High resolution imaging and 3D analysis of Ag nanoparticles in cells with ToF-SIMS and delayed extraction Open Access

Beside conventional applied methods such as light- and electron-microscopy, mass spectrometry imaging becomes increasingly important for cellular analyses as specific and untargeted chemical information is obtained as part of the image.^1–3 Compared to desorption electrospray ionization and matrix-assisted laser desorption/ionization (MALDI), time of flight secondary ion mass spectrometry (ToF-SIMS) provides the best lateral resolution, and in contrast to MALDI,⁴ real 3D information is accessible. The lateral resolution of ToF-SIMS is only exceeded by NanoSIMS, a special SIMS instrument, which works with a magnetic sector mass analyzer, applies oxygen or cesium primary ion beams (not pulsed), and offers high sensitivity as well as nanoscale resolution better than 50 nm.⁵ NanoSIMS is suitable for isotopic and trace element analysis but is also increasingly applied in life sciences.^6–8 Specifically in combination with transmission electron microscopy (TEM), NanoSIMS can provide very precise and detailed information on cell morphology and subcellular processes and, if combined with isotopic labeling, on the localization of different molecules within the cell compartments.^9,10 However, NanoSIMS analyses are limited by the number of simultaneously detectable elemental masses, time consumption, and the need for extensive sample preparation. Therefore, ToF-SIMS is a more versatile technique, providing a broad mass range and the possibility of 3D analysis.

Nevertheless, using common measurement modes, ToF-SIMS provides either a submicrometer lateral resolution or a mass resolution of several thousands.^11,12 However, the delayed extraction of the secondary ions enables the combination of high lateral and high mass resolutions. A lateral resolution of 400 nm and a mass resolution of 10 000 at m/z 385.4 using delayed extraction were shown by Vanbellingen et al. on tissue sections.¹¹ The principle of delayed extraction is commonly used in the MALDI community to reduce the uncontrolled decay of molecular ions in the extraction phase.^13–15 Collision and fragmentation of the generated ions in the field free space are minimized, and as an additional benefit, the mass resolution increases.¹⁶ Therefore, the principle of delayed extraction is also a promising strategy for ToF-SIMS analysis, preferable for instruments equipped with a reflectron analyzer. Ion extraction is decoupled from the ion generation process by switching off the extraction voltage for several hundred ns after the primary ion pulse. A field free emission of secondary particles in a plume is obtained, and the long primary ion pulse, commonly required for high lateral resolution imaging, does not restrict the mass resolution anymore. In addition, Lee et al. described the positive effect of delayed extraction for the analysis of conductive samples by topography.¹⁷ The field free emission allows secondary ions from the side of the substrate to drift away from the sample surface before the extraction voltage is switched on. Therefore, the extraction is less influenced by topographic field effects and the amount of collectable ions from the side is increased; less shadow effects appear, and sharper images are obtained. However, in general, a significant reduction of detectable secondary ions has to be considered, as only ions directly located below the analyzer entrance are collected and fast ions of lower masses can leave the extraction zone within the field free delay time.^11,17,18 To conclude, ToF-SIMS combined with delayed extraction is a very promising candidate for subcellular analyses with a high lateral and mass resolution.¹⁹ 

