Over the last decades, Raman spectroscopy (RS) and surface-enhanced Raman spectroscopy (SERS) have been applied for analyzing biological particles such as cells [1,2], bacteria [3,4], and extracellular vesicles (EVs) [5,6]. Such biological particles contain a wide range of biomolecules, for example, nucleic acids, proteins, carbohydrates, and lipids. Therefore, their Raman spectra have a large number of characteristic bands, which lead to the Raman-band overlap among the biomolecules. When analyzing biological samples, it is essential to discriminate the spatial origin of the signal to analyze the surface structure and interior content of the particles [7]. For example, EVs are currently one of the rapidly growing research topics in the life sciences. Increasing amounts of evidence suggest that EVs play an important role in cell-to-cell communication in various physiological processes. EVs are layered structures. The EV membrane structure defines their targeting properties, and the functionalities depend on the interior content [5].

In this Letter, we present a novel method called distance-controlled SERS, which is based on SERS wavelength dependence and SERS distance dependence of gold nanospheres (AuNSs). SERS enhancement is mainly contributed by the electromagnetic (EM) enhancement mechanism (10⁴–10⁶) and the chemical enhancement mechanism contributes 10³–10⁵ [8,9]. The EM enhancement mechanism is caused by the excitation of a localized surface plasmon resonance (LSPR). LSPR occurs when the incident EM field E₀ coherently drives the electron cloud of a SERS-active surface [8]. The oscillating electron cloud radiates an enhanced EM field E with gain factor g. If there are molecules in this enhanced EM field, the molecules strongly radiate Raman scattering field E’. The Raman field will be further enhanced in the same LSPR principle of the SERS-active surface with the gain factor g’. They finally yield a SERS enhancement factor (EF) given by [8,10]

In this study, biological samples are modeled by biotin-coated polystyrene (PS@bio) beads. We study the Raman spectra of 100-nm PS@bio beads conjugating with streptavidin-modified AuNSs (strep@AuNSs). Biotin–streptavidin (bio–strep) binding is known for its high binding affinity and hence the PS beads conjugate AuNSs through the bio–strep complex (PS@bio–strep@AuNSs) [13]. In this conjugation model, the biotin–streptavidin complex acts as an exterior layer while the PS bead acts as an interior. Using the finite-element method (FEM) in COMSOL Multiphysics® software, we model the PS@bio–strep@AuNSs to study distance-controlled SERS and compare modeling results with measurements. Experimentally, we use dry PS@bio–strep@AuNSs samples conjugated with 10-nm, 20-nm, 40-nm, 60-nm, and 80-nm AuNSs. We aim to find the size-optimized AuNSs that weaken the Raman signal of the interior and strengthen the Raman signal of the exterior layer. Computationally and experimentally, we found that PS@bio–strep@60-nm AuNSs produced the optimal signal from the exterior layer.

The biotin-coated PS beads purchased from Nanocs Inc. had a 100-nm diameter and were suspended in 1% aqueous solution. The streptavidin-modified AuNSs with varying diameters (10 nm, 20 nm, 40 nm, 60 nm, 80 nm) purchased from Nanocs Inc. with the same concentration of 0.1 mg/mL were spherical and monodispersed in a 50 mM phosphate buffer, pH 7.4 with 10% glycerol, 0.1% BSA, and 0.01% sodium azide. The size standard deviation of the AuNSs was in the range of 15%. The equilibrium time of the biotin–streptavidin binding is 5 hours [14,15]. Therefore, the PS beads and the AuNSs were mixed with the ratio 10:1 and stored at 4°C for 5 hours to form the conjugations PS@bio–strep@AuNSs. Samples of 0.5-µL PS@bio–strep@AuNSs were pipetted on a stainless-steel surface and dried at room temperature. In COMSOL Multiphysics software, the PS@bio–strep@AuNSs samples were modeled by a PS bead, a single AuNS, and a bio–strep layer, as illustrated in Fig. 1(a). We assumed that the bio–strep complex was formed around the PS bead. The dimensional size of the bio–strep complex is 4.2 nm × 4.2 nm × 5.8 nm, which was measured by Weber et al. [16]. Li and Zhang showed that the thickness of the bio–strep complex is 3.6 ${\pm} $ 0.2 nm after the equilibrium time [15]. In this work, we chose a 4-nm bio–strep layer for our model, which was in line with the work by Li and Zhang [15]. This layer had a refractive index of 1.42 [15]. The diameter of the PS was 100 nm and its refractive index was 1.60 [17]. The complex refractive index of AuNSs was modeled by the Brendel–Bormann model from Rakic’s study [18]. The PS@bio–strep@AuNSs model was in a homogeneous medium with a refractive index of 1.2. In Fig. 1(b), we simulated the extinction cross section of a single AuNS in different dielectric media with refractive indices ranging from 1 to 1.33 to study the effect of the local refractive index on the extinction cross section of AuNSs. Theoretically, the refractive index of the media is approximately 1 in a dry state (in air) while the refractive index is approximately 1.33 in the liquid state (in solution). Here, we chose a homogeneous medium with a refractive index of 1.2 to match the experimental conditions described above. In the simulations, it was also assumed that the local refractive index surrounding PS@bio–strep@AuNSs increased due to chemical residuals from the buffer solution [19].

