All-in-one Superparamagnetic and SERS-active Niosomes for Dual-targeted in Vitro Detection of Breast Cancer Cells

In vitro and in vivo biosensing through surface-enhanced Raman scattering often suffer from signal contamination diminishing both the limit of detection and quantication. However, overcoming the lack of specicity requires excessive nanoparticle concentrations, which may lead to adverse side effects if applied to patients. We propose encapsulation of iron oxide (Fe x O y ) and gold (Au) nanoparticles (NPs) into the bilayer structure of transferrin-modied niosomes. This approach enables achieving greatly enhanced and contamination-free SERS-signals in vitro as well as a dual-targeting functionality towards MCF-7 breast cancer cells. An in-depth characterization of Fe x O y NPs- and AuNPs-loaded niosomes (AuNPs/Fe x O y NPs/NIO) after magnetic downstream processing reveals dened hybrid niosome structures, which show a long-term SERS-signal stability in various media such as MCF-7 cell culture medium. In vitro 2D-SERS imaging unveil a successful incorporation of a non-toxic dose of hybrid NPs into MCF-7 cells, which leads to strong and almost contamination-free SERS-signals. The measured signal-to-noise ratio of the in vitro signal exceeds the values required by DIN 32645 for the successful validation of a detection method.


Introduction
Nowadays, the survival rate for diseases such as breast cancer is signi cantly increased in the early stages of diagnosis and can be further improved by the availability of highly sensitive detection methods that allow an accurate monitoring of the disease progression [1][2][3]. Although widespread screening programs can crucially reduce the mortality rate, the current limitations in sensitivity and speci city of the existing imaging techniques explain an insu cient detection of smaller cancerous tissues, which results in complications and the need for invasive therapeutic procedures (e.g. resection of the breast) as well as disease recurrence [4][5][6][7]. Consequently, there is an increasingly pressing demand for accurate and coste cient cancer detection methods.
An emerging technique, the so called surface-enhanced Raman spectroscopy (SERS), provides outstandingly speci c and sensitive detection of analytes such as biomarkers, cells and tissues, with accuracy reaching single-molecule level [8][9][10][11]. Using specially designed analytical nanoparticles (NPs) as SERS-tags, the target moieties can be selectively bound by the NPs and detected indirectly through the Raman signature of a Raman reporter molecule [12][13][14]. In contrast to the established bioimaging methods, SERS offers multiple advantages such as exceptionally high sensitivity, intrinsic molecular ngerprint information, resistance to photobleaching and -degradation as well as a non-invasive longterm multiplex monitoring within various applications [11,[15][16][17][18]. Despite proving its e cacy in several breast cancer-related in vitro, in-vivo and ex-vivo studies [19][20][21][22], the clinical establishment of SERS still remains challenging due to several drawbacks: i) the slow uptake rate of bare SERS-tags, ii) uncertain biocompatibility, iii) interfering species (e.g. a protein corona) adsorbing onto the metallic surface after in vitro/in vivo administration. These drawbacks increase the noise signal and induce signal contamination [23][24][25][26][27]. Unspeci c absorption of biomolecules onto NPs in biological environments may lead to unforeseen physicochemical NP interactions followed by agglomeration and major signal uctuations [28][29][30]. This results in a deteriorated detection limit which is detrimental for the successful validation of a diagnostic procedure according to the German standard method DIN 32645 and other international guidelines, thus hindering the establishment of a reliable detection method [31].
To overcome these obstacles, an encapsulation of the prepared SERS-tags becomes progressively common. By using different inorganic (e.g. silica [32,33]) or organic (mostly liposomes/lipids [34][35][36], polyethylene glycole (PEG) [37,38]) substances, single or multiple NPs can be incorporated in a shell of the respective substance. Especially an encapsulation of multiple functionalized plasmonic NPs into one carrier NP improves the biocompatibility, impedes the leaching of Raman-active molecules, reduces undesired NP-NP interactions and prevents signal contamination [18,35,39]. In addition, a shell-based spatial con nement of several NPs enhances the generation of plasmonic hot spots at interparticle gaps which yields in a drastic enhancement of SERS signal intensity [40][41][42]. Even though plasmonic NPsloaded liposomes are continuously being developed for implementation in biomedical research [43][44][45], prevailing drawbacks still have to be addressed for a successful applicability. Most of the encapsulation procedures lack e ciency and require multiple post-synthetic puri cation steps to separate bare metallic NPs from the hybrid NPs, often leading to low yields and hence insu cient SERS signal intensities [46,47]. Moreover, the low physical and chemical stability of liposomes can result in a degradation of the structural integrity and leakage of the liposome structure throughout encapsulation and application [45,48,49]. Therefore, a novel method to assemble stable hybrids of encapsulated metal NPs is eminently needed.
To circumvent these case problems, another type of organic NPs, the so-called niosomes, can be utilized, which are also capable of entrapping optically active NPs to detect and visualize cancer [51]. Niosomes are vesicular carriers made of non-ionic surfactants forming a bilayer structure that allows the simultaneous encapsulation of hydrophilic and lipophilic compounds [52][53][54]. Due to the surfactants, niosomes have a chemically and physically stable structure which enables long storage times [55].
Additionally, their non-ionic nature makes them highly biocompatible and biodegradable as well as coste cient [56,57]. The surface of niosomes can easily be modi ed with various biomarkers to address a targeting functionality [54]. These outstanding properties entitle niosomes a particular promising material for various future applications, with clinical trials already indicating a successful nanocarrier capability [55][56][57].
Herein, we present a concurrent encapsulation of 5-thio-2-nitrobenzoic acid (TNB)-functionalized gold nanoparticles (AuNPs) and superparamagnetic iron-oxide nanoparticles (Fe x O y NPs) to accomplish the synthesis of the rst niosomes with magnetic and SERS-active properties ( Figure 1). The hydrophilic AuNPs and hydrophobic Fe x O y NPs were synthesized, functionalized and characterized as described in previous publications [62,63]. PEG-maleimide-functionalized niosomes (NIO) and simultaneous encapsulation of the inorganic NPs were prepared using the thin-lm hydration method similar to a recently published work [51], followed by a magnetic puri cation procedure. Finally, the niosome surface was decorated with transferrin (Tf) to augment the uptake by malignant breast cancer cells via Tfreceptor-mediated endocytosis [51,64,65]. Throughout the hybrid niosome assembly, we investigated mainly two parameters to further enhance the resulting SERS-signal intensity: the addition of an ionic substance (HCl) to the AuNP solution before encapsulation and the overall AuNP mass fraction impact.
The successful hybrid niosome (AuNPs/Fe x O y NPs/NIO/Tf) preparation was veri ed by investigating various particle properties using several methods such as i) dynamic light scattering (DLS), ii) SERS, iii) transmission electron microscopy (TEM), iv) UV/Vis Spectroscopy and iiv) Zeta potential measurements. Moreover, by suspending the hybrid niosomes in different (complex biological) media, the long-term SERS-signal integrity and occurrence of possible signal contaminations was analysed over a period of 46 days. Ultimately, after verifying the biocompatibility within in vitro cytotoxicity studies, the hybrid niosomes were used for a 2D-SERS mapping of malignant MCF-7 breast cancer cells, hereby investigating the SERS-activity and signal integrity of the hybrid niosomes as well as con rming a successful detection according to national and international guidelines, such as the German standard method DIN 32645. Consequently, the hybrid niosomes revealed highly advantageous properties, combining excellent chemical stability, enhanced targeting functionality and remarkable SERS signal intensities, leading to great potential of the AuNPs/Fe x O y NPs/NIO/Tf as a sensitive SERS-based cancer detection platform technology.
Results And Discussion