In 2007, Fletcher et al. published the first example for 3D imaging of a single cell using ToF-SIMS with C₆₀⁺ as ion beam.²⁰ Mass peaks of cholesterol, other lipids, and fatty acids were used for the first three dimensional imaging of an oocyte. Soon after, Breitenstein et al. used a more focused Bi₃⁺ ion beam for analysis and C₆₀⁺ for sputtering animal cells.²¹ Different signals were gained to identify and reconstruct the nucleus and other compartments of the cell. One year later, Vaidyanathan et al. were able to localize pigmented antibiotics in bacteria using C₆₀⁺.²² The detection of pharmaceutical compounds and their metabolites in single cells using ToF-SIMS without delayed extraction was achieved by Passarelli et al. in 2015.²³ Using Ar₂₀₀₀⁺ for sputtering, an unlabeled drug compound was detected and localized in the membrane and subsurface region of the cell. 3D images of the drug metabolite and the cell regions were generated. Besides, the localization of nanoparticles in cells and tissue is the subject of several studies.^24–31 In 2014, Angerer and Fletcher detected TiO₂ nanoparticles in single cells as particles of about 5 μm in diameter distributed throughout the cells.²⁴ 3D reconstruction is reported to be challenging due to differences in the sputter rates between natural organic and TiO₂-rich regions. Beyond that, in 2015, Kokesch-Himmelreich et al. showed for the first time that it is also possible to detect organic nanoparticles in cells by ToF-SIMS.³¹ In 2017, Veith et al. used ToF-SIMS for direct detection of SiO₂ nanomaterials in rat lung tissue.³⁰ The same nanoparticle distribution was observed using fluorescence microscopy. Therefore, labels are not mandatory to detect nanomaterials in tissue. This direct detection makes ToF-SIMS even more attractive for nanotoxicological studies. In the same year, functionalized gold nanoparticles were visualized in single cells by Bloom et al. using C₆₀⁺ cluster-SIMS and information on the drug pathway and mechanism was obtained.²⁵ The functionalization compound and the gold nanoparticles did not appear in the same cell regions, and so, cleavage after cell uptake was proposed. In 2016, Nees et al. investigated the effects of silver nanoparticles (AgNPs) on dog kidney cells, using laser postionization secondary neutral mass spectrometry (laser-SNMS) and ToF-SIMS in combination with Ar-cluster ion sputtering and Bi₃⁺ primary ions for analysis.²⁷ Laser-SNMS was used due to the reported low ionization efficiency of silver. The Ar-cluster ion beam allowed a high sample removal rate in combination with low molecular degradation. AgNPs were mainly detected close to the cell nucleus but were not removed during the abrasion of the cell due to different sputter rates of organic and inorganic compounds.

AgNPs are widely used due to their antibacterial effect, for example, in wound dressings, catheters, and antibacterial coatings for implants to prevent infections.³² Nevertheless, their effect on the human body is still not completely understood. Exact localization of the AgNPs in single cells after in vitro exposition is one step to understand their impact on the human body.^33–35 To estimate possible health risks through AgNP in coatings of implants, the influence of AgNP on cell stress, viability, proliferation, and differentiation of primary human mesenchymal stem cells (hMSCs) was examined in two studies in 2014 by Pauksch et al.^28,29 Below the cytotoxic concentration, no influence on cell differentiation but an increase in cell stress was observed. In addition, AgNP accumulation in the cells was detected using ToF-SIMS.

The objective of the present study is to use the delayed extraction mode for analyses at the cellular level and to determine the benefits and limitations of this mode. We use hMSCs incubated with AgNP as a model system and intend a more accurate visualization of the AgNP within the cells in 2D and 3D. Therefore, we optimize the delayed extraction mode for cellular analyses and compare the achieved resolution as well as the measurement conditions with the spectrometry and imaging mode. Additionally, we perform 3D analyses with delayed extraction and use Ar-clusters and O₂⁺ to remove the cells with the aim of a detailed 3D visualization of the AgNP inside the cells.

Two thousand hMSCs per cm² were grown on silicon wafers with (111) orientation and antimony as a dopant (CrysTec Kristalltechnologie, Germany) in the presence of 12 μg of AgNP/g growth medium at 37 °C and 6% CO₂ for 4 days. Before the incubation, the hMSCs were kept under the same conditions in growth medium without AgNP for one day. As controls, hMSCs were grown without any AgNP incubation. The growth medium was composed of MesenPro RS (Gibco) with 20% PANSera ES (Pan Biotech), 1% GlutaMAX (Gibco), and 0.2% Genta/Ampho (Life Technologies). The AgNPs were provided by ras materials GmbH (Regensburg, Germany) as a colloidal dispersion with a silver content of 10 w/w% in a stabilizing matrix containing 4% polyoxyethylene glycerol trioleate and 4% polyoxyethylene sorbitan monolaurate (Tween20). AgNPs have been characterized in earlier studies, using ultraviolet-visible spectroscopy, dynamic light scattering, and zeta-potential measurements.^28,29 Depending on the cell medium and incubation time, a hydrodynamic diameter from 5 to >100 nm was measured.²⁸ Cells were chemically fixed for 1 h at 4 °C with 2% paraformaldehyde and 1% glutaraldehyde in 0.1 M sodium phosphate buffer.³⁶ Fixed cells were stored in Milli-Q water at 4 °C to enable a measurement period of at least four weeks.³⁷ The cells were air dried prior to confocal and ToF-SIMS analyses.

