Chloroauric acid (HAuCl₄·3H₂O, 99%), cetyltrimethylammonium bromide (CTAB, 99%), silver nitrate (AgNO₃, >99%), sodium borohydride (NaBH₄, 99%), L-ascorbic acid (AA, 99%), trisodium citrate (C₆H₇O₅Na₃), tetraethyl orthosilicate (TEOS), sodium hydroxide (NaOH), (3-dimethylaminopropyl) ethyl-carbodiimide monohydrochloride (EDC), malachite green isothiocyanate (MGITC), phosphate-buffered saline (PBS), N-hydroxysuccinimide (NHS), and 2-(N-morpholino) ethane sulfonic acid (MES) were purchased from Aladdin Chemical Reagent Co. Ltd., China. Thiolcarboxylic polyethylene glycol (HS-PEG-COOH, MW∼400), dithiothreitol (DTT), Interleukin 5 (IL-5), a monoclonal antibody to human IL-5 (detector antibody, dAb), a monoclonal antibody to human IL-5 (capture antibody, cAb), Interleukin 8 (IL-8), immunoglobulin E (IgE), and eosinophil cation protein (ECP) were purchased from Sangon Biotech Co. Ltd., China. Ultrapure water (18 MΩ cm^-1) was used throughout the experiment.

Patients with asthma were recruited from the Second Affiliated Hospital of Fujian Medical University. Subjects with asthma (n = 10) had objective evidence of variable airflow obstruction as indicated by one or more of the following: 1) methacholine airway hyperresponsiveness (PC20 FEV1 < 8 mg/mL), 2) >15% improvement in FEV1 10 min after 200 mg inhaled salbutamol, or 3) peak expiratory flow (>20% maximum within-day amplitude from twice-daily peak expiratory flow measurements over a period of 14 days).

The protocol for sputum processing was modified from that recommended by Woolhouse et al. (2002) Sputum samples were treated with freshly prepared 0.1% DTT at a ratio of 4:1 (v/w). After treatment, each sample was rocked for 15 min. An additional 4 volumes of NaCl were added, which were then rocked for a further 5 min. The sputum was centrifuged at 1,200 rpm for 5 min and the supernatant was stored at −80°C until required.

AuNRs were prepared by the seed-mediated growth process, according to the report by Sau and Murphy (2004). The Au seed solution was prepared by adding a freshly ice-cold NaBH₄ solution (0.60 mL, 0.01 M) to a mixed aqueous solution composed of HAuCl₄ (0.25 mL, 0.01 M) and CTAB (7.50 mL, 0.1 M). This seed solution was stirred rapidly for 2 min and then kept at room temperature (25°C) for 2 h. To prepare the AuNRs growth solution, HAuCl₄ (4 mL, 0.01 M) and AgNO₃ (0.75 mL, 10 mM) were mixed with CTAB (95 mL, 0.1 M), followed by the addition of a freshly prepared AA solution (0.64 mL, 0.1 M). In total, 0.4 mL of Au seed solution was added to the AuNRs growth solution with rapid stirring for 3 min, then kept at room temperature for 3 h. The resulting AuNRs were separated and purified by centrifugation at 10,000 rpm for 30 min.

AuNRs@SiO₂ was prepared by the sol-gel method with a slight modification (Jiang et al., 2020). The prepared AuNRs solution was centrifuged at 7,500 rpm for 20 min and redispersed in ultrapure water. MGITC solution (100 μL, 10 μM) was added dropwise to mix well with AuNRs solution, centrifuged at 7,500 rpm for 15 min, and redispersed in 10 mL of ultrapure water. NaOH (0.1 mL, 0.1 M) was added to the mixture with stirring for 10 min 40 μL of TEOS/methanol (V/V: 1/4) solution was injected three times at 30-min intervals. The reaction was allowed to be gently stirred for 24 h, and the resulting colloidal solution was centrifuged twice at 4,500 rpm for 15 min.

HS-PEG-COOH (20 μL, 20 μM) was added dropwise to mix well with the AuNRs@SiO₂ solution, centrifuged at 4,500 rpm for 15 min, and redispersed in 1 mL of PBS. In order to activate the carboxyl groups, MES (50 μL, 0.25 M) and EDC (15 μL, 0.1 M) were added to the mixture and reacted for 20 min at room temperature with gentle inversion, followed by the addition of NHS (10 μL, 0.5 M). After mixing for 20 min, the mixture was purified by centrifugation at 4,500 rpm for 15 min and redispersed in 1 mL of PBS. Detection antibody (10 μL, 5 μg/mL) and capture antibody (10 μL, 5 μg/mL) were added to the mixture (500 μL). Both were stirred for 3 min at room temperature, followed by incubation at 37°C for 1 h. The two separate reaction parts were mixed together to obtain a biosensor solution for IL-5 detection.