Raman properties of the AuNPs
In general, a reproducible application of SERS can only be realised by plasmonic nanoparticles with homogeneous particle properties. Therefore, an in-depth characterization of the AuNP plasmonic properties is essential. To investigate the in uence of ripening time and core colloid size on the SERS activity, the synthesis of four nal AuNP solutions was carried out using two core colloids with a size difference of 4.5 nm (Figure S1 a) after speci c core colloid ripening times (10 and 42 days as well as 1 and 51 days, respectively). Representative TEM images of the four synthesized AuNP samples con rm equally sized particles of approx. 38-40 nm with no morphological variances (Figure 2 a).
All TNB-functionalized AuNPs were measured analogously via SERS by adding HCl as stated in section 4 (characterization methods) to characterize the Raman enhancing ability. (nitro scissoring vibration), 1067 cm -1 (succinimidyl stretch overlapping with aromatic ring modes), 1341 cm -1 (symmetric stretch of the nitro group) and 1559 cm -1 (aromatic ring stretching modes) [66]. For further SERS intensity comparisons between the obtained TNB spectra, the intensity of the dominating nitro group stretch signal mode at 1341 cm -1 was used and referred to as the "maximum Raman intensity".
After different nal colloid ripening times (up to 120 days), SERS-spectra were measured (Figure 2 c). For all AuNPs, a longer nal colloid ripening time results in a higher maximum Raman intensity. By comparing the AuNPs synthesized from the core colloids with the same size but different core colloid ripening times, a superior maximum Raman intensity after identical nal colloid ripening times as a consequence of a longer core colloid ripening time becomes obvious. Apart from the ripening times, the core colloid particle size shows strong impact on the Raman intensity: AuNPs with a similar core colloid ripening time (AuNP 1.1 and AuNP 2.1, AuNP 1.2 and AuNP 2.2) still differ in their maximum Raman intensity at identical nal colloid ripening times. The fact that larger core colloids lead to increased Raman intensities can be attributed to the high crystallinity of the core particles, which predominantly contribute to the Raman enhancing effect in contrast to the partly amorphous AuNP shell [62]. Since the limit of detection (LOD) and limit of quanti cation (LOQ) crucially depend on the signal-to-noise ratio, these results emphasise that larger core colloids with longer core and nal colloid ripening times are essential to achieve higher Raman intensities and consequently to gain more sensitivity. For the following experiments, AuNPs which reached a maximum Raman intensity of approx. 2200 s -1 were used.

