X-ray Fluorescence Uptake Measurement of Functionalized Gold Nanoparticles in Tumor Cell Microsamples

Quantitative cellular in vitro nanoparticle uptake measurements are possible with a large number of different techniques, however, all have their respective restrictions. Here, we demonstrate the application of synchrotron-based X-ray fluorescence imaging (XFI) on prostate tumor cells, which have internalized differently functionalized gold nanoparticles. Total nanoparticle uptake on the order of a few hundred picograms could be conveniently observed with microsamples consisting of only a few hundreds of cells. A comparison with mass spectroscopy quantification is provided, experimental results are both supported and sensitivity limits of this XFI approach extrapolated by Monte-Carlo simulations, yielding a minimum detectable nanoparticle mass of just 5 pg. This study demonstrates the high sensitivity level of XFI, allowing non-destructive uptake measurements with very small microsamples within just seconds of irradiation time.

were calibrated with signals of residual non-deuterated solvents. NMR numeration of synthesized compounds does not match with IUPAC nomenclature and only serves for NMR assignment. High-resolution mass spectrometry (HRMS) analysis was performed using Agilent 6224 ESI-TOF (110-3200 m/z). IR measurements were performed on FT/IR-4100 (Jasco). Elemental analysis was conducted on an EuroEA Elemental Analyzer a HEKAtech HAT oxygen analyzer (Fa. EuroVector/Hekatech). Automated purification steps on RP silica were carried out with a puriflash® 430 (Interchim).  Although the integral of the multiplet at 1.41 -1.22 ppm is too high, indicating 16 instead of 15 protons, the distinct 13 C{ 1 H}-NMR spectrum as well as the elemental analysis confirm the purity of the compound.

