FULL PAPER www.advancedscience.com Multiparametric Profiling of Engineered Nanomaterials: Unmasking the Surface Coating Effect Audrey Gallud, Mathilde Delaval, Pia Kinaret, Veer Singh Marwah, Vittorio Fortino, Jimmy Ytterberg, Roman Zubarev, Tiina Skoog, Juha Kere, Manuel Correia, Katrin Loeschner, Zahraa Al-Ahmady, Kostas Kostarelos, Jaime Ruiz, Didier Astruc, Marco Monopoli, Richard Handy, Sergio Moya, Kai Savolainen, Harri Alenius, Dario Greco, and Bengt Fadeel* 1. Introduction Despite considerable efforts, the properties that drive the cytotoxicity of engineered nanomaterials (ENMs) remain poorly understood. Here, the Considerable progress has been made authors inverstigate a panel of 31 ENMs with different core chemistries and a in recent years with respect to hazard assessment of engineered nanomaterials variety of surface modifications using conventional in vitro assays coupled (ENMs), and advanced methods including with omics-based approaches. Cytotoxicity screening and multiplex-based high-throughput and high-content screen- cytokine profiling reveals a good concordance between primary human ing platforms as well as omics-based ap- monocyte-derived macrophages and the human monocyte-like cell line proaches are gaining traction.[1] However THP-1. Proteomics analysis following a low-dose exposure of cells suggests a careful study design and appropriate data analysis tools are required to maximize the nonspecific stress response to ENMs, while microarray-based profiling benefit of these emerging technologies.[2] reveals significant changes in gene expression as a function of both surface Indeed, despite the great promise of high- modification and core chemistry. Pathway analysis highlights that the ENMs throughput approaches,[3] the limitation in with cationic surfaces that are shown to elicit cytotoxicity downregulated DNA most studies is the small number of ENMs, replication and cell cycle responses, while inflammatory responses are making it difficult to draw conclusions re- upregulated. These findings are validated using cell-based assays. Notably, garding the material properties that drive the biological responses to ENMs.[4] More- certain small, PEGylated ENMs are found to be noncytotoxic yet they induce over, another challenge that is often over- transcriptional responses reminiscent of viruses. In sum, using a looked is the fact that biology is complex. multiparametric approach, it is shown that surface chemistry is a key Indeed, it appears that toxicologists often determinant of cellular responses to ENMs. The data also reveal that apply the same “paradigm” for all ENMs cytotoxicity, determined by conventional in vitro assays, does not necessarily rather than seeking to understand biologi- cal responses to ENMs in all their complex- correlate with transcriptional effects of ENMs. ity (which is not to say that ENMs necessar- ily trigger novel responses—the pathways A. Gallud, M. Delaval, H. Alenius, B. Fadeel P. Kinaret, V. S. Marwah, D. Greco Institute of Environmental Medicine Faculty of Medicine and Health Technology Karolinska Institutet Tampere University Stockholm 171 77, Sweden Tampere 33720, Finland E-mail: bengt.fadeel@ki.se P. Kinaret, V. S. Marwah, H. Alenius, D. Greco Institute of Biotechnology University of Helsinki Helsinki 00790, Finland V. Fortino Institute of Biomedicine The ORCID identification number(s) for the author(s) of this article University of Eastern Finland can be found under https://doi.org/10.1002/advs.202002221 Kuopio 70211, Finland © 2020 The Authors. Published by Wiley-VCH GmbH. This is an open J. Ytterberg, R. Zubarev access article under the terms of the Creative Commons Attribution Department of Medical Biochemistry & Biophysics License, which permits use, distribution and reproduction in any Karolinska Institutet medium, provided the original work is properly cited. Stockholm 171 77, Sweden DOI: 10.1002/advs.202002221 Adv. Sci. 2020, 7, 2002221 2002221 (1 of 18) © 2020 The Authors. Published by Wiley-VCH GmbH www.advancedsciencenews.com www.advancedscience.com involved may very well be conserved).