Stratification of tumour cell radiation response and metabolic signatures visualization with Raman spectroscopy and explainable convolutional neural network.

Reprogramming of cellular metabolism is a driving factor of tumour progression and radiation therapy resistance. Identifying biochemical signatures associated with tumour radioresistance may assist with the development of targeted treatment strategies to improve clinical outcomes. Raman spectroscopy (RS) can monitor post-irradiation biomolecular changes and signatures of radiation response in tumour cells in a label-free manner. Convolutional Neural Networks (CNN) perform feature extraction directly from data in an end-to-end learning manner, with high classification performance. Furthermore, recently developed CNN explainability techniques help visualize the critical discriminative features captured by the model. In this work, a CNN is developed to characterize tumour response to radiotherapy based on its degree of radioresistance. The model was trained to classify Raman spectra of three human tumour cell lines as radiosensitive (LNCaP) or radioresistant (MCF7, H460) over a range of treatment doses and data collection time points. Additionally, a method based on Gradient-Weighted Class Activation Mapping (Grad-CAM) was used to determine response-specific salient Raman peaks influencing the CNN predictions. The CNN effectively classified the cell spectra, with accuracy, sensitivity, specificity, and F1 score exceeding 99.8%. Grad-CAM heatmaps of H460 and MCF7 cell spectra (radioresistant) exhibited high contributions from Raman bands tentatively assigned to glycogen, amino acids, and nucleic acids. Conversely, heatmaps of LNCaP cells (radiosensitive) revealed activations at lipid and phospholipid bands. Finally, Grad-CAM variable importance scores were derived for glycogen, asparagine, and phosphatidylcholine, and we show that their trends over cell line, dose, and acquisition time agreed with previously established models. Thus, the CNN can accurately detect biomolecular differences in the Raman spectra of tumour cells of varying radiosensitivity without requiring manual feature extraction. Finally, Grad-CAM may help identify metabolic signatures associated with the observed categories, offering the potential for automated clinical tumour radiation response characterization.

Metabolic profiling of murine radiation-induced lung injury with Raman spectroscopy and comparative machine learning.

Radiation-induced lung injury (RILI) is a dose-limiting toxicity for cancer patients receiving thoracic radiotherapy. As such, it is important to characterize metabolic associations with the early and late stages of RILI, namely pneumonitis and pulmonary fibrosis. Recently, Raman spectroscopy has shown utility for the differentiation of pneumonitic and fibrotic tissue states in a mouse model; however, the specific metabolite-disease associations remain relatively unexplored from a Raman perspective. This work harnesses Raman spectroscopy and supervised machine learning to investigate metabolic associations with radiation pneumonitis and pulmonary fibrosis in a mouse model. To this end, Raman spectra were collected from lung tissues of irradiated/non-irradiated C3H/HeJ and C57BL/6J mice and labelled as normal, pneumonitis, or fibrosis, based on histological assessment. Spectra were decomposed into metabolic scores via group and basis restricted non-negative matrix factorization, classified with random forest (GBR-NMF-RF), and metabolites predictive of RILI were identified. To provide comparative context, spectra were decomposed and classified via principal component analysis with random forest (PCA-RF), and full spectra were classified with a convolutional neural network (CNN), as well as logistic regression (LR). Through leave-one-mouse-out cross-validation, we observed that GBR-NMF-RF was comparable to other methods by measure of accuracy and log-loss (p > 0.10 by Mann-Whitney U test), and no methodology was dominant across all classification tasks by measure of area under the receiver operating characteristic curve. Moreover, GBR-NMF-RF results were directly interpretable and identified collagen and specific collagen precursors as top fibrosis predictors, while metabolites with immune and inflammatory functions, such as serine and histidine, were top pneumonitis predictors. Further support for GBR-NMF-RF and the identified metabolite associations with RILI was found as CNN interpretation heatmaps revealed spectral regions consistent with these metabolites.

A Direct Immunoassay Based on Surface-Enhanced Spectroscopy Using AuNP/PS-b-P2VP Nanocomposites.

