*Article* **Profiling of Sub-Lethal in Vitro Effects of Multi-Walled Carbon Nanotubes Reveals Changes in Chemokines and Chemokine Receptors**

**Sandeep Keshavan 1, Fernando Torres Andón 1,2,3, Audrey Gallud 1,4, Wei Chen 5,6, Knut Reinert 7, Lang Tran <sup>8</sup> and Bengt Fadeel 1,\***

	- Knut.Reinert@fu-berlin.de

**Abstract:** Engineered nanomaterials are potentially very useful for a variety of applications, but studies are needed to ascertain whether these materials pose a risk to human health. Here, we studied three benchmark nanomaterials (Ag nanoparticles, TiO2 nanoparticles, and multi-walled carbon nanotubes, MWCNTs) procured from the nanomaterial repository at the Joint Research Centre of the European Commission. Having established a sub-lethal concentration of these materials using two human cell lines representative of the immune system and the lungs, respectively, we performed RNA sequencing of the macrophage-like cell line after exposure for 6, 12, and 24 h. Downstream analysis of the transcriptomics data revealed significant effects on chemokine signaling pathways. *CCR2* was identified as the most significantly upregulated gene in MWCNT-exposed cells. Using multiplex assays to evaluate cytokine and chemokine secretion, we could show significant effects of MWCNTs on several chemokines, including CCL2, a ligand of CCR2. The results demonstrate the importance of evaluating sub-lethal concentrations of nanomaterials in relevant target cells.

**Keywords:** multi-walled carbon nanotubes; nanoparticles; chemokines; macrophages; transcriptomics

### **1. Introduction**

Nanotoxicology is a scientific discipline aimed at assessing the potential adverse effects of engineered nanomaterials (NMs) as well as enabling the safe use of NMs [1]. Nanotoxicology research has made great strides in the past ten to fifteen years, and efforts to pin down the mechanism(s) of NM toxicity may ultimately inform regulatory frameworks with the goal of exploiting nanotechnology in a safe and sustainable manner [2]. However, much attention has been focused on the same basic paradigms, including the so-called oxidative stress paradigm [1,2]. This has certainly provided considerable insight into the biological and toxicological effects of NMs, but there may not be a one-size-fits-all mechanism with which to explain NM toxicity in different cells or tissues. The use of global omics-based approaches enables the exploration more broadly of biological mechanisms that influence the toxicity and efficacy of NMs [3]. Considerable progress has been made

**Citation:** Keshavan, S.; Andón, F.T.; Gallud, A.; Chen, W.; Reinert, K.; Tran, L.; Fadeel, B. Profiling of Sub-Lethal in Vitro Effects of Multi-Walled Carbon Nanotubes Reveals Changes in Chemokines and Chemokine Receptors. *Nanomaterials* **2021**, *11*, 883. https://doi.org/ 10.3390/nano11040883

Academic Editors: Werner Blau and Angelina Angelova

Received: 11 February 2021 Accepted: 26 March 2021 Published: 30 March 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

in the last few years in terms of applying transcriptomics and proteomics coupled with computational analysis to address the NM effects [4].

The nanomaterial repository of the Joint Research Centre (JRC) of the European Commission (EC) provides a collection of exhaustively characterized NMs that have been applied as benchmark materials in a number of projects, including the large, ECfunded FP7-NANOREG project, a pan-European project aimed at a common approach to the regulatory testing of nanomaterials [5]. Hence, in FP7-NANOREG, we performed cytotoxicity screening and cytokine profiling of nineteen NMs using the human monocytelike THP-1 cell line and obtained evidence that diverse NMs can be grouped based on their pro-inflammatory potential [6]. Similarly, in the EC-funded FP7-MARINA project on hazard assessment and risk management of NMs, a comprehensive study was performed on a panel of metal oxides from the JRC repository [titanium dioxide (TiO2) (NM103 and NM104), zinc oxide (ZnO) (NM110 and NM111) and silicon dioxide (SiO2) (NM200 and NM203)] using a range of cellular assays representing different target organs or systems (immune system, respiratory system, gastrointestinal system, reproductive organs, kidney and embryonic tissues) [7]. The results enabled a hazard ranking of the NMs. The study also revealed cell type-specific response to NMs; overall, the most sensitive cells studied were the murine alveolar macrophages (MH-S). In the EC-funded FP7-ENPRA project, which also focused on hazard assessment of NMs, a panel of NMs procured from the JRC repository were investigated using the human hepatoblastoma C3A cell line, and silver (Ag) and ZnO were found to be more potent with respect to cytotoxicity and cytokine secretion, whereas the multi-walled carbon nanotubes (MWCNTs) and TiO2 displayed less toxicity [8]. The conclusion that the effects of NMs are cell type-dependent was also demonstrated in a study of 23 NMs using a panel of ten different cell lines [9]. In fact, even when assessing NM toxicity towards cells originating from the same organ, the outcome was dependent on the specific cell line used, and its origin (human or mouse), as illustrated by the fact that the three lung epithelium-derived cell lines used differed in terms of oxidative stress [9].

CNTs have received particular attention due to their apparent similarities with other fiber-like materials, although inadequate or limited evidence of carcinogenicity exists for most CNTs [10]. Notwithstanding, in a recent study of seven different CNTs and two carbon nanofibers (CNFs), all the tested materials except one highly aggregated MWCNT sample induced genotoxicity in human bronchial epithelial cells (BEAS-2B) [11]. There was a tendency for CNTs/CNFs with increasing length and diameter to display slightly greater toxicity. Di Cristo et al. [12] performed a comparative study of the toxicity of three benchmark MWCNTs obtained from the JRC repository (NM400, NM401, and NM402) using two murine macrophage cell lines (RAW264.7 and MH-S) and found that long and needle-like NM401, but not short and tangled NM400 or NM402, affected cell viability in a dose-dependent manner in both cell models. For this reason, we selected NM401 as representative MWCNTs for our studies. For comparison, we chose Ag and TiO2, which have been shown in numerous previous studies to possess high and low toxicity potential, respectively. The three NMs [Ag (NM300K), TiO2 (NM104), and MWCNTs (NM401)] were tested using two human cell lines, the lung adenocarcinoma cell line A549 (often used as a model of the lung epithelium) and THP-1. Following cytotoxicity screening using established protocols [6], we performed genome-wide transcriptomics analysis by applying RNA sequencing, coupled with pathway analysis [13,14]. The results were then corroborated by cytokine–chemokine profiling.

#### **2. Materials and Methods**

#### *2.1. Nanomaterials*

The NMs used in this study are classified as representative test materials and include a (random) sample from one industrial production batch. This ensures that the sample has been homogenized and is sub-sampled into vials under reproducible (GLP) conditions, and that the stability of the sub-samples is monitored. Thus, to the extent possible for industrial materials, all the sub-samples should be identical and differences in test results between

laboratories for the same endpoint should not be due to differences in the NMs tested [15]. For detailed physicochemical characterization of the selected NMs [Ag (NM300K), TiO2 (NM104), and MWCNTs (NM401)] performed at the JRC, refer to: Rasmussen et al. [15] and references therein. For the dispersion of the NMs, the SOP developed under the ECfunded NANOGENOTOX joint action was used, as described previously [6]. Briefly, stock dispersions were prepared by pre-wetting the NM powders in 0.5 vol% ethanol followed by dispersion in sterile-filtered 0.05% w/v bovine serum albumin (BSA)-water (Milli-Q® water, Sigma-Aldrich, Stockholm, Sweden). Both the water and the BSA (obtained from Sigma-Aldrich, Stockholm, Sweden) were endotoxin-free. The samples were then dispersed by sonication (16 min at 400 W) before being added to cell cultures.

#### *2.2. Human Cell Lines*

Human THP-1 acute monocytic leukemia cells and A549 lung adenocarcinoma cells were obtained from the American Type Culture Collection (ATCC) (Manassas, VA, USA). The cells were mycoplasma tested regularly using MycoAlert® mycoplasma detection kit (Lonza, Basel, Switzerland). THP-1 cells were maintained in RPMI-1640 medium (Sigma-Aldrich, Stockholm, Sweden), supplemented with 10% heat-inactivated fetal bovine serum (FBS), 2 mM L-glutamine, 1 mM Na-pyruvate, 5.0 × <sup>10</sup>−<sup>5</sup> <sup>M</sup> <sup>β</sup>-mercaptoethanol, 100 U/mL penicillin, and 100 mg/mL streptomycin. A549 cells were cultured in DMEM (ThermoFisher, Stockholm, Sweden), supplemented with 10% heat activated FBS, 2 mM L-glutamine, 1 mM Na-pyruvate, 100 U/mL penicillin, and 100 mg/mL streptomycin. To induce differentiation into macrophage-like cells, THP-1 cells were stimulated for 3 days with 150 nM phorbol myristate acetate (PMA) (Sigma-Aldrich, Stockholm, Sweden).

#### *2.3. Endotoxin Testing*

NMs were tested for endotoxin contamination using the LAL test (Limulus Amebocyte Lysate Endochrome, Charles River Endosafe, Charleston, SC, USA) according to the manufacturer's instructions. The NMs were all endotoxin-free (<0.5 EU/mL) (data not shown).

#### *2.4. Cell Viability Assay*

Cell viability was monitored with the lactate dehydrogenase (LDH) release assay using the CytoTox96® non-radioactive cytotoxicity kit (Promega, Stockholm, Sweden), as previously described [16]. The samples were analyzed using the Tecan Infinite® F200 plate reader operating with Magellan v7.2 software (Männedorf, Switzerland). The percentage of cell viability was calculated based on the ratio between the absorbance of each sample and the negative control sample.