[5] From a practical stand- gene expression.[11,12] Such studies serve to complement, but can- point, there is a trade-off between simple and robust assays not replace, toxicological studies using conventional assays. which are amenable to high-throughput screening[6] versus com- In the present study, we conducted comprehensive in vitro prehensive, omics-based approaches that promise a wealth of testing of a panel of 31 ENMs, including cytotoxicity screening information—though the latter studies may ultimately not be and cytokine profiling, using primary human monocyte-derived informative if the results are not validated. The seminal profil- macrophages (HMDMs) and the human monocytic THP-1 cell ing study by Shaw et al.[7] provides a case in point. The authors line. Furthermore, proteomics and transcriptomics assessment conducted high-content screening of a large panel of ENMs of of THP-1 cells exposed to a low dose (EC10 ) of ENMs was per- varying core compositions and surface modifications and were formed along with detailed bioinformatics analysis of the data. able to derive robust structure-activity relationships by using this Finally, validation experiments were performed to verify the im- approach. The cell-based assays applied were reflective of cell pact of ENMs on specific cellular pathways. Taken together, these death and oxidative stress, two of the most commonly studied results have provided a systematic overview of nano–bio interac- endpoints in nanotoxicology. Hence, by definition, such stud- tions and served to shed light on the role of chemical composition ies, though they provide a solid basis for hazard screening, can- and surface modifications of ENMs. Our results also show that not provide information on alternative cell fates, such as low- ENMs that appear biologically inert when assessed using con- dose and/or long-term effects on cell function in the absence of ventional toxicity assays can still yield striking low-dose effects cell death.[8–10] Multi-omics approaches, combining, for instance, on cells. proteomics and transcriptomics methods, coupled with bioinfor- matics analysis of the data, afford an opportunity to study the bi- 2. Results ological responses to ENMs in an unbiased manner, taking into account all the changes that occur at the level of protein and/or 2.1. Experimental Design Faria et al.[13] recently put forward a suggestion for a “minimum information standard” for publications on nano–bio interactions, focusing on three categories namely material characterization, T. Skoog, J. Kere Department of Biosciences & Nutrition biological characterization, and experimental protocols (in other Karolinska Institutet words, materials, models, and methods). The aim is to improve Huddinge 141 83, Sweden reproducibility and to enable comparisons between publications M. Correia, K. Loeschner in the field. The true value of such reporting requirements lies in National Food Institute their adoption by the scientific community and one should seek Technical University of Denmark Kongens Lyngby 2800, Denmark to place a reasonable burden on scientists in order to achieve Z. Al-Ahmady, K. Kostarelos compliance.[14] In the present study, conducted in the frame of Faculty of Biology the European Commission-funded NANOSOLUTIONS project, Medicine & Health close attention was paid to the selection of materials, models, University of Manchester and methods. In brief, we prepared an extensive panel of ENMs Manchester M20 4GJ, UK comprising eight different core chemistries encompassing both Z. Al-Ahmady metal, metal oxide, and carbon-based materials and varying School of Science & Technology Nottingham Trent University diameters and surface modifications, i.e., amino/ammonium-, Nottingham NG1 8NS, UK carboxyl/carboxylate-, or poly(ethylene glycol) (PEG)-terminated K. Kostarelos surfaces versus pristine ENMs. The three surface modifications Catalan Institute of Nanoscience and Nanotechnology (ICN2) provide stability to ENMs in aqueous media, while they convey Barcelona 08193, Spain different charges. Hence, amino/ammonium-terminated