A biosensor was developed for directly detecting human immunoglobulin G (IgG) and adenosine triphosphate (ATP) based on stable and reproducible gold nanoparticles/polystyrene-b-poly(2-vinylpyridine) (AuNP/PS-b-P2VP) nanocomposites. The substrates were functionalized with carboxylic acid groups for the covalent binding of anti-IgG and anti-ATP and the detection of IgG and ATP (1 to 150 µg/mL). SEM images of the nanocomposite show 17 ± 2 nm AuNP clusters adsorbed over a continuous porous PS-b-P2VP thin film. UV-VIS and SERS were used to characterize each step of the substrate functionalization and the specific interaction between anti-IgG and the targeted IgG analyte. The UV-VIS results show a redshift of the LSPR band as the AuNP surface was functionalized and SERS measurements showed consistent changes in the spectral features. Principal component analysis (PCA) was used to discriminate between samples before and after the affinity tests. Moreover, the designed biosensor proved to be sensitive to different concentrations of IgG with a limit-of-detection (LOD) down to 1 µg/mL. Moreover, the selectivity to IgG was confirmed using standard solutions of IgM as a control. Finally, ATP direct immunoassay (LOD = 1 µg/mL) has demonstrated that this nanocomposite platform can be used to detect different types of biomolecules after proper functionalization.

Exploring Intensity Distributions and Sampling in SERS-Based Immunoassays.

Surface-enhanced Raman scattering (SERS) is a sensitive, widely used spectroscopic technique. However, SERS is perceived as poorly reproducible and insufficiently robust for standard applications in analytical chemistry. Here, we demonstrated that reliable SERS immunoassay quantification at low concentrations (pM range) can be achieved by careful experimental design and appropriate data analysis statistics. A SERS-based immunoassay for IgG in human serum (3.1-50.0 ng mL-1 or 20.6-333 pM) was developed as a proof of concept. Calibration curves were created using the population median of the band area, centered at 592 cm-1, of a SERS reporter (Nile Blue A). Histograms of 7200 SERS spectra show lognormal distributions. SEM images of the sensor platform confirm a correlation between the number of SERS probes (ERLs) at the surface and the SERS intensity response. The IgG immunosensor reported here presented a limit of detection of 1.11 ng mL-1 or 7.39 pM and a limit of quantification of 9.04 ng mL-1 or 60.30 pM, within a 95% confidence level. The % error of the predicted versus the actual response of a quality control (QC) sample was 0.13%. The percent error of the QC sample decreases exponentially with the number of measurements. Randomly selected spatially separated measurements provided lower QC % error than a larger number of measurements that were closely spaced. We propose that it is necessary to describe the measured populations using an appropriate sample size for good statistics and consider the interrogation of sufficiently large and well-separated areas of the sensor surface to achieve a reliable sampling.

Digital plasmonic holography with iterative phase retrieval for sensing.

Propagating surface plasmon waves have been used for many applications including imaging and sensing. However, direct in-plane imaging of micro-objects with surface plasmon waves suffers from the lack of simple, two-dimensional lenses, mirrors, and other optical elements. In this paper, we apply lensless digital holographic techniques and leakage radiation microscopy to achieve in-plane surface imaging with propagating surface plasmon waves. As plasmons propagate in two-dimensions and scatter from various objects, a hologram is formed over the surface. Iterative phase retrieval techniques applied to this hologram remove twin image interference for high-resolution in-plane imaging and enable further applications in real-time plasmonic phase sensing.

Intensity Fluctuations in Single-Molecule Surface-Enhanced Raman Scattering.

