2.1.

Chemicals

Tetraethyl orthosilicate (TEOS) was ordered from Wako Pure Chemical Industries (Germany). Poly(vinylpyrrolidone) (Mw 55,000) (PVP), 3-mercaptobenzoic acid, 4-mercaptophenol, 4-mercaptobenzoic acid, thiophenol, 6-mercaptopyridine-3-carboxylic acid, 2-mercapto-4-methylthiazol-5-acetic acid, (3-aminopropyl)triethoxysilane (APTES), 4,7,10-trioxa-1,13-tridecanediamine, $n$-hydroxysulfosuccinimide sodium salt (sulfo-NHS), PAMAM dendrimer generation 4.5 polyamidoamine, ethanol (99.8%), ammonia (30 wt. %), phosphate-buffered saline (PBS), 2-($n$-morpholino)ethanesulfonic acid (MES), bovine serum albumin (BSA), and propidium iodide solution ($1\mathrm{mg}/\mathrm{mL}$ in water) were purchased from Sigma–Aldrich (Germany). EDC-hydrochloride was supplied from AppliChem GmbH (Germany). A silver-enhancing kit was received from BBI international, Plano GmbH (Germany). Polyvinyl alcohol (Mw 30,000–70,000) (PVA) was obtained from VWR (Germany). EpCAM (VU1D9) mouse mAb (Alexa Fluor®488) was supplied from New England BioLabs GmbH (Germany). Sterile syringe filter (220-nm pore size) made of polyethersulfone from Rotilabo® was used. Deionized water was used in all procedures. For cell culture DMEM (Dulbecco’s Modified Eagle Medium), pen/strep (penicillin/streptomycin), EDTA (ethylenediaminetetraacetic acid), and erythrocytes lysing buffer were purchased from Invitrogen, and FCS (fetal bovine serum) was ordered from Biochrom AG (Germany).

2.2.

Multicore Surface-Enhanced Raman Spectroscopy Labels

The preparation of MSLs was previously reported.³⁴ In brief, aggregation of gold NPs around 40 nm in diameter was controlled with a linear water-soluble diamine. Raman reporter molecules were covalently bound at the metal surface. In our approach, a reporter molecule concentration of $1:{10}^{-6}$ (%vol.) in water was applied. MSLs with the reporter molecule 3-mercaptobenzoic acid, 4-mercaptobenzoic acid, and thiophenol were used for cell experiments. Different charged polymers such as PVP and PVA prevent further aggregation and support the growth of a silica shell. Before preparation of the silica shell, silver was deposited at the surface of gold aggregates with reporter molecules. Then, $600\mu \mathrm{L}$ MSL was filtered with a syringe filter (220-nm pore size) to remove larger clusters. After that the silica shell was completed by the use of TEOS and ammonia.

2.3.

Surface Functionalization

The silica shell was functionalized in an improved way with APTES as follows. After 20-min incubation with TEOS, $1\mu \mathrm{L}$ of APTES was added under vigorous stirring to the mixture for further 10 min. The mixture was then washed twice with ethanol and once with water. Sonication was used between the washing steps to suspend the MSL. After a further washing step, water was removed and the pellet was suspended in a mixture of $50\mu \mathrm{L}$ PAMAM dendrimer, 3.7 mg EDC, and 4.2 mg sulfo-NHS aliquot with MES buffer ($\mathrm{pH}=4.5$) to complete the final volume to 1 mL and activated for 30 min by shaking before using. The reaction mixture with the MSL was shaken for another 30 min. To remove all unreacted dendrimer solutions, the suspension was washed with MES buffer by centrifugation (8000 rpm, 5 min) two times. $20\mu \mathrm{L}$ of EpCAM (alexa®fluor 488) antibody was added to the activated MSL and was carefully shaken for 3 h. Finally, all unreacted proteins were removed from the solution (6000 rpm, 5 min) and the sample was washed by centrifugation three times with block buffer ($\mathrm{pH}=7.4$, $\mathrm{PBS}+0.1\%$ BSA).

2.4.

Cell Cultivation and Reagents

For cell experiments, the cell lines MCF-7 (breast adenocarcinoma) and human foreskin fibroblasts (HFFs) as well as primary leukocytes from healthy volunteers were used. MCF-7 cells and HFFs were cultured in $\mathrm{DMEM}+10\%$ $\mathrm{FCS}+1\%$ pen/strepat 37°C under a 5% ${\mathrm{CO}}_2$ atmosphere. For splitting the confluent culture of MCF-7 cells, trypsin and EDTA were used. Leukocytes from whole blood samples were prepared by lysis of 8-mL patient blood with 40-mL erythrocyte lysis buffer for 10 min. White blood cells were collected by centrifugation (10 min, 1200 rpm). In a second step, the leukocyte pellet was resuspended with 15-mL erythrocyte lysis buffer and incubated for 10 min. After centrifugation, the cell pellet containing leukocytes was suspended in PBS + 2 mM EDTA buffer.