Synthesis of magnetic and SERS-active niosomes
The concurrent encapsulation of the TNB-functionalized AuNPs (in the presence or absence of HCl while AuNP preparation) and Fe x O y NPs was performed following the procedures shown in Figure 1. To verify the Raman enhancing ability of the encapsulated AuNPs as well as the successful magnetic separation process, SERS-spectra were recorded before magnetic puri cation (red spectra) and after removing the supernatant (yellow spectra) from the magnetically separated product (blue spectra) ( Figure 3).
For the hybrid niosome samples, a distinct Raman intensity after the magnetic separation procedure was measured, indicating the presence of TNB-functionalized AuNPs. Comparing the sample with the respective negative control (blue spectra Figure 1   s -1 ) for the hybrid niosome sample and negative control (blue spectra Figure 1 c with d) was obtained. In contrast to the samples without HCl, the substantially higher Raman intensity of the HCl-treated AuNP sample implicates the successful encapsulation of more AuNPs due to the acid-induced reduction of the AuNP surface charge, hence lowering the repulsion forces and distances from each other throughout the encapsulation process.
Apart from the preceding addition of an ionic substance (e.g. HCl), the Raman intensity can be increased by using higher AuNP amounts in the encapsulation procedure (Figure 4 a). While AuNP amounts below 20 µg result in almost negligible Raman intensities, raised amounts (40 µg, 50 µg and 80 µg) lead to a sharp intensity increase. This disproportionate Raman intensity surge can be attributed to an enlarged amount of encapsulated AuNPs per niosome and thus a larger number of generated interparticle plasmonic hotspots generated. If Figure 4: Measured maximum Raman intensities of a) hybrid niosomes after addition of different AuNP mass fractions and b) hybrid niosomes and negative controls (with a utilized AuNP mass of both 50 µg and 80 µg) in dependence of a previous HCl addition.
the amount of AuNPs is further incremented (for example to 200 µg), only a smaller rise in signal intensity was measured, suggesting a saturation of niosomes with encapsulated gold nanoparticles thus indicating that the average number of AuNP per niosome can hardly be increased. Comparing 50 µg and 80 µg hybrid niosomes samples with added HCl (Figure 4 b), the 80 µg sample shows a maximum Raman intensity either due to an ampli ed amount of encapsulated AuNPs, lower repulsion forces between incorporated AuNPs or both. When using 80 µg AuNPs instead of 50 µg, the difference between the mean intensities of the respective AuNP(HCl)/Fe x O y NPs/NIO to the respective negative control (MIX AuNP(HCl) with Fe x O y /NIO) grows signi cantly. This implies that in contrast to the lower amount, the addition of 80 µg AuNPs before encapsulation process leads to an ampli ed maximum Raman intensity of the resulting hybrid niosomes, whereas the amount of AuNPs, that remains in the solution after magnetic separation and still generates low Raman signals, is almost identical. Furthermore, even after removing a signi cant amount of AuNPs in the magnetic separation procedure, the AuNP (80 µg, HCl)/Fe x O y NPs/NIO surpasses the maximum Raman intensity of a plain 80 µg AuNP dispersion, indicating a successful SERS-active niosome synthesis. Since the maximum Raman intensity of hybrid niosome samples with less than 80 µg AuNPs is quite low and the use of AuNP amounts above 80 µg often resulted in particle agglomeration (data not shown), for all further experiments, unless otherwise stated, an amount of 80 µg AuNPs with previously added HCl was utilized for the hybrid niosomes preparation (in the following simply referred to as AuNPs/Fe x O y NPs/NIO).
Repeated measurements of the hybrid niosomes SERS-activity during a period of 46 days reveal hardly any SERS-signal intensity uctuations over the rst 11 days (Figure S1 b). The increase in signal intensity by approx. 37 % within the following 35 days is probably due to a continuous agglomeration of the entrapped AuNPs. Stable SERS-activity is a good indication for a long-term stability of the hybrid niosome suspension.