coat- J. Ruiz, D. Astruc ings provide a positive charge, PEGylation is an example of neu- ISM UMR CNRS No. 5255 tral surface charge, and carboxyl/carboxylate coatings provide a University of Bordeaux negative charge. The purpose was to address the hypothesis that Talence 33 405, France surface functionalization is the key driver of the biological (cel- M. Monopoli lular) responses to ENMs. As our model, we employed primary Department of Pharmaceutical & Medicinal Chemistry HMDMs versus human monocyte-like THP-1 cells (nondifferen- Royal College of Surgeons in Ireland (RCSI) tiated). Both cell types were maintained in the same cell culture Dublin 2, Ireland medium in the presence of 10% fetal bovine serum (FBS). Cyto- R. Handy School of Biological & Marine Sciences toxicity screening and cytokine profiling was performed in both University of Plymouth models, while omics methods were applied only in THP-1 cells. Plymouth PL4 8AA, UK The reason for choosing a cell line as opposed to primary cells for S. Moya omics was to minimize variability between experiments. Over- Soft Matter Nanotechnology Laboratory all, the objective was to study the impact of ENMs on immune CIC biomaGUNE Donostia-San Sebastián 20014, Spain cells as the immune system represents the first barrier against K. Savolainen foreign intrusion and previous studies have shown that mono- Finnish Institute of Occupational Health cytes/macrophages are particularly susceptible to ENMs.[15] With Helsinki 00032, Finland regard to methods, we first explored the impact of the panel of Adv. Sci. 2020, 7, 2002221 2002221 (2 of 18) © 2020 The Authors. Published by Wiley-VCH GmbH www.advancedsciencenews.com www.advancedscience.com ENMs by using conventional cell viability assays to establish a subsequent experiments were performed with endotoxin free dose response and cytokine/chemokine arrays (using a subtoxic ENMs. dose) followed by proteomics and transcriptomics assays cou- pled with detailed bioinformatics analyses. It is important to note that the cells were exposed to equipotent doses (i.e., EC10 ) for 2.3. Cytotoxicity Screening the cytokine and omics profiling studies to allow for a compar- ison between the ENMs (with minimal variations in cell death). The ENMs were evaluated for cytotoxicity using HMDMs and Functional assays were then performed to verify the microarray the undifferentiated monocytic THP-1 cell line. To this end, cells results. were exposed for 24 h to freshly dispersed ENMs at doses up to 100 µg mL−1 and the Alamar Blue assay was applied as a proxy for cell viability (based on cellular metabolic activity).[18] For 2.2. ENM Characterization HMDMs, the data were generated with cells derived from three individual donors (each experiment performed in triplicate). As ENMs of different chemical composition were pre- shown in Figure S3 (Supporting Information), a greater cyto- pared with three types of surface functionalization: car- toxicity was observed for CuO-core and CuO–NH2 in compari- boxyl/carboxylate groups (COOH/COO− ), amino/ammonium son to CuO–COOH and CuO–PEG. Regarding MWCNTs, only groups (−NH2 /−NR3+ ) or PEG. Additionally, in the case of TiO2 MWCNT–PEG induced cell death, around 50% at the highest spheres and rods, CuO NPs, and multiwalled carbon nanotubes concentration. For TiO2 NPs, neither spheres nor rods affected (MWCNTs), nonfunctionalized variants (designated hereafter cell viability, while for CdTe QDs, all were found to be cytotoxic as “core” ENMs) were synthesized. ENMs were subjected to a irrespective of the surface modification. Furthermore, for Ag NPs thorough characterization of physicochemical properties (Table and Au NPs (5 and 20 nm), only the ENMs with ammonium sur- S1, Figures S1 and S2, Supporting Information). Morphology face functionalization were cytotoxic. Finally, no effect on cell vi- and primary particle size were determined by using transmis- ability was observed for the NDs. sion electron microscopy (TEM) (Figure S1a–h, Supporting For the THP-1 cell line, the data obtained were from three Information). The spherical and rod-shaped