Toward the Nanoscale Chemical and Physical Probing of Milk-Derived Extracellular Vesicles using Raman and Tip-Enhanced Raman Spectroscopy

,


Introduction
Milk is an essential resource for human nutrition, with numerous bioactive compounds such as proteins, amino acids and growth and immunological factors 1 .In recent years, it has been shown that milk, especially bovine milk, is rich in extracellular vesicles (EVs), which are nano-to micrometer-sized structures released from the cell into the extracellular space and play a very important role in cell-cell communication and numerous biological processes 2,3 .The high bioavailability and the ability to cross biological barriers are some of the characteristics of mEVs that make them highly promising for drug delivery and theranostic applications 4,5 .As a consequence of the increasing interest in EVs, a variety of analytical techniques have been employed to extensively study them EVs, in order to gain a comprehensive overview of their morphology, functionality, and chemical composition 6,7 .Raman spectroscopy is a non-destructive technique which can provide a unique fingerprint of the chemical composition of the sample through the analysis of molecular vibrational modes.Raman spectrum is obtained when the sample is irradiated with a monochromatic laser and contains all the characteristic peaks corresponding to the vibrational frequencies of the molecules present on the sample.In particular, Raman spectroscopy on EVs can identify proteins, lipids and nucleic acids 8,9 .In numerous works, Raman has been used to assess the purity of EVs preparations 10,11 or to detect biomarkers under pathological conditions 12,13 .Despite its unique features, the main limitation of Raman spectroscopy is represented by the sized of spot of the laser, which hampers the use of this technique in the analysis of nanoscale materials, and to its sensitivity, which requires a sufficient volume of the sample to produce a detectable signal.Tip enhanced Raman spectroscopy (TERS) is an advanced technique that combines Raman spectroscopy with atomic force microscopy (AFM) in order to increase the spatial resolution and the sensitivity of Raman spectroscopy.In TERS, the presence of a metallic coating on the tip and its nanometer size result in the amplification of the Raman signal though plasmonic effect 14 .This allows the acquisition of Raman spectra from nanosized volumes of the samples enabling their nanoscale chemical analysis 15 .TERS has been used to study nuclear acids, proteins, lipid membranes and cells [16][17][18] .Trough TERS, it is possible to achieve optimal spatial resolution for the study of biological samples, which would allow understanding and analysis at the molecular level, while standard Raman spectroscopy enables only the investigation of bulk samples 19 .In a recent study, TERS has been performed on micrometer extracellular vesicles released from red blood cells, suggesting that this technique is promising for the investigation the molecular heterogeneity of the surface of EVs 20 .In this work, we demonstrated the potentiality of TERS to obtain detailed local information on nanometer sized mEVs, which can be eventually used to obtain a complete biochemical characterization and a comprehensive view of the specific molecular vibrations of the bonds present on the sample.After a preliminary characterization of the mEVs using AFM and transmission electron microscopy (TEM), conventional Raman spectrum of mEVs films was collected and analyzed in order to obtain their 'vibrational signature'.Finally, TERS was used to obtain Raman spectra of isolated mEVs.Also, different locations on the surface of an isolated mEV were selected and probed to retrieve information about the compositional variations on the membrane, which paves the way to the nanoscale chemical mapping of the surface of a single mEV.

Materials and methods mEVs isolation
Preparation and isolation of mEVs from bovine milk collected from a local farm (Rome, Italy) was performed following methodologies already described in literature [21][22][23][24] .After verifying the quality of isolation using standard Western blot, mEVs were preserved for subsequent TEM and AFM analyses by immersing them in 0.1% paraformaldehyde (PFA) in PBS for 30 minutes at room temperature.

Transmission electron microscopy
For TEM analysis, preserved mEVs were stained with 2% uranyl acetate for 10 minutes and placed on carbon-coated grids with a mesh size of 200 for observation.The analysis was performed using a JEM-F200 Multi-purpose Electron Microscope (JEOL, Japan) operating at 80 keV.TEM image study of mEVs was performed by ImageJ and statistical analysis was performed on no less than 200 mEVs.