The morphology and dimensions of the cells were determined using an S neox optical profiler confocal microscope (SENSOFAR, Spain). 3D measurements were carried out in 100 nm steps and with 50-times magnification. Data analysis was performed using the respective software “sensoscan 6.3,” where maximum heights of the cells were measured and 3D images were created.

For TEM examination, the hMSCs were also incubated with 12 μg of AgNP/g growth medium for 4 days. First, cells were fixed chemically for 30 min on ice with 2% paraformaldehyde in 0.1 M sodium phosphate buffer (pH 7.2–7.4) with 2% glutardialdehyde and 0.02% picric acid, followed by a 20 min fixation with 1% osmium tetroxide in 0.1 M sodium cacodylate buffer (pH 7.2–7.4). Afterward, the samples were dehydrated and embedded in Epon. For the examination, ultrathin sections (80–100 nm) were mounted on collodion-coated copper grids. Analysis was done using a Leo 912 transmission electron microscope (Carl Zeiss AG, Oberkochen, Germany) equipped with a TRS Sharpeye slow scan dual speed CCD camera (Albert Troendle Prototypentwicklung, Moorenweis, Germany).

The ToF-SIMS measurements were performed using a TOF.SIMS 5–100 instrument (ION-TOF GmbH, Muenster, Germany) equipped with a 25 kV Bi-cluster primary ion gun, with a 20 kV gas cluster ion beam (GCIB) and a dual source column for sputtering. All analyses were carried out with Bi₃⁺ primary ions, in the positive ion mode and with a cycle time of 65 or 100 μs. Sputtering was done with 10 keV Ar-clusters of different sizes (Ar_n⁺) and 500 eV oxygen-ions (O₂⁺). Measurements with Ar_n⁺ were performed without a flood gun, whereas for sputtering with O₂⁺, the flood gun was used. The crater size was always 150 μm larger than the analysis field. Data analysis was performed with the “surfacelab 6.7” software (ION-TOF GmbH) using the signals CH₃⁺ (15.02 u), Na⁺ (22.99 u), Si⁺ (27.97 u), K⁺ (38.96 u), C₃H₅⁺ (41.04 u), C₃H₇⁺ (43.05 u), and C₅H₁₂N⁺ (86.10 u) for internal mass calibration. As silver signals, the ¹⁰⁷Ag⁺ (106.91 u) peak and the ¹⁰⁹Ag⁺ (108.91 u) peak were examined.

For the 2D measurements of cells with a high mass and lateral resolution, the delayed extraction mode was obtained after starting with the existing setting for the fast imaging mode (burst alignment mode). The initial imaging mode works with a well-focused, unbunched primary ion beam and long pulses of 100 ns, which results in a lateral resolution of about 400 nm, but at the expense of less detectable ions and only the nominal mass resolution (m/Δm of several hundreds). The parameters used for the optimization of the delayed extraction mode were the delay time, voltage of the analyzer lens, the X/Y analyzer deflection plates, as well as the surface and virtual drift potential (VDP) (cf. Fig. 1). The well–known cell signal at m/z = 184.08, the phosphatidylcholine head group,³⁸ was optimized to gain maximum counts and high mass resolution by aligning the above–mentioned parameters.

For each measurement of a single cell, the parameters of delay time, lens voltage, and potentials had to be adjusted slightly. The delay time varied in a range of 40 ns. The lens voltage was set between 2790 and 3025 V and the VDP between 128.2 and 158.2 V. The settings for the X analyzer deflection plate varied in the range of 1.0%–3.0% and for the Y analyzer deflection plate between 16.8% and 19.4%.