Additionally, TEM images were recorded to display the successful encapsulation of the inorganic NPs ( Figure 5). A representative TEM image of TNB-functionalized AuNPs that exhibit a strong contrast to the background can be seen in Figure  contrast AuNPs, smaller low-contrast Fe x O y NPs are located, with both NP types being surrounded by an organic shell structure, thereby forming the hybrid niosome. Most frequently, four AuNPs were found to be incorporated per niosome ( Figure 5 d -f) whereas the observed number varies between one and up to 15 AuNPs ( Figure 5 g -i). It should be noted that the size of the hybrid niosomes in the pictures is in the range of 150-250 nm, which can be attributed to the destruction of the spherical 3D structure due to sample drying and the electron beam focus, leading to the formation of a planar and therefore larger quasi-2D layer.
To get a deeper insight into the SERS activity of AuNPs/Fe x O y NPs/NIO, the resulting electromagnetic eld (E-eld) enhancement, which is created by the electromagnetic radiation of metallic NPs, was calculated using nite-difference time-domain simulations (FDTD)  Figure 7: Measured UV/Vis spectra a) before and after magnetic puri cation obtained from hybrid niosomes and negative controls and b) hybrid niosomes with different AuNPs mass fractions as well as with and without previously added HCl.
Since the majority of the AuNPs in the negative control sample is removed after magnetic puri cation, a AuNP-speci c maximum is hardly visible. The large optical density (OD) discrepancy of the magnetically puri ed dispersions can thus be attributed to the successful AuNP encapsulation and hence raised mass fraction in the solution. Although the AuNPs/Fe x O y NPs/NIO sample after magnetic separation contain a lower AuNP amount than the supernatant sample ( Figure S4 a and b), Figure 3 showed that the exhibited SERS-intensity after magnetic separation is several times higher than the resulting SERS-intensity of the supernatant, which is a clear evidence for encapsulation and thereby induced SERS-activity.
Furthermore, the spectra of AuNPs/Fe x O y NPs/NIO samples with different AuNPs mass fractions were compared (Figure 7 b). Regardless of a previous HCl addition, a higher OD and thus an increased number of encapsulated AuNPs is obtained after induction of a higher AuNP mass fraction within the encapsulation procedure. This relation is in good agreement with the SERS results displayed in Figure 4.
A further analogy to the results of the SERS measurements becomes comprehensible when comparing UV/Vis spectra of the same AuNP mass: adding HCl before the encapsulation process also leads to an elevated OD suggesting a more e cient encapsulation of AuNPs. Moreover, compared to the samples without included HCl, the spectra Figure 8: a) DLS measurements and b) Zeta potentials of the hybrid niosomes and negative control.
of the HCl-treated AuNP samples reveal a more steadily decreasing curve pro le after the maximum at approx. 526 nm, which becomes clearer when comparing the rst derivative of the curves: after the minimum at around 560 nm, the slope of the samples without added HCl increases faster than the slope of the samples with HCl-treated AuNP, which even display smaller minima at approx. 730 nm (Figure S4 c). This can be attributed to the HCl-induced proximity among AuNPs within the hybrid structure and the resulting coupling of their surface plasmon resonance [67,68].
DLS measurements con rm that the average particle size of the hybrid niosomes and negative control after magnetic separation is approximately 120 nm which predominantly indicates the presence of niosomes (Figure 8 a). This hydrodynamic diameter is in the size range of bare Fe x O y NPs/NIO, which is expected to ensure a awless uptake by cancer cells via receptor-mediated endocytosis [69][70][71].
Furthermore, it is shown that the Zeta potential of the AuNPs/Fe x O y NPs/NIO and the negative control are almost equal after magnetic separation and correspond to the value of Fe x O y NPs/NIO, thus proving a successful magnetic separation procedure (Figure 8 b). Moreover, the nal Zeta potential of -35 mV indicates an electrostatically well-stabilized suspension. The Zeta potential of the supernatants resemble the AuNP-speci c Zeta potential due to the abundance of free AuNPs ( Figure S5 a and b).

