Digital SERS bioanalysis of single-enzyme biomarkers

Significance

Single-enzyme reactions in digital bioanalysis have been observed using bright fluorogenic substrates encapsulated in microchambers, but their broad emission spectrum has limited molecular selectivity and multiplexing capability. Raman scattering spectroscopy provides molecular fingerprints with sharp and distinct peaks, but low sensitivity has hindered its application to digital bioanalysis. Here, we developed plasmonically active, uniform microchamber arrays, enabling highly sensitive, molecular-selective, and multiplexed detection of single-enzyme reactions using surface-enhanced Raman scattering (SERS) spectroscopy. SERS-based digital bioanalysis enabled precise quantification of acetylcholinesterase in cerebrospinal fluid, demonstrating its clinical potential for type-specific dementia diagnosis based on minute enzyme-level differences.

Abstract

Digital bioanalysis enables highly sensitive detection of biomolecules at the single-molecule level, making it a widely used technique in biomedical research. However, conventional approaches typically rely on fluorescence detection of single-enzyme reactions, which limits molecular selectivity and the ability to analyze multiple targets simultaneously. To address these limitations, we developed a digital bioanalysis platform based on surface-enhanced Raman scattering spectroscopy and microchamber arrays decorated with silver nanoparticles. This platform achieves a million-fold amplification of Raman signals from products generated by single-enzyme reactions, enabling precise digital counting of enzyme biomarkers with high molecular selectivity and multiplexing capability. We applied this platform to detect and distinguish two closely related enzyme biomarkers, acetylcholinesterase (AChE) and butyrylcholinesterase. By leveraging the sharp and distinct Raman spectral signatures of the reaction products, the platform achieved multiplexed biomarker quantification with femtomolar-level sensitivity. As a proof-of-concept, the platform successfully quantified AChE in human cerebrospinal fluid within 8.5 min, highlighting its potential utility in clinical diagnostics, particularly for differentiating types of dementia based on subtle differences in enzyme levels. Hence, this study presents a valuable alternative to fluorescence-based digital bioanalysis by offering enhanced molecular selectivity and multiplexing capability. Its application extends the scope of digital bioanalysis and broadens its capacity to quantify multiple biomarkers in complex biological samples with high precision and efficiency.

The COVID-19 pandemic has reaffirmed the critical role of in vitro diagnostics, particularly liquid biopsies, in modern medicine (1, 2). Rapid and accurate diagnosis leads to the effective selection of drugs and therapies (2); therefore, there is a need to further enhance the versatility of liquid biopsy. An ideal liquid biopsy requires assay sensitivity, throughput, and molecular selectivity to accurately discriminate the target of interest. In general, fluorogenic and colorimetric assays are primarily used for liquid biopsies because of their sensitivity and throughput, allowing the detection of a wide range of biomarkers with pM sensitivity (3). Several biomarkers, particularly protein-based biomarkers, are present in body fluids at concentrations ranging from fM to aM (4), leaving room for exploring promising biomarker candidates with high sensitivity.

Digital bioanalysis has made significant advances in both sensitivity and throughput (5–7) and has garnered considerable attention in recent years. In digital bioanalysis, the target biomolecules are fractionated into microdroplets or microchambers at the single-molecule level and detected by coupling them to the fluorogenic reaction of enzymes. Biomolecule presence is then detected as a binary fluorescence signal (presence as “1” or absence as “0”) based on a defined threshold, leading to the term “digital” bioanalysis. The number of target biomolecules is individually counted based on binarized signals, a process known as digital counting (SI Appendix, Fig. S1), enabling highly quantitative analysis of various biomolecules from nucleic acids to proteins (4, 8–10). While fluorogenic reactions are highly effective in terms of assay sensitivity and throughput, their relatively low molecular selectivity and multiplexing capability makes it challenging to differentiate between slightly different molecular structures of substrates and/or products and to perform multiplex assays. Incorporating multiplexing capability in liquid biopsy expands the range of diagnostic panels, as well as improves diagnostic accuracy for various diseases such as cancer and dementia (11, 12), therefore, developing alternative detection modalities is urgently required to enhance the versatility of digital bioanalysis.

