Surface enhanced Raman scattering artificial nose for high dimensionality fingerprinting

Abstract

Label-free surface-enhanced Raman spectroscopy (SERS) can interrogate systems by directly fingerprinting their components' unique physicochemical properties. In complex biological systems however, this can yield highly overlapping spectra that hinder sample identification. Here, we present an artificial-nose inspired SERS fingerprinting approach where spectral data is obtained as a function of sensor surface chemical functionality. Supported by molecular dynamics modeling, we show that mildly selective self-assembled monolayers can influence the strength and configuration in which analytes interact with plasmonic surfaces, diversifying the resulting SERS fingerprints. Since each sensor generates a modulated signature, the implicit value of increasing the dimensionality of datasets is shown using cell lysates for all possible combinations of up to 9 fingerprints. Reliable improvements in mean discriminatory accuracy towards 100% are achieved with each additional surface functionality. This arrayed label-free platform illustrates the wide-ranging potential of high-dimensionality artificial-nose based sensing systems for more reliable assessment of complex biological matrices.

Fig. 1. Schematic illustration of artificial-nose-empowered surface-enhanced Raman spectroscopy.

‘Functionalized Array for Surface-Enhanced Raman Spectroscopy (FASERS)’ represents an array of plasmonic surfaces for label-free SERS presenting different self-assembled monolayers. A range of molecular interactions takes place within complex biological media at each unit sensor where mildly selective SERS enhancement of the constituents gives multiplexed spectral datasets. The increased data dimensionality obtained enables facile identification of closely related samples.

Fig. 2. Fabrication and characterization of FASERS substrates.

a Schematic illustration and representative SEM images of gold film-coated polystyrene beads (PS)–Si₃N₄ (Au-nanopillars) SERS-active substrates that are used in FASERS. SEM images were obtained from top view and angled view from 45° tilted stage without extra metal coating. Scale bar, 400 nm. b Normalized SERS intensities of two prominent peaks of 4-aminothiophenol (ATP) with varying concentration on non-functionalized Au-nanopillars. Data represent mean ± 1 s.d. from nine obtained spectra (N = 3, n = 3 spectra) and fitted with a sigmoidal curve. Inset graph shows the mean of nine spectra from a 1 mM ATP solution. c SERS spectra of functionalized Au-nanopillars in PBS (10 mM, pH 7.4) with various SAM-forming molecules: non-functionalized (bare), 1-propanethiol (3CH₃), 3-mercapto-1-propanol (3OH), 3-mercaptopropionic acid (3COOH), 3-amino-1-propanethiol (3NH₂), 1-undecanethiol (11CH₃), 11-mercapto-1-undecanol (11OH), 11-mercaptoundecanoic acid (11COOH), 11-amino-1-undecanethiol (11NH₂). Solid lines and grey shaded area represent mean and ±1 s.d. of nine obtained spectra (N = 3, n = 3 spectra). d Representative AFM images of several of the functionalized gold film-coated Si wafers (Au–Si) with various SAM-forming molecules (from top-left to bottom-right): Bare, 3NH₂, 11CH₃, 11COOH. Scale bar, 400 nm. Average of the mean roughness (R_a) of each surface was noted in the image with ±1 s.d. (n = 3 scans). Inset images indicate the range of water contact angles observed at each surface. e Representative high-resolution XPS spectra of C 1s, S 2p and Au 4f for 3OH (top), 3COOH (middle) and 3NH₂ (bottom) SAM-functionalized Au–Si.

Fig. 3. Physicochemical SERS fingerprints of molecular analytes using FASERS.

a Chemical structures of the four model analyte molecules. b–e Series of SERS spectra obtained from functionalized Au-nanopillars with various SAM-forming molecules: non-functionalized (bare), 1-propanethiol (3CH₃), 3-mercapto-1-propanol (3OH), 3-mercaptopropionic acid (3COOH), 3-amino-1-propanethiol (3NH₂), 1-undecanethiol (11CH₃), 11-mercapto-1-undecanol (11OH), 11-mercaptoundecanoic acid (11COOH), 11-amino-1-undecanethiol (11NH₂). b 500 µM p-phenylenediamine (p-PDA) in phosphate buffer (10 mM, pH 5.0), c 500 µM 4-aminophenylacetic acid (4-APA) in phosphate buffer (10 mM, pH 7.5), d 100 µM Rhodamine 6 g (R6G) in phosphate buffer (10 mM, pH 5.0), and e 500 µM folic acid (FA) in phosphate buffer (10 mM, pH 7.5). Solid lines and grey shaded areas represent mean and ± 1 s.d. of nine obtained spectra (N = 3, n = 3 spectra). f–i Peak analysis on the prominent peaks of the solutions; f p-PDA, g 4-APA, h R6G and i FA. Data represent the peak intensities determined from the nine spectra (N = 3, n = 3 spectra). ****p < 0.0001, ***p < 0.001, **p < 0.01 and *p < 0.05 based on one‐way ANOVA and Tukey’s honest significance test.