MUA-AHX-GPI:
MUA-AHX-NHS (171 mg, 399 μmol, 1.00 eq.) and GPI (247 mg, 794 μmol, 1.99 eq.) were separately dried under oil pump vacuum for 4 h. Subsequently, GPI was suspended in 10 mL of DMF, cooled to 0 °C and triethylamine (0.58 mL, 4.2 mmol, 11 eq.) was added while stirring. To the cloudy suspension, a solution auf MUA-AHX-NHS in 7 mL DMF was added dropwise while stirring over a period of 15 min. Afterwards, the reaction mixture was stirred at room temperature for another 18 h. The solvent was removed under reduced pressure and the resulting colourless oil was purified by reversed phase silica gel chromatography (C18, H2O/CH3CN = 98:2 → 0:100 (+ 0.01% FA), (v/v), UV (254 nm)). The target molecule MUA-AHX-GPI (202 mg, 323 μmol, 81%) was obtained as colourless solid. The 1 H-NMR spectrum shows an additional signal at 2.68 ppm (t, 3 J(H,H) = 7.2 Hz, 0.23 H). This triplet corresponds to the -CH2-S-S-CH2-group of the disulfide product species. With an integral of the -CH2-SH group of the target molecule of 1.77, the assumption can be made, that the corresponding disulfide of MUA-AHX-GPI was formed in a ratio of 1:15 in relation to the product. The appearance of the disulfide product can be caused by contact with air. Because the product is a mixture of diastereomers, some carbon atoms give additional signals (labelled with a and b).
The presence of the disulfide dimer is supported by the appearance of a signal at 39.8 ppm (CH2-S-S), which couples with the previously described triplet at 2.68 ppm ( 1 H-NMR) in the HSQC spectrum. It is further confirmed by the appearance of additional signals with low intensity from internal CH2 groups which also appear as shoulder in the range of 30.6 -30.2 ppm.
Other signals in the 13         After reaction overnight the conjugates were purified and concentrated by repeated centrifugations (30-90 min depending on the volume, 20,000 g).
The concentrated AuNP-conjugates could then be resuspended in the desired buffers for stability tests and cell uptake experiments. Figure S18 shows exemplary transmission electron microscopy (TEM) measurements of different batches of AuNPs used in this study, underlining their low dispersity and reproducibility of the mean particle diameter. To convert the AuNP concentration to the weight concentration of gold, we assume ideal sphericity of the nanoparticles so the volume of an individual particle with diameter dc is = 6 3 . With the density of gold ρ = 19.32 g·cm -3 and the number density of particles = A • NP (NA = Avogadro's number) we obtain the weight concentration of gold = · · in g·L -1 or mg·mL -1 respectively. As example, for NP = 12.5 nM and dc = 12.0 nm we obtain = 0.13 ± 0.2 mg·mL -1 . Note that the dispersity of the AuNPs, even if it is as low as 5% as in this study, affects the calculation of the particle concentration, as well as the calculation of the particle mass, so a direct measurement of the gold weight concentration, as with ICP-MS, is more accurate. We also note that CNP only refers to the mass of the AuNP core, neglecting the mass of the surface coating [6].
We tested several ligand coatings, three coatings based on poly(ethylene glycol) (PEG), which are discussed in the main text, and several coat-  [7][8][9]. This motif was linked to 11-mercaptoundecanoic acid (MUA) via an amide bond (custom synthesis, ABX advanced biochemical compounds GmbH, Germany) to yield MUA-PSMA-I ( Figure S22). The mercaptodecane-spacer was used in all ligands to obtain a high grafting density of the ligands on the AuNP [10,11]. MUA coated AuNPs were used as a control ( Figure S19). In MUA-PSMA-I, the binding motif is located close to the AuNPs' surface (1-2 nm distance) yielding a small conjugate, however its binding ability could be diminished because the size of the conjugate could hamper the insertion of the motif into the binding pocket of the receptor. To account for this, we synthesized a set of additional ligands with an additional 6-aminohexanoic acid (AHX) based spacer: MUA-AHX-PSMA-I with the same binding motif and just an additional AHX spacer, MUA-AHX-GPI with an alternative binding motif, and MUA-AHX-Glu with a terminal glutamic acid as another control with no binding motif ( Figure S22).
Another approach allowing even more flexibility of the motif is the use of a longer poly(ethylene glycol)-based spacer [12,13]. To this end we used -carboxypoly(ethyleneglycol)--(11-mercaptoundecanoic acid) (M = 818 g/mol, PEGMUA1kCOOH) (Iris Biotech, Germany, Figure S19), a thiolated PEG-ligand with terminal carboxylic acid groups (-COOH), to coat the AuNPs. The PSMA-I motif was then coupled to the PEG functionalized AuNPs via EDC-coupling ( Figure S20). To this end, the terminal carboxylic acid groups were activated by addition of 1-ethyl-3-(3-dimethylaminopropyl)-carbodiimide (EDC) and N-hydroxysulfosuccinimide (sulfo-NHS) in ratios of 1:80000:160000 (AuNPs:EDC: sulfoNHS). Samples were purified by centrifugation (twice) after which the particles were redispersed in buffer solution (phosphate buffered saline, PBS, 10 mM, pH 7.6). The reaction with PSMA-I was facilitated by heating up to 70° C for 2 h with a 1000fold excess of PSMA-I relative to the AuNPs. Non-reacted PSMA-I was removed by centrifugation (thrice) and the particles (PEG-PSMA-I, Figure  S19) were redispersed in water. The PEG-functionalized particles without coupled PSMA-I were used as a control (PEGMUA1kCOOH, Figure S19). As another PEG-control we used particles functionalized with a mixture of a larger PEGMUA-Ligand (M ~ 2kDa, ~ 25 % of the ligand mixture, Figure  S19) with no terminal carboxylic acid groups, and 11-mercaptoundecanoic acid (MUA) (M = 218 g/mol, ~ 75 % of the ligand mixture, Figure S19): PEG-MUA2k/MUA ( Figure S21). Such mixed ligand layers have been demonstrated to allow tuning of the particles' surface charge without compromising particle stability [11]. For particles with 75 % MUA, high unspecific uptake was observed in PC3 cells [11]. These were therefore used as another control of negatively charged PEGylated nanoparticles.

Particle characterization
The stability of the functionalized AuNPs was monitored with dynamic light scattering (DLS) as described [14,6]. Figure S23 shows the number weighted distributions of the hydrodynamic diameters of the samples after preparation including purification and concentration. All samples were colloidally stable in water. MUA coated particles were prepared in aqueous solutions with pH 9 to increase the electrostatic colloidal stabilization (by deprotonation of the terminal carboxylic acid groups), because the steric stabilization provided by this small ligand is known to be limited for AuNPs with diameters of dc ~ 12 nm as used in this study [11]. At lower pH (~5.5 in ultrapure water or ~ 7.4 in PBS) we observed strong indications of agglomeration by DLS and UV/vis absorption spectroscopy. Shift and broadening of the plasmon peak indicate agglomeration of the nanoparticles as well as the shift and limited reproducibility of the apparent hydrodynamic diameter measured by DLS [6,14]. Absorbance spectra of MUA in different media at different waiting times are shown in Figure S24. Another destabilization behavior was observed for the sample MUA-AHX-Glu. These particles were stable in water as well as in PBS, but in cell medium (DMEM) they agglomerated fast, leading to sedimentation (Figure S25). The other samples exhibited higher stability in water, PBS and cell medium. MUA-AHX-PSMA-I were colloidally stable in water, but exhibited notable agglomeration in DMEM ( Figure S26). MUA-PSMA-I were colloidally more stable in the different media, underlining that a longer ligand does not necessarily provides a better stabilization ( Figure S28).