Raman spectroscopy is a powerful tool for probing molecular vibrations by scattered light to identify specific molecules with high molecular selectivity. The distinct peaks in the scattering spectrum effectively reveal subtle differences in molecular structures of substrates and/or products, offering superior molecular selectivity and multiplexing capability compared to other detection methods (13). However, Raman spectroscopy has an inherent low sensitivity and throughput, making it difficult to detect minute signals from target molecules. Surface-enhanced Raman scattering (SERS) spectroscopy enhances sensitivity by exploiting the optical amplification effect of metallic nanostructures, allowing the detection of enzymes with sensitivity at sub-nM to pM levels (14). While the SERS-based approach shows promise, the variability in signal amplifications limits the quantification of trace biomolecules, indicating the need for improvements in sensitivity and throughput.

In this study, we develop a SERS-based platform for digital bioanalysis, termed “digital SERS,” to enable the precise quantification of biomarkers with high sensitivity, molecular selectivity, and multiplexing capability. In digital SERS, a microchamber array was uniformly decorated with silver (Ag) nanoparticles. The target enzyme biomarkers were stochastically encapsulated in each chamber at single-molecule level with substrates, and the Raman signal of products, generated by single-enzyme reaction, was enhanced by Ag nanoparticles with high reproducibility and uniformity. The presence or absence of the target enzyme biomarkers in each chamber was precisely determined based on the binary SERS signal of the reaction products, leading to highly sensitive, molecular-selective, and multiplex digital counting of target enzyme biomarkers. This feature has facilitated the pioneering achievement of digital bioanalysis using a detection method other than a fluorogenic assay. Furthermore, we successfully demonstrated the multiplex digital counting of acetylcholinesterase (AChE) and butyrylcholinesterase (BuChE), two challenging biomarkers to differentiate simultaneously with a fluorogenic assay. In addition, a proof-of-concept study using clinical samples from patients with cognitive disorders demonstrated the platform’s potential as a liquid biopsy tool to support patient stratification.

Results

Fabrication and Evaluation of the SERS Device.

A microchamber device for SERS, termed SERS device, was fabricated to enhance the reproducibility of SERS signals (Fig. 1A). The microchamber array was produced in a fluoropolymer film coated on a glass substrate using photolithography (10) with a diameter of 3.5 µm and a height of 1.7 µm (16 fL). Subsequently, the microchamber array was treated with 3-aminopropyltriethoxysilane (APTES) to facilitate electrostatic adsorption of Ag nanoparticles to the bottom of the chambers (Fig. 1A). Laser scanning reflected light microscopy (SI Appendix, Fig. S2) revealed that the aggregation state of the Ag nanoparticles was finely controlled by adjusting the concentration of trifluoroacetic acid (TFA) to generate numerous SERS hotspots (15) (Fig. 1A and SI Appendix, Fig. S3). Bright spots, originating from Ag nanoparticle aggregates at the chamber bottom, were uniformly distributed along the chamber arrays across a wide field of view measuring 1.68 mm × 1.50 mm, encompassing approximately 45,000 chambers, with a defect rate of only approximately 0.04% (SI Appendix, Fig. S4). The Ag nanoparticles within the chamber were examined using scanning electron microscopy (Fig. 1B and SI Appendix, Fig. S5). Approximately 1,000 Ag nanoparticles with a diameter of 40 nm were selectively adsorbed on the chamber bottom with particle aggregations, indicating the presence of numerous SERS hotspots at the nanoscale gaps and junctions between the nanoparticles (16) in each chamber. By optimizing the fabrication process, the SERS device was successfully produced with a yield of 97.9% (990/1,011, SI Appendix, Fig. S6).

Fig. 1.

Fabrication and evaluation of SERS device. (A) Schematic of the SERS device, where Ag nanoparticle aggregates were deposited onto the bottom of the microchamber through APTES and TFA treatments. (B) Electron microscopic images of the SERS device. (Scale bars: 200 nm.) (C) Schematic illustration of sealing choline compounds into the SERS device using oil. (D) Raman imaging of the SERS device with a line-shaped laser. (E) SERS image at 764 cm⁻¹ of a 100 µM thiocholine solution (Scale bar: 20 µm), with (F) the corresponding intensity histogram. (G) SERS spectral image extracted from each chamber in Fig. 1E with (H) a representative SERS spectrum. (I) SERS images and (J) spectra of thiocholine solution at varying concentrations, ranging from 0 to 300 µM. SERS spectrum at 0 M was subtracted from that of thiocholine at 0.3 µM to 300 µM, to remove the background noise. (K) Signal intensity against concentrations of choline compounds, comparing SERS and Raman analyses with and without Ag nanoparticles. Peak intensities of thiocholine (764 cm⁻¹), acetylthiocholine (1,690 cm⁻¹), and choline and acetylcholine (716 cm⁻¹) were used for SERS, while the intensity at 2,988 cm⁻¹ was used for Raman analysis. The red, purple, brown, and light blue solid lines in the circle plot represent the fitting curves for SERS with thiocholine, acetylthiocholine, choline, and acetylcholine (power-law, scaling exponent of 0.4, 0.3, 0.2, and 0.3), respectively, while the same colors of solid lines in the rectangle plot represent linear fittings for Raman analysis. Corresponding color or black dotted lines represent the background mean + 3 SD for SERS or Raman analysis, respectively. The LoDs were defined as the concentrations where the fitting curves intersect with the corresponding dotted lines.