Administration of hybrid niosomes to complex media
After the successful synthesis, magnetic puri cation and characterization of the AuNPs/Fe x O y NPs/NIO, an investigation of the niosomal shielding effect on the SERS-signal integrity against external in uences (e.g. high osmotic pressure, biological molecules, acidic environment etc.) that usually affect NPs after administration in organisms was realized. For this purpose, three different media were prepared: phosphate-buffered saline (PBS), a saturated bovine serum albumin (BSA) suspension and the cell culture medium (CCM) used to cultivate MCF-7 breast cancer cells. After the addition of the respective medium, there is initially a slight reduction in the SERS signal intensity ("0 days"), which is mainly due to the dilution of the respective sample. If PBS is added to plain TNB-functionalized AuNPs, a strong increase in the SERS-signal intensity is initially recorded due to a reduction of the repulsive forces between the AuNPs, thus resulting in an exalted generation of interparticle SERS-hotspots (Figure 9 a). This is followed by a rapid decrease within the rst hours, which nally leads to a complete and persistent loss of SERS-signal due to particle agglomeration. The hybrid niosomes, on the other hand, show no major uctuations as a result of the niosomal shielding, with a decrease in signal intensity of approximately 35% only after an extended storage. This is probably caused by a partial loss of the structural integrity of some niosomes due to the high osmotic pressure of the PBS suspension. Nevertheless, qualitative identi cation of the symmetrical stretching vibrations of the nitro group is still clearly possible even after 46 days (Figure 9 b). After addition of the BSA solution to plain AuNPs, a steady decrease in signal intensity was observed, followed by a stable signal intensity until the 46 th day (Figure 9 a). In contrast, the signal uctuations of the hybrid niosomes are again negligible throughout the entire measurement period, and an enhanced SERS signal intensity with only a small proportion of signal contamination is seen in the spectrum at longer storage times (Figure 9  These results suggest that the niosomes act as a protective layer against a wide range of different molecules and pH-shifts found in biological environments, thus serve to maintaining the SERS-activity over a long storage period. In particular, the results in CCM show that after administration of the particles in vitro, no signal contamination by the cultivation medium itself and therefore no in uence of the medium on cell imaging can be expected.