#### *2.5. Transmission Electron Microscopy*

Cellular uptake/localization of NMs was monitored by TEM as described [17]. Briefly, cells were fixed in 2% glutaraldehyde in 0.1 M sodium cacodylate buffer containing 0.1 M sucrose and 3 mM CaCl2, pH 7.4. Cells were then washed and post-fixed in 2% osmium tetroxide in 0.07 M sodium cacodylate buffer containing 1.5 mM CaCl2, pH 7.4, at 4 ◦C for 2 h, dehydrated in ethanol followed by acetone, and embedded in LX-112, an epoxy derivative. Sections were contrasted with uranyl acetate followed by lead citrate and were examined in a Tecnai 12 Spirit Bio TWIN TEM (FEI, Eindhoven, The Netherlands) at 100 kV. Digital images were obtained using a Veleta camera (Olympus, GmbH, Münster, Germany).

#### *2.6. Cytokine and Chemokine Analysis*

TNF-α released in the culture medium was measured using an ELISA kit following the instructions provided by the manufacturer (MabTech, Nacka, Sweden). The absorbance of the reaction product was measured using a spectrophotometer (Tecan Infinite® F200, Männedorf, Switzerland) and the results for each sample were calculated using a standard curve of recombinant human TNF-α protein. Lipopolysaccharide (LPS) (100 ng/mL;

Sigma-Aldrich, Stockholm, Sweden) was used as a positive control for TNF-α release. Results are expressed as ng/mL of released cytokine, based on three independent experiments. Furthermore, profiling of cytokines and chemokines released by THP-1 cells was performed by using the U-PLEX chemokine panel 1 (human) (K15047K-1) and the V-PLEX pro-inflammatory panel 1 (human) (K15049D-1), respectively. We employed the Meso Scale Discovery (MSD) (Rockville, MD, USA) multi-plex electrochemiluminescence (ECL) platform to quantify cell supernatant concentrations of the indicated biomarkers, according to the manufacturer's instructions. As a positive control, cells were exposed to 0.1 μg/mL LPS (Sigma-Aldrich, Stockholm, Sweden) for 24 h. The samples were analyzed on the MSD Meso SECTOR® S600 instrument and the data were analyzed using MSD Discovery Workbench® software (v. 4.0) (Rockville, MD, USA). Samples with values below the lower limit of detection (defined as 2.5 S.D. above the background) were excluded from further analysis. The cytokine and chemokine expression data retrieved from the multi-plex assay were further analyzed using hierarchical clustering analysis, as described previously [6]. Complete linkage and Euclidean distances were employed as metrics to draw association dendrograms between cytokines/chemokines and the different treatment conditions. The cluster analysis and the corresponding heatmaps were prepared using R 3.2 [6].

#### *2.7. Western Blotting*

For protein detection, cells were harvested and lysed at 4 ◦C in RIPA buffer [50 mM Tris HCl (pH 7.4), 150 mM NaCl, 1% Triton X-100, 0.25% sodium deoxycholate, 0.1% SDS, 1 mM EDTA] supplemented with protease and phosphatase inhibitors plus 1 mM DTT (Sigma Aldrich, Stockholm, Sweden) as described previously [18]. Thirty μg total protein were loaded into each well of a NuPAGE 4–12% Bis-Tris gradient gel (ThermoFisher, Stockholm, Sweden) and subjected to electrophoresis. The proteins were then transferred to a Hybond low-fluorescent 0.2 μm PVDF membrane (Amersham, Buckinghamshire, UK), blocked for 1 h in Odyssey® Blocking Buffer (PBS) (LI-COR), and stained overnight at 4◦C with antibodies against NLRP12 (Abcam, Stockholm, Sweden) and GAPDH (ThermoFisher, Stockholm, Sweden) as loading control. The membranes were then probed with the goat anti-rabbit IgG (H+L) HRP-conjugated antibody (ThermoFisher, Stockholm, Sweden) or the goat anti-mouse IRDye 680RD antibody (LI-COR Biotechnology, Lincoln, NE, USA) and proteins were detected using Clarity™ ECL substrates (BioRad, Hercules, CA, USA) and Super RX-N film (FujiFilm Nordic AB, Stockholm, Sweden), or the LI-COR Odyssey® CLx scanner operating with Odyssey® Image Studio software (LI-COR Biotechnology).

#### *2.8. RNA Sequencing*

Total RNA was extracted from cells harvested at 0 h, 6 h, 12 h, and 24 h of exposure to NMs (25 μg/mL) using the TRIZOL reagent (Life Technologies, Stockholm, Sweden) according to the manufacturer's recommendations. Total RNA was quantified by NanoDrop™ (NanoDrop Technologies, ThermoScientific, Stockholm, Sweden) and RNA quality was assessed using the Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA, USA). Three biological replicates of each sample were submitted for RNA sequencing [19]. In brief, the sequencing was performed using 1 μg total RNA following the Illumina® mRNA-Seq library preparation protocol (Illumina, San Diego, CA, USA). To this end, poly(A) RNA was isolated by two rounds of oligo (dT)25 Dynabeads™ (Invitrogen, Stockholm, Sweden) purification. Then, the chemically fragmented mRNAs were purified by Agencourt RNA-Clean XP SPRI beads (Agencourt-Beckman Coulter, Beverly, MA, USA) and converted to first strand cDNA, followed by second strand cDNA synthesis. The paired-end sequencing library was prepared from purified double stranded cDNA using the NEBNext® DNA Library Prep Kit (Illumina, San Diego, CA, USA). The purified ligated product was PCR amplified and the prepared libraries were quantified and quality-assessed and sequenced on the Illumina® HiSeq 2000 platform (Illumina, San Diego, CA, USA).

#### *2.9. Pathway Analysis*

Canonical pathway analysis was done using the Ingenuity Pathway Analysis (IPA) software (content version 24718999) (Ingenuity Systems, Redwood City, CA, USA). The significance of the pathways was estimated through the curated ingenuity knowledge database using a causal analysis approach [20], complemented by hierarchical cluster analysis, as described in Reference [21]. Significant pathways were filtered by *p*-values < 0.001 and activation z-scores > 2 or >2 (data not shown), representing a significant deactivation or activation, respectively. Data were integrated using hierarchical clustering on quantile-normalized data.

#### *2.10. Statistical Analysis*

One-way analysis of variance (ANOVA), followed by a Dunnett's or Sidak's multiple comparison test analysis was used for the analysis of statistical significance, and *p* < 0.05 was considered significant. For nonparametrically distributed data, the two-tailed Mann-Whitney test was used. Statistical tests were performed using GraphPad Prism 8 (San Diego, CA, USA).

#### **3. Results**

#### *3.1. Characterization of Benchmark Nanomaterials*

The test materials [Ag (NM300K), TiO2 (NM104), and MWCNTs (NM401)] were obtained from the JRC nanomaterial repository and detailed characterization has been provided by the JRC (refer to Table S1 for an overview and see references therein). The MWC-NTs are characterized by their rigid and needle-like appearance (length: 4048 ± 2371 nm; diameter: 67 ± 24 nm). The TiO2 NMs are rutile, with an average diameter of 25 nm and have been extensively studied in the FP7-MARINA project [7]. The Ag NMs were provided in a colloidal suspension with a primary particle size of 15 nm. The latter NMs were studied extensively in the FP7-NANOREG project (see, for instance, Bhattacharya et al. [6]).

#### *3.2. Cytotoxicity Assessment of Nanomaterials*

We utilized two human cell lines: the lung cell line A549 and the monocyte-like cell line THP-1. The latter cells were differentiated into macrophage-like cells using PMA [6]. The cells were exposed for 24 h to the three different NMs at concentrations ranging from 1 to 100 μg/mL and cell viability was monitored using the LDH release assay. The TiO2 NMs were cytotoxic for THP-1 cells only at the highest dose while Ag NMs and MWCNTs triggered a dose-dependent cytotoxicity at doses above 10 and 25 μg/mL, respectively (Figure 1A). In contrast, the NMs were not cytotoxic towards the A549 lung cell line, apart from a minor effect noted for the Ag NMs at high doses (Figure 1B).

**Figure 1.** Cytotoxicity screening of representative nanomaterials (NMs). Human macrophage-differentiated THP-1 cells (**A**) and lung epithelium-derived A549 cells (**B**) were exposed to TiO2 NMs (NM104), Ag NMs (NM300K), and MWCNTs (NM401) for 24 h and cell viability was evaluated by using the LDH release assay. The results shown are mean values ± S.D. of three independent experiments. \* *p* < 0.01; \*\* *p* < 0.001.

#### *3.3. Cellular Uptake of Benchmark Nanomaterials*

We monitored cellular uptake by performing TEM imaging of THP-1 cells after exposure for 24 h at 25 μg/mL (a dose at which the cell viability remained >50% for all three NMs). The Ag NMs could not be visualized with certainty at this time-point, possibly due to dissolution of the NMs, as shown in previous studies [22]. The MWCNTs, on the other hand, were found to damage the microtome (Supplementary Figure S1). Therefore, results are shown for TiO2 NMs. As seen in Figure 2A, macrophage-like THP-1 cells readily internalized large clusters of NMs in the absence of ultrastructural signs of cell death, consistent with the results of the LDH release assay.