The SERS device was evaluated using various choline compounds (acetylcholine/choline and acetylthiocholine/thiocholine), which are substrates or reaction products of AChE—a target biomarker (Fig. 1C). Solutions with specified amounts of the choline compounds were introduced into the device. Subsequently, oil was used to seal the chambers, and the samples were examined using slit-scanning Raman microscopy with line-shaped laser illumination (Fig. 1D) (17). SERS signals associated with the stretching vibrations of the choline head or carbonyl groups were observed across the device (SI Appendix, Fig. S7), displaying high uniformity in both peak intensity and spectral shape (SI Appendix, Fig. S8); for instance, the coefficient of variation of the peak intensity was 7.7% for thiocholine at 764 cm⁻¹ (Fig. 1 E−H) without blinking (SI Appendix, Fig. S9 and Movie S1), by virtue of averaging SERS signal from numerous hot spots in each chamber. Power-law relationships between signal intensity and choline compound concentration were identified (Fig. 1 I−K and SI Appendix, Fig. S10), with the limit of detection (LoD) for thiocholine being 130 nM at 764 cm⁻¹, representing a million-fold enhancement compared to the absence of Ag nanoparticle decoration (226 mM) (Fig. 1K and SI Appendix, Fig. S11). This enhancement is similar to that of the fluorogenic products at the sub-µM level, demonstrating the remarkable sensitivity of the SERS device (Fig. 1K and SI Appendix, Fig. S10). The LoDs for other compounds were 6, 8, and 28 mM for acetylcholine (716 cm⁻¹), acetylthiocholine (1,690 cm⁻¹), and choline (716 cm⁻¹), respectively (Fig. 1K and SI Appendix, Fig. S10). The thiol moiety in thiocholine facilitated adsorption on the Ag surface to produce a strong Raman signal primarily due to chemical enhancement (16), resulting in a lower LoD. Therefore, the disparity in LoDs between the substrates and reaction products was amplified by over 10,000-fold through the introduction of a thiol group at the enzymatic reaction site for AChE. This enhancement highlights the potential of this approach for highly sensitive, molecular-selective, and multiplex detection of target biomolecules in the field of digital bioanalysis.

Digital Counting of Biomarkers Using SERS.

Biomarkers were digitally counted using a SERS device. AChE, an enzyme that hydrolyzes choline esters, is a recognized biomarker in blood tests and has recently gained attention owing to its role in neuronal signal transduction, making it a potential indicator for diagnosing neurodegenerative diseases (11). Acetylthiocholine, commonly used in colorimetric assays for AChE, was used as the substrate (18). Upon interaction with AChE, acetylthiocholine was enzymatically cleaved into thiocholine and acetate, with thiocholine adsorbing on the Ag surface, generating a signal at 764 cm⁻¹ that can be detected by slit-scanning Raman microscopy (Fig. 2A). The AChE/substrate solution was introduced to the SERS device, followed by the application of oil to seal the chambers where AChE molecules were stochastically fractionated (Fig. 2B). While the overall spectral shape remained uniform throughout the device, specific chambers exhibited a prominent peak at 764 cm⁻¹ (Fig. 2 E−G), as shown in the SERS image at 764 cm⁻¹, demonstrating the stochastic appearance of bright chambers (Fig. 2 C and D). By digitally counting the bright chambers with intensities exceeding 4.5 times the background noise at 764 cm⁻¹ (Fig. 2H), the count correlated with the AChE concentration (Fig. 2 I and J and SI Appendix, Fig. S12), aligning well with the theoretical values based on the Poisson distribution (SI Appendix, Fig. S13). In addition, time-lapse imaging indicated a thiocholine production rate of 3,580 s⁻¹ (Fig. 2K, SI Appendix, Fig. S14, and Movie S2), closely resembling the turnover rate of previously reported AChE activity (19). Collectively, each bright chamber corresponded to a single AChE molecule, demonstrating the feasibility of digital counting of AChEs. The LoD of AChE in digital SERS was 13 fM, marking a 77-fold enhancement over the bulk SERS assay (1.0 pM) and a 146-fold improvement over the colorimetric assay (1.9 pM) (Fig. 2J), supporting the remarkable sensitivity of digital bioanalysis.