In vitro experiments
Since intracellular SERS detection vitally depends on the e cient internalization of the SERS nanoprobes, it is highly important to develop methods that can effectively deliver the SERS-active NPs into the cell.
Considering that a disordered iron metabolism is a hallmark of cancer cells, the Tf-receptor is commonly used as a prominent marker of cancer cells and also identi ed as a therapeutic target [72]. Remarkably, Tf-receptor 1 (CD71) [73,74], but also other iron-regulating and -regulated proteins such as transcription factors [75] are overexpressed in human breast cancer cells. Therefore, apart from a successful iron-oxide based magnetic targeting with an external magnetic eld [63,76], a second, cell-speci c targeting functionality was necessary. As the noisomes are to be used to label breast cancer cells, the Tf-receptor system is a preferable route for internalization of niosomes. Thus, Tf was coupled to the maleimide groups of the hybrid niosomes. The successful Tf-functionalization of the hybrid niosomes was investigated using Zeta potential measurements, which clearly showed a decrease in the absolute value of the Zeta potential due to the coupling of positively charged Tf ( Figure S6) similar to the results reported in other works [51,77].
Since the synthesized hybrid niosomes will be primarily used for the detection of living breast cancer cells, particularly uncontrolled necrotic signal cascades should be avoided. Hence, Figure 10: Viability of MCF-7 cells after in vitro administration of hybrid niosomes with or without Tf-functionalization and incubation under in uence of an external magnetic eld.
cytotoxicity studies were carried out using MCF-7 breast cancer cells ( Figure 10). From the results of our experiments, we conclude that a hybrid niosomes volume of at least 14 µL per 10k MCF-7 cells leads to no signi cant reduction in viability and therefore can be assumed as a safe volume for in vitro administration. However, for hybrid niosome volumes larger than 14 µL, surface modi cation with Tf leads to slightly reduced viability. This interference can be attributed to an increased uptake of the functionalized hybrid niosomes due to a successful Tf-receptor-mediated endocytosis.
In vitro 2D SERS images clearly demonstrate that the dyadic targeting composed of a magnetic manipulation as well as Tf-based endocytosis resulted in the accumulation of AuNPs/Fe x O y NPs/NIO in MCF-7 cells after 24 hours of incubation ( Figure 11). Strong localized TNB-signals were detected, possibly indicating an accumulation in speci c cell compartments, as well as a non-speci c accretion throughout the cytosol of the cell. The oscillation of the TNB nitro group in the in vitro SERS-spectra is located at approx. 1341 cm -1 as in the TNB-reference spectra. Apart from a minor background noise, no signal contamination can be detected, which indicates the structural integrity of the hybrid niosomes. The calculated signal-to-noise ratios (SNR) are always clearly above the prescribed SNR minimum of 9.0 according to national and international guidelines, such as DIN 32645. In particular, the highest occurring SNR in the center of each region of interest is 3-4 times higher than the minimum requirement of 9, implicating a veri ed SNR as speci ed by the DIN standard. In vitro tests with HeLa cells also show an analogous outcome ( Figure S7). These results prove the shielding effect of the niosomal protective layer and thus the great potential of the synthesized hybrid niosome system to be used for analytically labelling Tfoverexpressing cancer cells such as breast carcinoma cells.

Conclusion
In this study, we presented the rst successful preparation and in vitro application of superparamagnetic and SERS-active niosomes. The synthesized AuNPs/Fe To validate SERS-based imaging and monitoring of tumorous tissues for ex vivo diagnosis or even for staining tumor cells in vivo using the presented hybrid niosomes, extensive and elaborate in vivo experiments are mandatory. Since we chose TNB as the Raman-active compound, the hybrid niosomes possess the inherent capability to establish a quanti cation principle using the isotope dilution (ID) technique, which has already been used in combination with SERS [78][79][80]. Therefore, apart from ex vivo quanti cation procedures to investigate biomarker concentrations in human uids (blood, cerebrospinal uid etc.), an in vivo ID-SERS approach could aim to establish a correlation between the isotope-based Raman-shift and 2D-SERS measurements to draw conclusions about the volume and shape of the tumor. In essence, the results from this study bear great potential to establish a novel and highly sensitive imaging technique which allows monitoring and prospectively targeting of smallest cancerous breast tissues.