A total of 10 μL of IL-5 at different concentrations (0, 0.05, 0.1, 0.5, 1, 5, 10, and 50 pg/mL) were added to the biosensor solution and shaken at 37°C for 15 min. The mixtures were then tested by Raman spectroscopy. To examine the stability of the biosensor, IL-5 (10 μL, 10 μg/mL) was added to 1 mL of biosensor solution with 50% diluted serum and 50% diluted sputum, followed by incubation at 37°C for 15 min. The mixtures were then measured by Raman spectroscopy for 1 week. To examine the specificity of the biosensor, ECP, IgE, IL-8, and IL-5 (10 μL, 10 μg/mL) were added to 1 mL of biosensor solution with 50% diluted serum and 50% diluted sputum, followed by incubation at 37°C for 15 min. The mixtures were then measured by Raman spectroscopy. To evaluate the response of the biosensor, 10 μL of IL-5 with different concentrations (0.5, 1, 5, and 10 pg/mL) were added to 1 mL of biosensor solution with 50% diluted sputum, followed by incubation at 37°C for 15 min. The samples were then analyzed by Raman spectroscopy.

Sputum IL-5 was measured by ELISA (Shanghai Jianglai Industrial Limited by Share Ltd., Shanghai, China). The concentration detection range was 1.56–100 pg/mL. The detection limit was 0.38 pg/mL. Sputum IL-5 was measured by SERS. In total, 1 mL of sputum supernatant was added to 1 mL of biosensor solution, followed by incubation at 37°C for 15 min. The samples were then analyzed by Raman spectroscopy.

Transmission electron microscopy (TEM) images were obtained using a JEM-1400 (100 kV) microscope. The UV-vis absorption spectra were obtained using a TU-1810 UV-vis spectrophotometer. The zeta potential was measured with a Malvern Zetasizer ZEN3600 instrument. Energy-dispersive X-ray spectroscopy (EDS, Thermo Scientific Talos F200X) was used to analyze the composition. Raman spectra were obtained using a compact Raman system with 785 nm excitation and a 1 s exposure time (40 mW, Advantage Raman Series, DeltaNu).

The surface enhancement factor (EF) was calculated using Eq. 1 (Dong et al., 2021). E F = I S E R S / N S E R S I N R S / N N R S = I S E R S / C S E R S × V S E R S I N R S / C N R S × V N R S = I S E R S / C S E R S I N R S / C N R S (1) where I _SERS is the intensity of the SERS spectrum of the MGITC-labeled biosensor assembly at 1,172 cm⁻¹, and I _NRS is the intensity of the same peak of the normal Raman spectrum of the MGITC aqueous solution. N _NRS and N _SERS represent the number of MGITCs irradiated by a laser beam in normal Raman scattering (NRS) and SERS, respectively. V _NRS and V _SERS are the volumes of solution excited by a laser beam in NRS and SERS, respectively. C _NRS is the concentration of MGITC in the water solution. C _SERS denotes the concentration of MGITC in the SERS biosensor solution. The strongest Raman intensity at 1,172 cm⁻¹ peak was chosen for EF calculation.

The rate of the surface enhancement factor (REF) was calculated based on Eq. 2: (Lee et al., 2021): R EF = EF Assembly EF S i n g l e = I S E R S A s s e m b l y / N S E R S A s s e m b l y I N R S / N N R S I S E R S s i n g l e / N S E R S s i n g l e I N R S / N N R S = I S E R S A s s e m b l y I S E R S s i n g l e (2) where EF _Assembly is the EF of the SERS spectrum of the MGITC-labeled biosensor assembly, and EF_Single is the EF of the same peak of the normal Raman spectrum of the biosensor single solution.

As shown in Figure 1A, the AuNRs prepared by the seeded growth method exhibited uniform size and good monodispersity. The length of the long axis was 55 ± 5 nm, and the short axis length was 20 ± 2 nm, with an aspect ratio of 2.75 ± 0.5. As shown in Figure 1B, the silicon dioxide layer of AuNRs@SiO₂ was extremely thin (2∼3 nm). However, it did not completely cover the surface of the nanorods: the silicon dioxide layer on the side of the gold nanorods was more obvious, while almost no coating was observed on the head. The UV-Vis absorption spectrum of AuNRs@SiO₂ is shown in Figure 1C. It can be seen from the figure that the AuNRs had two obvious peaks of surface plasmon resonance (SPR): 516 nm (SPR on the horizontal axis) and 650 nm (SPR on the vertical axis). After the AuNRs were coated with silica, the SPR in the horizontal axis direction was maintained, while the SPR peak in the vertical axis direction showed a red shift to 660 nm.