 

Fig. 1. Simulations of the interaction between the gold nanosphere and the PS bead. (a) Sideview (xz plane) of 3D model of a polystyrene bead (PS) conjugating a gold nanosphere (Au) via biotin–streptavidin complex (bio–strep layer). Inset shows how the point probes were placed for calculating SERS EF. (b) Log-scale plot of extinction cross section as a function of the diameter of a gold nanosphere in different dielectric media at 532-nm excitation.

Download Full Size | PDF

In the model, the AuNS conjugated the PS bead in the z-direction. To avoid an infinite edge in the model, there was a 0.5-nm space between the AuNS and the bio–strep layer. The model was simulated with a 532-nm excitation wavelength corresponding to the further Raman spectroscopy experiments. The electric field was linearly polarized in the z-direction and propagated in the x-direction. As illustrated in Fig. 1(a), 32 point probes for calculating SERS EF were placed along the z-direction from the outer surface of the bio–strep layer to the PS region. The number of probes distributed on the bio–strep layer and on the PS region was equal. The whole model was surrounded by a spherical perfect matching layer (PML) with a thickness of half of the excitation wavelength. The model was meshed by the physics-controlled tool with a smallest feature of 0.5 nm. We modeled PS@bio and PS@bio–strep@AuNSs with different AuNS diameters of 10 nm, 20 nm, 40 nm, 60 nm, and 80 nm corresponding to the experimental samples.

Raman spectra of PS@bio and PS@bio–strep@AuNSs samples were measured with time-gated Raman spectroscopy. An example of a time-gated Raman spectrum is shown in Fig. S1 of Supplement 1. The data processing of the Raman spectra is described in Fig. S2 of Supplement 1. The processed Raman spectra are shown in Fig. 2. All other spectra can be found in Fig. S3 of Supplement 1. The dry samples formed a so-called coffee-ring shape. The Raman spectra were measured at five different locations at the edge of the coffee ring and then averaged. In Table 1, the Raman bands of PS were referenced from Anema et al. [20] and the bands of the bio–strep complex were referenced from Focsan et al. [21] and Galarreta et al. [22]. The fingerprint bands of PS and bio–strep complex were based on two criteria. First, the intensity was high. Second, there was no overlap of Raman bands between PS and bio–strep complex. For the PS, only the 627-cm^–1, 996-cm^–1, and 1610-cm^–1 bands met these criteria. The fingerprint bands of the bio–strep complex were 1462 cm^–1 and 1562 cm^–1. We chose the 996-cm^–1 band from PS as a reference band because of its highest intensity, as shown in Fig. 2.

 

Fig. 2. Comparison between Raman spectra of PS@bio and PS@bio–strep@60-nm AuNSs. The fingerprint Raman bands 627 cm^–1, 996 cm^–1, and 1610 cm^–1 are from PS. The fingerprint Raman bands 1462 cm^–1 and 1562 cm^–1 are from biotin–streptavidin complex.