2.5.

Cell Detection

After the cells were grown for 1 day, cells were transferred into block buffer ($\mathrm{PBS}+0.1\%$ BSA) for 30 min. Anti-EpCAM antibody functionalized and nonfunctionalized MSL were added to the cells in a defined concentration, volume, and time. In detail, $100\text{-}\mu \mathrm{L}$ nonfunctionalized MSL in two different concentrations ($\mathrm{OD}=0.11$ and 0.15 at 534 nm) were added to $500\mu \mathrm{L}$ of MCF-7 cells or leukocytes in block buffer for 1 h. Incubation was carried out during mixing of cells. Cells were washed by centrifugation (5 min, 300 rpm) with block buffer three times. MCF-7 cells and leukocytes in suspension were also treated with EpCAM functionalized MSL (MSL-EpCAM) in three different concentrations ($\mathrm{OD}=0.11$, 0.13, and 0.15 at 534 nm) in the same manner. Adherent MCF-7 cells and HFF grown on ${\mathrm{CaF}}_2$ slides were transferred into 2.5-mL block buffer and treated with $500\text{-}\mu \mathrm{L}$ MSL-EpCAM ($\mathrm{OD}=0.13$ at 534 nm) for 1 h. Furthermore, MCF-7 cells were incubated with MSL-EpCAM for 30 min, 1, and 2 h in the same manner. To test successful removal of free MSL by centrifugation, buffer without cells was measured.

2.6.

Spectroscopy and Microscopy

Raman spectra were collected using two different microscopic Raman systems. Single Raman spectra of whole single cells ($40\times$ objective, $\mathrm{P}=40\mathrm{mW}$) and NP suspensions ($10\times$ objective, $\mathrm{P}=40\mathrm{mW}$) were collected with the RXN1 microprobe (Kaiser Optical Systems). The transmissive holographic grating covers the spectral range from 0 to $3550{\mathrm{cm}}^{-1}$ at a spectral resolution of $4{\mathrm{cm}}^{-1}$. The instrument contains a built-in multimode laser at 785 nm (Invictus). The laser wavelength was calibrated using cyclohexane, the intensity axis was calibrated using a white light source, and the wavenumber axis was calibrated using a helium neon source according to the manufacturer’s instructions. Cell maps were measured using an Alpha 300R Plus confocal Raman microscope with a single-mode diode laser at 785 nm (WITec, Ulm, Germany) and a $60\times /\mathrm{NA}=1$ water-immersion objective (Nikon, Japan). Spectra were collected with a $0.5{\mathrm{cm}}^{-1}$ step size, an exposure time of 0.5 s using a $300/\mathrm{mm}$ grating, and 25-mW laser power. The spectral resolution was about $6{\mathrm{cm}}^{-1}$ and the spectral window ranged from 300 to $3200{\mathrm{cm}}^{-1}$. For collection of Raman spectra ${\mathrm{CaF}}_2$ slides and quartz glass cuvettes were used. Raman spectra were baseline-corrected and intensity-normalized (vector normalization) using the Opus software (Bruker). Raman maps were analyzed with WITec Project software.

Fluorescence images were measured with an AxioImager A1 from Zeiss (filter set No. 09: STX-excitation: 450 to 490 nm, STX-emission: 515 to 565 nm). Absorption spectra were collected using the V-670 (Jasco, Japan) spectrophotometer. Transmission electron microscopy (TEM) images were acquired with a CEM 902A (Zeiss, Germany) at 300 mesh-grids (formvar/carbon-film). Centrifuges Heraeus Fresco 17 (Thermo Scientific) and UniGen 15DR (Herolab) were used for sample preparation.

3.1.

Structural and Optical Characterization of MSL

The final structure of MSLs is shown in Fig. 1(a) and consists of aggregated gold NPs, a monolayer of reporter molecules, a thin silver shell, a silica shell, and antibodies. The optimized procedure included to encapsulate MSL with APTES instead of APTMS. Because of the slower reaction times of APTES, a more controlled growth of the silica shell is possible.