To compare the resolution of the delayed extraction mode with common measurement modes, 2D images were taken in the spectrometry (low-current bunched) and fast imaging (burst-alignment) mode. All depicted images were recorded from the same cell with a field of view of 230 × 230 μm² and in the random scanning mode. Before the first measurement, the cell was cleaned by sputtering with Ar₁₅₀₀⁺ (10 keV, GCIB dose of 5.76 × 10¹³ ions/cm²). For the delayed extraction mode, images with 1024 × 1024 pixel and a primary ion dose density of 1.01 × 10¹² ions/cm² were recorded. In the imaging mode, the same pixel and primary ion dose density of 1.02 × 10¹² ions/cm² were employed. In the spectrometry mode (width =1 ns), images with 256 × 256 pixel and a primary ion dose of 1.16 × 10¹² ions/cm² were acquired. The achieved lateral resolution was determined five times by a line scan on the total ion image for each measurement mode. Therefore, small features having interfaces with high intensity contrast were chosen and the average of the results was taken. Mass resolving power was defined by the full width at half maximum (FWHM) definition for the signals C₄H₈N⁺ (m/z 70.07), C₅H₁₂N⁺ (m/z 86.10), and C₅H₁₅NPO₄⁺ (m/z 184.08). Besides, the counts of the same signals were evaluated for a defined cell region of 30 × 30 μm².

For 3D reconstruction, cells were depth profiled with Ar₁₅₀₀⁺ (10 keV) in the interlaced mode. For the depicted control cell [cf. Figs. 3, 4(a), and 4(b)], cultivated without AgNP, a total GCIB dose of 6.27 × 10¹⁵ ions/cm² (2.41 × 10¹⁴ ions/cm² per cycle) was needed until the cell was fully ablated. For another cell, cultivated with AgNP [cf. Figs. 6, 7(a), and 7(b)], a total GCIB dose of 5.58 × 10¹⁵ ions/cm² (2.53 × 10¹⁴ ions/cm² per cycle) was applied. In this case, 23 sputter cycles were necessary to remove the cell. Analyses were performed in the delayed extraction mode with the following parameters: 1024 × 1024 pixel, 10 shots/pixel, and three frames (the region was scanned three times prior to the next sputter interval). The cells were reconstructed in 3D by unfolding the structure using the Si⁺-signal. Additionally, depth analysis and 3D reconstruction using O₂⁺ (500 eV) as sputter species, with a total dose of 6.56 × 10¹⁷ ions/cm² (3.30 × 10¹⁵ ions/cm² per cycle), were carried out for another cell, cultivated with AgNP [cf. Figs. 4(c) and 4(d)]. Two hundred scans were performed in the spectrometry mode with the following parameters: 128 × 128 pixel, 1 shots/pixel, and 20 frames (the region was scanned 20 times prior to the next sputter interval). To evaluate the “healing effect” of Ar_n⁺-clusters for organic signal intensities, a cell, first ablated with O₂⁺, was depth profiled with Ar₁₅₀₀⁺ (10 keV). The intensities of the signals C₄H₈N⁺ and C₅H₁₂N⁺ were evaluated for each scan in a defined cell region.

The sputter yields of silver with Ar_n⁺-clusters of different sizes were determined. For this, an approximately 150 nm thick silver layer was deposited on a silicon wafer using pulsed laser deposition. Ten thousand pulses of a KrF-Excimer laser (248 nm, 20 Hz, 8 J/cm²) were used at room temperature with a substrate-target distance of 5 cm. The silver layer was sputtered with Ar_n⁺ until an increasing Si⁺-signal was detected. Cluster sizes of n = 100, 500, 1000, 1250, 1500, 1750, and 2000 were tested. For each cluster size, at least four measurements were performed and the resulting crater depths were determined by confocal microscopy.

The best lateral resolution achieved was 360 nm with the imaging mode and delayed extraction as shown in Fig. 2. This value corresponds well to the reported 400 nm of Alain Brunelle's group on tissue sections of rat cerebellum.¹¹ 

The spectrometry mode was applied to gain a high mass resolution. With low current, an improved lateral resolution of 3.6 μm compared to 10 μm in the high current (HC) mode was achieved. Therefore, the width of the primary ion pulse was reduced to 1 ns at the expense of significant current and thus intensity loss. The fast imaging mode (burst-alignment mode) was optimized for high lateral resolution.

As shown in Fig. 2, the delayed extraction mode combines the high lateral resolution of the imaging mode with the high mass resolution of the spectrometry mode. Unfortunately, the peak shape with delayed extraction is not as good as the peak obtained in the spectrometry mode. The observable peak tail is an effect of delay time as described by Vanbellingen et al.¹¹ The delay time was optimized for maximum counts and mass resolution at the expense of peak shape. Therefore, the tail is a consequence of a too long delay after the primary ion pulse and reflects the energy dispersion by a longer flight time. However, all mass spectra recorded with delayed extraction showed no masses below m/z = 20, as light mass ions escape from the extraction region during the delay time. Therefore, only ions of m/z > 20 were used for mass calibration, and in doubtful cases, peak identification was done by the help of spectra recorded in the spectrometry mode. However, an internal calibration standard would be helpful, especially in the case of totally unknown sample systems.