To further explore the cellular impact of the three NMs, we monitored TNF-α production. TNF-α is a prototypic pro-inflammatory cytokine that is strongly induced by LPS. As shown in Figure 2B, while LPS triggered significant secretion of TNF-α, TiO2 NMs had no effect, despite the considerable cellular uptake of these NMs. Furthermore, Ag NMs triggered some TNF-α production at low doses, but not at higher doses, whereas the MWCNTs triggered TNF-α production at high doses, albeit less than LPS. These results thus reveal a marginal effect of the NMs on TNF-α production in macrophage-like cells and suggest (indirectly) that these NMs are endotoxin-free, as TNF-α is a potent inducer of LPS [23].

**Figure 2.** Cellular impact of representative NMs. (**A**) TEM images of untreated THP-1 cells (**a**,**a'**) and cells exposed to TiO2 NMs for 24 h at 25 μg/mL (**b**,**b'**). Scale bars: 5 μm (**a**,**b**) and 1 μm (**a'**,**b'**). Refer to Figure S1 for additional findings derived from the TEM imaging. (**B**) THP-1 cells were exposed to TiO2 NMs (NM104), Ag NMs (NM300K), and MWCNTs (NM401) for 24 h and TNF-α production was measured by ELISA. The results are mean values ± S.D. of three independent experiments. \*\*\* *p* < 0.001.

#### *3.4. Transcriptomics Analysis of Nanomaterials*

To investigate the effects of the selected NMs in more detail, we applied RNA sequencing. THP-1 cells were selected as a model. The cells were exposed for 6 h, 12 h, and 24 h to NM300K, NM104, and NM401 at 25 μg/mL in order to determine the kinetics of the transcriptomics responses. Samples were sequenced using the Illumina® HiSeq 2000 sequencing platform. RNA sequencing revealed that significant numbers of differentially expressed genes (DEGs) were affected by the NMs (NM300K: 313, NM104: 674; NM401: 124). Only DEGs with a significance level of <0.05 (FDR) and absolute fold-change ≥2 were included in the subsequent analysis. We focused the IPA analysis on immune cells and immune cell lines. The heatmap in Figure 3 shows the hierarchical cluster analysis of the top canonical pathways identified in THP-1 cells exposed to NM300K, NM104, and NM401 at various time-points. The color coding in the heatmap depicts the *p*-values for the pathways

shown. The samples corresponding to the 24 h exposure to Ag (NM300K) and MWC-NTs (NM401) are grouped together. To further refine the analysis, we analyzed each NM separately. The results for TiO2 and Ag are shown in Supplementary Figures S3 and S4, respectively, while the results for MWCNTs are reported in Figure 4. The analysis shows clear time dependence insofar as the changes in gene expression are more robust (based on *p*-values) at 24 h when compared to 6 h and 12 h. We found that cell cycle related pathways were affected by all the tested NMs, and several pathways related to immune cell function were affected by MWCNTs (Figure 4) and Ag NMs (Supplementary Figure S3), but not by TiO2 NMs (Supplementary Figure S2). Based on the activation z-scores, cell cycle pathways were deactivated, while immune related pathways were activated (data not shown).

**Figure 3.** Pathway analysis of transcriptomics data. Macrophage-differentiated THP-1 cells were exposed to TiO2 NMs (NM104), Ag NMs (NM300K), and MWCNTs (NM401) for 6 h, 12 h, and 24 h (25 μg/mL) and samples were subjected to RNA sequencing. The heatmap shows the canonical pathway analysis of the transcriptomics data. The significance values indicate the probability of the association of the DEGs with the respective pathway. The cutoff for the *p*-value was *p* < 0.001 for at least one of the conditions.

**Figure 4.** Canonical pathways affected by MWCNTs. The heatmap shows the results of the canonical pathway analysis of the transcriptomics data obtained from THP-1 cells exposed to MWCNTs (NM401) at 25 μg/mL at the indicated time-points. The corresponding analysis for NM401 and NM300K exposed cells is shown in Figures S2 and S3. The significance values indicate the probability of the association of DEGs with the respective pathway. The cutoff for the *p*-value was *p* < 0.001 for at least one of the conditions.

> Our analysis showed that the top-most upregulated gene at every time-point in cells exposed to MWCNTs (NM401) was *CCR2*. The log2 ratio differential expression values for *CCR2* were 7.938 (6 h), 8.158 (12 h), and 13.712 (24 h). *CCR2* is a chemokine receptor encoding gene, and the corresponding receptor binds chemokine (C-C motif) ligand 2 (CCL2), also referred to as monocyte chemoattractant protein-1 (MCP-1). Figure 5A provides a graphic depiction of the network involving *CCR2* (at 24 h). It is notable that not only *CCR2*, but also *CXCR2* was significantly upregulated in cells exposed to MWCNTs. *CXCR2* is another chemokine receptor encoding gene and the corresponding protein serves as a receptor for CXCL8 (previously known as IL-8). The only downregulated gene in the network was *NLRP12,* encoding a member of the Nod-like receptor (NLR) family of proteins that have been shown to play a role in inflammasome activation [24]. However, *NLRP3* was not affected (data not shown). To validate the RNA sequencing, we checked the expression of the NLRP12 protein in THP-1 cells exposed for 24 h at 25 μg/mL, and we could confirm that the protein expression was decreased following exposure to NM401 compared to the control (Figure 5B).

**Figure 5.** Impact of MWCNTs on the chemokine signaling network in macrophage-like cells. (**A**) The figure depicts the transcriptomics results for MWCNT-exposed (25 μg/mL) THP-1 cells at 24 h. The color coding shows upregulated (red) and downregulated genes (green). Data were analyzed and visualized by using the IPA software tool (Qiagen, Inc., www.qiagenbioinformatics.com/products/ ingenuity-pathway-analysis). (**B**) Western blot assay for the expression of NLRP12 in THP-1 cells exposed or not to MWCNTs (25 μg/mL) for 24 h. GAPDH was included as a loading control.

#### *3.5. Cytokine-Chemokine Profiling of Nanomaterials*

To further corroborate the transcriptomics results and in order to add the biological context, we performed multiplex assays for the detection of cytokines (IFN-γ, IL-1β, IL-2, IL-4, IL-6, IL-8, IL-10, IL-12p70, IL-13, and TNF-α) and chemokines (Eotaxin, Eotaxin-2, Eotaxin-3, IL-8, IP-10, MCP-1, MCP-2, MCP-3, MCP-4, MDC, MIP-1α, MIP-1β, and TARC). To this end, THP-1 cells were exposed for 24 h to the different NMs (25 μg/mL). LPS (0.1 μg/mL) was included as a positive control. Three independent experiments were conducted and samples with values below the lower limit of detection were excluded from the subsequent analysis. The results confirmed the notion that the NMs did not elicit a pronounced induction of pro-inflammatory cytokines (at 25 μg/mL) (Figure S4). Hence, TNF-α and IL-8 (CXCL8) were not upregulated, while a modest induction was noted for IL-6. NM401 also triggered a modest induction of IL-1β. In contrast, several chemokines

were significantly upregulated in response to NM401 (Figure S5). In particular, MCP-1 (CCL2) was upregulated, along with TARC (thymus and activation regulated chemokine, also known as CCL17) and MDC (macrophage derived chemokine, also known as CCL22). It is notable that the magnitude of these responses was similar to that of LPS. NM104 and NM300K, on the other hand, did not show such effects.

To further probe the responses of macrophage-like cells to the three NMs, we performed hierarchical cluster analysis to draw association dendrograms between cytokine and chemokine responses, respectively. LPS exposed samples (supernatants from exposed cells) were identified as separated from the other samples, both with respect to cytokine (Figure 6A) and chemokine responses (Figure 6B). However, the NM401 exposed samples clustered closer to the LPS samples in terms of chemokine responses, whereas NM104, and to some extent, NM300K, segregated with the untreated control samples (Figure 6B).

**Figure 6.** Profiling of cytokine and chemokine production in NM-exposed cells. Macrophagedifferentiated THP-1 cells were exposed to TiO2 NMs (NM104), Ag NMs (NM300K), and MWCNTs (NM401) for 24 h at 25 μg/mL and cytokine and chemokine production was monitored in the supernatants using multiplex assays. As a positive control, cells were exposed to 0.1 μg/mL LPS. The experiment was performed three times and results for individual biomarkers are reported in Figures S4 and S5. Hierarchical cluster analysis was performed to draw association dendrograms between cytokine (**A**) and chemokine (**B**) responses to NMs. Association clusters for exposures and biomarkers are represented by dendrograms at the **left** and at the **top** of the heatmap, respectively.

#### **4. Discussion**

Using representative NMs from the JRC nanomaterial repository, we have shown that THP-1 cells are more susceptible to NM-induced cell death than A549 cells. The differences between the two cell types could be because epithelial cells are less phagocytic when compared to macrophages. However, we did not perform a quantitative analysis of NM internalization in the present study. It is notable that A549 cells were also found to be refractory to other NMs when compared to primary human lung epithelial cells [25]. Our main finding was that several chemokines and chemokine receptors were significantly affected following a sub-lethal exposure (a dose at which the cell viability remained >50%) of THP-1 cells to MWCNTs. However, we did not observe the induction of proinflammatory cytokines, such as TNF-α or IL-8, at the same concentration (25 μg/mL), though MWCNT triggered TNF-α secretion was observed in THP-1 cells when tested at higher doses.

The cytotoxicity of MWCNTs towards THP-1 cells could be related to the shape of these NMs. Indeed, previous investigations have shown that long and needle-like NM401, but not short and tangled NM400 or NM402, elicited a dose-dependent loss of cell viability [12]. Furthermore, in vivo studies have demonstrated that "rod-like" MWCNTs are prone to triggering pulmonary responses with fibrosis and granuloma formation [26–28]. However, other properties or features of MWCNTs in addition to their geometric characteristics may also play a role, including the presence of metallic impurities [29–31]. We also found that Ag NMs (NM300K) triggered a loss of cell viability in THP-1 cells, while a minor effect was observed in A549 cells at the highest concentrations tested, i.e., 75 and 100 μg/mL. For the latter NMs, cellular uptake and subsequent dissolution of the particles within cells has been identified as one of the key determinants of cytotoxicity [32,33]. Hence, for different NMs, different physicochemical properties may come into play.