features of the two independent experiments each performed in triplicate (Figure different TiO2 NPs could be clearly distinguished. Dynamic light S4, Supporting Information). Regarding CuO NPs, CuO-core scattering (DLS) measurements showed, overall, that ENMs in and CuO–NH2 induced a greater cytotoxicity in comparison to suspension were in the range of 20–300 nm in diameter, except CuO–COOH, while CuO–PEG had a minor impact on cell viabil- CuO-core and CuO–COOH displaying an average hydrodynamic ity only at the highest concentration. MWCNT-core, –NH2 , and diameter size of about 1 µm, and the degree of agglomeration –COOH were found to be cytotoxic for THP-1 cells exposed to was found to be acceptable for all ENMs with PDI values < 0.5. 50 or 100 µg mL−1 , while MWCNT-PEG did not affect cell viabil- However, DLS measurements were not conducted for MWCNTs, ity. Neither TiO2 spheres/rods nor the NDs affected cell viability. and the results were unreliable for the very small ENMs includ- Regarding QDs, CdTe–NH2 was highly cytotoxic for THP-1 cells ing Au NPs, quantum dots (QDs), and nanodiamonds (NDs) when compared to COOH- or PEG-modified QDs. Similar pat- (≤ 5 nm). Zeta potential measurements showed, overall, that terns were observed for Ag NPs and Au NPs (5 and 20 nm), for the surface charge was altered with the surface chemistry (in which only the ammonium modified ENMs triggered cell death. Milli-Q water). MWCNTs, Ag NPs, Au NPs, and QDs were also To summarize, the results obtained using HMDMs and THP- characterized by using ultraviolet (UV)–visible spectroscopy, and 1 cells showed good concordance (Figure 1a,b). This is notable QDs were evaluated with photoluminescence spectroscopy. The not least as differentiated macrophages are adherent when cul- photoluminescence spectra shown in Figure S2f (Supporting In- tured ex vivo, while undifferentiated monocytes grow in suspen- formation) revealed that the amino-modified QDs did not display sion. However, some differences were also observed. Hence, QDs fluorescent properties, unlike the other QDs. MWCNTs were were more prone to induce cell death in HMDMs irrespective also examined with Raman spectroscopy and the characteristic of surface modification when compared to THP-1 cells. Over- D and G bands were clearly observed (Figure S2g, Supporting all, while certain ENMs appear to be inherently more cytotoxic Information). In addition, X-ray photoelectron spectroscopy (such as, QDs and CuO NPs), the ENMs with amino/ammonium (XPS) and Fourier transform infrared spectroscopy (FTIR) (data groups (−NH2 /−NR3+ ) tended to be more cytotoxic, as seen, not shown) confirmed the presence of the expected functional for instance, in the case of the Ag and Au NPs (5 and 20 nm). groups on the different ENMs. Overall, our analysis confirmed Moreover, PEGylation markedly reduced the cytotoxic potential the expected properties of the 31 ENMs. of some ENMs, such as the CuO NPs. Our studies also confirmed ENMs were then evaluated for possible endotoxin content that TiO2 NPs (regardless of shape and surface modification) ap- by using the Limulus Amebocyte Lysate (LAL) chromogenic peared inert, at least in the case of monocytes/macrophages, in assay.[16] In the case of MWCNTs and Au ENMs, interference with line with our previous studies of TiO2 NPs using primary human the LAL assay was noted. To overcome this issue, we used the macrophages and dendritic cells.[19] macrophage-based TNF-𝛼 expression test (TET), as described.