Atomic force microscopy
For AFM analysis, preserved mEVs were diluted in water and deposited on calcium fluoride slide (Crystran Ltd, UK).AFM topographical images were acquired in standard contact mode with DNP tips, with spring constant 0.06 (N/m) (Bruker Inc.) with Cypher Video-Rate AFM (Oxford Instruments, UK) The software Gwyddion (www.gwyddion.net)was used to analyze the obtained images.To reduce inaccuracies caused by temperature drift and sample tilt, the pictures underwent minimum processing such as deleting line coupling, compensating for substrate tilt, and removing a mean plane.Using Gwyddion's segmentation tools, statistical analysis on the number and size of the isolated mEVs was performed on no less than 200 isolated mEVs.

Raman spectroscopy
The mEVs were analyzed using a confocal inVia TM Raman Spectrometer (Renishaw, UK), with a 250 mm focal length.5 μL of mEVs suspension was deposited on a calcium fluoride slide (Crystran Ltd, UK) and left to dry in the air.They were analyzed at room temperature in the 600-3200 cm -1 spectral range.The signal was dispersed by a holographic grating of 1800 lines/mm and collected by a Peltier-cooled CCD detector.The excitation line at 532.1 nm was produced by a Nd:YAG continuous-wave diode-pumped solid-state laser (Renishaw) and focused on the sample through a short distance working objective N PLAN 50x, with NA = 0.75 (Leica Microsystems).

Tip-enhanced Raman spectroscopy
TERS characterization was performed using a Renishaw inVia TM confocal Raman spectrometer with 250 mm focal length and a Mitutoyo 50× magnification objective (NA 0.42 M PLAN, APO SL type, WD 0.5 mm) coupled to a Bruker Innova TM AFM 25 .The excitation line is the same as that used for Raman analysis (λ = 532.1 nm).TERS tips (Next-Tip S.L.) consisted of commercial AFM tips with an elastic constant of 45 N/m and nominal resonance frequency of 335 kHz, on which Ag and Au nanoparticles were deposited upon ultra-high vacuum conditions with a final curvature radius of 5 nm.

Preliminary characterizations of mEVs
Preliminary characterizations of mEVs were performed to assess their dimensions and size distribution.Figure 1(a) shows two typical TEM images which confirmed that mEVs have spherical shape.The statistical analysis on size distribution, reported in Figure 1(b), indicated an average size of 86 ± 34 nm.Analogous analysis on AFM images, a typical of which is reported in Figure 1(c), gave a result of 94 ± 38 nm, in very good agreement with TEM results.Both TEM and AFM results demonstrated that mEVs were monodispersed and confirmed their structural integrity and purity.

Raman spectroscopic analysis on mEVs
Raman analysis of mEVs was performed collecting spectra in two typical spectral range regions of biological structures, i.e., 600-1800 cm -1 and 2600-3200 cm -1 .In order to have a sufficient amount of sample, the area of interest was selected on the "coffee ring" resulting at the borders of the drop on the substrate.Acquisitions were repeated in several hundred different points in the same region to confirm reproducibility.The characteristic peaks of the Raman bands are shown in Figure 2, which was obtained by using the 532.1 nm excitation wavelength with a laser power of 50 mW, 1 s of exposure time, and 2 accumulations.As for the peaks in the 600-1800 cm -1 region, these include the peak at 1003 cm -1 for phenylalanine and the peak at 1064 cm -1 assigned to the C-C stretch in lipids.The 1126 cm -1 and 1263 cm -1 peaks correspond to C-N stretches and amide III in proteins, respectively.Additionally, the peak at 1296 cm -1 is related to CH2 deformation in lipids, while the peak at 1440 cm -1 is assigned to CH2 and CH3 deformation in lipids and proteins.Finally, the peak at 1655 cm -1 represents amide I and Please do not adjust margins Please do not adjust margins C=C bond stretching in acyl chains.In addition, other peaks of lower intensities were identified, including 718 cm -1 with 876 cm -1 assigned  Please do not adjust margins Please do not adjust margins to symmetric and asymmetric stretching of choline N + (CH3)3 group, the peak at 1740 cm -1 for (C=O) ester stretching present in lipids, and the peak at 1522 cm -1 of carotenoids.As for the peaks in the 2600-3200 cm -1 region, these include those at 2850 cm -1 and 2882 cm -1 , which represent symmetric and asymmetric CH2 stretching in lipids, and the peak at 2930 cm -1 , which is related to symmetric CH3 stretching of proteins and lipids.The complete peak assignment is reported in Table 1.The identified peaks are in agreement with the outcomes of previous Raman studies on extracellular vesicles [26][27][28][29][30][31][32] .This establishes a Raman fingerprint to be used as a reference for the subsequent TERS analysis.
Table 1.Summary of the identified peaks and bands in the standard Raman spectrum with their assignments according to data reported in literature 26- 32 .