Fig. 2.

Digital SERS of AChE. (A) Schematic of AChE hydrolysis of acetylthiocholine to thiocholine and acetate, followed by SERS detection of thiocholine adsorbed on the surface of Ag nanoparticles. (B) Schematic of digital counting of AChE. (C) SERS image at 764 cm⁻¹ of the AChE/acetylthiocholine mixture (10 pM AChE and 2 mM acetylthiocholine). (D) Magnified view (Scale bar: 50 µm) of area i) from (C) with SERS spectra of the chambers indicated by red and brown circles (E). (F) SERS spectral image extracted from chambers in area ii) from (C), with (G) a magnified view of the yellow-dotted square and (H) an intensity histogram. (I) SERS images at AChE concentrations varying from 0 to 100 pM. (J) Comparison of digital SERS (red), bulk SERS (green), and colorimetric analysis (black) for AChE detection. The red, green, and black solid lines represent the fitting curves for digital SERS (linear), bulk SERS (power-law, scaling exponent of 0.4), and colorimetric analysis (linear), respectively. The corresponding dotted lines represent the background mean + 3 SD. The LoDs were defined as the concentrations where the fitting curves intersect with the corresponding dotted lines. (K) Time-lapse SERS images. Red and blue colors represent peak intensity at 764 cm⁻¹ (thiocholine) and averaged peak intensity at 2,783 to 3,038 cm⁻¹ (CH₃ stretching in choline compounds), respectively, with a time course analysis of thiocholine concentration calculated from the peak intensity at 764 cm⁻¹ of four chambers indicated by (i–iv) in the SERS image.

Owing to its inherently high molecular selectivity, digital SERS has the potential to identify various biomarkers using different substrates and facilitate multiplex detection. For example, butyrylthiocholine—a specific substrate of BuChE—can be used for the digital counting of BuChE (Fig. 3 A and B), which is a well-known biomarker that is difficult to distinguish from AChE in fluorogenic and colorimetric reactions due to their similar enzymatic characteristics (20). The LoD of BuChE in the digital counting was 7.9 fM (Fig. 3C), comparable to that of AChE. Despite successful digital counting of BuChE, the reaction of butyrylthiocholine with BuChE produced thiocholine (Fig. 3A). This is the same reaction product formed when acetylthiocholine reacts with AChE (Fig. 2A), making it difficult to distinguish between AChE and BuChE using the current substrate sets. To address this issue, the digital counting of AChE was performed using MATP+ (Fig. 3A), a specific substrate of AChE (21) with a slightly different structure at the choline head group compared to acetylthiocholine. The Raman peak of the reaction product of MATP+ was observed at 706 cm⁻¹ (Fig. 3B), distinctly different from that of thiocholine at 764 cm⁻¹, owing to the inherent high molecular selectivity of SERS. The LoD of AChE using MATP+ was 22 fM (Fig. 3C), similar to that obtained using acetylthiocholine. Leveraging the high molecular selectivity of digital SERS, multiplex digital counting of AChE and BuChE was performed using a mixture of MATP+ and butyrylthiocholine as substrates. The SERS image, superimposed at 706 and 764 cm⁻¹, demonstrated the simultaneous single-molecule detection of AChE and BuChE (Fig. 3 D−H). The two-dimensional (2D) intensity plots at 706 and 764 cm⁻¹ extracted from individual chambers exhibited high orthogonality (Fig. 3I), confirming the feasibility of multiplex digital SERS for AChE and BuChE.

Fig. 3.

Multiplex digital SERS. (A) Schematics of BuChE hydrolyzing butyrylthiocholine and AChE hydrolyzing MATP+, followed by SERS imaging of reaction products adsorbed onto Ag nanoparticle surfaces. (B) SERS images at 764 cm⁻¹ of the BuChE/butyrylthiocholine mixture (10 pM BuChE with 2 mM butyrylthiocholine), and at 706 cm⁻¹ of the AChE/MATP+ mixture (10 pM AChE with 2 mM MATP+). (Scale bars: 50 µm.) (C) Plot of BuChE or AChE concentrations versus the number of bright chambers in the SERS images of BuChE/butyrylthiocholine mixture (red) or AChE/MATP+ mixture (green). Solid and dotted lines represent linear regressions and background mean + 3 SD, respectively. The LoDs were defined as the concentrations where the fitting curves intersect with the corresponding dotted lines. (D) Dual-color SERS image at 706 cm⁻¹ (green) and 764 cm⁻¹ (red) of a mixture containing 50 pM AChE, 10 pM BuChE, 1.5 mM MATP+, and 4 mM butyrylthiocholine (Scale bar: 50 µm), with SERS spectra of the chambers indicated by green, red, yellow, and brown circles (E). (F–H) SERS images extracted from each chamber in Fig. 3D. (I) 2D plots of the peak intensities at 706 and 764 cm⁻¹ of the SERS images with 1.5 mM MATP+, 4 mM butyrylthiocholine, and with or without 50 pM AChE and 10 pM BuChE.