[17] High levels of TNF-𝛼 secretion were found in macrophages ex- posed to Au–COOH (5 nm) and Au–COOH (20 nm), and this 2.4. Cytokine Profiling was blocked by polymyxin B, indicating the presence of endo- toxin (data not shown). Therefore, new carboxylated Au ENMs To evaluate the immune responses following ENM exposure, were prepared and these were endotoxin-free (< 0.5 EU). The we quantified the production of a panel of human cytokines, Adv. Sci. 2020, 7, 2002221 2002221 (3 of 18) © 2020 The Authors. Published by Wiley-VCH GmbH www.advancedsciencenews.com www.advancedscience.com Figure 1. Cytotoxicity screening showed good concordance between primary macrophages and the monocyte-like cell line. Heatmaps of cytotoxicity results obtained using A) HMDMs versus B) THP-1 cells A). The percentages of cell viability following 24 h exposure to the 31 ENMs (see Table S1, Figures S1 and S2, Supporting Information) are shown; the corresponding cytotoxicity results obtained using HMDMs and THP-1 cells are reported in Figures S3 and S4 (Supporting Information). The results are based on experiments using cells from three different donors (HMDMs) or biological replicates (THP-1), each experiment performed in triplicate. chemokines, and growth factors by using a multiplex array, as which groups mainly the cytokine responses in HMDMs. To bet- previously described.[18] HMDMs and THP-1 cells were exposed ter visualize the responses in the two cell models, the results for to a dose corresponding to a maximum of 10–15% cell death HMDMs are shown in Figure 2a, while results in THP-1 cells are (EC10 ) at 24 h. Cells incubated in medium alone or with LPS shown in Figure 2b. Again, noncytotoxic Au-20-PEG (Figure 1) (100 ng mL−1 ) for 24 h were used as negative and positive con- stands out as being the most inflammogenic NP. trols for immune activation, respectively. Overall, the cytokine responses were more pronounced in HMDMs as compared to THP-1 cells (Figure 2a,b). The results are reported for 23 of the 2.5. Proteomics Analysis 27 biomarkers in the multiplex array; the remaining cytokines were below the detection limit (Figure S5, Supporting Informa- To further probe the biological responses to ENMs, THP-1 cells tion). Notably, surface-modified Au NPs (5 and 20 nm) were were exposed for 24 h at the EC10 doses of the 31 ENMs. Stau- found to trigger the production of the Th1-type cytokines, IFN-𝛾, rosporine (0.3 × 10−6 m) and LPS (100 ng mL−1 ) were used as pos- and IL-12, as well as the Th2 cytokine, IL-10, and several proin- itive controls for cell death induction and immune activation, re- flammatory mediators including IL-6 and TNF-𝛼. Au NPs also spectively. LC-MS/MS-based proteomics analysis was performed elicited the secretion of the chemokines, MIP-1𝛼 (CCL3) and as previously described.[21] The heatmap in Figure S7 (Support- RANTES (CCL5), but not MIP-1𝛽 (CCL4). We performed hier- ing Information) shows the unsupervised clustering of the top- archical cluster analysis[20] to highlight associations between the 100 proteins identified for all ENMs. No distinct clusters could different ENMs in relation to their cytokine profiles in the two be identified suggesting a nonspecific response to ENMs. Never- cell models (Figure S6, Supporting Information). Three clus- theless, in order to extract as much information as possible, path- ters were identified: cluster I which groups the most inflammo- way analysis was performed using the Ingenuity Pathway Analy- genic ENMs, such as PEGylated Au NPs (5 and 20 nm); clus- sis (IPA) software tool.[22] Figure 3 shows the analysis of the dif- ter II which groups mainly the THP-1 responses; and cluster III ferent metal ENMs with activation (red) or inactivation (blue) of Adv. Sci. 2020, 7, 2002221 2002221 (4 of 18) © 2020 The Authors. Published by Wiley-VCH GmbH www.advancedsciencenews.com www.advancedscience.com Adv. Sci. 2020, 7, 2002221 2002221 (5 of 18) © 2020 The Authors. Published by Wiley-VCH GmbH www.advancedsciencenews.com www.advancedscience.com the most significant canonical pathways (p < 0.001). Notably, the we focused our attention on the six metal ENMs that displayed pathway, “EIF2 signaling,” was found to be predominantly upreg- the most pronounced effects (see above) and performed gene on- ulated by all the ENMs. EIF2 (eukaryotic translation initiation fac- tology (GO) enrichment analyses.[25] The top-10 biological pro- tor 2) is an essential factor for protein synthesis.