Tip-enhanced Raman spectroscopic analysis on mEVs
TERS analyses were first performed in the same region on the coffee ring area obtaining the different characteristic peaks of the Raman bands at low and high wavenumbers.First of all, tip-in and tip-out spectra were acquired.This is a standard procedure when carrying out TERS measurements in order to assess the actual amplification that can be obtained.Basically, this procedure consists of taking a measurement in which the tip is very close to the sample so that the plasmonic effect amplifies the signal 33 (tip-in measurement) and then comparing it with a measurement in which the tip is very far away (ideally infinitely far from the sample) so that there is no amplification effect (tip-out measurement).By comparing these two measurements it is possible to understand which and how much signal is being amplified by the TERS effect and how much is a residual of the classical Raman signal.A typical result is reported in Figure 3, which shows that the peaks present at 1556 cm -1 and 2328 cm -1 are not related to the TERS measurement but to a residual of the Raman signal.In particular, they are attributable to the gaseous components present in the atmosphere (O2 and N2, respectively).These artifacts are present due to the long working distance of the objective and, although they do not represent interesting signals for characterization purposes, they can be used as indicators of the correctness of the measurement itself.In a typical TERS experiment, the topography of the sample is first acquired in order to collect the image of the surface where the locations of interest for Raman investigation are selected.
The same area of the coffee ring has been used to acquire multiple TERS spectra in order to assess the repeatability of the method.Indeed, while in 'macroscale' methods the evaluation of repeatability can be considered a well-established procedure, in nanoscale techniques its assessment may be quite an issue and requires some specific attentions.In particular, TERS spectra were acquired in almost the same point after a fixed time (e.g., approximatively one hour).In order to have a good SNR, to avoid fluctuations due to drift,  Please do not adjust margins

