*3.2. SEM Measurements*

SEM images were acquired from the FEI Nova NanoSEM 450 instrument (Hillsboro, OR, USA) operating at an accelerating voltage of 10 kV and under high vacuum.

### *3.3. Chemometrics—Principal Component Analysis*

Principal component analysis (PCA) is a multivariate procedure that can reduce the dimensionality of original raw data to several principal components (PCs). It is an effective technique that gives the possibility to categorize SERS spectra that are readily distinguishable via visual empirical analysis. The calculated PCs contain the most significant information from the whole introduced data set. The PCA was performed using the commercial Unscrambler® software (CAMO software AS, version 10.3, Oslo, Norway). The SERS data of all analyzed cells (leucocytes, HeLa, and PC3 cells) were optimized for PCA using the following steps: (i) smoothing with a Savitzky–Golay filter (Oslo, Norway), (ii) background correction (concave rubber band correction; the number of baseline points was 34 and the number of iterations was 10), and (iii) normalization using OPUS software (Bruker Optic GmbH, 2012 version, Ettlingen, Germany). The PCA was completed based on the NIPLAS algorithm, validation (random with 20 segments), significance 0.05, and a SERS spectra number of 120.

### **4. Results and Discussion**

### *4.1. Preparation of the SERS Platform*

In this study, we present a novel SERS platform prepared with the use of the electrospinning technique [42] for the label-free analysis of CTCs in blood samples. Fabrication of MBSP consisted of two steps:

(i) Electrospinning of polymer mats with desired parameters, i.e., diameter of the polymer fibers and diameter of the pores;

(ii) Sputtering of thin (usually tens of nanometers) layer of SERS-active metal, e.g., gold, silver, or their alloy via the PVD method.

The basic scheme of the utilized method for step (i) is shown in Figure 1.

**Figure 1.** (**a**) Basic layout of the setup utilized for electrospinning, which consists of a high voltage power supply (HV), two syringe pumps, and a grounded collector; (**b**) a photo of experimental setup used in experiments.

To prepare SERS-active platforms there was a need to coat polymer mats with metal NPs or metal islands. In the case of thin metal islands, it could be done by PVD or vacuum evaporation. In our study the layer of Ag:Au alloy was sputtered on the polymer fibers via the PVD method. The Ag:Au alloy ensured the combination of very high enhancement of the Raman signal provided by Ag with the chemical stability offered by Au [43]. In order to create a platform that provided optimal enhancement of the Raman signal, three different thicknesses of Ag:Au alloys (20, 40, and 80 nm) were tested.

The Ag:Au alloy layer of 20 nm deposited on the polymer mat was too thin to cover the platform and thus to obtain the SERS signal of *p*-MBA (*p*-mercaptobenzoic acid) or tumor cells. As a result, the recorded SERS spectra were derived from the polymer. The polymer mat covered with 40 nm of Ag:Au alloy showed the greatest SERS enhancement. No SERS signals from the polymers were observed. A similar level of enhancement was achieved for the 80 nm Ag:Au layer. Therefore, in the present study, the polymer mats covered with the 40 nm layer were used for all experiments as the most cost-effective. Moreover, the process of sputtering of 40 nm of Ag:Au alloy took only 4 min compared to 8 min for the 80 nm layer.

The morphology of the created SERS substrates named Au:Ag/MBSP SERS was examined by scanning electron microscopy (SEM). The SEM images of (P(LLA-CL)) covered with 40 nm of the Ag:Au alloy layer are presented in Figure 2 at smaller and larger magnifications, respectively. As can be seen in Figure 2a, the arrangemen<sup>t</sup> of fibers with a diameter of *ca*. 1.5 μm within mats was irregular with the slots between fibers of *ca.* 15 μm that were small enough not to let the tumor cells (with the diameter of 20–28 μm) pass through the (P(LLA-CL)) mat. Additionally, the SEM image shown in Figure 2b reveals that the obtained layer of Ag:Au consisted of semi-spheres with diameters ranging from 40 to 55 nm, and their size was responsible for the enhancement factor of the presented Au:Ag/PBSP SERS substrates, and determined the SERS efficiency of these surfaces.

**Figure 2.** SEM images of electrospun polymer mat coated with gold layer (40 nm) at (**a**) lower and (**b**) higher magnification.

The SERS platform designed in such a way worked also as a filter, which allowed the separation of circulating prostate cancer (PC3) and cervical carcinoma (HeLa) cells re-suspended in human blood plasma at a concentration of about 40 cells in 1mL of blood.

The main advantage of the proposed method is the fact that it does not require use of separate techniques to perform filtration, enrichment, and examination of tumor cells circulating in blood. Additionally, by combining these three basic steps in detecting cancer cells in one single process, the transfer of the cells from one place/method to another is eliminated. Therefore, the proposed strategy prevents contamination of the samples and disintegration of cell structures, and leads to improvement of the accuracy of analysis and reduction of the time of analysis.