It is interesting to consider whether the MWCNTs might emulate some other substrate(s) to which macrophages are programmed to respond. This is presently a matter of conjecture, but we have previously reported that SWCNTs induced chemokine secretion in primary human monocyte-derived macrophages through a Toll-like receptor (TLR) dependent signaling pathway [16], and a very recent study has provided evidence that the phosphatidylserine (PS) receptor Tim4 (T cell immunoglobulin mucin 4) contributes to the recognition of MWCNTs by murine peritoneal macrophages and plays a role in granuloma formation [34]. Thus, it appears that CNTs might "hijack" pattern recognition receptors that are otherwise deployed by macrophages to respond to microbes or dying cells.

Greco and co-workers recently employed toxicogenomics approaches to study the impact of carbon-based nanomaterials, including MWCNTs on various human cell lines [35,36]. The authors demonstrated that A549 cells are less sensitive than THP-1 cells, as evidenced by the magnitude of the molecular events [35]. In a follow-up study, the authors exposed THP-1 cells to long and rigid MWCNTs and studied genome wide transcription and gene promoter methylation in tandem [36]. Interestingly, among the 220 genes that were found to be affected both at the expression and methylation level, several chemokine encoding genes were identified. In the present study, both CCL17 and CCL22 were significantly upregulated in THP-1 cells exposed to MWCNTs. These chemokines are known to be highly expressed in the thymus and to a lesser extent by dendritic cells and macrophages in secondary lymphoid tissues [37]. Furthermore, both chemokines signal through CCR4, and both have been implicated in type 2 immune responses and were shown to play a role in asthma and in atopic dermatitis [38]. Moreover, we found that CCL2 was significantly induced by MWCNTs. CCL2 is a chemokine, which mediates monocyte chemotaxis and is involved in monocyte infiltration in inflammatory diseases such rheumatoid arthritis [39]. CCL2 acts as a ligand for CCR2, and it is notable that the gene encoding the latter receptor was the most highly upregulated gene in the present study. In a previous in vitro study, we provided evidence that the secretion of CCL3 and CCL5 by primary human monocyte-derived macrophages exposed to endotoxin-free SWCNTs occurred through a TLR2/4-MyD88-NF*κ*B signaling pathway [16]. Furthermore, other investigators have

shown that CCR5 (the receptor for CCL3, CCL4 and CCL5) plays an important role in the resolution of pulmonary inflammation in mice exposed toi SWCNTs [40]. Snyder-Talkington et al. [41] investigated in vivo responses to MWCNTs by microarray analysis and could show that several chemokine encoding genes were deregulated in the lungs of mice. Using small airway epithelial cells, the authors reported concordant results in vitro with regard to CCL2. Other investigators have shown, using the murine macrophage-like cell line J774A.1, that long (>20 μm) MWCNTs elicited CCL2 (MCP-1) secretion, even at a relatively low concentration [42]. Taken together, single- and multi-walled CNTs prominently affect chemokine signaling in immune-competent cells and especially CCL2 has been highlighted in several studies. The secretion of chemokines such as CCL2 may play a role in the granuloma formation that has been reported in the lungs following pulmonary exposure or in the pleural or abdominal cavity following intra-pleural or intra-peritoneal instillation of MWCNTs [43,44]. Moreover, Sydlik et al. [45] evaluated the biocompatibility of graphene oxide (GO) following implantation in subcutaneous and intraperitoneal tissues in mice and demonstrated a typical "foreign body" reaction (i.e., granuloma formation). The authors found that cells retrieved from these sites secreted significant amounts of monocyte chemotactic protein-1 (MCP-1) (CCL2) and macrophage inflammatory protein-1β (MIP-1β) (CCL4), providing further evidence for a role of these inflammatory chemokines in granulomatous tissue reactions. This also shows that CCL2 (MCP-1) upregulation is not specific for CNTs and is more likely part of a conserved response towards offending pathogens or xenobiotics. Indeed, we previously reported that *CCL2* was upregulated almost 90-fold in the lungs of rats exposed to CuO NMs and these findings were corroborated at the protein level [46]. Moreover, welding-related NMs (essentially, oxides of Fe, Mn and Cr) were found to induce the production of CCL2 in THP-1 cells [47]. Hence, CCL2 may be a particularly sensitive biomarker of immunological perturbations triggered by a variety of NMs.

In addition, we found that *NLRP12* was downregulated in THP-1 cells exposed to MWCNTs, and this was confirmed at the protein level. NLRP12 belongs to the NLR family of proteins involved in inflammasome activation [48]. However, unlike NLRP3, which plays an important role in IL-1β activation in response to a variety of stimuli including MWCNTs [49], NLRP12 seems to attenuate inflammation by dampening NF-κB signaling [50,51]. NLRP12 may also maintain intestinal homeostasis by modulating the gut microbiome [52]. Further studies are needed to investigate the role(s) of NLRP12 for MWCNT-triggered immune responses, and it is worth noting that NLRP12 may impinge on neutrophil function [53]. Neutrophils are a somewhat neglected cell type in nanotoxicology [54].

#### **5. Conclusions**

Using a combination of transcriptomics approaches and conventional biological assays, we have shown that sub-lethal doses of MWCNTs trigger a deregulation of chemokines and chemokine receptors in a human macrophage-differentiated cell line. These results support the emerging view that CNTs may elicit or interfere with chemokine signaling, and further our understanding of the toxicity of this class of materials [55,56]. However, it is noted that we have only studied one type of MWCNTs, and one cannot extrapolate the findings to other types of single or multi-walled CNTs. Indeed, grouping all CNTs into one material category is scientifically unjustified and may hinder innovation [57,58]. Therefore, further studies are needed to fully address the physicochemical properties that are responsible for the observed biological or toxicological effects of CNTs [10].

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/nano11040883/s1, Table S1: Physicochemical characterization of the selected NMs. Figure S1: TEM imaging of cells exposed to NM401. Figure S2: Pathway analysis of transcriptomics data from cells exposed to NM104. Figure S3: Pathway analysis of transcriptomics data from cells exposed to NM300K. Figure S4: Cytokine release in cells exposed to selected NMs. Figure S5: Chemokine release in cells exposed to selected NMs.

**Author Contributions:** Conceptualization, coordination: B.F.; funding acquisition: B.F. and L.T.; investigation and data analysis (in vitro studies): F.T.A. and S.K.; investigation and data analysis (omics analysis): A.G., W.C. and K.R.; writing, original draft: B.F.; writing, editing: B.F. and S.K. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the European Commission through the FP7 project MARINA (grant agreement no. 263215) and the Horizon 2020 project BIORIMA (grant agreement no. 760928).

**Data Availability Statement:** The transcriptomics data were deposited at the NCBI BioProject database (PRJNA286067). Any other results are available from the authors upon reasonable request.

**Acknowledgments:** We thank the Joint Research Centre (JRC) for providing NMs, and Lars Haag, Electron Microscopy Core Facility, Karolinska Institutet, for assistance with TEM imaging. We also wish to thank Anil Sharma, Mayo Clinic, and visiting scientist at Karolinska Institutet, for helpful discussions, and Olesja Bondarenko, Karolinska Institutet, for assistance in collecting samples for omics analysis.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **Serum Lowers Bioactivity and Uptake of Synthetic Amorphous Silica by Alveolar Macrophages in a Particle Specific Manner**

**Martin Wiemann 1,\*, Antje Vennemann 1, Cornel Venzago 2, Gottlieb-Georg Lindner 3, Tobias B. Schuster <sup>2</sup> and Nils Krueger <sup>2</sup>**

	- <sup>3</sup> Evonik Operations GmbH, Brühler Straße 2, 50389 Wesseling, Germany; gottlieb-georg.lindner@evonik.com
	- **\*** Correspondence: martin.wiemann@ibe-ms.de; Tel.: +49-251-9802340

**Abstract:** Various cell types are compromised by synthetic amorphous silica (SAS) if they are exposed to SAS under protein-free conditions in vitro. Addition of serum protein can mitigate most SAS effects, but it is not clear whether this is solely caused by protein corona formation and/or altered particle uptake. Because sensitive and reliable mass spectrometric measurements of SiO2 NP are cumbersome, quantitative uptake studies of SAS at the cellular level are largely missing. In this study, we combined the comparison of SAS effects on alveolar macrophages in the presence and absence of foetal calf serum with mass spectrometric measurement of 28Si in alkaline cell lysates. Effects on the release of lactate dehydrogenase, glucuronidase, TNFα and H2O2 of precipitated (SIPERNAT® 50, SIPERNAT® 160) and fumed SAS (AEROSIL® OX50, AEROSIL® 380 F) were lowered close to control level by foetal calf serum (FCS) added to the medium. Using a quantitative high resolution ICP-MS measurement combined with electron microscopy, we found that FCS reduced the uptake of particle mass by 9.9% (SIPERNAT® 50) up to 83.8% (AEROSIL® OX50). Additionally, larger particle agglomerates were less frequent in cells in the presence of FCS. Plotting values for lactate dehydrogenase (LDH), glucuronidase (GLU) or tumour necrosis factor alpha (TNFα) against the mean cellular dose showed the reduction of bioactivity with a particle sedimentation bias. As a whole, the mitigating effects of FCS on precipitated and fumed SAS on alveolar macrophages are caused by a reduction of bioactivity and by a lowered internalization, and both effects occur in a particle specific manner. The method to quantify nanosized SiO2 in cells is a valuable tool for future in vitro studies.