[23] As expected, cess GO terms for Ag-NR3+ , Au-5-NR3+ , Au-20-NR3+ , Au-20-PEG, staurosporine upregulated the “mitochondrial dysfunction” path- QD–PEG, and QD–COOH were sorted by adjusted p-value (Ta- way, while LPS triggered the pathway related to phagocytosis ble S2, Supporting Information). GO terms were displayed ac- in macrophages and monocytes (Figure 3). The corresponding cording to up- or down-regulated DEGs. For Ag-NR3+ , Au-5- heatmaps for the metal oxide and carbon-based ENMs are shown NR3+ , and Au-20-NR3+ , the most significantly downregulated GO in Figure S8a,b (Supporting Information), respectively. Overall, terms corresponded to DNA replication, DNA repair, and cell cy- the proteomics data indicated a general stress response in cells cle transition. With regard to the upregulated GO terms, all six exposed to ENMs at EC10 doses. ENMs showed a significant impact on cytokine-mediated signal- ing and inflammation/neutrophil-mediated immunity. For com- parison, CuO-core and CuO–NH2 were found to affect metal 2.6. Transcriptomics Analysis detoxification and ion homeostasis (data not shown), suggesting a very different cellular response. We also determined the top- We then performed genome-wide expression analysis of THP-1 20 up- and downregulated genes for the six metal ENMs (Table cells. To this end, THP-1 cells were exposed for 24 h at the EC10 S3, Supporting Information). These results do not necessarily doses of the 31 ENMs, and total RNA was harvested for microar- match the GO enrichment analysis, as the latter was performed ray analysis, as described previously.[24] Figure 4a depicts the hi- on the full set of significantly up- or downregulated DEGs. Sev- erarchical clustering of the differentially expressed genes (DEGs) eral metallothionein-encoding genes were found to be upregu- found to be significantly regulated by ENMs in THP-1 cells (av- lated in Ag-NR3+ exposed cells. For five out of the six ENMs erage gene expression values derived from three replicates; the (Au-5-NH3+ , Ag-NH3+ Au-20-PEG, QD–PEG, QD–COOH), SER- cluster analysis before averaging of the replicates is shown in Fig- PINB2 was found to be one of the most strongly downregulated ure S9, Supporting Information). Notably, the strongly cytotoxic genes (Table S3, Supporting Information). SERPINB2 encodes a Au-5-NR3+ and Ag-NR3+ ENMs clustered together on the basis serine protease inhibitor abundantly expressed in macrophages of the deregulated DEGs. A second, closely related cluster was and implicated in the regulation of innate and adaptive im- formed by the noncytotoxic ENMs, Au-20-PEG, QD–PEG, and mune responses.[26,27] Moreover, AHRR encoding the aryl hydro- QD–COOH, while Au-20-NR3+ (also cytotoxic) was also closely carbon receptor repressor[28] was found to be one of the most clustered yet formed a separate node in the dendrogram (Fig- downregulated genes in cells exposed to surface-modified MWC- ure 4a). However, other cytotoxic ENMs such as CuO-core or NTs (data not shown). To explore the transcriptomics data fur- CuO–NH2 did not cluster with the ammonium-modified Au or ther, we utilized the KEGG (Kyoto Encyclopedia of Genes and Ag ENMs. Overall, these results serve to illustrate that cytotoxi- Genomes) database that integrates genomic, chemical, and func- city profiling of ENMs does not necessarily align with the gene tional information.[29] Figure 5a shows the results for the “hu- expression profiles of the ENMs (following a low-dose exposure). man diseases” category for the 23 ENMs (out of 31) for which Indeed, the Au-5-NR3+ , Au-20-NR3+ , and Ag-NR3+ ENMs were significant DEGs were identified. Results are plotted based on all strongly cytotoxic, while the Au-20-PEG, QD–PEG, and QD– activation scores (p < 0.001). Interestingly, the three transcrip- COOH ENMs were noncytotoxic (as shown in Figure 1b; and Fig- tionally “active,” yet noncytotoxic ENMs (i.e., Au-20-PEG, QD– ure S4, Supporting Information), yet these six ENMs all displayed COOH, and QD–PEG) clustered together, with activation of NF- closely related gene expression changes (Figure 4a). 