2.1.

Reagents and Chemicals

Myoglobin (Mb) from horse skeletal muscle and Amyloid $\beta$ (1-42) peptide ($\mathrm{A}\beta 42$) were purchased from Sigma-Aldrich and used as received. $\mathrm{A}\beta 42$ solution was prepared by dissolving $\mathrm{A}\beta 42$ powder in 50 mM NaOH and finally diluting in phosphate-buffered saline (PBS) at the desired final concentration.

For SERS experiments, a colloidal solution of cetrimonium bromide-capped gold nanobipyramids (AuBPs, $1\mathrm{mg}/\mathrm{ml}$ in water) having a diameter and length of 35 and 105 nm, respectively, was used (Nanoseedz, Hong Kong). Milli-Q water was used throughout our experiments. Fresh solutions of biomolecules were considered for the measurements.

2.2.

Optical Fiber-Based Light Delivery System for Raman Measurements

An HCPCF-based light delivery system was assembled for carrying out the Raman measurements. The optical setup consisted of a commercial confocal Raman microspectrometer (Labram HR, Horiba Jobin-Yvon, Japan) coupled with a homemade fiber holder that holds the HCPCF (Fig. 1). The excitation laser produced by a solid-state diode laser source emitting at 532 nm was focused into the fiber core using a $50\times$ long working distance microscope objective (NA 0.55, Olympus). The Raman scattered light from the sample propagated back through the fiber and was collected in a backscattering geometry using the same objective, was analyzed by a diffraction grating ($600\text{grooves}/\mathrm{mm}$), and acquired by a charged coupled device (Synapse CCD Detector) embedded in the spectrometer.

Fig. 1

Schematic of the fiber-enhanced Raman spectroscopic setup. (M, mirror; NF, notch filter; and O, microscopic objective).

All experiments were performed with an incident power measured on the sample of 5 mW, whereas the spectral acquisition times for ethanol, Mb, and $\mathrm{A}\beta 42$ were 5, 10, and 15 s, respectively. Before measurements, the spectroscopic system was calibrated on the $520.8{\mathrm{cm}}^{-1}$ Raman peak of a silicon wafer.

Commercial hollow core crystal fibers (Model HC-1060, Thorlabs) were used in FERS experiments. Both ends of a 5-cm-long fiber were first cleaved, then the fiber was meticulously vertically aligned to the microscope objective using a fiber holder mounted on an $XYZ$ stage. The chamber of the fiber holder was filled with the sample solution ($30\mu \mathrm{l}$), which was then allowed to flow down through the microcavities of the fiber for 2 min before the measurements. Photodegradation of the analytes during Raman spectra acquisition was cautiously prevented by assuring a smooth continuous flow of analyte solution through the fiber during the measurements. This was achieved by pumping the analyte solution using a syringe pump in to the sample holder at a regular interval of time. Control measurements were carried out by direct Raman sampling, in which the HCPCF was replaced with an open cuvette filled with the analyte solution and the laser was focused onto the exposed solution in the open cuvette.

3.1.

Fiber-Enhanced Raman Sensing

In order to compare the vibrational signal response produced by our FERS system with that obtained by a conventional Raman spectroscopy setup, Raman measurements were carried out both with the analyte solution inside the HCPCF and with a conventional direct sampling method. The HCPCF (Fig. 2) used in the experiments shows a core diameter of $10\mu \mathrm{m}$, and its pitch diameter is about $2.6\mu \mathrm{m}$. The microstructured air holey cladding region is periodically arranged around the central core with a total dimension of about $50\mu \mathrm{m}$ with an air filling fraction of $>90\%$.

Fig. 2

SEM image of the cleaved end of the HCPCF used in the experiments.

The background Raman spectrum of the optical fiber was preliminarily measured in order to assess the absence of any interference with the Raman spectra of the analytes. We verified that once the HCPCF is filled with water, due to a change in the refractive index, its transmission spectrum gets shifted from NIR (centerd at 1060 nm) to the visible range with maximum transmission in the region from 500 to 900 nm³⁶ (Fig. 3).