Around 20 years ago, the first reports of single-molecule surface-enhanced Raman scattering (SM-SERS) caused a revolution in nanotechnology. Several researchers were quick to recognize the importance of a technique that can provide molecular vibrational fingerprinting at the SM level. Since then, a large amount of work has been devoted to the development of nanostructures capable of SM-SERS detection. A great effort has also been geared toward elucidating the different mechanisms that contribute to the effect. The understanding of the concept of plasmonic SERS hotspots, the role of chemical effects, and the dynamics of atomic and cluster rearrangements in nanometric domains has significantly advanced, driven by new computational and experimental methods used to study SM-SERS. In particular, SERS intensity fluctuations (SIFs) are now recognized as a hallmark of SM-SERS. Interpretation of SM-SERS data must take into consideration temporal and spatial variations as a natural consequence of the extreme localization inherent to surface plasmon resonances. Further analysis of variations in spectral signature, due to either molecular reorientation or photo (or thermal) processes, pointed to a new area that combines the power of SERS fingerprinting at the SM level to modern concepts of catalysis, such as hot-electrons-driven chemistry. This large body of work on the fundamental characteristics of the SM-SERS effect paved the way to the interpretation of other related phenomena, such as tip-enhanced Raman scattering (TERS). Despite all the fundamental progress, there are still very few examples of real applications of SM-SERS. In recent years, our research group has been studying SIFs, focused on different ways to use SM-SERS. The obvious application of SM-SERS is in analytical chemistry, particularly for quantification at ultralow concentrations (below 1 nM). However, quantification using SM-SERS faces a fundamental sampling problem: the analytes (adsorbed in very small amounts, i.e., low surface coverage) must find rare SERS hotspots (areas with intense electric field localization that yields SERS). This limitation leads to strong temporal and spatial variations in SERS intensities, which translates into very large error bars in an experimental calibration curve. We tackled this problem by introducing the concept of "digital SERS". This approach provided a roadmap for SERS quantification at ultralow concentrations and a potential pathway for a better understanding of the "reproducibility problem" associated with SERS. In this Account, we discuss not only the analytical applications but also other implementations of SM-SERS demonstrated by our group. These include the use of SM-SERS as a tool to probe colloidal aggregation, to evaluate the efficiency of SERS substrates, and to characterize the energy of localized resonances. SERS involves a series of random processes: hotspots are rare; surfaces/clusters constantly reconstruct; and molecules diffuse, adsorb, and desorb. All these pathways contribute to strong fluctuations in SERS intensities. Our work indicates that a statistical view of the effect can lead to interesting insights and the potential to fulfill the promise of this SM technique for real-world applications.

Raman spectroscopy detects metabolic signatures of radiation response and hypoxic fluctuations in non-small cell lung cancer.

BACKGROUND: Radiation therapy is a standard form of treating non-small cell lung cancer, however, local recurrence is a major issue with this type of treatment. A better understanding of the metabolic response to radiation therapy may provide insight into improved approaches for local tumour control. Cyclic hypoxia is a well-established determinant that influences radiation response, though its impact on other metabolic pathways that control radiosensitivity remains unclear. METHODS: We used an established Raman spectroscopic (RS) technique in combination with immunofluorescence staining to measure radiation-induced metabolic responses in human non-small cell lung cancer (NSCLC) tumour xenografts. Tumours were established in NOD.CB17-Prkdcscid/J mice, and were exposed to radiation doses of 15 Gy or left untreated. Tumours were harvested at 2 h, 1, 3 and 10 days post irradiation. RESULTS: We report that xenografted NSCLC tumours demonstrate rapid and stable metabolic changes, following exposure to 15 Gy radiation doses, which can be measured by RS and are dictated by the extent of local tissue oxygenation. In particular, fluctuations in tissue glycogen content were observed as early as 2 h and as late as 10 days post irradiation. Metabolically, this signature was correlated to the extent of tumour regression. Immunofluorescence staining for γ-H2AX, pimonidazole and carbonic anhydrase IX (CAIX) correlated with RS-identified metabolic changes in hypoxia and reoxygenation following radiation exposure. CONCLUSION: Our results indicate that RS can identify sequential changes in hypoxia and tumour reoxygenation in NSCLC, that play crucial roles in radiosensitivity.

Digital Protocol for Chemical Analysis at Ultralow Concentrations by Surface-Enhanced Raman Scattering.