Fig. 1

(a) Structure of multicore surface-enhanced Raman spectroscopy (SERS) labels (MSL). (b) SERS spectra of aggregated gold nanoparticles with reporter molecule 3-mercaptobenzoic acid without (black), thin (green), medium (red), and thick silver shell (blue). (c) Transmission electron microscope (TEM) images of MSLs, (d) silica coated MSL after incubation with tetraethyl orthosilicate (TEOS) for 30 min (e) and after incubation with TEOS for 20 min followed by APTES for 10 min. The blue arrow marks the transition between both silanes. (f) Absorbance spectrum of MSL.

A thin silver shell further enhances the SERS intensities of Raman reporter 3-mercaptobenzoic acid bound to the gold NPs [Fig. 1(b)]. The decrease of the SERS intensities with increasing thickness of the silver shell might be due to the lower penetration of radiation. A silica shell prevents adsorption and enhancement of interfering molecules, and provides stability of MSLs against further aggregation for more than 3 months. After storage for 1 year in the dark at ambient temperature, MSLs gave SERS signals of similar intensities (data not shown). The aggregation was controlled with the aim that between two and four gold NPs were present in the silica shell [Fig. 1(c)]. The silica thickness is around 12 nm after a 30-min incubation time [Fig. 1(d)]. Adding APTES to the mixture for the last 10 min gave a significant increase to overall 22 nm [Fig. 1(e)]. This is attributed to the higher reactivity of APTES compared to TEOS. In TEM images, the transition between the two silanes is visible [Fig. 1(e), blue arrow] Silica provides an effective binding site for conjugation of the amine-containing APTES. The hydrolysis of APTES to silanols and the further condensation generates a surface with organo-functional alkoxysilane molecules. APTES-functionalized surfaces have been shown to be nontoxic,³⁸ and therefore constitute a good and efficient substrate for conjugation of biomolecules. NP aggregates show good SERS enhancement with near-infrared (NIR) excitation^39,40 that is consistent with their plasmonic properties with absorption maxima near 534 and 783 nm [Fig. 1(f)]. The slope of the latter band is typical for aggregated gold NPs and enables efficient excitation by NIR lasers at 785 nm. Gaps in gold NP clusters show an enhanced electromagnetic field⁴¹ called “hot spots” and enable short measurement times down to milliseconds for adsorbed molecules.

3.2.

Antibody Conjugation

The conjugation of the antibody to MSL is shown in Fig. 2. Dark field (c, d) and fluorescence images (a, b) of dried MSL are compared after conjugation of anti-EpCAM antibodies labeled with a fluorescent dye. A polyamidoamine (PAMAM) dendrimer was used to enlarge the surface for covalently binding of antibodies at the silica surface. PAMAM has a specific size and shape and a highly functionalized terminal surface. The term “dendrimer” was derived from its tree-like branching structure. The G4.5 dendrimer contains 128 terminal carboxyl groups on the outer surface. For cross linking of secondary amines (APTES with IgG antibody) with carboxylates (PAMAM dendrimer), EDC and sulfo-NHS enable formation of a stable amide bond. The efficiency of EDC-mediated coupling is increased in the presence of sulfo-NHS. Fluorescence images indicate the success of antibody conjugation at the MSL surface after several washing steps. The positions of MSL agree well in the dark field and fluorescence image. Virtually all MSLs are conjugated with a high number of antibodies (Fig. 2).

Fig. 2

Fluorescence images (a, b) and dark field images (c, d) of anti-EpCAM antibody (Alexa®488) conjugated MSL. Every dot consists of MSL clusters. Little dots may be single MSL. In images (a) and (c) the sample margin is marked. The sample shows the same greenish and bluish backgrounds as the glass slide, respectively.

3.3.

Cell Detection by Nonfunctionalized MSL

The specificity of nonfunctionalized MSLs (MSL-non) for tumor cell detection was tested by incubating MSLs-non at two different concentrations with living MCF-7 cells in suspension. Raman spectra of about 20 single cells were measured in 1 s with 785-nm excitation. MSLs-non are not expected to bind specifically to cells because of the missing antibody, and subsequently they should not show SERS spectra of the Raman reporter 3-mercaptobenzoic acid. The percentage values indicate the fraction of cells with a SERS spectrum from MSL (Table 1). At low concentrations of MSL-non (${\mathrm{OD}}_{534\mathrm{nm}}=0.11$), MSL-non did not bind to MCF-7 cells. By increasing the MSL-non concentration (${\mathrm{OD}}_{534\mathrm{nm}}=0.15$), an unspecific interaction was detected in 70% of cells, this means 30% of measured cells show no SERS signal. Forthcoming experiments will investigate the conditions (such as incubation time and concentration) in more detail if MSLs-non constitutes an alternative route for 100% tumor cell detection.

