Dynamic protein coronas revealed as a modulator of silver nanoparticle sulphidation in vitro

Proteins adsorbing at nanoparticles have been proposed as critical toxicity mediators and are included in ongoing efforts to develop predictive tools for safety assessment. Strongly attached proteins can be isolated, identified and correlated to changes in nanoparticle state, cellular association or toxicity. Weakly attached, rapidly exchanging proteins are also present at nanoparticles, but are difficult to isolate and have hardly been examined. Here we study rapidly exchanging proteins and show for the first time that they have a strong modulatory effect on the biotransformation of silver nanoparticles. Released silver ions, known for their role in particle toxicity, are found to be trapped as silver sulphide nanocrystals within the protein corona at silver nanoparticles in serum-containing cell culture media. The strongly attached corona acts as a site for sulphidation, while the weakly attached proteins reduce nanocrystal formation in a serum-concentration-dependent manner. Sulphidation results in decreased toxicity of Ag NPs.


Protein
Accession Number

Supplementary Table 2. Quasi-spherical silver NPs main hard corona components.
Major components of silver quasi-spherical nanoparticles serum protein hard coronas, with proteins highlighted in yellow being common to all coronas and making up between 50 and 65 % of the hard corona.

Formation and separation of hard and soft protein coronas
When nanoparticles come in contact with a biological environment, they interact with biomolecules which form what are now known as coronas around the particles. The main constituents of these coronas are proteins, which is why they are often referred to as "protein coronas"; they are classified as "hard" (or long-lived, slowly-exchanging) and "soft" (or short-lived, rapidly-exchanging) depending on various parameters which will be briefly discussed here.
In a medium such as serum thousands of different types of proteins are present, in concentrations spanning over several orders of magnitude 2 . Upon introduction of a nanoparticle in such an environment, the most abundant proteins will rapidly reach the particle surface and bind to it. However, they may not be the biomolecules with the highest affinity for that surface, therefore over time they will be replaced by proteins with higher affinity, but lower mobility, which take longer to reach the nanoparticle. This behaviour is known as the Vroman effect 3,4 . The affinity of a protein for a certain nanoparticle surface depends on parameters such as particle type 5,6 , surface chemistry 7,8 , particle size 7,9-11 and shape 6 .
Furthermore, corona formation is an equilibrium process, which also depends on the initial concentration of biomolecules in the system, as well as the mean residence time of each protein at the particle surface and inter-protein interactions such as, for instance, cooperative binding. These parameters have been discussed in detail elsewhere 12,13 , and they are important in distinguishing between hard and soft coronas 14 .
In practice, in experimental settings such as the ones described in this paper and in several protein corona studies 7,10,15 , nanoparticles are incubated in plasma/serum-containing media for a certain amount of time. Following incubation, several rounds of washing take place through centrifugation, removal of supernatants and resuspension of particles in water or a protein-free buffer. The first washing step removes the unbound proteins and all the loosely-bound proteins whose mean residence time at the particle surface is shorter than the centrifugation step (15 minutes in our case). Resuspension of the pelleted particles in water or buffer changes the biomolecule equilibrium which, for some proteins, means they will detach from the nanoparticles and move into the bulk. Centrifugation results, again, in removal of the proteins that have become free, as well as the loosely-bound biomolecules with residence times shorter than the centrifugation step. The hard corona is comprised of all the proteins that are still bound to the particle surface after repeated washing, while the soft corona proteins are those which were bound (to the particle surface or the hard corona 15 ) at the end of the incubation, but were removed during several rounds of centrifugation and resuspension. The centrifugation time and speed, as well as the number of repeats of this washing process depend on the particle type and of the protein concentration in the incubation medium. The proteins that cannot be removed through repeated washing by centrifugation and resuspension in water form what is known as the hard corona. These proteins can only be detached from the surface of nanoparticles using harsh treatments such as boiling in a mixture of surfactant and reducing agents or denaturation in concentrated urea solutions or enzymatic digestion.
The long-lived nature and strong-binding of the hard corona have been proven by experts in the field in studies showing this hard corona provides a protein fingerprint that can be used to trace particles passing from one environment to another, as the hard corona conserves many of the components acquired in the initial incubation environment 16,17 . Furthermore, it has been shown that the hard protein corona is retained during intracellular trafficking 18 and is only degraded when the particles are exposed to the low pH harsh conditions inside lysosomes 19 .