Download Full Size | PDF

Table 1. Comparison Raman Bands of Polystyrene and Biotin–Streptavidin Complex

View Table

AuNSs affect the intensity of observable Raman scattering due to two opposite mechanisms. First, SERS enhancement boosts the signal especially in the vicinity of AuNS. However, AuNSs also absorb light reducing the intensity of Raman scattering. Therefore, the total observable signal depends on both the absorbance of AuNSs and the distance between the scattering point and the AuNS. In Fig. 3(a), the SERS EF was calculated based on Eq. (1), where the SERS EF is approximately proportional to |E|⁴ [8]. The average SERS EF in the bio–strep layer was much higher than in the PS region. This shows that AuNSs strongly enhanced the signal from the exterior layer in dry PS@bio–strep@AuNSs samples. However, the absorption of AuNSs decreased the optical intensities (both excitation and emission) in the “high-volume” originated signal. As the result, the signal from the interior (large volume of the PS) was decreased when conjugating the PS with AuNSs. The signal from the interior was lowest when AuNSs gave the highest SERS EF. In Fig. 3(a), the highest computational SERS EF occurred at 60-nm AuNSs. This is in agreement with the experimental observations shown in Fig. 3(b), where the 996-cm^–1 band from the PS in PS@bio–strep@60-nm AuNSs was the lowest. The fingerprint Raman bands from the bio–strep complex (1462 cm^–1 and 1562 cm^–1) and the PS (627 cm^–1 and 1610 cm^–1) were further analyzed to prove our predictions in simulations. In Fig. 4, the ratios between the fingerprint bands and the reference band (996 cm^–1 from the PS) were calculated and then normalized by the ratio of the PS@bio sample. The ratios of the PS fingerprint bands (627 cm^–1 and 1610 cm^–1) did not change when the PS was conjugated with AuNSs since the reference band in the ratios was a PS band. However, the ratios of the fingerprint bands of the bio–strep complex (1462 cm^–1 and 15602 cm^–1) changed and had a maximum for the sample PS@bio–strep@60-nm AuNSs.

 

Fig. 3. Comparison between computational SERS EF and observations from measured SERS spectra. (a) Log-scale plot of computationally obtained spatially averaged SERS EF as a function of the diameter of AuNS in the biotin–streptavidin layer (solid line), the PS bead (dashed line) in PS@bio–strep@AuNSs models in the media with the refractive index of 1.2 at 532-nm excitation. Here, 0-nm diameter indicates the PS@bio model. (b) Intensity of 996-cm^–1 band from PS as a function of AuNS diameter in PS@bio–strep@AuNSs dry samples. Here, 0-nm diameter indicates the PS@bio dry sample. The original spectra are shown in Fig. S3 of Supplement 1.

Download Full Size | PDF

 

Fig. 4. Normalized ratio between fingerprint bands of PS (627 cm^–1 and 1610 cm^–1) and of biotin–streptavidin layer (1462 cm^–1 and 1562 cm^–1) and the 996-cm^–1 reference band from PS as a function of AuNS diameter in PS@bio-strep@AuNSs dry samples. Here, 0-nm diameter indicates the PS@bio dry sample.

Download Full Size | PDF

In conclusion, this Letter demonstrates distance-controlled SERS with PS@bio–strep@AuNSs models. We aimed to extract the Raman information from the bio–strep complex. In Fig. S4 of Supplement 1, we normalized the Raman spectra of PS@bio and PS@bio–strep@60-nm AuNSs and then took the absolute difference between the two normalized spectra. The Raman signals from the bio–strep complex were very low because of the low signal-to-noise ratio. However, we experimentally observed the effect of distance-controlled SERS probes on the Raman spectra of the PS and the bio–strep complex. The observations were in good agreement with the models in COMSOL Multiphysics software. The highest signal from the bio–strep complex was observed for PS@bio–strep@60-nm AuNSs, which matched the highest computational SERS EF. In Fig. 3(b) and Fig. 4, there are large error bars in the measurements. This is due to the non-uniform distribution of PS@bio–strep@AuNSs in dry coffee-ring samples. Furthermore, large variations in Raman intensities are the major issue in SERS due to the probability of molecules being found in SERS hotspots [23]. In this work, the concentration of biotin and streptavidin was small, leading to large variations in the signal. The developed characterization method might be applicable, for example, as a quality control tool in drug delivery systems relying on EVs or in EV-based diagnostics. In therapy applications, EVs carry a wide array therapeutically relevant ingredients into targeted recipient cells [24], while in diagnostic applications, the composition of EVs can be attributed to a specific disease, such as cancer [25,26]. In both envisioned applications, the compositional information and its spatial origin (surface or interior) can be analyzed by taking an EV-containing sample followed by partitioning it into smaller sub-samples. By applying Au-nanoparticles with different sizes on each sub-sample, the depth-dependent compositional information can be obtained by comparing the differences in Raman spectra.