Assignments
Please do not adjust margins to the modulation of the chemical properties of the surface at the nanometer scale, as well as to possible modifications induced by the laser irradiation, repeatability tests were performed on different (but close) locations of the coffee ring.The choice of the coffee ring instead of a single EV allowed us to have a significant amount of material from extracellular vesicles, so that the average spectrum is obtained, thus avoiding the local fluctuations due to nanoscale non homogeneity of the chemical properties of the EVs membrane.Also, this allowed us to probe different points which have the same average properties but are sufficiently far from each other, so that each measurement is not affected by possible modification of the sample due to local heating produced by exposure to laser irradiation.An example of two TERS spectra acquired following this procedure is reported in Figure S1.The two spectra were collected using different number of accumulations (i.e., 2 and 4 subsequent TERS spectra).Notably, spectra collected with increased number of accumulations did not show changes in the spectral content in terms of detected peaks, but only an increase in the SNR, as expected.Therefore, the number of accumulations has been kept as low as possible (but always at least 2) to avoid excessive heating of the sample and tip. Figure S2 shows a zoom of the spectra acquired in the region of main interest for the present work (2800-3100 cm -1 ).
The two measurements definitely do not show any significant difference -with the exception, obviously, of a higher SNR in correspondence of higher number of accumulations.Thus, repeated TERS measurements on the same sample acquired after a relatively short time present the very same peaks, confirming the repeatability of our TERS measurements.However, the issue of repeatability of TERS measurements at the nanoscale requires further discussion.First of all, nanoscale characterizations are in principle able to detect variations at the nanometer scale on the sample surface.Therefore, in order to assess the repeatability of the measurements at the nanometer scale, a careful control and evaluation of instrumental drift must be carried out.A second point which must be considered is that all TERS spectra are obtained selecting a certain number of accumulations.The higher the number of accumulations, the higher the SNR.Nevertheless, a larger number of accumulations requires a longer acquisition time, during which the laser is illuminating the sample increasing the risk of thermal degradation of the sample itself.For the sake of completeness, it must be also noted that prolonged exposure to the laser radiation may also degrade the coating of the tip, which results in a variation of the performance of the tip and, definitely, in the outcome of TERS measurements.Therefore, a reasonable compromise must be selected.The last point which must be considered when evaluating the repeatability of TERS -especially on biological samples -is the effect of heating and thermal degradation of the sample.When multiple spectra are acquired in the same point of the surface, local heating may occur modifying the local chemical properties and thus introducing artefacts in the TERS spectra.While this is generally important in Raman spectroscopy, it may be extremely relevant in TERS where laser-induced heating effects can be particularly intense due to the highly confined electromagnetic field at the sample surface 16 .
A typical TERS spectrum, as the one reported in Figure 4, which was obtained with 25 mW power, 10 s exposure time, and 2 accumulations, shows clear peaks in the same two Raman bands which can be easily rationalized also on the basis of the standard Raman spectra previously acquired.For instance, in the range 600-1800 cm -1 , the peaks at 702 cm -1 and 876 cm -1 represent the characteristic peaks of cholesterol and asymmetric stretching of choline, respectively.The peaks at 1064 cm -1 and 1296 cm -1 are related to C-C stretch and CH2 deformation in lipids.Furthermore, two distinct bands relating to the peaks 1440 cm -1 and 1455 cm -1 were detected, which represent the scissoring and bending of CH2 and CH3.Further peaks at 1652 cm -1 and 1665 cm -1 can be associated Please do not adjust margins Please do not adjust margins to the acyl chains and Amide I.As for the 2600-3200 cm -1 Raman shift region, the presence of characteristic peaks which identify proteins and lipids, i.e. the symmetric and asymmetric CH2 stretching at 2850 cm -1 and 2882 cm -1 , respectively, and the peak at 2930 cm -1 of the symmetric CH3 stretching.In this region, two other peaks related to 2900 cm -1 and 2921 cm -1 are also identifiable which fall within the general band of the CH stretching mode.As can be seen in the spectrum obtained in TERS configuration, some peaks differ in intensity and width from those obtained in standard Raman spectra.Indeed, in the range 600-1800 cm -1 almost all the peaks obtained in Raman are obtained, even if with lower intensity.In particular, in the region 1400-1800 cm -1 some very distinct and close peaks are observed such as 1440 cm -1 and 1456 cm -1 or as 1652 cm -1 and 1665 cm -1 , which are not resolved with standard Raman spectroscopy.Furthermore, in the high Raman shifts region, same information and peaks present in standard Raman spectroscopy (2850 cm -1 , 2882 cm - 1 and 2930 cm -1 ) were collected, but with two additional peaks (2900 cm -1 and 2921 cm -1 ), which confirms the possibility to recognize major CH vibrational bonds present in mEVs.
It is worth noting that, as clearly seen in the presented Raman and TERS spectra, confocal Raman signal has undoubtedly a much higher signal-to-noise ratio (SNR) than TERS.This is due to the fact that the signal is collected from a bigger volume of material, which is proportional to the area illuminated by the electromagnetic radiation, i.e., to the spot of the laser which is approximately 1 m.As a direct consequence, the main drawback of standard confocal Raman is that its spatial resolution is limited by the spot of the laser itself and that a significant volume of material must be probed to obtain an adequate SNR.This is the reason for performing confocal Raman on the 'coffee-ring', in order to have the massive presence of vesicles and thus a sufficient amount of material to obtain a good SNR.Conversely, TERS has generally a poorer SNR because the enhancement of the electromagnetic field due to the tip allows one to collect the signal from a nanometer sized volume.So, notwithstanding the plasmonic amplification, only a smaller volume of the sample contributes to the signal resulting in a lower SNR.Nevertheless, this allows TERS to probe the sample with nanometer spatial resolution, which makes it suitable for nanoscale measurements, e.g., enabling not the study of the average properties of a huge amount of mEVs as confocal Raman, but allowing the investigation of single EV, e.g., to assess the variation of the properties among the various vesicles of the population as well as to investigate the variation of the chemical properties on the surface of a vesicle.In the following, this possibility of probing a single EV in different locations is explored.