Cytokine production and quantification
As discussed in the main paper, both pristine and partially-sulphidated Ag NPs increase TNFα and MIP-2 production 3-fold. Here we see that IL-1β is only slightly increased by a high concentration (50 μg ml -1 ) of Ag NPs after partial-sulphidation. IL-6 production is increased at 50 μg ml -1 particle dose by both partially and completely-transformed NPs, in agreement with the observations for TNFα which, together with IL-6, is an early marker of inflammation 20 . It should be noted that for the Ag NPs 50 μg ml -1 is a lethal dose, with about 12 % cell survival ( Figure 5), so cytokine release profiles may not be comparable to those at sub-lethal values. IL-18 concentrations are only measurable for the sulphidated NPs treatments and an increase as compared to controls is observed only for the highest concentration (100 μg ml -1 ) of partially transformed Ag NPs. Calibration curves for all measured cytokines are provided in Supplementary Fig. 19.

Incubation in cell culture media
Silver nanocubes were incubated (37 °C, 5 % CO 2 ) in RPMI-1640 medium, without any added serum, for 1 hour and 24 hours. Transmission electron microscopy (TEM) images can be seen in Supplementary Fig 1, showing that at 24 hours silver sulphide is present. At 1 hour, however, there is no nano-Ag 2 S at the surface of the metal nanoparticles. We have previously shown that, upon incubation in serum-containing RPMI-1640 cell culture medium, the polyvinylpyrrolidone (PVP) coating around the nanocubes is already replaced by proteins after 1 hour 1 . As such, the sulphidation of silver nanoparticles in foetal bovine serum (FBS) containing cell culture medium occurs with proteins and not polymer present at the surface of the Ag nanoparticles (NPs).

SDS-PAGE of serum hard coronas
Silver nanocubes with diameters ranging from 50 to 88 nm (as determined by SPIP   Table 3.

Nanoparticle Tracking Analysis for particle size measurement
Cubic and quasi-spherical Ag NPs were incubated (24 hours and 7 days) in RPMI-1640 cell culture medium supplemented with 1 % or 10 % FBS. The silver concentration during incubation was the same as for the TEM and MS studies. After incubation, unbound proteins were removed by centrifugation and the NPs were re-suspended in MilliQ water and then diluted such as to ensure a concentration of ≈ 10 9 particles ml -1 . Measurements of Ag NPs hydrodynamic diameter were performed using a Nanoparticle tracking analysis equipment (NTA, NanoSight LM10-HS, NanoSight Ltd., UK) with the NanoSight software, version 3.0.
The results are presented in Supplementary Fig. 6. Recordings of 60 seconds each were acquired in triplicate for every sample.

UV-Vis spectra of silver nanocubes in BSA
UV-vis spectra of silver nanocubes incubated in 0.4 mg ml -1 or 4 mg ml -1 BSA were collected in the range of 300 to 800 nm using a Shimadzu UV-visible-NIR UV-3600 spectrophotometer. Measurements were performed in triplicate for each of the three samples prepared for the two protein concentrations, before and after washing of unbound and looselybound proteins by centrifugation of particles followed by re-suspension in phosphate buffered saline. As previously published, blue shifting of the peak position upon washing would indicate soft corona removal 21 , if that corona was initially present. Here, we see no such shifts ( Supplementary Fig. 8), proving the absence (as expected) of BSA soft coronas.

Ion release of silver nanocubes in cell culture media
Silver nanocubes (2 and 10 μg ml -1 ) were incubated in RPMI-1640 supplemented with 1 % or 10 % FBS for 1 or 7 days. After incubation, undissolved particles were separated by spinning down the suspension (30 minutes, 16000 g). The released silver ions, which remained in the liquid phase, were analysed by flame atomic absorption spectroscopy (F-AAS) on a PerkinElmer Analyst 300 atomic absorption spectrometer mounted with a silver lumina hollow cathode lamp (PerkinElmer, Denmark), after dilution of the supernatant with 5 % HNO 3 . Triplicate samples were prepared for each incubation condition and duplicate F-AAS samples were measured for each incubation sample. For a given set of conditions (particle size, layer thickness), FDTD simulations of plasmon shifts of Ag NPs depending on increasing RIs due to accumulation of Ag 2 S result in a calibration curve. Comparing experimentally measured shifts with that calibration curve, we assess the percentage of sulphide present in a layer of known thickness and, hence, known volume. Considering the density of Ag 2 S we calculate the amount of sulphide around one Ag NP for the given conditions, which we then multiply by the total number of silver nanoparticles, thus obtaining the total amount of sulphide. We express this as percentage of transformed silver, knowing the mass we introduced in the system at the beginning of the incubation.