Since the preparation of AuNRs requires the use of a high concentration of CTAB as a surfactant to maintain the stability of the nanorods, these AuNRs tend to aggregate with the reduction of CTAB during the subsequent modification process. Moreover, CTAB has strong biological toxicity, which greatly limits the application potential of AuNRs in the fields of biological research, medical imaging, and treatment. Therefore, in this paper, SiO_2, which has better biocompatibility, was used for surface modification. AuNRs@SiO₂ could not only improve the stability and dispersibility of the nanoparticles but also have excellent light transmittance. The coated AuNRs would have less impact on the light-absorbing properties and could also provide a large number of connection sites, making the subsequent modification and biomolecule coupling much easier.

It is worth noting that the thickness of the silica layer was found to be highly related to the thickness of the CTAB layer on AuNRs. TEOS can be easily attracted by positively charged CTAB during the hydrolysis process, so that the TEOS is hydrolyzed and coated on the surface of AuNRs (Pérez-Juste et al., 2005). Through the seed-mediated growth method, the {100} planes on the side of AuNRs were generally wrapped by a thick layer of CTAB, while the {110} planes on the head were relatively less (Hill et al., 2009). Thus, when a small amount of TEOS was added, the TEOS was preferentially attracted to the side of AuNRs for hydrolytic coating. In contrast, the head of AuNRs had less CTAB, so it could not attract enough TEOS, which made it difficult to form a silicon dioxide layer.

This work then suggests the use of EDS spectroscopy to explain how SiO₂ is applied to AuNRs (Figure 1D). The results displayed in the figure, which shows the distribution of Au, Si, and O elements, demonstrate that the distribution ranges of these elements remain mostly unchanged along the length of the AuNRs. However, across the width of the AuNRs, the distribution ranges of Si and O elements are wider than those of Au. This suggests that SiO₂ primarily coats the sides of the AuNRs, while there is less or no coating at their ends. As a result, the heads of the AuNRs are comparatively bare for the upcoming synthetic modification. This suggests that there is more SiO₂ encapsulation on the sides of the AuNRs and less or no encapsulation on the end of the AuNRs.

As mentioned above, during the preparation of AuNRs@SiO₂, a small amount of SiO₂ was preferentially coated on the sides of the AuNRs, while the heads were relatively bare. Therefore, when the bifunctional molecule SH-PEG-COOH was added, it tended to couple to the heads of the AuNRs through the strong Au-S chemical bonds. Furthermore, to modify the IL-5 antibody on the AuNRs, the carboxyl groups in SH-PEG-COOH needed to be activated. With the activation by EDC-NHS, the carboxyl group of SH-PEG-COOH was chemically bonded with the amino group on the surface of the antibody, forming a stable amide bond (Yuan et al., 2021). Thus, the IL-5 antibody was attached to the heads of the AuNRs. In the construction of an anti-IL-5-coated AuNR biosensor, when IL-5 antigen was added, the specific recognition could trigger the end-to-end self-assembly of the AuNRs.

As shown in Figure 2, when IL-5 solution was added to the AuNR sensor, the AuNRs formed an ordered structure by end-to-end self-assembly. At a low concentration of IL-5 (e.g., 0.1 pg/mL), the IL-5 antigen was not able to bind all the AuNRs (Figure 2A), so the self-assembled structure was shorter. With the increase in IL-5 concentration, the end-to-end self-assembled structure gradually elongated. When the other conditions remained unchanged, the number of end-to-end structures was directly related to the concentration of IL-5 added, indicating that the formation of the self-assembled AuNR structure was caused by the specific binding between IL-5 and antibody.