TERS characterization of single mEV
The potentiality of TERS in the analysis of mEVs in terms of spatial resolution was demonstrated by selecting an isolated mEV and probing different locations on its surface.The study was focused in particular on the signal corresponding to the CH stretching as it resulted in the peaks with the highest intensity.Therefore, the analysis was limited to the range 2750-3050 cm -1 .Figure 5(a) shows a typical isolated mEV.On the basis of its size, it can be considered representative of typical mEVs.Different spectra were collected in correspondence of 9 points on the surface, indicated with letters from (a) to (i) in Figure 5(b), using 2.5 mW power, 3 s exposure time and 4 accumulations.The limitation to 9 points was ascribed to the presence of thermal drift (particularly overheating by the laser impacting the sample), which, by increasing the temperature of the sample, causes the zone of interest to shift (i.e., a single vesicle might shift a few hundred nm and cause the analysis to be performed on the substrate or on totally different zones).In order to solve this problem, it is therefore necessary to limit the duration of the analysis (i.e. the number of points) and the power that impacts on the sample, thus limiting the temperature rise and hence the thermal drift.
The spectra acquired in correspondence of the 9 points selected on the surface have been averaged obtaining the spectrum reported in Figure 5(c), which is characterized by the presence of two more intense peaks at 2882, 2906 and 2930 cm -1 .Really, in the case of biological materials, the analyzed spectral region is characterized by an abundancy of peaks which therefore are difficult to be rationalized.As for the two peaks at 2882 and 2930 cm -1 , they can be associated to CH2 and CH3 stretching, as already found using standard Raman spectroscopy.Nevertheless, when analyzing the single spectra obtained in the different points, a higher number of peaks can be detected.Indeed, the fitting procedure (carried out using Origin2021 Pro) indicated an average of 15 peaks in each collected spectrum.Differences among the collected spectra indicate the dependence of the Raman signal on the actual location of the mEV probed using the plasmonic tip.This demonstrates the capability of TERS to detect difference in the chemical composition of mEVs with nanometer lateral resolution.The main information from this technique is related to the surface of the sample (the very first few nanometers on the sample) so, ideally, only the signal from the lipidic membrane is observed and, thanks to the correlation with the topographic information, highly spatially resolved structural and chemical information can be obtained.Nonetheless, the identification of specific organic compounds from the local Raman spectrum is very difficult as the accurate assignment of the detected peaks is hampered by the high number of molecules which present peaks in the CH stretching zone.The accurate rationalization of the acquired spectra, however, would require a deeper analysis which far exceeds the scopes of this study.Limiting the analysis to the peaks ascribable to the CH2 and CH3 bonds (2882 and 2930 cm -1 , respectively), it is possible to demonstrate the capability of nanoscale chemical mapping on single mEVs of TERS.As an example, Figure 6 shows the variation of the intensity of the two peaks at 2882 and 2930 cm -1 in correspondence of some of the investigated points.
In correspondence of the point d in Figure 6 the intensity of the 2882 cm -1 peak was 10% higher than that of the 2930 cm -1 one.In other points the two peaks have almost the same intensity.Finally, in some points like (a) or (i) the intensity of the 2882 cm -1 peak is about 50% of that of the 2930 cm -1 peak.This can be associated to a not uniformity in the distribution of CH2 and CH3 bonds on the surface, which seems to suggest a certain heterogeneity of mEVs surface at the nanoscale.Also, it is worth noting that the possibility to localize certain functional groups on the mEV surface depends also on the