Furthermore, it has to be noted that the optimization of the extraction delay is related to maximum counts, mass resolution, and improved peak shape but strongly depends on the mass of interest.^11,17,18 Probably due to the choice of lower m/z-values for optimization and due to topographic effects, we only attained a mass resolving power of about m/Δm ≅ 4700 at m/z = 184.08, as shown in Table I, significantly less than Vanbellingen et al., who reported a value of 10 000 at m/z = 385.4 for tissue sections.¹¹ 

Nevertheless, the disadvantage of all imaging modes is the loss of intensity compared to the HC spectrometry mode working with several dozen picoamperes of primary ion current. Applying the delayed extraction, the yield of the detected secondary ions is further decreased. The extent of decrease strongly depends on the mass of interest and the corresponding applied extraction delay, which is, especially for samples with topography, always a compromise between the mass resolution, peak intensity, peak shape, and influence of topography.^17,18 According to Lee et al., the use of delayed extraction can improve the analysis of samples with topography significantly.¹⁷ Also, secondary ions from the side are detectable, and less shadow effects appear. Consequently, Lee et al. stated a less reduction of the total ion yield for samples with topography compared to totally flat samples. It was further shown that also conductive or insulating properties of the samples have to be considered.¹⁸ The positive effect of less reduced total ions is not observable for insulating samples. This is in line with our measurements. We observed a reduction of the total ion yield from 1.69 × 10⁶ to 5.83 × 10⁵ that even exceeds the decrease of 50% reported by Vanbellingen et al. on flat tissue sections.¹¹ In contrast, for typical cell signals, we attained improved signal intensities as shown in Table I. Specifically, the phosphatidyl head group at m/z = 184.08 showed a significant higher intensity than without delayed extraction, as this mass was used to find the optimum of delay time with respect to maximum counts. Obviously, higher primary ion dose and longer measurement time are needed to obtain images that achieve the same intensity scale with delayed extraction as images recorded in the HC spectrometry mode, as reported by Vanbellingen et al.¹¹ However, the use of delayed extraction simplifies cellular analyses, providing high mass resolution and sharp images.

In order to take advantage of delayed extraction for 3D investigations of cells, we used Ar_n⁺-clusters (10 keV) to remove the cells in layers and analyzed with delayed extraction in-between.

To visualize the cellular structure, we chose several mass peaks like C₄H₈N⁺, C₅H₁₂N⁺, and C₅H₁₄N⁺, which are well known as amino acid and head-group related fragments^38–41 and showed good quality images. The cell nucleus could be visualized nicely with C₂H₈N⁺ (m/z = 46.06), a possible amino acid fragment, but not assigned and mentioned in the literature so far. In contrast, the mass peak at m/z = 136.1, adenine, used by Fletcher et al. did not show any representative nucleus structure in our measurements.

As Ar_n⁺-clusters are reported to be suitable for sputtering organic materials and do not cause much fragmentation,^27,42–45 the intensity of the cellular signals remained stable at a high level, until the cell was completely ablated as shown in Figs. 3 and 4(b).

The fast imaging mode with delayed extraction allowed the combination of nicely shaped reconstructions and clear identification of signals from different cell compartments as shown in Fig. 4(b). The associated confocal microscopy 3D image in Fig. 4(a) reveals the cellular dimensions. The cell nucleus could be visualized quite nicely by C₂H₈N⁺ (m/z = 46.06). Attributed to the different sputter rates of nucleus and other cell compartments, the cell could not be reconstructed realistically along the z-axis as shown in Fig. 4(b).

The intracellular nanoparticles could not be removed during Ar_n⁺-sputtering, which was expected and will be discussed later on. In contrast, O₂⁺ is reported to be more suitable for the removal of inorganic compounds than Ar_n⁺-clusters. Figure 5(b) presents the depth profiles of cells incubated with AgNP and sputtered with Ar₁₅₀₀⁺ (10 keV) and O₂⁺ (500 eV), respectively. Although the intensities are not comparable due to different measurement parameters, it is demonstrated that silver is not removed during sputtering with Ar_n⁺-clusters. Similar observations were made on nanoparticles in cells in other studies.^24,27 The signal intensity of ¹⁰⁷Ag⁺ increases with cell removal, which supports our assumption that the AgNP persists and accumulates on the substrate when the cell is removed.