High-Speed Digital SERS.

The proof-of-principle experiments for digital SERS described in the previous section used slit-scanning Raman microscopy. While this method provides detailed spectral information, it requires 4.5 h to capture a complete image of a SERS device measuring several square millimeters. To expedite image acquisition for diagnostic applications, a high-speed Raman imaging system was developed using wide-field illumination and a wavelength-tunable narrow-linewidth bandpass filter positioned within the imaging optics (Fig. 4A). The SERS signal emanating from the reaction product area under wide-field illumination was selectively collected through an angular-dependent bandpass filter with a transmission bandwidth of less than 2 nm (<65 cm⁻¹) (Fig. 4B). Two sets of SERS images were acquired with the filter angle adjusted to the peak top (signal) and bottom (background), which were then subtracted to eliminate background noise to improve LoD (SI Appendix, Fig. S15). A representative wide-field image of the SERS device, containing a mixture of AChE and MATP+, is shown in Fig. 4C. The acquisition time for this image was 8.4 min, achieved by tiling 50 area images, covering a field of view measuring 1.50 mm × 1.25 mm and comprising approximately 34,000 chambers. In the magnified image (Fig. 4D), distinct bright chambers indicating the presence of a single AChE molecule were clearly visible when the filter angle was set at approximately 23.3°, aligning with the transmission band of the filter covering the peak of the MATP+ product at 706 cm⁻¹ (552.76 nm) (Fig. 4E). The LoD for AChE was determined to be 32 fM (Fig. 4 F and G), a value nearly equivalent to that achieved with slit-scanning Raman microscopy but with an acquisition time 32 times faster. By adjusting the filter angle to approximately 19.8°, the system also successfully detected AChE with acetylthiocholine at an LoD of 20 fM, aligning with the transmission band of the filter covering the peak of the acetylthiocholine product at 764 cm⁻¹ (554.52 nm) (Fig. 4G). Leveraging the high-throughput capabilities of the system, high-speed multiplex digital SERS of AChE and BuChE were conducted using a mixture of MATP+ and butyrylthiocholine as substrates (Fig. 4 H and I).

Fig. 4.

High-speed digital SERS. (A) Optical system for wide-field SERS imaging. (B) Schematic of the transmission wavelength adjusted by the angle of narrow bandpass filter, along with representative SERS images of the AChE/MATP+ mixture (100 pM AChE with 2 mM MATP+). (C) Wide-field SERS image of the AChE/MATP+ mixture (10 pM AChE with 2 mM MATP+), with (D) a magnified image (Scale bar: 30 µm) of area (i). (E) Transmission bands of the filter at three different angles, along with a representative SERS spectrum of the product of the AChE/MATP+ mixture. (F) SERS images of the AChE/MATP+ mixture at different AChE concentrations (0 to 100 pM). (G) Plot of the number of bright chambers (count) at different AChE concentrations. Green and red represent the AChE/MATP+ (23.3°) and AChE/acetylthiocholine mixtures (19.8°) with a wide-field imaging system, while blue and orange represent the AChE/MATP+ (706 cm⁻¹) and AChE/acetylthiocholine (764 cm⁻¹) mixtures with a slit-scanning imaging system, respectively. Solid lines represent fitting curves and dotted lines represent the background mean + 3 SD. The LoDs were defined as the concentrations where the fitting curves intersect with the corresponding dotted lines. (H) Dual-color SERS image of a mixture containing 50 pM AChE, 5 pM BuChE, 1.5 mM MATP+, and 4 mM butyrylthiocholine. (I) Transmission bands of the filter at five different angles, with representative SERS spectra of the products from the AChE/MATP+ (green) and BuChE/butyrylthiocholine (red) mixtures. Filter angles of 23.3° and 26.0° for AChE/MATP+ (green and light green) and 19.8° and 16.9° for BuChE/butyrylthiocholine (red and light red) were used to obtain the SERS image in (H).