Table 1

Percentage of single MCF-7 cells and leukocytes with surface-enhanced Raman spectroscopy (SERS) spectrum of reporter molecule for concentrations of MSL-non and MSL-EpCAM between OD534  nm0.11 and 0.15.

3.4.

Cell Detection by Functionalized MSL

The specificity of anti-EpCAM antibody-functionalized MSL (MSL-EpCAM) for tumor cell detection was tested in an analogous manner. As EpCAM is over-expressed in MCF-7 cells, the MSL-EpCAM is expected to bind specifically. At low concentration of MSL-EpCAM (${\mathrm{OD}}_{534\mathrm{nm}}=0.11$), SERS spectra of the reporter molecule were detected for 75% of the MCF-7 cells in suspension after antibody-antigen binding (Table 1). Compared with the results of MSL-non, this observation suggests true-positive cell detection. After increasing the concentration to ${\mathrm{OD}}_{534\mathrm{nm}}=0.15$, SERS spectra were obtained for 100% of the MCF-7 cells, probably due to a mix of specific and nonspecific binding of MSL-EpCAM to MCF-7 cells. As described above, already 70% of cells bound MSL-non at this concentration. On the one hand, only 75% of MCF-7 cells were specifically detected at ${\mathrm{OD}}_{534\mathrm{nm}}=0.11$ concentration; therefore, the MSL-EpCAM concentration needs to be increased for 100% specificity. On the other hand, the concentration should be below ${\mathrm{OD}}_{534\mathrm{nm}}=0.15$ to avoid nonspecific interactions. Consequently, the MSL concentration was adjusted. In further experiments, the concentration of MSL-EpCAM was set to an optical density near ${\mathrm{OD}}_{534\mathrm{nm}}=0.13$. After incubation of MCF-7 cells with MSL-EpCAM for 1 h, approximately 100% of ${10}^6$ MCF-7 cells were detected (Table 1).

An uptake of MSL-EpCAM (${\mathrm{OD}}_{534\mathrm{nm}}=0.13$) into the cytoplasm of living adherently grown MCF-7 cells was not observed even after prolonged incubation for 2 h. Dark field images at three depths resolved MSLs only at the upper cell membrane [Fig. 3(a)] but not in cytoplasm (b) and at lower membrane (c). The results in Fig. 3 are consistent with fluorescence images in Fig. 4 and SERS images in Fig. 5 that confirm the localization of MSL-EpCAM near the cell membrane. The large size of MSL over 120 nm and low incubation times might explain the lack of internalization of MSL by cells and the low toxicity in viability tests (see Fig. 6). Incubation times of only half an hour lead to incomplete binding of MSL-EpCAM to MCF-7 cells, and thus give similar results as for using concentrations of MSL-EpCAM below ${\mathrm{OD}}_{534\mathrm{nm}}=0.13$.

Fig. 3

Dark field images at three focal planes of MCF-7 cells after incubation with MSL-EpCAM for 2 h (${\mathrm{OD}}_{534\mathrm{nm}}=0.13$) (a-c). (a) White circles mark MSLs visible at the upper cell membrane, (b) but not in the cytoplasm, (c) lower membrane. MSLs are not evident within the cell, but at the outer membrane.

Fig. 4

(a, c, e, g) Dark fields and (b, d, f, h) fluorescence images of (a, b, e, f) MCF-7 and (c, d, g, h) human foreskin fibroblast (HFF) cells after incubation with fluorescent labeled (a, b, c, d) EpCAM antibodies (Alexa®488) and (e, f, g, h) MSL-EpCAM. Arrows mark MSL-EpCAM at MCF-7 cells in the same region of the fluorescent and dark field images (e, f).

Fig. 5

Bright field image of (a) MCF-7 and (b) HFF cells. Chemical map of the C-H deformation band at $2880{\mathrm{cm}}^{–1}$ of (c) MCF-7 and (d) HFF cells. (e, f) Chemical map of the MSL band at $1073{\mathrm{cm}}^{–1}$. Representative Raman spectra of MSL and cell from the Raman images (position indicated by blue and green star).

Fig. 6

Viability test with propidium iodide. Fluorescence and dark field images of MCF-7 cells after incubation with (a, b) MSL-EpCAM or (c, d) pure EpCAM as reference for 1 h. Nuclei of dead cells are colored in red after 1 h incubation time.