With O₂⁺ (500 eV) as sputter species, the intensity of the ¹⁰⁷Ag⁺ signal decreases when the silicon substrate is reached, which accounts for a notable abrasion of the inorganic particles. However, O₂⁺ (500 eV) sputtering of the cell leads to a dramatic drop of the organic signal intensities right after the first sputter cycle. Only the shape of the cell, but no inner compartments, could be visualized [cf. Fig. 4(d)].

Therefore, we considered a combination of both the sputter species, O₂⁺ (500 eV) and Ar₁₅₀₀⁺(10 keV), in order to benefit from both: An improved ablation rate of AgNP and less cellular damage. Due to the healing effect of Ar_n⁺-clusters, we expected an enhancement of the signal intensities of the organic cell fragments after O₂⁺ (500 eV) sputtering, according to Mihara et al., who used Ar-clusters to clean a cross-section of polymer multilayer films from the redeposited material.⁴⁷ Besides, Miyayama et al. reported that Ar_n⁺-clusters are useful to remove the primary ion beam-induced damage of a polyimide film.⁴⁵ We also observed an increase in the signal intensities during Ar_n⁺-cluster bombardment, but the recovery of the signals took a sputter dose of about 6.4 × 10¹⁴ ions/cm² Ar₁₅₀₀⁺-clusters (10 keV), which is equivalent to a cell-abrasion of at least 40–50 nm. Taking into account that the mean height of the cells was about 500 nm, this approach is not suitable for the depth profiling and 3D reconstruction.

Furthermore, we observed no beneficial effect using O₂⁺ (500 eV) as sputter species with regard to the ionization probability of silver as predicted by Kollmer et al.⁴⁸ and Storm et al.⁴⁹ The latter stated a rather low ionization probability for silver in a very early study⁴⁹ under O⁻ bombardment. In contrast, in our experiment, O₂⁺ was applied for sputtering and Bi₃⁺ for analysis, and we hoped for a slight enhancement of the ionization probability of silver. Anyhow, the intensities of the silver signals inside the cells were equally low in all measurements. Consequently, the ionization efficiency of silver is not increased by one of the sputter species. However, silver could be clearly identified in the spectra of the cells incubated with AgNP [cf. Fig. 5(a)]. The isotopes ¹⁰⁷Ag⁺ and ¹⁰⁹Ag⁺ show the required isotopic pattern and could unambiguously be assigned.

In Fig. 6(a), a cell and the distribution of silver in- and outside the cell after several sputter cycles during depth analysis with Ar₁₅₀₀⁺ (10 keV) are shown [cf. Fig. 6(b)]. In the beginning, silver seems to be distributed homogeneously over the cell and the wafer with an increased signal intensity in the cellular edge region. This might be due to a real accumulation of AgNPs in the outer cell region or, as implied by the last image after complete removal of the cell, due to an enhanced ionization probability of silver at the silicon interface, which will be discussed in more detail later on. The images in-between reveal a slight enrichment of silver particles in the intracellular region although the intensities of the silver signals are quite low because of the poor ionization probability of silver within the cellular matrix.²⁷ 

The overlay of the cellular signal C₄H₈N⁺ and the silver signals in Fig. 7(a) suggests the presence of the AgNP in the cellular region. But obviously, we cannot provide a clear proof of the NP being intracellularly incorporated and not only deposited on the cells or just adhered to the silicon surface by ToF-SIMS measurements only. Anyhow, TEM images recorded from cellular sections demonstrate the intracellular uptake of the AgNP unambiguously, as already reported by Pauksch et al. [cf. Fig. 7(c)].^28,29

In contrast to reports by Nees et al.,²⁷ Haase et al.,⁵⁰ and others,^28,29 no agglomeration of intracellular silver is visible. As inorganic AgNPs remain unaffected by the Ar_n⁺-clusters,²⁵ silver accumulates during depth profiling, which results in an interference of silver signals from different cellular levels. As a consequence, the localization of agglomerates and 3D reconstruction of AgNP are not possible, as demonstrated in Fig. 7(b).