Clinical Validation Using Liquid Specimens.

Trace levels of AChE in human cerebrospinal fluid (CSF) were monitored using digital SERS to investigate their potential as biomarkers for diagnosing dementia. Dementia, affecting over 40 million people globally (22), is caused by several cognitive disorders, such as Alzheimer’s disease (AD), and cerebrovascular disease. Because the pathogenesis differs depending on the causes, accurate diagnosis, particularly in the prodromal phase [mild cognitive impairment (MCI)], is crucial for appropriate therapy and drug treatment. For AD, which constitutes the majority of patients with cognitive disorders, precise early diagnosis methods have been established using CSF biomarkers, such as amyloid-β and tau proteins (23). Conversely, effective biomarkers for diagnosing vascular cognitive disorders (VCD), such as vascular dementia are yet to be established in clinical practice (24), despite VCDs representing 20% of dementia cases. Cholinergic deficiency in the brain and CSF has been reported in patients with VCD, potentially affecting AChE levels in CSF due to cholinergic neuron loss (25). Therefore, we evaluated the potential of AChE in CSF as a biomarker for VCD using samples from 15 patients with AD (26), 11 patients with VCD (27), and 10 individuals without AD and VCD as controls (Fig. 5A and SI Appendix, Table S1). Fig. 5B shows a representative image of the SERS device with a mixture of CSF from patients with cognitive disorders and MATP+ as a substrate obtained using the aforementioned high-speed imaging system. Similar to the purified sample, distinct chambers originating from individual AChE molecules were observed, enabling the digital quantification of AChE in CSF. Comparing AChE levels among the VCD, AD, and control groups revealed significantly lower levels of AChE in patients with VCD compared to patients with AD and controls (Fig. 5C). The area under the curve of ~0.8 for VCD diagnosis based on CSF AChE levels (Fig. 5D) highlights its potential as a diagnostic biomarker for VCD.

Fig. 5.

Digital SERS of AChE in CSF of patients with cognitive disorders. (A) Schematic of the sample preparation of CSF for digital SERS of AChE. (B) Representative wide-field SERS images of CSF from patients with VCD, AD, or individuals without VCD and AD as controls (patients with MCI not meeting the diagnostic criteria for VCD or AD). (Scale bar: 50 µm.) (C) CSF AChE levels of patients with VCD, AD, or control, analyzed by digital SERS. *P < 0.05, **P < 0.01. (D) Receiver operating characteristic curves for differentiating between VCD and AD, VCD and control, as well as AD and control groups.

Discussion

Digital SERS has enabled the quantification of biomolecules at the single-molecule level with high sensitivity, molecular selectivity, multiplexing capability, and throughput, owing to the development of a SERS device and a high-speed Raman imaging system. The SERS device significantly enhanced the reproducibility of the SERS signals, while the high-speed Raman imaging system reduced imaging time and collectively enhanced the versatility of digital bioanalysis. The molecular selectivity of digital SERS for minute structural differences has driven various technological advances, from identifying biomarkers challenging to distinguish using conventional fluorogenic assays to multiplexing detection targets. The use of substrates with Raman tags, which has made remarkable progress in recent years, increases the likelihood of discriminating biomarkers catalyzing similar reactions at the isozyme level and expanding the number of targets to 10 or more (28–30). Furthermore, the range of target biomolecules can be broadened by integrating other systems, such as enzyme-linked immunosorbent assay, PCR, and CRISPR-Cas, into digital bioanalysis (4, 10, 31–33). These efforts will pave the way for the ultimate multiplex digital panel diagnostic technology capable of analyzing numerous biomarkers across various biomolecule types with single-molecule resolution and molecular fingerprinting.