Fig. 3

Transmission spectra of the HCPCF filled with air (red dashed line) and with water (blue continuous line). The excitation line at 532 nm (green dashed line) and the spectral range (green band) considered in our Raman experiments are also shown.

3.2.

FERS of Ethanol and Myoglobin

Initially, the Raman spectrum of a 25% (v/v) ethanol–water solution was recorded. Figure 4(a) shows a comparison between the Raman spectral profiles of ethanol as obtained by direct and HCPCF-mediated sampling. The Raman peak at $880{\mathrm{cm}}^{-1}$, which is attributed to the skeletal CCO symmetric stretching^36,47 of ethanol, appears in both spectra. Nevertheless, compared to the sample measured inside the HCPCF, its peak intensity collected by conventional sampling appeared 10-fold lower. Moreover, the Raman peaks at 1049, 1089, and $1277{\mathrm{cm}}^{-1}$, which are assigned to skeletal CCO deformations, CO stretching, and deformation wagging of ethanol, respectively, were only detected in the sample measured in HCPCF. Subsequently, in order to test the system performances for biomolecule analysis, we examined the vibrational spectra of Mb, here taken as a model of a biomolecule extensively investigated in Raman studies^48,49 [Fig. 4(b)]. Prominent vibrational peaks at 1129, 1170, 1306, 1372, 1584, and $1636{\mathrm{cm}}^{-1}$, which are attributed to $\upsilon 22$, $\upsilon 30$, $\upsilon 21$, $\upsilon 4$, $\upsilon 19$, and $\upsilon 10$ of the heme group, respectively,^48,50 were observed by direct sampling of a highly concentrated (1 mM) protein solution. Further measurements were carried out by decreasing the protein concentration down to $50\mu \mathrm{M}$. Similar to the case of ethanol, intense Raman signals were collected by FERS, while most of them were undistinguishable or noisy in the spectrum obtained by the conventional Raman scheme. These results offered us a first test bench proving the efficiency of our FERS-based setup.

Fig. 4

Comparison between Raman spectra acquired by direct sampling (blue, black lines) and using the HCPCF setup (red lines) of (a) a 25% ethanol solution and (b) Mb at 1 mM (black lines) and $50\mu \mathrm{M}$ (blue, red lines).

3.3.

FERS of $\mathrm{A}\beta 42$

Once we clarified the potential offered by FERS with HCPCF, we verified its efficacy in the Raman detection of $\mathrm{A}\beta 42$. Initially, measurements of $\mathrm{A}\beta$ with a concentration of $100\mu \mathrm{g}/\mathrm{ml}$ were performed both under direct sampling and by the HCPCF-based setup [Fig. 5(a)]. Characteristic Raman modes of $\mathrm{A}\beta$ peaking at 830/850, 1003, 1457, and $1660{\mathrm{cm}}^{-1}$ can be identified by both approaches and ascribed to the Fermi doublet of Tyr, the aromatic Phe ring breathing, ${\mathrm{CH}}_2$ scissoring vibrations, and the amide I region, respectively^44–48 (Table 1). Additional peaks were detected in the FERS spectrum ascribed to the backbone CCN stretching vibration ($960{\mathrm{cm}}^{-1}$), the amide III band (1230 to $1300{\mathrm{cm}}^{-1}$), and the aromatic Tyr ring stretching ($1613{\mathrm{cm}}^{-1}$). In accordance with previous results on ethanol and Mb, the spectral features of $\mathrm{A}\beta$ measured using HCPCF appeared sharp and intense in such a way that a signal enhancement of $\sim 20$-fold at the minimum was observed as compared to the direct sampling modality. Nonetheless, apart from some exceptions (e.g., amide I, amide III, the ${850\text{-}\mathrm{cm}}^{-1}$ band of Tyr, and the ${1003\text{-}\mathrm{cm}}^{-1}$ band of Phe), the peaks of the $\mathrm{A}\beta$ solution measured by conventional Raman showed a low signal-to-noise ratio or were mostly imperceptible. These outcomes validated the potential of FERS with HCPCF for efficient detection of $\mathrm{A}\beta$.