**Keywords:** nanomaterials; synthetic amorphous silica; in vitro testing; NR8383 alveolar macrophage; ICP-MS analysis of cell bound SiO2

### **1. Introduction**

Synthetic amorphous silica (SAS) form a major group of industrially relevant nanomaterials (NMs) [1–3]. They are produced either from aqueous solutions of sodium silicate to form colloidal silica, silica gels, and precipitated silica, or may be synthesized from the gaseous phase of SiCl4 to form fumed (pyrogenic) silica [3,4]. Due to the production process, pyrogenic and precipitated silica form indivisible aggregates, which have no physical boundaries among their primary structures. These aggregates have external dimensions that are highly variable in nature with some particles being in the nano-size range [3,5]. Since most SAS materials come as dry powders, a non-intentional uptake into the body may occur via inhalation [6,7]. Animal studies have shown that SAS can induce transient inflammatory responses in the rat lung [6–10] whereas fibrogenic or genotoxic effects, as induced by crystalline silica, such as quartz or cristobalite, were not induced even at high lung burden [11,12].

**Citation:** Wiemann, M.; Vennemann, A.; Venzago, C.; Lindner, G.-G.; Schuster, T.B.; Krueger, N. Serum Lowers Bioactivity and Uptake of Synthetic Amorphous Silica by Alveolar Macrophages in a Particle Specific Manner. *Nanomaterials* **2021**, *11*, 628. https://doi.org/10.3390/ nano11030628

Academic Editor: Eleonore Fröhlich

Received: 24 January 2021 Accepted: 26 February 2021 Published: 3 March 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

To describe and predict the bioactivity of the multitude of SAS, in vitro assays using different cell types and incubation conditions have been published [13–17]. Recently, the effects of SAS from all major production processes were tested using rat alveolar macrophages (NR8383) in vitro [18]. A major finding was that SAS, irrespective of the production method, elicit highly uniform responses, e.g., with respect to the release of lactate dehydrogenase (LDH), glucuronidase (GLU), tumour necrosis factor alpha (TNFα) or induction of H2O2 release [18]. In these experiments, cells were exposed to SAS under protein-free standard conditions, i.e., in the absence of serum proteins. This is a wellestablished testing procedure reflecting the circumstance that particles entering into lung alveoli are primarily protein-free entities. Furthermore, the protein-free in vitro conditions allow to study particle effects linked, e.g., to surface properties of nanoparticles [19]. However, for most SAS, the half maximal effective concentration (EC50) for several endpoints is comparatively low under protein-free conditions (e.g., 10–20 μg/mL for the release of LDH) [18], a finding which does not correspond to the in vivo effects of SAS especially when effects of amorphous and crystalline silica are compared [12]. On the other hand, the bioactivity of SAS in vitro is strongly reduced when particles (i) are administered in the presence of serum [20], (ii) are protein-treated prior to exposure [21–24] or (iii) are dispersed by extensive ultrasonic energy in the presence of low concentrations of albumin [25]. All aforementioned protein-treatments inevitably lead to the formation of a protein corona [26,27]. However, they also influence particle dispersion, gravitational settling and uptake by cells. Studies with fluorescent silica probes showed a decreased gravitational settling in the presence of protein, whereas larger precipitates of SAS settled onto the cells in the absence of protein [21,22,28]. The direct contact with the cell membrane and/or the uptake of particles is a prerequisite at least for poorly soluble particles to elicit cellular effects. Therefore, a lower rate of gravitational settling can reduce the effects of nanoparticles under submersed in vitro conditions [29]. Interestingly, the addition of protein has no uniform effect on the uptake of SAS particles and this appears to be especially relevant for macrophages. In RAW264.7 macrophages, addition of FCS enabled the uptake of SAS but mitigated cytotoxic effects [21]. In contrast, albumin-coating of NM-200, a precipitated SAS and of the pyrogenic NM-203 led to more cytotoxicity in THP-1 cells, but lowered the cytotoxicity in RAW264.7 cells [25]. Another unexpected result was provided by Binnemars–Postma and co-workers, who showed that human M1 macrophages ingested more silica nanoparticles in the absence of serum, whereas the presence of serum increased the uptake of SAS by M2 macrophage [30]. Although not yet fully elucidated, these discrepancies may be due to the presence of surface receptors involved in particle uptake, such as the scavenger receptor expressed in RAW264.7 cells [31]. Together, these results show that the effects of FCS on bioactivity and uptake of SAS are cell-type specific and need to be explored more thoroughly, especially when different types of macrophages are compared.

In the present paper, we aim to close this gap for NR8383 alveolar macrophages from rat lung using fumed and precipitated SAS with small and large specific surface areas, which were chosen to represent the large number of different SAS on the market. The well-established NR8383 cell line is widely used to analyse the bioactivity of particles in the lung: in the so-called macrophage model, NR8383 cells are exposed to settling particles under submersed conditions. The particles' bioactivity in the lung is then predicted from a set of assays carried out with the cell culture supernatant [19,32–36]. However, because SAS particles are small, their gravitational settling followed by cellular uptake may be incomplete counteracting a reliable dosimetry. Although elaborated models may predict the fraction of gravitationally settled of SAS in a time-dependent manner [37], there is still a need for quantitative measurements of SAS in cells. The quantification of SAS in cells requires a sensitive and valid analytical method. Since the mass spectrometric measurement of silicon suffers from N2 interference, an instrument with a high mass resolution is needed. Moreover, the conventional solubilisation of silica particles in organic matter with hydrofluoric acid (HF) is subject to critical handling guidelines and requires specialized

labs and personnel [38]. Recently, Bossert et al. (2019) proposed a HF-free hot alkaline lysis followed by acid treatment to dissolve silica particles; silicon was then detected by ICP-OES or by a colorimetric method with a limit of detection (LOD) being in the range of 40–100 mg/L [39]. Of note, cells exposed in vitro to approximately 10 μg SAS per mL likely underscore this LOD by at least one order of magnitude, unless the cell number is scaled up to very high amounts. To meet more conventional dimensions of cell culture testing, i.e., testing of several million cells per well, here we present a sufficiently sensitive method for SAS quantification, combining alkaline dissolution, acid neutralization and high resolution ICP-MS. By this, the amount of cellular SAS was measured and compared with the subcellular distribution of SAS in NR8383 cells in the absence and presence of foetal calf serum (FCS). This way the effect of FCS, known to reduce the bioactivity of SAS on macrophages in vitro, could be analysed for the first time on the basis of quantitative uptake data.

#### **2. Materials and Methods**

*2.1. Materials*

The materials were provided by Evonik Operations GmbH (Hanau-Wolfgang, Germany) as dry powders and had been extensively characterized in a previous publication [18]. The main data are summarized in Table 1.

**Table 1.** Material properties of the synthetic amorphous silica (SAS) used in the study.


(1) As measured by transmission electron microscopy. (2) Specific surface area measured by N2 adsorption. (3) Zeta potential and point of zero charge. (4) Measured according to enhanced OECD 105 Test Guideline on solubility. (5) Measured in 5% solution; (6) a measure for the number of silanol groups on the surface of silica according. For details, see [18].

#### *2.2. Preparation of Particle Suspensions*

The main goal of the study was to compare effects of SAS in the absence and presence of foetal calf serum, after particles had been dispersed according to a previously established protocol [18]. In brief, particles were suspended in sterile H2O (Aqua ad injectabilia, Braun Melsungen, Germany) at a concentration of 2 mg/mL. Suspensions were vortexed, stirred with a magnetic bar for 90 min and passed through a sterile polyamide gauze with a nominal pore width of 5 μm (Bückmann, Mönchengladbach, Germany) (see [4] for filtration characteristics). 5 mL of each filtrate was then transferred to a 20 mL glass vial and subjected to an ultrasonic dispersion energy 270 J/mL [4]. Dispersed masses amounted to 70–100% the original masses, as determined by gravimetric analysis. The stock aqueous stock suspensions of SIPERNAT® 160, SIPERNAT® 50, AEROSIL® OX50 and AEROSIL® 380 F were adjusted to 360 μg/mL, and stored at 4 ◦C for up to 4 weeks. Immediately before experiments, aqueous stock suspensions were mixed with an equal volume double concentrated KRPG buffer (see below) or F-12K media to obtain a physiologic medium composition.

#### *2.3. Cultivation of NR8383 Macrophages and Cell Culture Assays*

NR8383 cells (ATCC, Manassas, VA, USA; ATCC® Number: CRL-2192TM) were maintained in F-12K cell culture medium (Sigma-Aldrich, Taufkirchen, Germany) supplemented with 15% foetal calf serum (FCS), 1% penicillin/streptomycin and 1% L-glutamine (all from PAN Biotech, Aidenbach, Gremany) under cell culture conditions (37 ◦C and 5% CO2) [32]. For the assay, 3 × <sup>10</sup><sup>5</sup> cells were seeded per well of a 96-well plate, and covered with 200 μL F-12K cell culture medium plus 5% (*v*/*v*) FCS to foster cell adherence. The

next day, the medium was replaced by serum-free or FCS (10%)-containing F-12K medium containing increasing concentrations of each material (11.25, 22.5, 45 or 90 μg/mL). After 16 h, supernatant was retrieved to determine LDH, GLU and TNFα. To measure the release of H2O2, materials were equivalently diluted in KRPG buffer (129 mM NaCl, 4.86 mM KCl, 1.22 mM CaCl2, 15.8 mM NaH2PO4, 5–10 mM glucose; pH 7.3–7.4).

