The SERS signal of 4-MBA-labeled GNBPs was significantly enhanced compared to the weak signal of 4-MBA. The enhancement factor of GNBPs can be calculated according to the equation AEF=(I_SERS/C_SERS)/(I_RS/C_RS) (Yu and Cheng, 2022). Where, I_SERS is the SERS signal intensity of GNBPs at C_SERS concentration and I_RS is the Raman signal intensity under non-SERS conditions at C_RS concentration. After calculation, it is known that the enhancement factor of GNBPs is 6.67×10⁵. Therefore, GNBPs have a good SERS enhancement effect.

The SEM image of MBs is shown in Figure 3A, which shows that the surface of MBs is rough and homogeneous in size with an average diameter of 323.9 nm. GNBPs were assembled on the surface of MBs and more aggregation of GNBPs was observed between adjacent MBs particles (as shown in Figures 3B–C), indicating that the MBs enrichment strategy and the CHA reaction could lead to significant aggregation of GNBPs and dual-amplification of the SERS signal. The EDS spectrum in Figure 3D further verifies that GNBPs are successfully aggregated on the surface of MBs. To investigate the uniformity, SERS mapping was performed and scheme of color ranging from red (maximum intensity) to blue (minimum intensity) was employed. As shown in Supplementary Figure S1, although some areas still existed as blue or red, most of them were uniformly green and the RSD was calculated to 8.29%, proving the great uniformity of the proposed SERS sensor. For an intuitive illustration, finite-difference-time-domain (FDTD) simulation was performed to confirm the spatial electric field distribution which irradiated by a beam of linearly polarized light. The complex refractive index of gold was adopted from the refractive index database in the simulation software package Au (Gold)-CRC. All geometric parameters for simulations were consistent with the average actual size. As shown in Supplementary Figure S2, high density hotspots were generated around the tips of GNBP and MBs enrichment strategy can reduce the gaps between the GNBPs, forming more hotspots to enhance the electric field intensity. This is clear evidence of the amplification of the field through interparticle plasmon coupling, resulting from the further aggregation of GNBPs caused by magnet-induced aggregation effect. Thus, the composite structure can generate numerous hotspots for the SERS signal enhancement.

The CHA reaction efficiency determines the amount of GNBPs captured by MBs. Figure 4A optimizes the amount of SERS probe. When the SERS probe volume is less than 4 μL, the characteristic peak intensity of 1,080 cm¹ is significantly enhanced with increasing volume, which is due to increase of the CHA reaction efficiency caused by the increase of molecular collision chance in solution. When the SERS probe volume was greater than 4 μL, the intensity showed a decreasing trend with the increase of volume due to the fact that as the SERS probe contributes more to the background signal than to the SERS signal. Therefore, the optimal SERS probe amount is 4 μL. Similarly, the amount of capture probe can be optimized to 3 μL (Figure 4B). The reaction temperature is critical for hairpin DNA hybridization in CHA reactions. As shown in Figure 4C, the intensity of the 1,080 cm¹ characteristic peak increases with increasing temperature and reaches a maximum at 37°C. Figure 4D optimizes the pH of the reaction and it can be observed that the reaction efficiency is best when the pH is 7.5; therefore, the optimum pH for the reaction is 7.5.

Five different batches of prepared SERS sensors were selected for the detection of CYFRA21-1, and the detection reproducibility of the SERS sensors was verified. As shown in Figure 5A, the five SERS spectra did not differ significantly in shape. Figure 5B shows the intensity of the five SERS spectra at 1,080 cm¹ characteristic peak with a relative standard deviation (RSD) of 4.07%, indicating that the prepared SERS sensor has excellent reproducibility. To investigate the specificity of the SERS sensor, CEA, AFP and SCC-Ag were selected as interferents. CYFRA21-1 solution (1 μg/mL) and CEA, AFP, SCC-Ag solution (10 μg/mL) and blank control were detected using the SERS sensor, respectively. As shown in Figures 5C, D, CYFRA21-1 exhibited significant characteristic peaks compared with the interferer and blank control, indicating the good specificity of the prepared SERS sensor.

CYFRA21-1 was dispersed into PBS buffer and diluted to different concentrations (1 pg/mL, 10 pg/mL, 100 pg/mL, 1 ng/mL, 10 ng/mL, 100 ng/mL, and 1 μg/mL). Detection of different concentrations of CYFRA21-1 was performed with the prepared SERS sensor and the working curve of CYFRA21-1 versus the amount of change in SERS signal intensity was plotted based on the signal of the 4-MBA at the characteristic peak at 1,080 cm¹. As shown in Figure 6A, the SERS signal intensity gradually increased with the increase of CYFRA21-1 concentration. The concentration-intensity correction curve in logarithmic coordinates was obtained by taking the logarithm of CYFRA21-1 concentration as the horizontal coordinate and the SERS signal intensity at the 1,080 cm¹ characteristic peak as the vertical coordinate (Figure 6B). It can be observed that the SERS signal intensity at the characteristic peak of 1,080 cm¹ showed a good linear relationship with the logarithm of CYFRA21-1 concentration in the range of 1 pg/mL∼1 μg/mL. The linear regression equation was y = 2,863.40x+34,998.40 (R ² = 0.9808) and the limit of detection for CYFRA21-1 in PBS was calculated to 0.69 pg/mL. Similarly, to investigate the detection performance of CYFRA21-1 in serum with the SERS sensor, CYFRA21-1 was dispersed into serum and diluted to different concentrations (1 pg/mL, 10 pg/mL, 100 pg/mL, 1 ng/mL, 10 ng/mL, 100 ng/mL, and 1 μg/mL). As shown in Figures 6C, D, the SERS signal intensity showed a good linear relationship with the logarithm of CYFRA21-1 serum concentration. The linear regression equation was y = 2,642.83x+32,233.65 (R ² = 0.9802). The detection limit of the SERS sensor for CYFRA21-1 in serum was calculated to be 0.76 pg/mL. As shown in Table 3, the sensitivity of the SERS sensor based on magnetic sphere enrichment and CHA signal amplification was superior to other protein marker detection methods. These results indicate that the SERS sensor can be used for the quantitative analysis of CYFRA21-1 with excellent sensitivity.

In order to investigate the accuracy of the SERS sensor in clinical application, CYFRA21-1 expression levels in the serum of healthy individuals and LC patients were measured via the SERS sensor. Figures 7A, B show the average SERS spectra of serum SERS and the signal intensity at the characteristic peak of 1,080 cm¹ in 30 healthy individuals and 30 LC patients, respectively, and it can be seen that the intensity at the characteristic peak of 1,080 cm¹ in LC patients is significantly higher than that in normal individuals. The expression level of CYFRA21-1 was calculated by substituting the intensity at the 1,080 cm¹ peak in the SERS spectra of 30 healthy subjects and 30 LC patients into the linear regression equation (Supplementary Table S1). To verify the accuracy of the SERS assay results, the SERS assay results were linearly fitted to the ELISA results, and the Pearson correlation coefficient was calculated to be 0.97, indicating that the results of the two assays were highly correlated (Figures 7C). Therefore, the SERS sensor has good accuracy and broad application prospect for clinical diagnosis of LC.