Fig. 5

Raman spectra of $\mathrm{A}\beta 42$ acquired by conventional Raman spectroscopy (blue lines) and with an HCPCF (red lines): (a) $100\mu \mathrm{g}/\mathrm{ml}$ and (b) $40\mu \mathrm{g}/\mathrm{ml}$. Shaded regions identify amide I (red) and amide III (blue) bands of $\mathrm{A}\beta$.

Table 1

Assignments of main Raman peaks of Aβ42.49

Afterward, we spent effort in demonstrating the sensitivity of our FERS setup upon decreasing the $\mathrm{A}\beta$ concentration. Successive FERS measurements were performed on a $40\text{-}\mu \mathrm{g}/\mathrm{ml}$ solution of $\mathrm{A}\beta$ [Fig. 5(b)]. In this case, the vibrational spectra measured with the FERS system appeared strongly enhanced compared to those obtained by conventional Raman, which were noisy without any distinct Raman peaks detected.

Aimed at further improving the optical performances of our setup, we integrated the FERS system with AuBPs within the fiber core to take advantage of a supplementary SERS effect. The choice of using nonspherical nanoparticles here is motivated because of their well-established superior SERS efficiency generated by effective SERS hot spots at the sharp ends.^51,52 AuBPs were adhered to the fiber core by electrostatic-mediated surface immobilization between positively charged nanoparticles and the negative surface charge of the inner silica core of the fiber. In short, the nanoparticle solution was allowed to enter into the fiber core by dipping the HCPCF in 2 ml of particle solution for about 5 min. The fiber was then dried at 40°C in an oven for 5 min to remove the solvent and extensively rinsed with Milli-Q water to remove unbound nanoparticles. The $\mathrm{A}\beta$ solution ($40\mu \mathrm{g}/\mathrm{ml}$) was then allowed to flow through the particle-coated fiber core as in the previous measurements. Figure 6 depicts a comparison between spectra obtained by SERS-active HCPCF and bare HCPCF. We demonstrated that an additional $\sim 10$ times signal enhancement could be attained using the AuBPs-functionalized fiber. All the prominent characteristic Raman peaks of $\mathrm{A}\beta$ were identified in the SERS spectrum, suggesting that the native structure of the inspected species was almost preserved.⁵³ The appearance of prevailing peaks as, e.g., Tyr modes at 830, 850, and $1613{\mathrm{cm}}^{-1}$, Phe modes at 1003 and $1205{\mathrm{cm}}^{-1}$, as well as $\mathrm{C}─\mathrm{N}$ vibrational modes from aminoterminated amino acids within the 1080-to ${1150\text{-}\mathrm{cm}}^{-1}$ region may represent an indication of the spatial proximity of these residues to the Au surface.⁴⁹ An overall signal amplification of $\sim 200$ times can be finally estimated as provided by both the use of a HCPCF and its integration with AuBPs within our Raman setup.

Fig. 6

Raman spectra of $40\mu \mathrm{g}/\mathrm{ml}$ $\mathrm{A}\beta 42$ achieved with a bare HCPCF (blue line) and with the AuBPs-coated HCPCF (red line).

This work represents a proof-of-concept demonstration of an effective Raman detection system of $\mathrm{A}\beta 42$ showing high sensitivity combined with well-resolved and intense spectral $\mathrm{A}\beta 42$ features. Thanks to the above characteristics, the proposed FERS-SERS method proves suitable for future implementations aimed at extending the limit of detection to match physiological requirements. $\mathrm{A}\beta 42$ in the CSF of healthy humans ranges around $1.2\pm 0.4\mathrm{ng}/\mathrm{ml}$ and decreases in the case of AD patients.^54,55 Future efforts will be spent to optimize the system by a comparison among different plasmonic nanoparticles, a tight control over their density, as well as their possible functionalization with biorecognition elements.