Fig. 4. Molecular dynamics simulation of analytes on SAM-functionalized gold surfaces.

a Snapshots of two model analytes, p-phenylenediamine (p-PDA) and 4-aminophenylacetic acid (4-APA), where the analyte–gold surface separation reaches a minimum. Analyte center-of-mass is used to measure the distance to the closest Au surface atom; % populations when analytes proximal (<0.6 nm) to a SAM/Au are calculated from MD generated equilibrium ensemble for each system: non-functionalized (bare), 1-propanethiol (3CH₃), 3-mercapto-1-propanol (3OH), 3-mercaptopropionic acid (3COOH), 3-amino-1-propanethiol (3NH₂), 1-undecanethiol (11CH₃), 11-mercapto-1-undecanol (11OH), 11-mercaptoundecanoic acid (11COOH), 11-amino-1-undecanethiol (11NH₂). Analyte orientation angles of b p-PDA and c 4-APA relative to the Au surface showing the direction of each analyte’s functional groups when proximal (<0.6 nm) to a SAM/Au. Angles < 90° indicate that the NH₂ group of the analyte is pointing towards the SAM/Au surface, whereas angles > 90° specify that the charged groups (NH₃⁺/COO⁻) are facing the SAM/Au. SAM and analyte protonation states are modeled based on the cognate experimental buffer conditions, i.e. p-PDA (pH 5.0) and 4-APA (pH 7.5).

Fig. 5. Artificial-nose-empowered statistical multivariate analysis of model biological systems.

a, b Series of SERS spectra from cell lysates obtained from functionalized Au-nanopillars with various SAM-forming molecules: non-functionalized (bare), 1-propanethiol (3CH₃), 3-mercapto-1-propanol (3OH), 3-mercaptopropionic acid (3COOH), 3-amino-1-propanethiol (3NH₂), 1-undecanethiol (11CH₃), 11-mercapto-1-undecanol (11OH), 11-mercaptoundecanoic acid (11COOH), 11-amino-1-undecanethiol (11NH₂). a Hs578Bst normal fibroblast-like cells, and b Hs578T breast carcinoma cells. Solid lines and grey shaded area represent mean and ± 1 s.d. of six obtained spectra (N = 3, n = 2 spectra). Coloured shaded bands refer to tentative assignments established in the literature^–. c–h Principal component analysis (PCA)–linear discriminant analysis (LDA) for the two cell lysates: c First principal component (PC1) loadings of SERS spectra from each functionalization. Overlapped grey lines indicate the background SERS signatures of the substrates in PBS (10 mM, pH 7.4). d Mean accuracy for discrimination of cancerous cell lysates (Hs578T) from non-cancerous cell lysates (Hs578Bst) using PCA–LDA models, calculated at each dimensionality from all possible cross-combinations of the nine PC1s. e Evaluation of predictive performance of the PCA–LDA models in (d) using leave-one-out cross-validation (LOOCV). CV-classification errors represent the misclassified fraction of the observations for each LOOCV model. Data represent mean ± 1 s.d. of all the models at each dimensionality. f Scatter plot of the first principal component (PC1) versus the second principal component (PC2) from non-functionalized Au-nanopillars. g Representative two-dimensional PCA scatter plots for the two cell lysates using PC1s of two different SAM functionalization. Blue dotted line was derived by LDA as a classification algorithm to separate the two groups. Red and green dotted boundaries represent confidence intervals of the ± 1 s.d. of each group. Inset is the calculated accuracy in cancerous cell lysates (Hs578T) discrimination for each model. h Representative three-dimensional PCA scatter plots with each axis corresponding to the PC1s of three different SAM-functionalized Au-nanopillars. Blue planes depict classification derived from LDA algorithm separating the two groups. Red and green ellipsoids represent ± 1 s.d. of each group.