Chemicals and reagents
Benzyl

Synthesis of Fe x O y NPs
The solvothermal synthesis of the Fe x O y NPs and the particle properties were stated in our previous publication [63]. In summary, oleic acid (3 mmol) and oleylamine (3 mmol), Fe(acac) 3 (1 mmol) and 1,2hexadecanediol (5 mmol) were added to a ask containing benzyl ether (10 mL) and heated to 100 °C for 30 min. Under nitrogen atmosphere, the mixture was rst re uxed at 200 °C and afterwards at 265 °C for 30 min each. The solution was cooled down to room temperature, followed by three washing steps with ethanol. After puri cation, the black precipitates were dispersed in chloroform. The notation Fe x O y is used in this report since the synthesized iron oxide NPs consist of both the magnetite Fe 3 O 4 and maghemite γ-

Synthesis and TNB-functionalization of AuNPs
The synthesis procedure and a comprehensive characterization regarding the AuNP synthesis, TNBfunctionalization and resulting particle properties (crystallinity, size, morphology etc.) was presented in our previous publication [62]. In short, the preparation of the core colloids was accomplished by re uxing a gold-(III) chloride solution (206 mL) while heating at 100 °C for 30 min and for another 15 min after addition of 50 mL citric acid solution. The synthesis of the nal colloids was carried out by appending a sodium-(L) ascorbate solution (233 mL) and a gold-(III) chloride solution (233 mL) to a ask containing a core colloid solution (12 mL). Subsequently, after keeping the nal solution at 30 °C for 1 h, the particles were puri ed via centrifuging and replacing the supernatant with HEPES buffer. By following this protocol, two different core colloids (AuNP 1 as well as AuNP 2) were synthesized. After speci c ripening times, each core colloid was used to synthesize two nal colloids (AuNP 1.1 and AuNP 1.2 as well as AuNP 2.1 and AuNP 2.2), respectively. The AuNP dispersions had a solid content of approx. 100 µg/mL (determined gravimetrically). The TNB-functionalization of AuNPs was performed by adding 10 µL of DTNB solution per 50 µg/mL of AuNP dispersion. After a 30 min incubation, the particles were puri ed twice by centrifugation and replacing the supernatant with HEPES buffer.

Hybrid niosome synthesis via niosomal incorporation of Fe x O y NPs and AuNPs
The synthesis of PEG-maleimide-functionalized niosomes was based on the thin-lm hydration method according to the protocol stated by Ag Seleci et al. [58].  [82,83]. For the light source, a continuous plane wave was utilized with the wavelength set to 633 nm. One single Yee cell was sized as 0.1 nm × 0.1 nm, whereas the remaining boundary conditions were PML. The E-eld amplitude was output at all points in the area of interest. A calculation of the total E-eld enhancement in one Yee cell was achieved by dividing the absolute square of the E-eld amplitude with the squared amplitude of the incident E-eld in the absence of the structure.

Preparation of Tf-conjugated hybrid niosomes
To prepare AuNPs/Fe x O y NPs/NIO/Tf, Traut's reagent (50 nmol) and Tf (10 nmol) were suspended in HEPES EDTA buffer (500 μL) at pH 8.5 (30 mM HEPES and 0.1 mM EDTA). The thiolation of Tf was completed after incubating for 2 h at room temperature. In addition, using a size exclusion chromatography (30 kDa, Sephadex G50, Cytiva), the thiolated Tf was washed with HEPES (pH 6.5) and concentrated to a nal volume of 200 µL. This was followed by an immediate application of hybrid niosome suspension (750 µL). A subsequent incubation for 24 h at 4 °C resulted in the formation of a thioether linkage [84] and the binding of the Tf-speci c thiol groups to the maleimide moieties of the niosome PEG chains.

SERS-measurements in different media
The SERS-signal uctuations of the AuNPs/Fe x O y NPs/NIO/Tf hybrid NPs and a negative control (plain TNB-functionalized AuNPs with the same AuNP mass content) in PBS (0.04 M), BSA (40 mg/mL) and MCF-7 cell culture medium (volumetric mixing ratio of NPs suspension : added medium was 2.5:1 for all samples) were examined over a period of 46 days. For a su cient comparability of the samples, the initial SERS signal intensities for both the hybrid niosomes and negative control were normalized.