Nanoscale Accepted Manuscript
Open A too large volume of material experiencing the plasmonic amplification of the TERS signal would result in a cross-talk among the Raman signal from two adjacent locations reducing sensitivity and resolution of the method.On the basis of the nominal radius of the tip used in this study, a negligible cross-talk among the adjacent points in Figure 6 should be expected.Nevertheless, the increase of the radius of the tip due to wear and melting due to the prolonged exposition to the laser radiation could reduce the actual resolution of the technique.More detailed studies would be needed to address this issue and quantify the actual resolution of the method.Also, while this study demonstrated the capability of TERS to detect specific molecular bonds on the surface of a single mEV, the possibility to identify specific molecules depends on their actual characteristic Raman peaks.In particular, in the present study, a variation of the CH2 and CH3 bonds has been revealed, without the possibility to assign this variation to the distribution of specific molecules.
The capability to select the location of the sample to analyse represents a relevant advantage of TERS.Indeed, in principle Raman and TERS characterizations can be dramatically affected by artefacts from the substrate, residuals of chemicals, or possible contaminants, which may result in the presence of spurious peaks in the same spectral region where the samples under investigation have their characteristic peaks and bands.Actually, this is the reason for the choice of the CaF2 substrate instead of glass ones.Also, the possible presence of spurious peaks induced by residuals of PBS, in which mEVs were kept after isolation, should be discussed.Really, from a practical point of view, in our experiment the presence of PBS can be considered marginal as the sample, before being deposited on the slide, was also further diluted with distilled water.However, preliminary Raman investigation of the bare PBS have been performed.When still in solution, no features are present in Raman spectrum coherently with the fact the salts are dissociated.Due to evaporation of water, salt crystals start to form on the sample.When acquiring Raman spectrum in correspondence of massive residuals of PBS after drying, the contribution to the spectrum of PBS is generally negligible.The only region in which it may be significant is a broad band between 800 cm -1 and 1000 cm -1 .Notably, in this region, mEVs show clear and narrow peaks which cannot be confused with those of PBS.Moreover, the band in the PBS spectrum is more intense between 900 cm -1 and 1000 cm -1 , where mEVs do not show peaks.Moreover, when performing Raman analysis on micrometer sized Please do not adjust margins Please do not adjust margins crystals, only a weak peak is observed closely below 1000 cm -1 which can be identified as corresponding to the HPO4 2-ion at 989 cm -1 34 .Actually, due to the dimension of the size of the laser beam, some crosstalk may be present also when analysing points of the sample near the microcrystals.Conversely, as far as TERS spectra are regarded, it must be observed that they show characteristic peaks in the range 1300-1700 cm -1 and in the range 2800-3000 cm -1 where no contribution from PBS is observed.Indeed, this is due to the fact that the high lateral resolution of TERS allowed us to probe areas of the mEVs sample were PBS residuals were not present.This confirms that one of the main advantages of TERS is that it allows one to perform local Raman spectroscopy at selected locations of interest of the sample.Therefore, such spectra can be selectively obtained on the samples of interest (e.g., the mEVs in the present work) avoiding residuals and other not interesting materials as they can be clearly distinguished in optical microscopy (if relatively big) or AFM images (if at the submicron scale).
Enabling one to perform Raman spectroscopy in selected points of the sample with nanometer lateral resolution, TERS is virtually capable to extend the application of Raman spectroscopy at the nanoscale.In the specific case of the EVs, of which mEVs are an example, Raman spectroscopy has been proposed as a tool to study populations and subpopulations of extracellular vesicles derived from different biological sources, e.g., to assess their purity or for diagnostic purposes 10,[35][36][37][38][39][40][41][42][43] .While confocal Raman involves the global analysis of mean properties of a high number of vesicles, TERS may enable the investigation of single EVs with high spatial resolution, e.g., to identify domains and specific fingerprint areas to be used as biomarkers.Also, using the approach described, TERS could be used to characterize and map proteins or lipids on the surface of single vesicles.Indeed, on the lipid membrane which constitutes the surface of extracellular vesicles there is a plenty of molecules (including not only lipids, but also proteins, nucleic acids, glycans), the variety of which reflects and may be used to identify the cell of origin and their activity 44 .Among the different EVs, the biochemical analysis of milk derived EVs and the description of their lipid and protein content have been recently reviewed 45 .Really, Raman spectroscopy have been extensively used in the study of both Please do not adjust margins Please do not adjust margins proteins 46 and 28 .Therefore, it has been quite natural to use TERS to extend such analysis at the nanometer scale.Indeed, TERS has been proposed to analyze protein structure 16,[47][48][49] .The possibility to study proteins in a complex structure like the membrane of extracellular vesicles, however, requires the capability to acquire local TERS spectra with sufficient SNR where characteristic peaks can be univocally identified.In other words, the prerequisite for the profiling of proteins on the EVs surface is the capability to obtain a Raman spectrum to be used as the 'fingerprint' of a certain protein.
Nevertheless, obtaining such a degree of information at the nanoscale could dramatically improve the knowledge of the properties of EVs in view of possible applications, e.g., in the biomedical field 20 .
Overall, the reported results demonstrated that TERS may represent a powerful technique to analyze the surface of EVs by acquiring complete local Raman spectra and to map the variation of intensity of the relevant peaks.Indeed, the latter corresponds to variations of the quantity of the molecules corresponding to those characteristic peaks, allowing one to map variations in the chemical composition on the surface of the sample.Such images should reflectqualitatively or semi-quantitatively, i.e., referring to relative composition -the distribution of a certain molecule or a certain group on the surface of the EVs.Notwithstanding these promising opportunities, the routinely and standardized application of TERS is currently limited by the spatial resolution of the technique, due to the radius of the coated tip, by the need for a higher SNR which must be obtained avoiding nevertheless a too intense and prolonged exposition to the laser radiation to avoid heating and degradation of the sample, as well as to the drift of the instrumentation when nanometer spatial resolution is required.Notwithstanding these limitations, the results of this study confirm that TERS can be effectively resolve heterogeneity in the chemical composition of single mEVs and in particular of their membrane at the nanometer scale, making this technique promising for advanced studies, e.g., to investigate the mechanisms of loading of mEVs at the molecular scale.