Obviously, the combination of oxygen and Ar-clusters is not useful for 3D analysis of AgNP in cells. Therefore, the use of Ar-clusters, varying in size and energy, might be a more promising approach. The challenge is to find parameters that allow simultaneous sputtering of organics and inorganics, while maintaining the cellular information. Seki et al. reported about a cluster size dependency of the sputter yield.⁵¹ With constant acceleration energy and decreasing cluster size, an increased ion yield was observed. In an initial approach, we started to vary the cluster size of 10 keV Ar-clusters. To evaluate the sputter efficiency of silver with 10 keV Ar-clusters, a silver-layer of about 150 nm was deposited on a silicon wafer and removed with Ar_n⁺-clusters of different sizes. The sputter yields of silver for the different cluster sizes are shown in Fig. 8(a). As expected, the sputter yield increases with the decreasing cluster size, which makes small clusters more appropriate for silver abrasion. These results fit quite well with reports of Seki et al.⁵¹ Although this is quite promising, it must be kept in mind that the sputter yield of silver in organic matrices such as cells is different from that of a pure silver layer. To evaluate this in the future, the AgNP should be embedded in organic materials like gelatin. In addition, there is a stronger impact on the ion yield by varying the energy/atom rather than just the cluster size as reported by Seah et al.⁵² and others,^51,53 as the sputter yield is proportional to the acceleration energy of the Ar_n⁺-clusters. Additionally, the impact of the smaller cluster size and higher energy/atom on cellular damage has to be evaluated. From the literature, it is well known that the energy per atom should be comparatively small to maintain the organic information with less damage.^43,44,54 Nevertheless, going to smaller Ar_n⁺-cluster sizes might be a promising approach to converge the different sputter rates of organic cell materials and inorganic nanoparticles. It might be a strategy to keep the advantage of high organic signal intensity and combine it with the successful removal of inorganic particles, therefore allowing a realistic 3D reconstruction of both.

Furthermore, the depth profile of the silver layer on a silicon wafer, shown in Fig. 8(b), demonstrates that the ionization efficiency of silver is increased by the presence of silicon or silicon oxide as reported for gold by Yang et al.⁵³ The intensity of the silver signals clearly increases when the silicon substrate is reached. This effect is also observed, when cells, cultivated on silicon wafers and incorporated with AgNP, are completely removed [Fig. 6(b)]. Reaching the silicon substrate, the ionization efficiency of the remaining silver increases and the silver becomes more visible in the image.

We employed the delayed extraction mode for 2D and 3D analyses of hMSCs incubated with silver nanoparticles. In comparison to the spectrometry and imaging mode, a combination of high lateral and high mass resolution was achieved by delayed extraction of the secondary ions. For 3D analysis, Ar-clusters (10 keV) of different cluster sizes and O₂⁺ (500 eV) were applied. The signal intensity of the intracellular silver was low and could not be increased by O₂⁺. Besides, Ar₁₅₀₀⁺ (10 keV) as sputter species in combination with delayed extraction enabled a detailed 3D visualization of the cell shape as well as of cell compartments such as the nucleus. We observed that Ar₁₅₀₀⁺ (10 keV) was not able to significantly sputter the AgNP inside the cells, but a model experiment with a defined silver layer revealed an increasing silver sputter yield of silver for smaller Ar-clusters. The use of smaller cluster sizes might be a promising compromise for depth profiling of cells with inorganic nanoparticles. A systematic study to identify the optimal Ar_n⁺-cluster size and acceleration energy for a compromise of reasonable ablation rate and less cellular damage is needed. An approximation of the sputter yields of organic and inorganic components might finally allow for a 3D reconstruction of cells with intracellular AgNP. To visualize the NP in the cells, a future approach could also be a FIB cut of the cells under gazing angle to stretch the crater wall virtually. Due to this, NP of 100 nm should be detectable by ToF-SIMS.⁵⁵ 

This work was funded by the Deutsche Forschungsgemeinschaft (DFG, Collaborative Research Centre Transregio 79–subprojects M5 and B7). The authors thank O. Dakischew for her help with the cell culture experiments and Boris Mogwitz for the preparation of Ag model layers. The authors also thank the coworkers of ION-TOF GmbH for fruitful discussions.