In vitro experiments
For the in vitro experiments, the cells were seeded on TPP® 96-well plates (Sarstedt, for cytotoxicity studies) or Poly-L-Lysine-(Sigma-Aldrich, for 2D-SERS measurements) coated cover glasses and cultured for 24 h in DMEM (Thermo Fisher Scienti c, Waltham, MA, USA) supplemented with 10% FCS (Thermo Fisher Scienti c), streptomycin, and penicillin (each 100 µg per ml, Thermo Fisher Scienti c) at 37 °C in a water saturated atmosphere containing 5% CO 2 . To assess the metabolic activity of MCF-7 cells after AuNPs/Fe x O y NPs/NIO/Tf administration, a CellTiter-Blue® cell viability assay was used to measure the reduction capacity of Resazurin (Promega Corp., Madison, WI, USA). 10 4 cells were seeded per well with 200 µL added cultivation medium and cultivated for 48 h. Afterwards, the cells were exposed with different concentrations of bare or Tf-loaded hybrid niosomes, incubated for 15 min by attaching a 1.3 T permanent neodymium magnet to achieve a magnetic guidance and subsequently incubated for 24 h. After diluting the CellTiter-Blue® reagent 1:10 with supplement-free MEM medium, 200 µl of the suspension was added to each well and incubated at 37 °C and 5 % CO 2 for 4 h. A uorescence spectrometer (GENios Microplate Reader, Tecan) was used to measure the uorescence (530 nm excitation/590 nm emission). For the 2D-SERS measurements, the AuNPs/Fe x O y NPs/NIO/Tf were added in a non-toxic concentration (14 µl). The cells were incubated under the in uence of a magnetic eld (1.3 T permanent neodymium magnet) for 15 minutes. Subsequently the cells were grown in a water saturated atmosphere containing 5% CO 2 for 24 h at 37 °C. Afterwards cells were xed with 4 % paraformaldehyde (PFA, Carl Roth) at room temperature (21 °C) for 20 min and subsequently washed twice with PBS and mounted with Shandon Immu-Mount (9990402 Epredia) for following SERS measurements.

Characterization methods
The DLS and Zeta potential measurements were accomplished using a Zetasizer Nano ZS from Malvern Instruments. To minimize uorescence, the samples were diluted by a factor of 10 4 -10 5 and subsequently measured with a 173° backscattering set-up thrice. Using the Zetasizer Nano software, the data was evaluated utilizing the intensity distributions to assess the hydrodynamic diameters. The Zeta potentials were attained using a capillary zeta cuvette (DTS1070C, Malvern Panalytical Ltd). The pH-values of samples were acquired with the pH-meter Seven Compact from Mettler Toledo. To obtain SERS spectra, a LabRAM Aramis Raman microscope equipped with a 600 grooves/mm holographic grating and a HeNe excitation laser (633 nm) in combination with the LabSpec 5 Software from Horiba was used. Each sample was irradiated with an exposure time of 200 · 1 s with a spot size around 550 nm. Prior to each acquisition, the spectral line position was calibrated according to the recommendations given in the ASTM guideline using the polystyrene ring breathing mode at 1001.4 cm −1 [85]. The SERS-spectra of plain AuNP solutions (in Section 3.1) were attained after adjusting their OD to 1 and adding HCl (4 µL, 1 M) to 250 µL of AuNP solution. For the 2D-SERS mappings, a 20x objective lens was used with an average image acquisition time of approx. 8 h. UV/Vis spectra were acquired using a LAMBDA 1050 dualbeam photometer from Perkin Elmer. The measurement was carried out in a Quartz Suprasil semi-micro cuvette (layer thickness 1.0 cm) from Hellma Analytics. TEM images were taken with a Tecnai G2 F20 TMP from FEI at 200 kV. For sample preparation, 20 µl of the solution were mixed with a staining solution (10 µl) containing 2 % aqueous phosphotungstic acid and applied onto a carbon lm on a 3.05 mm woven copper net with 300 mesh from Plano GmbH. Particle sizes were determined by measuring 50 nanoparticles, using the ImageJ software (Version 1.42q).        Measured UV/Vis spectra a) before and after magnetic puri cation obtained from hybrid niosomes and negative controls and b) hybrid niosomes with different AuNPs mass fractions as well as with and without previously added HCl.