### *4.2. SERS Investigations of Circulating Tumor Cells*

The Au:Ag/MBSP SERS platform worked both as a filter and as an efficient SERS support and allowed for: (i) separation of studied cells from the complex blood sample due to the sizes of MBSP pores (Figure 2) and sizes of particular blood components (Table S1), and (ii) enrichment of circulating tumor cells within a small and defined area of the SERS substrate. Figure 3a illustrates the experimental setup used for the detection of studied cells whereas Figure 3b demonstrates the filtration process.

As the amount of single CTCs in peripheral blood is small [44], the highly-efficient cell enrichment and single cell capturing were essential for further cell examination.

The proposed concept based on spiking blood samples obtained from healthy donors with a known number of HeLa and PC3 cells (40 cells in 1 mL of blood) may have in the future a practical potential in medicine.

In order to push the sample through the device, a constant pressure of approximate 80 × 10<sup>3</sup> Pa was applied. The whole filtration process took about 4 minutes. Since the pores in the Au:Ag/MBSP SERS platform had a diameter of *ca.* 15 μm and the sizes of blood components and analyzed CTCs did not exceed 15 and 28 μm, respectively (see Table S1, Supplementary Materials), the separation of CTCs from other blood components could be performed. The smaller components of blood plasma passed through the Au:Ag/MBSP SERS platform whilst the largest CTCs remained on the surface of the modified (P(LLA-CL)) mat. As mentioned before, the Ag:Au nanostructures present on the polymer mat fibers are responsible for amplification of the Raman signal of CTCs. Therefore, in the next step the spectroscopic fingerprints of captured CTCs were recorded to perform detailed molecular analysis and identification of studied cells.

**Figure 3.** (**a**) The scheme of capturing circulating tumor cells (CTCs) from the blood sample. The system involves: a vacuum pump, Buchner flask, and filter funnel. The surface-enhanced Raman spectroscopy (SERS) platform was placed on the filter funnel and a droplet of blood spiked with CTCs was put on the platform. After turning on the pump, the liquid was sucked through the mat to the flask, whereas the CTCs remained on the surface of the SERS platform. (**b**) Filtration process of the fluid with the CTCs. The setup consists of a ceramic filter and SERS-active platform placed in the very center. After pipetting a small amount of fluid (top) with CTCs, the vacuum pump is turned on and the blood passes through the mat and the ceramic filter to the Büchner flask, whereas the CTCs stay on the SERS-active platform (steps i–iv).

In order to collect the reference spectra of all studied cells (PC3, HeLa, and leucocytes as an example of healthy cells) the SERS measurements were performed directly from pre-cultures (see Figure S1, Supplementary Materials). Table 1 presents the main SERS bands observed in analyzed cell spectra and the corresponding bands assignments.

Figure 4 depicts the SERS spectra of the cells isolated from blood samples using the Au:Ag/MBSP SERS platform according to the procedure discussed above. As can be seen these spectra showed differences in the position of some bands and their relative intensities. However, the common bands corresponded to the main components of the eukaryotic cell [45]: nucleic acids, proteins, and lipids were clearly observed in all SERS spectra.

**Figure 4.** Averaged and normalized SERS spectra of (**a**) leucocytes, (**b**) cervical carcinoma (HeLa), and (**c**) prostate cancer (PC3) cells recorded on polymer-based SERS platform. Experimental conditions: excitation at 785 nm, laser power at 1.5 mW, and 45 seconds integration time. Each SERS spectrum was obtained by averaging at least 25 single spectra from different places on the SERS substrate.

All the spectral fingerprints depicted in Figure 4 corresponded with the reference SERS data in Figure S1. In Figure 4 one can observe that the vibrational modes of nucleic acids were present at 785 and 1093 cm<sup>−</sup>1. The week bands around 1268 and 1660 cm<sup>−</sup><sup>1</sup> were characteristic of amide I and amide III bands, respectively. In all recorded SERS spectra there appeared vibrational modes characteristic of phenylalanine (1003 cm<sup>−</sup>1), tyrosine (850 cm<sup>−</sup>1), and tryptophan (725 cm<sup>−</sup>1). As can be observed, the SERS spectrum of particular cells also had their own specific spectral features. For example, the band at 1345 cm<sup>−</sup>1, which corresponded to adenine and guanine, could be seen in PC3 cells, but not in the HeLa cells and leucocytes. Additionally, the relative intensities of some bands could also be used for differentiation of analyzed cells. To make identification of PC3 and HeLa cells, the ratio of the relative intensities of the bands at 658 cm<sup>−</sup>1/725 cm<sup>−</sup><sup>1</sup> could be used. In the SERS spectra of leucocytes, the most prominent bands appeared at 652 cm<sup>−</sup><sup>1</sup> (C–C twist of tyrosine) [46], 726 cm<sup>−</sup><sup>1</sup> (C–S in protein, CH2 rocking, adenine) [47], 1003 cm<sup>−</sup><sup>1</sup> (C–C of phenylalanine) [48], 1170 cm<sup>−</sup><sup>1</sup> (C–H in plane of tyrosine or nucleic acid) [49], 1458 cm<sup>−</sup><sup>1</sup> (nucleic acid nucleotides) [50], and 1618 cm<sup>−</sup><sup>1</sup> (ν(C=C), tryptophan, tyrosine) [51]. All these dissimilarities enabled recognition of circulating cells. The spectroscopic data revealed that the healthy leucocyte cells could be distinguished from tumor cells using bands at 1032 cm<sup>−</sup><sup>1</sup> (CH2CH3 bending modes of lipids) [51] and 1452 cm<sup>−</sup><sup>1</sup> (structural protein modes of tumors) [52]. The intensive band at 1452 cm<sup>−</sup><sup>1</sup> was assigned to overlapping asymmetric CH2 bending and CH2 scissors vibrations. The bands of phospholipids, elastin, and collagen were also identified in this region [41]. These differences reflected the changes in biochemical pattern of cancer cells (compared to healthy cells) as the result of carcinogenesis. Table 1 depicts all observed SERS bands with their assignments.