Single molecule surface-enhanced Raman spectroscopy (SM-SERS) has the potential to revolutionize quantitative analysis at ultralow concentrations (less than 1 nM). However, there are no established protocols to generalize the application of this technique in analytical chemistry. Here, a protocol for quantification at ultralow concentrations using SM-SERS is proposed. The approach aims to take advantage of the stochastic nature of the single-molecule regime to achieved lower limits of quantification (LOQ). Two emerging contaminants commonly found in aquatic environments, enrofloxacin (ENRO) and ciprofloxacin (CIPRO), were chosen as nonresonant molecular probes. The methodology involves a multivariate resolution curve fitting known as non-negative matrix factorization with alternating least-squares algorithm (NMF-ALS) to solve spectral overlaps. The key element of the quantification is to realize that, under SM-SERS conditions, the Raman intensity generated by a molecule adsorbed on a "hotspot" can be digitalized. Therefore, the number of SERS event counts (rather than SERS intensities) was shown to be proportional to the solution concentration. This allowed the determination of both ENRO and CIPRO with high accuracy and precision even at ultralow concentrations regime. The LOQ for both ENRO and CIPRO were achieved at 2.8 pM. The digital SERS protocol, suggested here, is a roadmap for the implementation of SM-SERS as a routine tool for quantification at ultralow concentrations.

Breast cancer subtype specific biochemical responses to radiation.

External beam radiotherapy is a common form of treatment for breast cancer. Among patients and across different breast cancer subtypes, the response to radiation is heterogeneous. Radiation-induced biochemical changes were examined by Raman spectroscopy using cell lines that represent a spectrum of human breast cancer. Principal component analysis (PCA) and partial least squares discriminant analysis (PLSDA) revealed unique Raman spectral features in the HER2 and Ki67 subtype. The changes in Raman spectral profiles to different doses of radiation (0-50 Gy) included variations in the levels of proteins, lipids, nucleic acids and glycogen. Importantly, the differences in radiation-induced changes on the normal breast epithelial cell line MCF10A could be discriminated within and across the various breast tumor cell lines. These results demonstrate a novel approach to uncover differences between breast cancer cell subtypes and surrounding normal tissues by their biochemical variations in response to radiation.

Recessed Gold Nanoring-Ring Microarray Electrodes.

A 6 × 6 recessed Au nanoring-ring electrodes microarray was fabricated over a glass substrate using focused ion beam milling. The electrochemical responses of this device to a reversible redox pair were examined. In redox-cycling mode, the lower ring acts as a generator and the upper ring as a collector. High collection efficiencies (close to 100%) and amplification factors (∼3.5) were achieved with this configuration. The redox-cycling behavior of this device was modeled using COMSOL Multiphysics. The effects of scaling the geometric parameters of the electrodes (ring height and radius), potential sweep rates, and inter-electrode gap distance were evaluated through simulations. The computational models showed that the attainable limiting current depends strongly on the ring radius, while it is almost independent of the ring height variations (for a particular inter-electrode gap). The effects of the scan rate and inter-electrode gap distance on the electrochemical characteristics of the device are also discussed. This study provides insights on the influence of the geometry on the performance of these arrays, which should guide the development of future applications.

Comparing the Electrochemical Response of Nanostructured Electrode Arrays.

The electrochemical responses from periodic 6 × 6 arrays of recessed gold nanorings were compared to the 6 × 6 recessed gold nanodiscs arrays. The nanostructured arrays were fabricated by focused ion beam milling and their electrochemical response from a reversible redox pair was obtained. Three-dimensional cyclic voltammetry simulations using COMSOL were performed on 6 × 6 periodic arrays of both recessed nanodiscs and nanorings to elucidate the differences in mass transport between these geometries. Specific mass transport properties near the electroactive surface of the electrodes were elucidated by analyzing the calculated concentration profiles of the redox species. Relative contributions from radial diffusion regimes inside the nanoholes play an important role on the electrochemical response of the recessed nanorings. Arrays of nanodiscs are common in different types of applications, particularly in biosensors. The results presented here suggest that the performance and sensitivity of electrochemical nanosensors can be simply improved by implementing electrodes with a geometry which offer greater current density while keeping the overall footprint of the sensor element constant.

SPR platform based on image acquisition for HER2 antigen detection.