Assays were carried out as described [32]. In brief, H2O2 was quantified with the Amplex Red® assay by photometrically measuring formed resorufin at 570 nm (reference value: 620 nm) with a plate reader (Tecan Infinite F200Pro, Tecan GmbH, Crailsheim, Germany); positive controls were run with 360 μg/mL zymosan (Sigma-Aldrich, Taufkirchen, Germany). Measurements were corrected for background absorbance of cell free-particle controls and converted into concentrations of H2O2 as described. LDH activity was measured with the Roche Cytotoxicity Kit (Sigma-Aldrich, Taufkirchen, Germany) according to the manufacturer's protocol. GLU activity was detected with p-nitrophenyl-D-glucuronide dissolved in 0.2 M sodium acetate buffer (pH 5) containing 0.1% Triton X-100. Both the LDH- and GLU-based values were corrected for cell-free absorption and normalised to the positive control (0.1% Triton X-100 in F-12K) which was set to 100%. Tumour necrosis factor α (TNF α) was determined with a specific enzyme-linked immosorbent assay (ELISA) for rat TNFα (Quantikine ELISA Kit) according to the manufacturer's protocol (Bio-Techne GmbH, Wiesbaden-Nordenstadt, Germany). The TNFα-forming capacity of NR8383 cells was tested with 0.5 μg/mL lipopolysaccharide (LPS, Sigma-Aldrich, Taufkirchen, Germany). Notably, aliquots for measuring LDH, GLU and TNFα were taken from the same well.

#### *2.4. Particle Size Determination under Cell Culture Conditions with Particle Tracking Analysis*

The hydrodynamic diameter was determined by optical tracking analyses using a NanoSight LM10 instrument equipped with a violet laser (405 nm), an Andor CCD camera, and particle tracking software NTA3.0 (all from: Malvern Instruments GmbH, Herrenberg, Germany). Starting with the aqueous particle stock suspension, dilutions were prepared in KRPG and F-12K medium in the absence and presence of 10% FCS. Concentration was uniformly set to 90 μg/mL. Suspensions were incubated under cell culture conditions (37 ◦C, 5% CO2, 100% humidity) for 90 min (KRPG) and 16 h (F-12K), respectively. Suspensions were further diluted to obtain measurable concentrations, approximately in the range of <sup>5</sup> × <sup>10</sup><sup>8</sup> particle/mL.

#### *2.5. Electron Microscopy of NR8383 Macrophages*

NR8383 cells were seeded onto small discs (diameter 6 mm) of Melinex film (Plano, Wetzlar, Germany) placed in the wells of a 96-well plate, subjected to particle treatment for 16 h as described in paragraph 2.4. Then, media were withdrawn and cells were immediately covered with 2.5% glutardialdehyde in 0.1 M sodium phosphate buffer (SPB, pH 7.3) for 60 min. Cells were washed three times with SPB, post-fixed in 1% OsO4, dehydrated in ethanol to the 70% step, and stained en bloc with uranium acetate (1%). Cells were dehydrated via ethanol/propylene oxide, and embedded in Epon 812 (Sigma Aldrich, Taufkirchen, Germany). Ultrathin sections (50–60 nm) were viewed with a Tecnai G2 electron microscope operated at 100 or 120 kV; images were taken with a Quemesa digital camera (Olympus Soft Imaging Solutions, Münster, Germany).

#### *2.6. Quantification of Cell-Associated SiO2 NP by High Resolution ICP-MS*

To quantitate the amount of cell-associated SiO2 NP, 2.8 × 106 NR8383 cells were seeded into each well of a 6-well plate and incubated in 6 mL F-12K medium containing 11.25 μg/mL of each SiO2 samples, both in the absence and presence of 10% FCS. After 16 h, the culture medium was completely withdrawn and replaced by 1 mL phosphate buffered saline (PBS). Cells were detached from the plates by vigorous pipetting and transferred into a 15 mL test tube (Falcon) pre-loaded with 6 mL fresh F-12K medium. Cells were spun down (200× *g*, 10 min), washed with 2 mL KRPG buffer, pelleted again (200× *g*, 10 min), and finally re-suspended in 130–230 μL KRPG buffer. A defined volume (90%) of this final suspension was dehydrated at 60 ◦C for 12 h for analysis.

To measure the SiO2 content, each sample was dissolved with 50 μL of 20% NaOH and heated to 120 ◦C for 2 h. Lysates were diluted with 5 mL ultrapure H2O. In further dilution steps, 4% (*v*/*v*) of a HNO3 solution was added. Measurements of Si were carried out with a double focusing magnetic sector field ICP-MS instrument in the medium resolution mode (Element 2™, Thermo Fisher Scientific, Meerbusch, Germany) equipped with a quartz spray chamber and a quartz injection device (sample loop: 500 μL, ESI-Fastvalve). 115In was used as an internal standard. To calibrate the ICP-MS signal for 28Si, a Si standard stock solution of 100 mg/L (Labkings, Hilversum, The Netherlands) was diluted with 4% (*v*/*v*) of a HNO3-solution and an equivalent amount of NaOH as mentioned above to obtain a calibration range of 0.5–100 μg/L. Final SiO2 concentrations were obtained by multiplying measured Si concentrations with the stoichiometric factor of 2.1393.

#### *2.7. Statistical Evaluation*

For in vitro testing, i.e., effects on LDH, GLU, H2O2, and TNFα, three independent repetitions were carried out; data were expressed as mean ± standard deviation (SD). To find significant differences, values from each concentration step were compared to the respective vehicle-treated control using 2-way analysis of variance (ANOVA) with Dunnett's multiple comparisons test. Calculations were carried out with GraphPad Prism software. A value of *p* ≤ 0.05 was considered significant. Calculation of hydrodynamic diameters were carried out with NTA 3.0 software.

#### **3. Results**

#### *3.1. Particle Characterization*

Two precipitated and two fumed SAS were selected for this study, whose major physical-chemical properties are shown in Table 1. A smaller and a larger particle type was selected for each group, also reflected by the BET values which, overall, span more than one order of magnitude. Acidity of the SAS powders was low but more pronounced for both AEROSIL®s. The SEARs No., which is a measure for the number of silanol groups at the particles' surface, was similar for three SAS but lower for AEROSIL® OX50. Solubility was very similar (112.1 to 117.9 mg/L) but slightly higher for AEROSIL® 380 F.

The addition of FCS to a SAS dispersion is likely to change the particles' surface charge and agglomeration behaviour which is highly relevant for in vitro testing. Therefore, the hydrodynamic diameter (HD) of all four SAS was measured with particle tracking analysis (PTA) in H2O, and in the absence and presence of 10% foetal calf serum (FCS) in KRPG and F-12K medium, to mimic testing conditions (Table S1). In the case of SIPERNAT® 160 and SIPERNAT® 50, the HD (mode values) obtained in KRPG and F-12K were only slightly increased (<15%) compared to H2O; HD values also hardly increased upon of 10% FCS. AEROSIL® 380 F particles were too small to be measured in H2O but measurable agglomerates were found in F-12K medium especially in the presence of FCS. The size of AEROSIL® OX50 agglomerates increased in KRPG only. Overall, the effects of FCS on the hydrodynamic diameter of the particle fraction measurable by PTA were low. Nevertheless, the addition of FCS led to the formation of agglomerates visible by light microscopy at the bottom of the culture vessels (Figure 1). This effect was pronounced for AEROSIL® 380 F (Figure 1a,b), but low for AEROSIL® OX50 (Figure 1e,f), SIPERNAT® 160 (Figure 1i,j) and SIPERNAT® 50 (Figure 1m,n).

#### *3.2. Quantification of Particle Uptake by NR8383 Cells*

To measure the cell-associated SAS fraction which comprises internalized plus surfacebound particulate matter, we administered the lowest concentration of SAS particles (11.25 <sup>μ</sup>g/mL) to a defined number of cells for 16 h (2.8 × <sup>10</sup><sup>6</sup> per well). As expected, this treatment led to a low amount of dead cells in the absence (4.2–10.6%) and nearly no dead cells in the presence of serum (1.5–2.5%, see Table S2). Table 2 shows the cell-associated SiO2

masses measured by ICP-MS: Control cells contained low, though measurable amounts of 28Si. All SAS-treated cells showed a cell-associated SiO2 mass above background level, which could not be lowered, e.g., by avoiding SiO2 containing cell culture material (data not shown). In the presence of 10% FCS, values for the precipitated SAS SIPERNAT® 50 and SIPERNAT® 160 were higher than those for the fumed SAS AEROSIL® OX50 and AEROSIL® 380 F and this difference was not obvious under FCS-free conditions (Table 2). The cell-associated amount of SiO2 was reduced in the presence of 10% FCS-free conditions. The effect was pronounced for AEROSIL® 380 F (−69.5%), AEROSIL® OX50 (−83.8%), and SIPERNAT® 160 (−62.3%), but comparatively small for SIPERNAT® 50 (−9.9%).

**Figure 1.** Sedimentation and uptake of precipitates by NR8383 cells which were formed in the presence of AEROSIL® 380 F (**a**–**d**), AEROSIL® OX50 (**e**–**h**), SIPERNAT® 50 (**i**–**l**) and SIPERNAT® 160 (**m**–**p**). Micrographs were taken 16 h after administration of particles (180 μg/mL) in the absence (w/o cells) and presence of cells (+NR8383). Foetal calf serum (FCS) led to the formation of agglomerates. In the absence of serum, many cells appeared deteriorated.