Finite-difference time-domain (FDTD) simulations and interpretation
FDTD simulations were employed to obtain spectra of 60 and 70 nm spherical particles, surrounded by a 4, 7, or 10 nm thick layer of Ag 2 S with various degrees of sulphide occupancy (0-100 %), as can be seen in Supplementary Fig. 16 A-C, G-I. Peak shifts at the maximum dipole absorbance were used to obtain calibration curves for each setting (particle size and layer thickness). The calibration curves were used to assess the amount of silver transformed into silver sulphide for various incubation times and serum contents, considering that all Ag NPs are of the same size. The results are presented in Supplementary Table 4.
Layer thicknesses were chosen based on the protein hard-corona model previously described 1 and on TEM observations of Ag 2 S at our nanoparticles. The shifts in the position of the maximum absorbance peak were employed for calibration curves ( Supplementary Fig. 16 D-F, J-L), which were subsequently used to estimate the amount of silver transformed into sulphide based on the experimentally observed peak shifts. Similar amounts of Ag 2 S were obtained for a given particle size regardless of the chosen layer thickness, with the numerical values for all the studied settings being presented in Supplementary Table 4. Overall, about 15-20 % of the silver in the quasi-spherical particles is transformed into sulphide at 24 hours incubation in 1 % or 10 % FBS, with the values increasing to 30-40 % at 7 days. In this case, the theoretical model employed for simulations was that of a 70 nm spherical nanoparticle, with the diameter chosen based on the experimental values for the metal core obtained from TEM images.
Supplementary Fig. 17 shows the simulated data and calibration curve for a 60 nm Ag NP with a 7 nm layer of Ag 2 S, at various degrees of layer occupancy by the sulphide, ranging from 0 % to 100 %. For a cube and a sphere of 60 nm diameter, the simulations in Supplementary Fig. 17C show a ≈ 10 nm blue peak shift upon rounding of the edges and corners. This suggests that our experimental shifts for the cubes are underestimated, as these particles gradually become more spherical upon incubation in serum-containing media. The shape changes that occur throughout the incubation do not allow for precise FDTD models, so the size and shape of the particle at the end of the incubation (7 days in 1 % FBS) based on TEM images were considered as a starting point for the simulations in Supplementary Fig   16A. We simulated a spherical nanoparticle with a diameter of 60 nm, similar to that of the cube turned into a sphere after 7 days in 1 % serum, and with increasing amounts of sulphide in the surrounding layer. The thickness of the layer (4, 7 or 10 nm) did not impact the final estimations of sulphide content, as can be seen in Supplementary Table 4. The experimentally obtained plasmon shifts from UV-vis spectra were increased by 10 nm to account for the underestimation introduced by the particle reshaping, and, using the calibration curves in Supplementary Fig. 17B and 16 (D-F) to calculate how much space around the particles is occupied by Ag 2 S, we estimated that about 15 % of the silver in the nanocubes is transformed into sulphide after 7 days incubation in 1 % FBS.

MTT assay
The cellular mitochondrial activity was measured using MTT assays, with minor modifications to a previously described method 23 . Briefly, the cells were seeded at 2 x 10 4 cells per well. They were exposed to silver ions, PVP-coated Ag nanocubes, partially and completely-sulphidated silver nanocubes, all at concentrations of 0, 2, 5, 10, 15, 25, 50 and 100 μg ml -1 in RPMI-1640 supplemented with 10 % FBS and incubated for 24 hours at 37 °C and 5 % CO 2 . Following incubation, the test medium was collected and used for cytokine analysis. The cells were incubated with 100 μl of MTT solution (0.5 mg ml -1 MTT diluted in phenol red free RPMI-1640 medium without FBS) for 2 hours at 37 °C and 5 % CO 2 .
Subsequently, the MTT solution was discarded and DMSO (100 μl) was added to the every