Further advancements in SERS devices and high-speed Raman imaging systems pose significant challenges for future research. The volume of microchamber in digital SERS represents a trade-off relationship between assay sensitivity and throughput, as in conventional digital bioanalysis based on fluorogenic reactions (5). To achieve fM sensitivity with an assay time below 1 h, the chamber volume of a few tens of fL, as used in this study, is reasonable. Even with this balanced chamber volume, assay sensitivity can further be improved by increasing the number of microchambers analyzed (SI Appendix, Fig. S16). Developing microchambers in the SERS device with a higher aspect ratio is an important direction to increase chamber density while keeping chamber volume, thereby enhancing assay sensitivity without compromising throughput. The three-dimensional decoration of Ag nanoparticles, which can increase the number of SERS hotspots, will further enhance signal amplification and sensitivity. Another potential strategy for the development of SERS devices involves chemically modifying Ag nanoparticles to function as sensors for analyzing various ambient conditions in the chamber, such as temperature, pH, ion concentration, and redox potential (34). This modification enables a more detailed analysis of biomarkers. For high-speed Raman imaging systems, a promising approach is the development of a new system that can expand the number of detection targets without prolonging the imaging time. By incorporating spectral information in the vacant spaces between chambers in a wide-field image, wide-field multiplex Raman imaging can be achieved. It is highly probable that this innovative technique maintains imaging speed while simultaneously capturing scattering spectra from individual chambers, thereby improving the multiplexing capabilities for detection targets.

The design of the substrates for digital SERS is another important aspect for future research. The ability to discriminate between substrates and products of enzymatic reaction is essential in digital SERS. Therefore, the reaction products must exhibit large differences in signal strength compared to the substrates. Current digital SERS employed affinity control of the products with a silver surface to generate significant signal differences. The thiol moiety in the products, generated upon cleavage of the thioester bond in the substrates, efficiently adsorbed onto the silver surface, producing a strong SERS signal via chemical enhancement (16). To expand the range of measurable enzyme species in digital SERS, it is important to explore other affinity-tunable substrates with different chemical structures/reactions, as well as other switching mechanisms, such as control of electronic resonance (29).

For the widespread adoption of digital SERS in liquid biopsy, it is crucial to consider the size and cost of the imaging system, along with the expenses related to consumables, including the SERS device. The current high-speed Raman imaging system was set up on a 120 cm × 70 cm optical table featuring a high-power laser source and an expensive optical microscope. While core hospitals and clinical testing centers can manage this configuration, further downsizing and cost reduction are necessary to align with contemporary medical requirements, such as point-of-care testing. The cost of the SERS device amounts to approximately $2 per device (SI Appendix, Table S2). The bottleneck in the current fabrication of SERS devices lies in the decoration of Ag nanoparticles within the microchamber array. Enhancing this process could facilitate mass production and further cost reduction. Despite a few remaining issues, further optimizing the imaging system, SERS device, and assay protocol will pave the way for the practical application of digital SERS. This includes super-multiplex liquid biopsy for the early detection of various diseases, such as cancer and dementia.

Materials and Methods

Fabrication of the SERS Device.

Microchamber arrays were fabricated on thin glass substrates using conventional photolithography. A 32 mm × 24 mm cover glass (No. 1, Matsunami Glass) was sonicated for 1 h in an 8 M KOH solution, incubated overnight, rinsed with distilled water, and dried using an air blow gun. Fluoropolymer CYTOP (9% CYTOP, AGC) was spin-coated on the glass substrate at 1,000 rpm for 30 s. Subsequently, it was baked at 80 °C for 10 min and 180 °C for 1 h. A positive photoresist (AZ P4620, AZ Electronic Materials) was spin-coated on the CYTOP layer at 7,500 rpm for 30 s and cured at 100 °C for 5 min. After allowing the photoresist to rehydrate at 25 °C for over 5 min under 60% humidity, the glass substrate was exposed to ultraviolet light using a mask aligner (LA610dt, Nanometric Technology) and a chrome photomask with 1.5 μm holes. Subsequently, it was incubated for 1.5 min in a developer (AZ300 MIF, AZ Electronic Materials). The uncovered area of the photoresist was eliminated through dry etching with O₂ plasma (RIE-10NR, Samco) under the following conditions: O₂ flow rate of 50 sccm, gas pressure of 10 Pa, RF power of 50 W, and etching time of 16 min. The microchamber array was produced by eliminating the remaining photoresist through sequential rinsing with acetone, 2-propanol, and distilled water. The diameter and height of the chambers were measured using a laser microscope (VK-X1100, Keyence).

Ag nanoparticles were attached to the bottom of the microchamber array. The array was sonicated for 20 min in a 4 M KOH solution, incubated for 20 min, rinsed with pure water, and dried using an air blow gun. Next, the array was incubated in a 0.2% APTES (02309-62, Tokyo Chemical Industry) solution in acetone for 60 min, rinsed sequentially with acetone, ethanol, and distilled water, and dried with compressed air. Subsequently, equal amounts of Ag nanoparticles with a specified diameter of 40 nm (730807, Sigma-Aldrich) and 0.3% TFA were mixed, dropped onto the array, and placed on an ice block for 40 s. The array, along with the droplet, was then left to incubate for over 2 d and stored in a refrigerator at 4 °C until further use. Finally, the glass bottom of the microchamber array was decorated with Ag nanoparticles.