(1) Amounts of SiO2 in the cell pellet after 16 h, as measured by ICP-MS; values measured in duplicates are separated by a slash. (2) Cellassociated SiO2 in percent of the total added mass (67.5 μg SAS per 6 mL medium; weigh in of each SAS was individually corrected for its water content measured as loss on ignition [18]. (3) Ratios were calculated from mean values of columns 2 and 4. (4) 28Si values from control cells were assumed to represent SiO2 and were converted equivalently.

#### *3.3. In Vitro Toxicity Determination of SAS and Electron Microscopic Study*

In vitro toxicity of SAS was measured in the absence and presence of FCS with the well-established alveolar macrophage assay. The activity of lactate dehydrogenase (LDH) and glucuronidase (GLU), as well as the concentration of tumour necrosis factor α (TNFα), were determined in the cell culture supernatant. Corundum and quartz DQ12 particles

were included as negative and positive particle controls, respectively. Numerical results are shown in Table S3. The subcellular distribution of particles was investigated by transmission electron microscopy (TEM) of cells treated with the lowermost particle concentration of the study (11.25 μg/mL) and matched the particle concentration used above for the quantification experiments.

#### 3.3.1. AEROSIL® 380 F and AEROSIL® OX50

Both AEROSIL®s showed a high biologic activity under FCS-free conditions (Figure 2a,b), indicated by the dose-dependent release of LDH, GLU and TNFα. As for most SAS particles the release of H2O2 was low and became significant at the highest doses only. In the presence of 10% FCS, all aforementioned responses were abolished (H2O2) or drastically lowered, as indicated by the flattened curves for the release of LDH, GLU and TNFα. The degree of reduction and the shift in the low observed adverse effect concentration (LOAEC) are provided in Table 3.

**Figure 2.** *Cont.*

**Figure 2.** In vitro response of NR8383 alveolar macrophages to AEROSIL® 380 F and to AEROSIL® OX50 in the absence or in the presence of 10% FCS (dashed lines). Lactate dehydrogenase activity (LDH), glucuronidase activity (GLU), H2O2 concentration, and tumour necrosis factor alpha (TNFα) were measured in the supernatant from NR8383 cells exposed to AEROSIL® 380 F (**a**) or AEROSIL® OX50 (**b**). Effects of zymosan and lipopolysaccharide (LPS) on the formation of H2O2 and TNFα, respectively, are indicated by vertical lines.

The TEM investigation of NR8383 cells laden with AEROSIL® 380 F in the absence of FCS cells is shown in Figure 3a–d. Although minor portions of the material were regularly found at the outer cell membrane (Figure 3b), larger assemblies occurred within phagosomes (Figure 3a,c). Small particle deposits were found in lysosomes (Figure 3d) and autophagosomes, together with condensed cellular material (Figure 3d). Of note, small and often branched aggregates/agglomerates of AEROSIL® 380 F particles occurred in the cytoplasm; neither mitochondria nor the cell nucleus were found to contain particles. In the *presence of FCS*, particles (aggregates/agglomerates) of AEROSIL® 380 F were not found at the outer cell membrane or within the cytoplasm. Additionally, heavily laden phagosomes were not found. Instead, cells contained large phagosomes filled with fine granular material of low-to-medium electron density (Figure 4a). Particles were mainly found in lysosomes (Figure 4b).

AEROSIL® OX50 was found as single particles or small groups thereof within endosomes, most likely presenting lysosomes, and phagosomes (Figure 5a–c). Although the presence of FCS did not lead to a major change of this pattern, smaller aggregates appeared to be more frequent (Figure 6a–c). Typical uptake-figures (Figure 6b) showed single particles close to a membrane invagination, suggesting that particles enter into cells via small endosomes. The material was also found in autophagosomes.

**Figure 3.** Electron microscopy of NR8383 cells laden with AEROSIL® 380 F (11.25 μg/mL) in the absence of FCS for 16 h. (**a**) Overview of a cell containing particle filled phagosomes (Ph) and branched particle assemblies in distinct regions of the cytoplasm (asterisk). The cytoplasm of this cell appears condensed. (**b**) Particles adhering to a section of the outer cell membrane (arrows). (**c**) Particle filled phagosomes; arrows point to the enclosing membrane. (**d**) An autophagosome (aPh) filled with condensed matter together with several electron dense lysosomes (arrow); both compartments contain small amounts of the typical, small electron dense AEROSIL® 380 F particles.

**Figure 4.** Electron microscopy of NR8383 cells laden with AEROSIL® 380 F (11.25 μg/mL) in the presence of FCS for 16 h. (**a**) Aspect from a cell with typical phagosomes (Ph) mainly filled with fine granular material, particle laden lysosomes and a clear vacuole (CV). (**b**) Boxed area from (**a**) showing several particle-laden lysosomes (arrows).

**Figure 5.** Electron microscopy of NR8383 cells laden with AEROSIL® OX50 F in the absence of FCS for 16 h. Electron lucent areas close to particles are interpreted as cutting artefacts. (**a**) Overview of a cells with particle-containing phagosomes (large arrows) and smaller endosomes (small arrows). (**b**) shows a typical particle-laden; (**c**) shows a higher magnification of a phagosome (large arrow) and three small endosomes (small arrows) and arrows point to membrane continuities.

**Figure 6.** Electron microscopy of NR8383 cells laden with AEROSIL® OX50 F in the presence of FCS for 16 h. Electron lucent areas close to particles are interpreted as cutting artefacts. (**a**) Overview of a cell with particle-containing phagosomes (large arrows) and smaller endosomes (small arrows). (**b**) A membrane invagination (arrow) underneath a particle attached to the cell membrane, interpreted as an early uptake figure. (**c**) A small particle-filled endosome; (**d**) shows two particle-containing autophagosomes (aPh).

#### 3.3.2. SIPERNAT® 50 and SIPERNAT® 160

The biological activity of SIPERNAT® 50 and SIPERNAT® 160 was very similar to that of the AEROSIL®s (Figure 7a,b), though the dose-dependent release of LDH, GLU and TNFα was more pronounced for SIPERNAT® 160. Again, 10% FCS abolished the H2O2 response and strongly reduced the cytotoxic effect (LDH, GLU) and also TNFα formation. There was a strong reduction of bioactivity in the presence of FCS. The shift in LOAEC upon FCS treatment are provided in Table 3.

The electron microscopic examination revealed no major differences with respect to the endosomal compartments crowed by SIPERNAT® 50 or SIPERNAT® 160 particles. Even the fine structure of agglomerates within phagosomes appeared indistinguishable (Figures S1 and S2). With respect to SIPERNAT® 160, endosomes with larger particle assemblies appeared less frequent (Figures S3 and S4).

**Figure 7.** In vitro response of NR8383 alveolar macrophages to SIPERNAT® 50 and SIPERNAT® 160 in the absence or in the presence of FCS (dashed lines). Lactate dehydrogenase activity (LDH), glucuronidase activity (GLU), H2O2 concentration and tumour necrosis factor alpha (TNFα) were measured in the supernatant from NR8383 cells exposed to SIPERNAT® 50 (**a**) or SIPERNAT® 160 (**b**). Effects of zymosan and lipopolysaccharide (LPS) on the formation of H2O2 and TNFα, respectively, are indicated by vertical lines.


**Table 3.** Apparent reduction of the bioactivity of SAS nanoparticles upon addition of FCS.

(1) Measured at 90 μg/mL by linear interpolation; (2) LOAEC with and without FCS as derived from Table S3. LDH: Lactate dehydrogenase, GLU: glucuronidase, TNFα: tumor necrosis factor α.

#### *3.4. Evaluation of Data Using the Cell-Associated SiO2 Mass as a Dose Metric*

Finally, we plotted the release of LDH, GLU and TNFα (Table S1) against the cellassociated particle mass. Except for AEROSIL® 380 F, which partly adhered to the cell surface in the absence of FCS (Figure 3), the cell-associated particle mass in fact reflects fully internalized particles. Because a meaningful determination of SiO2 uptake relative to administered SAS concentration had to rely on non-compromised cells (i.e., at low cytotoxicity), we background corrected the "% total" values from Table 3 and extrapolated them to the maximum theoretical cell burden at a given concentration step (i.e., 7.5, 15, 30 and 60 pg/cell; see Supplementary Information for calculation).

These uptake-corrected abscissa values were then plotted against the released enzyme activities the absence and presence of FCS (Figure 8). The FCS-mediated reduction in particle uptake is reflected by a shortening of the curves and the lowered slopes seen for the releases of LDH, GLU and TNFα indicate the FCS-mediated reduction of biological activity. The slope reductions (LDH, GLU and TNFα curves) appeared largely uniform for each singly SAS but differed in the order SIPERNAT® 50 > SIPERNAT® 160 = AEROSIL® 380 F. AEROSIL® OX50 was not evaluable due to a strong shortening of the respective curve. Resulting EC50 values were calculated for the FCS-free administration of SAS (Table S4) and will be discussed below. Overall, the curves shown in Figure 8 reveal that the reduced bioactivity of SAS in the presence of FCS is material dependent and due, at least in parts, to cellular processes secondary to particle uptake.

**Figure 8.** In vitro response of NR8383 alveolar macrophages to the cell-associated masses of SIPERNAT® 160, SIPERNAT® 50, AEROSIL® OX50 and AEROSIL® 380 F in the absence and presence of FCS. Values (from Table S1) for lactate dehydrogenase activity (LDH), glucuronidase activity (GLU), H2O2 concentration, and tumour necrosis factor alpha (TNFα) from were plotted against the cell-associated masses of SAS which were measured for a low SAS concentration and then extrapolated to higher values.