MUA-AHX-GPI
and PEG-PSMA-I were also colloidally stable in different media ( Figure S28) and the stability of PEGylated AuNPs (PEG-MUA2k/MUA) in different media was reported previously [11]. The characterization of the AuNPs demonstrates that apart from the different ligands on the nanoparticle surface, their major difference is their colloidal stability in cell medium, that can strongly differ, even when the particles are stable in water and PBS.

Cell culture and ICP-MS particle uptake protocols
Cell culture experiments were conducted based on protocols described previously [14,15]. PC3-PIP cells with (PC3+PSMA) and PC3 cells without (-PSMA) overexpression of the PSMA receptor were used. PC3+PSMA and PC3-PSMA cells were seeded into 6-well plates at a density of 2 × 10 5 cells/well in serum containing medium (10% fetal bovine serum, FBS), and were allowed to attach overnight. The next day, the old cell medium was removed and the cells were exposed to 2 mL fresh medium containing the according nanoparticles. The plate was incubated at 37 °C for 24 h or 48 h. After exposure, the nanoparticle solution was removed and cells were washed with 2 mL PBS three times. Then, 0.3 mL trypsin, ethylenediaminetetraacetic acid (EDTA) (0.01% trypsin-EDTA) was added to detach the cells from the plate bottom and transferred to Eppendorf tubes. After centrifugation at 300 rcf for 5 min, cells were resuspended in 1 mL PBS, and 10 μL of this solution was diluted 10 times to count the cell number. Cells were then collected again by centrifugation. For digestion, 75 μL HNO3 was added and the sample left overnight to lyse the cells, then 150 μL HCl was added to digest the AuNPs. Finally, the samples were further diluted (1:10) with 2 wt% HCl prior to measuring the elemental concentration of Au in the sample with ICP-MS. The Au-concentrations of all nanoparticle solutions used for uptake experiments were also determined with ICP-MS to calculate the uptake. By dividing the detected mass of elemental gold by the number of cells in the sample, the amount of internalized AuNPs per cell could be given as mAu [pg/cell]. Experiments were performed in independent triplicates, each experiment was with different generations of cells and incubations were done at different days.
ICP-MS determines the amount of Au in a sample solution as ppb (parts per billion), referring to 1 g Au per 10 9 g sample solution. The mass of the sample solution is assumed to be the mass of water only, with a density of 1 g/mL, and thus 1 ppb refers to 10 -9 g/mL = 1 ng/mL of Au. In a typical sample in the here used protocol there are around 500,000 cells. The cell pellet has around 50 μL volume. In order to digest the cell pellet, 75 μL of HNO3 is added to lyse the cells overnight. Then 150 μL of HCl is added to form aqua regia to digest the AuNPs. The samples are then diluted 10 times with 2% HCl prior to ICP-MS testing, in order to protect the ICP-MS machine from too high concentrations of acid which could destroy the instrument. The final sample volume thus is 2250 μL, leading to 222 cells/μL ≈ 200,000 cells/mL. A typical result for fully loaded cells was 1000 ppb, which corresponds to 5 pg Au/cell (see Figure S29-S37). In samples without added AuNPs as blank the detected value was around 2 ppb Au, corresponding to 0.01 pg/cell. This is the ICP-MS detection limit in the here used protocol. In other words, the minimum amount of Au in the sample needs to be 0.01 pg/cell · 500,000 cells = 5000 pg. If we consider fully loaded cells with 5 pg/cell, this would correspond to 1000 cells. With our ICP-MS detection protocol we thus would be able to see the minimum amount of 1000 fully loaded cells. This estimation fits well to previous related studies, where with different cells and NPs a detection limit of around 400 cells had been determined [16]. With an autosampler the measurement time per such sample is 150 s.
We note that the numbers given here refer to the used ICP-MS protocol, which was not designed to lead to the minimum possible amount of gold to be detected. IPC-MS can in principle detect Au levels as low as 0.1 μg/L (https://www.eag.com/resources/appnotes/icp-oes-and-icp-ms-detection-limit-guidance/; accessed on 16.3.2021). Taking the here used sample volume of 2250 μL this results in 225 fg ≈ 0.2 pg. This is a much lower value that the 5000 pg as obtained above under different conditions not optimized to determine the minimum amount of Au. The theoretical 0.2 pg ICP-MS limit also can't be directly compared to the 5 pg XFI limit described in the main article, as also this value depended on the used protocol and thus can't be as good as the theoretical limit.