MSL-EpCAM was also incubated with leukocytes and binding was monitored by SERS spectroscopy. It was observed that MSL-EpCAM did not bind nonspecifically to leukocytes for concentrations between ${\mathrm{OD}}_{534\mathrm{nm}}=0.11$ and 0.15 (Table 1). Although it is unlikely that the binding capacity of MSL-non to leukocytes is higher than MSL-EpCAM, the interaction of MSL-non to leukocytes will also be probed in the context of the specificity of MSL-non for tumor cell detection in forthcoming experiments.

As leukocytes are nonadherent cells in suspension, the specificity of MSL-EpCAM was further tested for HFF as an adherent cell line without EpCAM antigen. Cells were grown on ${\mathrm{CaF}}_2$ slides and incubated with fluorescently labeled EpCAM [Figs. 4(a)–4(d)] and MSL-EpCAM [Figs. 4(e)–4(h)]. Whereas dark field images show a number of cells, fluorescence images prove the binding of EpCAM to MCF-7, but not to HFF [Figs. 4(a)–4(d)]. For MSL-EpCAM, dark field and fluorescence images reveal NPs bound to MCF-7 [Figs. 4(e) and (f)] consistent with Figs. 4(a) and 4(b) but no NPs bound to HFF [Figs. 4(g) and 4(h)].

Raman images of single cells were collected to visualize cellular distribution of MSL-EpCAM whose reporter molecules cannot be detected by fluorescence. Chemical Raman images [Figs. 5(c) and 5(d)] were generated by plotting the intensities of the C-H stretching vibrations between 2800 to $3020{\mathrm{cm}}^{–1}$ for MCF-7 and HFF cells [Figs. 5(a) and 5(b)]. All cells are visualized at a similar intensity scale. The intensities of the reporter molecule band at $1073{\mathrm{cm}}^{-1}$ of thiophenol are also plotted for both cell types on the same scale [Figs. 5(e) and 5(f)]. High intensities were found near the cellular membrane of MCF-7 cells that is consistent with the specific binding of MSL-EpCAM. The Raman image in Fig. 5(e) shows a nonuniform binding of MSL-EpCAM that is consistent with dark field and fluorescence images in Figs. 3 and 4, respectively. The nonuniform binding pattern might point to different affinities or potential blocking of some antigen binding sites.

Low intensities in the Raman image of HFF cells demonstrate the absence of MSL-EpCAM binding. A typical spectrum of MSL-EpCAM from the Raman image of MCF-7 cells is dominated by spectral contributions of the reporter molecule (Fig. 5, blue star). For comparison, a typical Raman spectrum of HFF cells is shown with spectral contributions of proteins and lipids, but without the reporter molecule (Fig. 5, green star).

3.5.

Cell Viability

Viability tests were carried out with MCF-7 cells using propidium iodide during incubation with MSL-EpCAM in a suspension of living cells for 1 h. Propidium iodide stains necrotic and apoptotic cells in red. Figure 6 shows that the majority of cells are unstained confirming their viability and low toxicity of MSL-EpCAM at the concentration ${\mathrm{OD}}_{534\mathrm{nm}}=0.13$. Only a few cells were stained, which appears to be independent of the incubation with MSL-EpCAM [Figs. 6(a) and 6(b)] or pure EpCAM [Figs. 6(c) and 6(d)]. A few dead cells can always be expected in cell cultures because of occasional necrosis and apoptosis. The low toxicity of MSL-EpCAM is probably related to the large size of more than $120\mu \mathrm{m}$ and the lack of internalization as written above. This observation is consistent with a report that amorphous silica NPs below $100\mu \mathrm{m}$ induced cytotoxicity.⁴² The authors suggested that the size is decisive to produce biological effects. This hypothesis was confirmed in a murine toxicology study by administration of SERS-active NPs measuring 100 nm in diameter.⁴³

The shape of cells in Figs. 4–6 is slightly different, because of the applied preparation protocol. The cells in Figs. 4 and 5 grew adherently onto ${\mathrm{CaF}}_2$ slides, whereas the cells in Fig. 6 were treated with trypsin to detach them from slides and transfer them in suspension. Such treatment is known to change the shape of cells. Gold NPs are also used in PT therapies due to their known PT effects by laser irradiation. By using short measurement times of 1 s or less, exciting with a NIR laser and 40-mW laser power phototoxic effects of cells in buffer is expected to be small. Heat-induced cell damage would cause spectral changes in protein vibrational modes that have not been observed so far.