#### **4. Discussion**

In this investigation, we analysed the effect of FCS on the apparent bioactivity of fumed and precipitated SAS using an established alveolar macrophage model. For the first time, the uptake of SAS by alveolar macrophages was quantified with a high resolution ICP-MS technique. This enabled us to attribute the effect of FCS to both an influence on particle adhesion and subsequent uptake by cells, and an influence on the bioactivity of the cell-associated, i.e., ingested SAS material.

The reduction of particle uptake in the presence of FCS was largest for both AEROSIL®s (69.5 to 83.5%): TEM analyses strongly suggest that the lower content of the AEROSIL®s in cells in the presence of FCS was mainly due to less particles captured within large phagosomes, whereas lysosomes or autophagosomes contained similar loads of particles under both conditions and, therefore, appear to be of minor relevance for the mitigating effect of FCS on the SAS effects. This appears to be different from the changing numbers of autophagosomes and lysosomes observed in alveolar macrophages treated with crystalline silica [40]. The failure to form larger particle-filled phagosomes may by explained, at least in part, by an absence of binding of AEROSIL® 380 F to the cell surface in the presence of FCS. The lack of large particle-filled and afterwards disrupted phagosomes may have also prevented AEROSIL® 380 F particles from entering into the cytoplasm under protein-free

conditions, a mode of particle uptake strongly suggested by the patchy distribution pattern of SiO2 nanoparticles (Figure 3a). The reduced uptake of AEROSIL® 380 F in the presence of FCS is seemingly in contrast to the strongly increased aggregate/agglomerate size (up to several hundred micrometers, see Table S1) and to the large number of precipitates visible with phase contrast optics (Figure 1a,b). We assume, however, that these particulates mainly consist of precipitated proteins which were cleared from the culture bottom most likely via ingestion by the NR8383 cells. Evidence for this assumption comes from numerous large vacuoles filled with low-contrast material possibly representing protein but only few silica particles (Figure 4a). Overall, the presence of FCS led to an enlargement of the AEROSIL® 380 F particles' HD, a reduced uptake into phagosomes, followed by an altered subcellular localization and reduced the biological activity.

The influence of FCS on the uptake of both precipitated SAS was not uniform. While FCS reduced the uptake of SIPERNAT® 160, most likely again by reducing the formation of particle-filled phagosomes, it had nearly no influence on the uptake of SIPERNAT® 50. The lack of ultrastructural changes seen for SIPERNAT® 50 upon FCS treatment was in line with the quantitative ICP-MS measurement. Therefore, the pronounced reduction of biologic activity seen for SIPERNAT® 50 was more directly attributable to a change of biological particle properties. Of note, SIPERNAT® 50 exhibited the smallest primary structures, the largest specific surface area, and also the highest number of reactive silanol groups as reflected by SEAR's number (see Table 1). Reactive silanol groups of SAS are believed to contribute to the bioactivity of crystalline silica [41,42]. A protein corona, which will inevitably form around SAS in the presence of FCS [26,27], may keep biological structures at a distance from reactive structures at the SAS surface and this protective effect, as illustrated by the reduced slopes of LDH, GLU and TNFα release (Figure 8), may be larger for highly reactive SAS.

Interestingly, the reduction of biological activity by FCS found for ingested AEROSIL® OX50 appeared to be solely due to the reduction in particle uptake (Figure 8). Unlike all other SAS nanomaterials, the slope of the shortened curve reflecting the effect of the ingested AEROSIL® OX50 was not reduced, suggesting that FCS had a minor effect on the biologic activity of ingested AEROSIL® OX50. Although a protein corona formation around this material is highly likely, its effect may be relatively small, possibly because AEROSIL® OX50 had the smallest specific surface and the smallest SEAR's number. Of note, the addition of protein to nanoparticles does not change or attenuate their biologic activity in general, as shown, e.g., for CeO2 [43], TiO2 or Fe2O3 [24]. Therefore, the specific surface reactivity of a given material needs to be taken into account if the effect of protein coating has to be predicted.

It may also be speculated that the primary particle size and/or the specific surface (BET value) correlates to the protective effect of FCS, because the highest reduction in bioactivity (80–90%) was found for AEROSIL® 380 F (see Table 3). However, despite large differences in size and/or BET surface, the mitigating effects of FCS on all other SAS were highly similar (70–80%), arguing against a simple correlation of primary particles size and FCS-mediated reduction of bioactivity. On the other hand, there may be a size and/or surface-dependent influence of FCS on the uptake of SAS, which was more reduced for AEROSIL® 380 F and also for the SIPERNAT® 160 both of which exhibit the largest specific surface of the fumed and precipitated SAS, respectively (see Table 2: Ratio FCSfree/10% FCS). However, this effect may be indirectly caused via an influence of FCS on particle settling.

In a previous test of the four SAS with the macrophage model, we found that the EC50 values for the release of LDH span a comparatively narrow range (from 13.2 μg/mL (AEROSIL® 380 F) to 31.7 μg/mL (SIPERNAT® 50). As shown here, this range becomes larger when the EC50 values are expressed as pg per cell (2.03 pg/cell (AEROSIL® 380 F) to 28 pg/cell (AEROSIL® OX50); see compilation in Table S4). The disparity of the EC50 values based on the cellular dose is likely to be more relevant because it is measured directly and is not deduced from particle agglomeration and settling. However, the knowledge of

cell-associated particle burden helps not only to compare different experimental conditions. It is also urgently needed to better compare in vitro and in vivo data with the aim to refine in vitro tests by using adequate cellular doses. At present, in vivo experiments have mostly been evaluated for organ burden of nanoparticles. Values at the single cell level are rare, although some progress has been made, e.g., for silver laden phagocytes in lymph nodes, whose silver content has been estimated to reach up to 140 pg per cell [44].

The question of which type of protein coating in vitro adequately mimics the situation of particles in the lung is still unresolved. While it is beyond dispute that nanoparticles in body fluids such as blood or extracellular fluid carry a protein corona [13,43,45], protein corona formation in the lung parenchyma is more complex. At least in theory, a respirable (SAS) particle will first contact and adsorb biomolecules of the lung surfactant (phospholipids, various surfactant proteins) before it enters into the lung lining fluid with its multitude of different proteins [46–48]. During inhalation exposure, the dose rate is typically low and the binding of surfactant and protein components to inhaled particles may be complete and more or less well-structured. In contrast, the administration of a particle-containing fluid into the lung, i.e., a high dose rate, may locally disturb the lung's surfactant layer and lead to unconventionally coated or even uncoated particles. In the case of colloidal SAS, this leads to a more intense inflammatory reaction of the rat lung during the first days after particle administration compared to inhalation exposure [36]. It is also noteworthy that in the case of an acute lung inflammation upon SAS, the protein concentration of the lung lining fluid will rise [33], which may limit the bioactivity of SAS as observed in vitro.

Based on these considerations and on the findings of this study, the way of in vitro testing being most predictive for the in vivo outcome remains a matter of discussion. Depending on the starting conditions, alveolar macrophages in vivo may engulf uncoated as well as protein-coated particles. However, we suggest that the well-established alveolar macrophage assay with NR8383 cells, which was originally developed and validated as a protein-free approach [32] and, as such, has been successfully incorporated into a tiered grouping strategy for nanomaterials [49], may be expanded for effects of proteins on the in vitro bioactivity of nanomaterials. As shown here for SAS, protein-free and protein-supplemented exposure of cells may differ substantially and the effects of both treatments should be understood as corner points possibly spanning the full range of responses of alveolar macrophages in situ. In any case, the additional inclusion of proteincontaining assays needs to be supported by a quantification of nanomaterials' uptake to avoid unwarranted conclusions. To this end, the method introduced here is applied as a reliable tool to quantify SAS nanomaterials at the cell culture level.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/2079-499 1/11/3/628/s1, Figure S1. Electron microscopy of NR8383 cells laden with SIPERNAT® 50 in the absence of FCS for 16 h. Figure S2. Electron microscopy of NR8383 cells laden with SIPERNAT® 50 in the presence of FCS for 16 h. Figure S3. Electron microscopy of NR8383 cells laden with SIPERNAT® 160 in the absence of FCS for 16 h. Figure S4. Electron microscopy of NR8383 cells laden with SIPERNAT® 160 in the presence of FCS for 16 h. Table S1. Hydrodynamic diameter of SAS in H2O and cell culture media. Table S2. Trypan Blue Exclusion Test and cell numbers of SAS-treated cells used for mass spectrometric. Table S3. Numerical values from in vitro tests with the Alveolar Macrophage Model. Table S4. EC50 values for cell-associated SAS causing the release of LDH, GLU and TNFα. Table S5. EC50 values calculated for extrapolated intracellular SAS concentrations.

**Author Contributions:** M.W., T.B.S., G.-G.L. and N.K. designed the study, A.V. and M.W. conceived and performed the cell experiments and wrote the manuscript. C.V. performed the quantification of SAS. All authors contributed to the final manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** The study was sponsored by Evonik Industries GmbH. M.W. and A.V. received funding from the German Federal Ministry of Education and Research (BMBF), grant number 03XP0213A.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data presented in this study are available in the article or in the Supplementary Material.

**Acknowledgments:** The cell culture work of Oliver Gräb is gratefully acknowledged.

**Conflicts of Interest:** M.W. and A.V. declare no conflict of interest. G.-G.L., T.B.S., C.V., and N.K. are employees of Evonik Industries GmbH, a company which produces and sells amorphous silica products.

#### **References**