Under an appropriate light wavelength, AuNRs underwent a plasmon coupling effect to enhance the surrounding electromagnetic field, which exhibited the SERS effect (Dey et al., 2020). When the AuNRs formed a self-assembled structure, the electromagnetic field near the gap of the AuNRs was further enhanced, so that the Raman signal was amplified by an order of magnitude (Xie et al., 2019). The Raman-active molecule selected in this study was isothiocyanate malachite green (MGITC), which contains several characteristic fingerprint peaks in the Raman spectrum: 798, 912, 1,172, 1,295, 1,362, and 1,583 cm⁻¹ (Kang et al., 2020). In order to illustrate the sensitivity of the MGITC-labeled AuNR biosensor, different concentrations of IL-5 were added for testing, and the SERS spectrum was collected as shown in Figure 3A. When the concentration of IL-5 was 0 pg/mL, the sensor was not yet assembled, and the corresponding SERS signal in Figure 3A was very weak, showing only the SERS signal of MGITC in monodispersed AuNRs. With the increase of IL-5 concentration, the end-to-end self-assembly of AuNRs was triggered. The Raman signal of MGITC was gradually enhanced and amplified by orders of magnitude due to the SPR effect between adjacent AuNRs. It is worth noting that the number of assemblies increased with the increase in IL-5 concentration, which was consistent with the results in Figure 2. When the concentration of IL-5 was 50 pg/mL, the SERS signal was the strongest. The calculation formulas for EF and R_EF were then obtained. As a result, the EF, in this case, was 1.39 × 10⁵, and the R_EF was 33.1. When the IL-5 concentration was 0.05 pg/mL, the corresponding SERS signal was only slightly higher than that of the monodisperse AuNRs, and the EF and R_EF values were 6.21 × 10³ and 1.48, respectively. Therefore, the AuNR biosensor was able to effectively enhance the Raman signal of MGITC by inducing the formation of end-to-end self-assembly through biorecognition.

As shown in Figure 3B, the relationship curve between the SERS signal and IL-5 concentration was presented, where the SERS signal was obtained by taking the strongest and characteristic Raman peak of MGITC at 1,172 cm⁻¹. By linear fitting, the SERS signal showed a good linear relationship with the detected IL-5 concentration in the range of 0.1–50 pg/mL, after end-to-end self-assembly of AuNRs. The equation for the calibration curve can be described as follows: Y = 9665.57   X + 13164.96 (3) where Y is the SERS signal intensity of 1,172 cm⁻¹, and X represents the logarithmic value of IL-5 concentration. The correlation coefficient (R ²) of the equation is 0.994. In this paper, the linear range of the AuNR sensor for IL-5 concentration detection was 0.1–50 pg/mL, far lower than the detection range of ELISA detection kits commonly used for IL-5 detection (Tanaka et al., 2020; Yu and Li, 2022). Thus, the self-assembly of AuNRs with the enhanced SERS signal could be used to detect airway inflammatory factors, and its detection range could fully meet the requirements of clinical applications.

In order to further explore the feasibility of the developed AuNR biosensor in clinical applications, we investigated the stability, specificity, and recovery rate of this method in diluted sputum, respectively. As shown in Figure 4, the SERS spectrum was recorded, reflecting the changes in the characteristic peaks of MGITC within 1 week of the assembly system formed by IL-5-induced specific recognition. Within the first 3 days, the intensity of the characteristic peak at 1,172 cm⁻¹ in 50% diluted sputum decreased by 9.7% from 28,317 to 25,559. In the following days, the MGITC characteristic peaks showed little change, and the assembly system remained unchanged. The results indicated that the sensor prepared in this paper had good stability in diluted sputum.

As shown in Figure 5, different interference reagents and IL-5 were spiked into the sensor at 50 pg/mL, and the SERS spectra were collected. It can be seen that the SERS signal of added IL-5 was significantly enhanced, while the rest of the interference reagents showed no obvious change in the SERS signal. This result showed that the biosensor had good specificity. The antibody modified on the heads of the AuNRs showed specific biorecognition of IL-5 and did not bind with the interference reagent.

As shown in Table 1, a recovery rate experiment with the biosensor was also carried out. The intensity of the highest characteristic peak at 1,172 cm⁻¹ in the SERS spectrum was collected. The concentration of IL-5 could be obtained using Eq. 3, and the recovery rate of the detection method was calculated. In 50% diluted sputum, the method yielded IL-5 recoveries in the range of 94.3%–109.4%. These results showed that the AuNR biosensor still maintained good sensitivity in complex biological environments such as sputum, which would fulfill the sensitive detection of trace airway inflammatory factors.

In this study, sputum IL-5 was measured by an ELISA kit and SERS biosensor in 10 subjects with asthma. As shown in Table 2, there are three cases that tested negative for the ELISA, and the agreement between the two methods was good for the other cases. The sputum IL-5 of samples 1, 4, and 5 measured by the SERS biosensor were 0.91, 0.68, and 1.32 pg/mL, respectively, which were not in the detection range of the ELISA kit (1.56–100 pg/mL). Compared to the ELISA method, the SERS biosensor provided greater accuracy and higher detection sensitivity.