HER2 antigen is a marker used for breast cancer diagnosis and prevention. Its determination has great importance since breast cancer is one of the most insidious types of cancer in women. HER2 antigen assessment in human serum is traditionally achieved by enzyme-linked immunosorbent assay (ELISA method), but it has some disadvantages, such as suppressing the thermodynamic-kinetic studies regarding the antibody-antigen interaction, and the use of labeled molecules that can promote false positive responses. Biosensors based on surface plasmon resonance (SPR) are sensitive optical techniques widely applied on bioassays. The plasmonic devices do not operate with labeled molecules, overcoming conventional immunoassay limitations, and enabling a direct detection of target analytes. In this way, a new SPR biosensor to assess HER2 antigen has been proposed, using nanohole arrays on a gold thin film by signal transduction of transmitted light measurements from array image acquisitions. These metallic nanostructures may couple the light directly on surface plasmons using a simple collinear arrangement. The proposed device reached an average sensitivity for refractive index (RI) variation on a metal surface of 4146 intensity units/RIU (RIU = RI units). The device feasibility on biomolecular assessment was evaluated. For this, 3 ng ml-1 known HER2 antigen concentration was efficiently flowed (using a microfluidic system) and detected from aqueous solutions. This outcome shows that the device may be a powerful apparatus for bioassays, particularly toward breast cancer diagnosis and prognosis.

Surface plasmon enhanced up-conversion from NaYF4:Yb/Er/Gd nano-rods.

The surface plasmons that are enabled by grating coupling in two-dimensional gold nano-particle arrays (AuNPAs) affected the spectral characteristics of the up-conversion (UC) emission from Yb(3+)-Er(3+)-Gd(3+) co-doped sodium yttrium fluoride (NaYF4:Yb/Er/Gd) nano-rods. The red emission of NaYF4:Yb/Er/Gd nano-rods at 660 nm (excited with a 980 nm diode laser) was significantly enhanced by the interaction with the AuNPAs. The geometric characteristics of the gold nanoparticles influenced the position of the surface plasmon resonance, and their near field strengths. The intensity of the red emission normalized versus the green emission reached 1.4, measured against a reference film in the absence of the metallic nanostructures. The lifetime for the green and red emission decreased steadily as the periodicity decreased (relative to the reference), reaching about 6% reduction for the 350 nm AuNPA. A qualitative agreement was obtained between the experimental results and finite difference time domain (FDTD) calculations.

Improving the performance of gold nanohole array biosensors by controlling the optical collimation conditions.

An experimental investigation on how the bulk and surface sensitivities of gold nanohole arrays fabricated by interference lithography affect the degree of white light beam collimation is presented. The optical transmission response of nanohole arrays has been recorded by focused and collimated beam transmission spectra. The results show that both the bulk and surface sensitivities for the collimated case are much larger than for the focused case. In particular, the shape of the spectra was dependent on the degree of beam collimation. The results showed that improved sensing performance (around 3.5 times) and higher figure of merit (around 4.4 times) can be obtained by simply adjusting the incident/collection experimental conditions in transmission measurements.

A silver nanoparticle embedded hydrogel as a substrate for surface contamination analysis by surface-enhanced Raman scattering.

A surface enhanced Raman scattering (SERS) substrate, capable of extracting small amounts of organic species from surfaces of different types of materials with variable roughness, has been fabricated. The substrate consists of Ag NPs encapsulated in poly(vinyl alcohol) (PVA) hydrogels, commonly known as PVA "slime". Unlike traditional SERS substrates, such as colloidal suspensions, the resulting PVA slime SERS substrate presents good viscoelasticity, allowing it to conform to the surface of various materials of arbitrary roughness. Surfaces of different materials, including sandpapers, cotton, metal, and wood, previously contaminated with nile blue A (NBA) were analyzed with the PVA slime SERS substrate. Limits of detection (LOD) as low as 100 ppb (0.79 ng in a total amount on an area of ∼3 cm(2)) were achieved for all surfaces tested. Pesticides and Sudan red III on the glass surface have also been detected, with a LOD of 1.6 ng per ∼3 cm(2).

Periodic metallic nanostructures as plasmonic chemical sensors.

Periodic plasmonic nanostructures are being widely studied, optimized, and developed to produce a new generation of low-cost and efficient chemical sensors and biosensors. The extensive variety of nanostructures, interrogation approaches, and setups makes a direct comparison of the reported performance from different sensing platforms a challenging exercise. In this feature Article, the most common parameters used for the evaluation of plasmonic nanostructures will be reviewed, with particular focus on the advances in periodic plasmonic nanostructures. Recent progress in the fabrication methods that allow for the high-volume production of periodic plasmonic sensors at low cost will be described, together with an assessment of the state of the art in terms of periodic structures employed for chemical sensing.