Conclusions
In conclusion, TERS has been demonstrated as a powerful tool to investigate chemical analysis of mEVs and, more in general, of EVs at the nanometer scale at the level of the single vesicle.After obtaining standard Raman spectra on an extended area of mEVs to be used as their reference 'signature', TERS has been performed on the same area in order to assess its capability to identify the characteristic peaks.Then, variations of the distribution of specific bonds, i.e., CH2 and CH3, on the surface of a single mEV selected on the sample have been analyzed confirming the capability of TERS to perform chemical mapping at the nanoscale on the surface of single EVs.

FIGURE 1 .
FIGURE 1. Morphological characterization of milk-derived extracellular vesicles: (a) typical TEM images of mEVs and (b) statistics on their size on the basis of TEM analyses; (c) typical AFM image of mEVs and (d) statistics on their size on the basis of AFM analysis.

FIGURE 2 .
FIGURE 2. Standard Raman spectrum of mEVs obtained by averaging hundreds of acquisitions in correspondence of the border of the sample droplet on the substrate ("coffee ring").

FIGURE 3 .
FIGURE 3. Example of the comparison of TERS tip-in and tip-out measurements performed on the coffee ring in a typical TERS experiment.Nanoscale Accepted Manuscript
Access Article.Published on 22 March 2024.Downloaded on 3/30/2024 9:24:52 AM.This article is licensed under a Creative Commons Attribution 3.0 Unported Licence.View Article Online DOI: 10.1039/D4NR00845F Please do not adjust margins Please do not adjust margins dimension of the volume of material from which the local Raman signal is collected which limits the lateral resolution of the technique.

FIGURE 5 .
FIGURE 5. (a) image of a typical mEV and (b) the same mEV image where 9 different points of the surface were probed, indicated with letters from (a) to (i) (c) average spectrum obtained in the range 2750-3050 cm -1 from the acquired 9 spectra which are reported in (d).

FIGURE 6 .
FIGURE 6. Variation of the intensity of the TERS peaks at 2882 cm and 2930 cm -1 , which are associated to CH2 and CH3 bonds, on four nanosized regions of the surface of the mEV.