Raman Imaging with Wide-Field Raman Microscopy.

Raman imaging was performed using a custom-built wide-field Raman microscope equipped with a 532 nm laser (Finesse, Laser Quantum). The laser beam was expanded 83.3 times using a beam expander, shaped into a rectangle using a mask, and then focused onto the sample with a water-immersion objective lens (LWD Lambda S 20XC WI, Nikon), resulting in an illumination area of 125 µm × 300 µm. The scattered light was collected by the same objective lens and chromatically separated by a dichroic beam splitter (LPD02-532RU-25 × 36 × 2.0, Semrock) and an edge filter (LP03-532RU-25, Semrock), which isolated Raman scattering light. A narrow band-pass filter (LL02-561-25, Semrock) mounted on a motorized rotation stage (OSCM-25YAW, Sigmakoki) transmitted specific wavelength regions, and the transmitted light was detected using an sCMOS camera (OrcaFusion BT, Hamamatsu Photonics). The samples were scanned two-dimensionally using a motorized stage on an inverted optical microscope (Ti2-E, Nikon). Focus drift was minimized during imaging with a real-time feedback system (PFS, Nikon). The laser intensity at the sample was 1 µW/µm², and the exposure time was 3 s/area/filter angle. Stage scanning, camera exposure, and rotation of the motorized stage for the narrow bandpass filter were controlled using microscope software (NIS-elements, Nikon).

Digital SERS Assay.

The excess solution of Ag nanoparticles and TFA on the SERS device was removed, and Kapton tape with a 7 mm hole was attached to the SERS device. Subsequently, 50 µL of buffer A [20 mM Tris HCl (pH 8.0), 40 mM MgCl₂, 100 mM NaCl, 200 µM Triton X-100, and 100 µM potassium iodide] was dropped on the hole of the Kapton tape to hold a hemispherical droplet and left for 20 min at room temperature. The droplet was then removed from the SERS device and washed with 50 µL of buffer B [20 mM Tris HCl (pH 8.0), 40 mM MgCl₂, 100 mM NaCl, and 100 µM Triton X-100] three times. The droplet of buffer B was then replaced with 50 µL of the assay solution. Subsequently, approximately 40 µL of the assay solution was removed to form a flat liquid surface in the hole, and 56 µL of mineral oil (M8410, Sigma-Aldrich) was added to seal the chambers. After 45 min (AChE/acetylthiocholine and AChE/MATP+), 75 min (BuChE/butyrylthiocholine), or 3.5 h (AChE, BuChE/MATP+, Butyrylthiocholine, or CSF/MATP+) of incubation at room temperature, SERS images were acquired using slit-scanning Raman microscopy or wide-field Raman microscopy at laser intensity of 0.24 mW/µm² or lower, so as not to affect the spectral analysis of the choline head group induced by photodamage (SI Appendix, Fig. S17).

Details of the experimental procedures for Evaluation of the SERS device, Raman imaging with slit-scanning Raman microscopy, Raman/SERS evaluation of choline compounds, Sample preparation for purified enzymes and clinical CSF, Data analysis for digital SERS, Bulk assay, and Time-lapse and photodamage investigation of thiocholine in SERS device are described in SI Appendix, SI Materials and Methods.

Supplementary Material

Appendix 01 (PDF)

Movie S1.

Time-lapse observation of thiocholine in SERS device. SERS spectroscopic images were captured using a slit-scanning Raman microscopy. A line-shaped laser was placed at the centers of ten microchambers containing 100 μM thiocholine. With an exposure time of 45 ms (frame rate of 20 frames/s), 200 spectroscopic images were continuously recorded for 10 s. The laser intensity at the sample was 0.036 mW/μm² without stage scanning.

Movie S2.

Time-lapse observation of single AChE activity in SERS device sealed with 10 pM AChE and 2 mM acetylthiocholine. Red and blue represent the peak intensity at 764 cm^-1 (thiocholine) and the average peak intensity at 2783–3038 cm^-1 (CH₃ stretching in choline compounds), respectively. SERS image was captured using a slit-scanning Raman microscopy with laser intensity at the sample and an exposure time of 0.036 mW/μm² and 0.2 s/line. The images were processed with a median filter to eliminate outliers using the ImageJ Remove Outliers command (radius=3, threshold=50). The observations began 2 min after sealing. The time interval between images was 8 min.

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information.