Improved synthesis of gold and silver nanoshells.

Metallic nanoshells have been in evidence as multifunctional particles for optical and biomedical applications. Their surface plasmon resonance can be tuned over the electromagnetic spectrum by simply adjusting the shell thickness. Obtaining these particles, however, is a complex and time-consuming process, which involves the preparation and functionalization of silica nanoparticles, synthesis of very small metallic nanoparticles seeds, attachment of these seeds to the silica core, and, finally, growing of the shells in a solution commonly referred as K-gold. Here we present synthetic modifications that allow metallic nanoshells to be obtained in a faster and highly reproducible manner. The main improved steps include a procedure for quick preparation of 2.3 ± 0.5 nm gold particles and a faster approach to synthesize the silica cores. An investigation on the effect of the stirring speed on the shell growth showed that the optimal stirring speeds for gold and silver shells were 190 and 1500 rpm, respectively. In order to demonstrate the performance of the nanoshells fabricated by our method in a typical plasmonic application, a method to immobilize these particles on a glass slide was implemented. The immobilized nanoshells were used as substrates for the surface-enhanced Raman scattering from Nile Blue A.

Quantification of ovarian cancer markers with integrated microfluidic concentration gradient and imaging nanohole surface plasmon resonance.

Nanohole array-based biosensors integrated with a microfluidic concentration gradient generator were used for imaging detection and quantification of ovarian cancer markers. Calibration curves based on controlled concentrations of the analyte were created using a microfluidic stepped diffusive mixing scheme. Quantification of samples with unknown concentration of analyte was achieved by image-intensity comparison with the calibration curves. The biosensors were first used to detect the immobilization of ovarian cancer marker antibodies, and subsequently applied for the quantification of the ovarian cancer marker r-PAX8 (with a limit of detection of about 5 nM and a dynamic range from 0.25 to 9.0 µg.mL(-1)). The proposed biosensor demonstrated the ability of self-generating calibration curves on-chip in an integrated microfluidic platform, representing a further step towards the development of comprehensive lab-on-chip biomedical diagnostics based on nanohole array technology.

Optofluidic concentration: plasmonic nanostructure as concentrator and sensor.

The integration of fluidics and optics, as in flow-through nanohole arrays, has enabled increased transport of analytes to sensing surfaces. Limits of detection, however, are fundamentally limited by local analyte concentration. We employ the nanohole array geometry and the conducting nature of the film to actively concentrate analyte within the sensor. We achieve 180-fold enrichment of a dye, and 100-fold enrichment and simultaneous sensing of a protein in less than 1 min. The method presents opportunities for an order of magnitude increase in sensing speed and 2 orders of magnitude improvement in limit of detection.

Detection of SARS-CoV-2 in saliva by a low-cost LSPR-based sensor.

The SARS-CoV-2 pandemic started more than 3 years ago, but the containment of the spread is still a challenge. Screening is imperative for informed decision making by government authorities to contain the spread of the virus locally. The access to screening tests is disproportional, due to the lack of access to reagents, equipment, finances or because of supply chain disruptions. Low and middle-income countries have especially suffered with the lack of these resources. Here, we propose a low cost and easily constructed biosensor device based on localized surface plasmon resonance, or LSPR, for the screening of SARS-CoV-2. The biosensor device, dubbed "sensor" for simplicity, was constructed in two modalities: (1) viral detection in saliva and (2) antibody against COVID in saliva. Saliva collected from 18 patients were tested in triplicates. Both sensors successfully classified all COVID positive patients (among hospitalized and non-hospitalized). From the COVID negative patients 7/8 patients were correctly classified. For both sensors, sensitivity was determined as 100% (95% CI 79.5-100) and specificity as 87.5% (95% CI 80.5-100). The reagents and equipment used for the construction and deployment of this sensor are ubiquitous and low-cost. This sensor technology can then add to the potential solution for challenges related to screening tests in underserved communities.