**Table 1.** Assignment of SERS bands depicted in Figure 4 [49,53–57].

The reproducibility of recorded SERS signals is a crucial parameter, especially in the terms of real clinical applications. We calculated the reproducibility of the SERS signals of leucocyte, PC3, and HeLa cells (usually 40 SERS spectra for each type of cells measured on the same SERS substrate were considered). The calculated standard deviations (RSDs) were performed for the most prominent bands at 725, 1003, and 1035 cm<sup>−</sup>1. The achieved results were 6.2%, 8%, and 7.5%, respectively (Table S2, Supplementary Materials). Additionally, Figure S2 shows examples of representative SERS spectra of all studied cells collected from different points within the same Au:Ag/MBSP SERS platform. However, the changes in relative intensities of the same bands were observed and were related to the effect of molecular orientation in relation to the polarization of plasmon excitations in the metal substrate.

The SEM images (Figure 5) show different types of cells filtered from blood and immobilized on the SERS-active platforms.

**Figure 5.** SEM images of SERS platforms after filtration of (**a**) HeLa cells, (**b**) prostate cancer PC3 cells, and (**c**) leucocyte cells. White blood cells were not captured on the polymer based SERS-active platform due to their size, which was smaller than the diameter of pores.

It is evident that HeLa and PC3 cells of the size of *ca.* 26 and 28 μm, respectively, could be easily detected on the SERS platform. The leucocytes, which had smaller sizes, passed through the pores and could not be identified by SEM imaging.

### *4.3. Principal Component Analysis*

The principal component analysis (PCA) was also performed for the statistical analysis of all studied cells. The data sets including 300 spectra obtained from HeLa, PC3, and leucocyte were analyzed by PCA using the commercial Unscrambler®software (CAMO software AS, version 10.3, Oslo, Norway). It is obvious from looking at Figure 6 that the SERS spectra of different cell types could be easily distinguished (all SERS data were divided into three clusters corresponding to leucocytes (blue), PC3 (green), and HeLa (red) based on the significant PCs (PC-1, PC-2)). We could clearly see that they were distributed into separated regions, which indicated the possibility of differentiation of analyzed cells. The PC-1 and PC-2 were counted in the wavenumber region between 600 and 1700 cm<sup>−</sup><sup>1</sup> and differentiated the cancer cells from normal cells with a sensitivity of 82%.

It should be noticed that the cluster of leucocytes showed a relatively higher homogeneity in comparison to HeLa and PC3 cells, which probably reflects the molecular changes in the structure and biochemical composition of cancer cells [58].

As it can be seen from Figures 4 and 6b many spectral features present in SERS spectra were captured by the main PCs.

The prominent SERS bands at *ca.* 725, 1032, and 1452 cm<sup>−</sup><sup>1</sup> also had the largest weights in the variations and indicated the most significant differences among the three types of analyzed cells. The important contribution to PC-1 also gave bands at *ca.* 1589, 1553, 1470, 1368, and 1380 cm<sup>−</sup>1, analyzed in the previous section. To enhance the sensitivity of differentiation among analyzed cells, a further PCA calculation was made on the limited SERS data (the region between 700 and 750 cm<sup>−</sup>1) where one of the most prominent marker bands at 725 cm<sup>−</sup><sup>1</sup> was observed (Figure S3). In this case, the calculated PC-1 and PC-2 values increased and explained up to 98% of the total variance. This result illustrated that all studied cells were clearly separated into three clusters corresponding to the prostate cancer (PC3), cervical carcinoma (HeLa), and leucocyte cell lines, respectively (Figure S3A).

**Figure 6.** (**a**) The score plots of PC-1 versus PC-2 components for differentiation of leucocyte, HeLa, and PC3 cells. PCA was calculated for the whole region (600–1700 cm<sup>−</sup>1); (**b**) PC-1 loading plot.