Discussion of ICP-MS uptake results
Here we present the results from the ICP-MS measurement of the cells prepared as detailed in chapter 4 above. First the uptake of MUA-PSMA-I and MUA is compared ( Figure S29-S32).     Figure S29 and S30 but plotted for the different types of cells investigated here.
It was observed, that there was higher uptake of MUA than for MUA-PSMA-I for both cell lines. This shows that uptake is dominated not by specific targeting, but by colloidal stability, whereby colloidally less stable AuNPs sediment on top of the cells and thus are incorporated to a higher extent [14,17]. Also, the uptake of MUA-PSMA-I is higher in PC3-PSMA cells not overexpressing the PSMA receptor, indicating cell lineage effects.  For all cells and nanoparticles the typical concentration dependent endocytosis was seen, as shown in Figures S33 and S34. The highest uptake is observed for MUA-AHX-GPI, however the uptake was higher in PC3-PSMA cells not overexpressing the PSMA receptor. This was also the case for the other particles. To facilitate comparison, the uptake for each of the different nanoparticles is plotted for the two different cell types in Figures  S35-S37.   In summary in the data presented here, there is no indication that uptake is specific due to binding of PMSA binding ligands to PSMA receptor modified cells. Colloidal stability and cell linage effect and thus non-specific effects have determined uptake in the 2 data sets.
The PEGylated samples (PEG1kCOOH, PEG-PSMA-I and PEG-MUA2k/MUA) discussed in the main text are colloidally more stable, however the highest uptake was found for PEGMUA2k/MUA, i.e. nanoparticles with no PSMA binding ligand. Thus, the capability of XFI for measuring low gold concentrations in cells is demonstrated, but conclusions regarding specific uptake cannot be drawn at this point and thus are not discussed in the main article.
The ICP-MS results shown here indicate a saturation level of cells with AuNPs at around 5 pg/cell. In case of the XFI measurements in the main paper the maximum amount of AuNPs per cell was around 400 pg per 888 cells (cf. Figure 1) ≈ 0.45 pg/cell, which is one order of magnitude lower than the ICP-MS data. However, ICP-MS and XFI recordings were not done under the same conditions. First, the surface capping of the NPs was different. For XFI the incubation time was 16 h. ICP-MS was recorded after 24 h and 48 h incubation time and the data show that uptake was not saturated yet at 24 h. For XFI an incubation concentration of CNP = 0.13 mg/mL was used, which is higher than the maximum concentration of CNP ≈ 0.095 mg/mL as used for ICP-MS. Higher NP concentrations may impair cell viability. In addition, some of the NPs may only be attached to the outer cell membrane instead of being endocytosed [18]. These NPs would be wrongfully counted by ICP-MS as internalized NPs, but upon the gel embedding procedure used for XFI measurements might be detached from cells. In addition, some NPs may have been lost during storage time of the agaroseembedded cells before actual XFI measurements. Thus, there is a range of possible explanations to account for the different determined amount of Au per cell for the separate ICP-MS and XFI studies shown here

XFI significances and fit values
In order to underline the extrapolation given in the main text from the measured data to the optimized sensitivity limit of our XFI approach, Figure S38 below shows the statistical significance and chi-squared values for the measured data given in Fig. 1 in the main text. One could thus directly scale down the measured AuNP-mass to such values that correspond to a Z=3 level, which is already around the order of magnitude reached by the optimization of XFI towards the shown sensitivity limit. As of note, the detectable AuNP-mass scales directly with Z. Hence, if Z = 100, one could measure 33-times less AuNPs, reaching the statistical limit of Z = 3, without any optimization. If one applies, in addition, the optimization as discussed in the main text, one can reach the level of around 5 pg AuNP mass in the X-ray beam volume. Figure S38. Significance and Χ 2 /ndf fit values for measurements shown in Figure 1. The detection limit of Z=3 is marked with a red line, while the ideal Χ 2 /ndf=1 value is shown with a blue line. Significances are shown for the Lα and the Lβ region as well as the global Z value, Lα and Lβ combined.