𝜅B, Toll-like receptor (TLR) and NOD-like receptor (NLR) sig- Venn diagrams highlighted that the ammonium modified Au naling. Furthermore, these ENMs all showed activation of infec- NPs (5 and 20 nm) triggered the most pronounced responses tious disease pathways including the “influenza virus” infection with the greatest number of unique DEGs (Figure 4b). Thus, at pathway. The three transcriptionally “active” and cytotoxic ENMs least for Au NPs, the ammonium modification results in stronger (i.e., Ag-NR3+ , Au-5-NR3+ , Au-20-NR3+ ) did not show any over- biological responses overall, both with respect to cell death and lap with Au-20-PEG, QD–COOH, and QD–PEG. Instead, these in terms of gene expression changes. However, the PEGylated Au ENMs were characterized by inactivation of DNA replication and NPs evidently triggered significant changes in gene expression cell cycle pathways (Figure 5a), in line with the GO enrichment yet displayed no overt cytotoxicity. PEGylated CuO, on the other analysis (Table S2, Supporting Information). Moreover, upstream hand, clustered close to the negative control (cell culture medium regulator analysis using IPA identified TLR7 and TLR9 as po- alone) in the heatmap (Figure 4a). tential regulators of gene expression in cells exposed to Au-20- The results presented here represent the largest nano- PEG, QD–COOH, and QD–PEG (Figure 5b). Furthermore, Bru- transcriptomics dataset to date, providing a wealth of informa- ton tyrosine kinase (BTK) was identified as a potential upstream tion for further analysis of the biological responses to ENMs. To regulator with a negative activation score (Figure S10b, Support- explore the relevance of the observed gene expression changes, ing Information). Venn analysis showed a considerable overlap Figure 2. Cytokine profiling of HMDMs A) and THP-1 cells B) exposed to 31 different ENMs. The cells were exposed to the ENMs for 24 h at doses inducing a maximum of 10–15% of cell death (EC10 ). Cell culture medium alone was used as a negative control. The secretion of cytokines, chemokines, and growth factors into the supernatant was monitored by using the BioPlex Pro Human Cytokine Standard 27-Plex array (Figure S5, Supporting In- formation). The heatmaps represent the mean concentrations detected in 3 individual experiments. A comparison of results obtained in HMDMs and THP-1 cells is shown in Figure S6 (Supporting Information). Adv. Sci. 2020, 7, 2002221 2002221 (6 of 18) © 2020 The Authors. Published by Wiley-VCH GmbH www.advancedsciencenews.com www.advancedscience.com Figure 3. Pathway analysis of proteomics data. THP-1 cells were exposed to the 31 ENMs for 24 h at doses inducing a maximum of 10–15% of cell death (EC10 ) and samples were then subjected to proteomics analysis (refer to Figure S7, Supporting Information). STS (EC10 ) and LPS (100 ng mL−1 ) were used as positive controls for cell death induction and immune activation, respectively. Pathway analysis was then performed using the Ingenuity Pathway Analysis (IPA) software.[22] The heatmap shows the results for the metal ENMs. The corresponding results for metal oxide and carbon-based ENMs are reported in Figure S8a,b (Supporting Information), respectively. The significance values for the canonical pathways were calculated by Fisher’s exact test right-tailed and indicate the probability of association of the proteins with the respective pathway. The cutoff for p-values was p < 0.001 for at least one of the conditions. between QD–COOH and QD–PEG (Figure S10a, Supporting In- tive” metal ENMs, we decided to explore cytokine responses fur- formation). ther by using the IPA software tool. We previously provided evidence, in an unrelated study of 19 ENMs, that grouping could be achieved on the basis of cytokine responses in THP- 2.7. Validation Experiments 1 cells. Hence, by using multiplex arrays coupled with bioin- formatics analysis of the cytokine results, we found that the Since our transcriptomics analysis pointed to the upregulation ENMs could be segregated into two distinct groups based on of cytokine-mediated signaling for the most transcriptionally “ac- the activation or deactivation of the nuclear receptors, PPAR Adv. Sci. 2020, 7, 2002221 2002221 (7 of 18) © 2020 The Authors. Published by Wiley-VCH GmbH www.advancedsciencenews.com www.advancedscience.com Adv. Sci. 2020, 7, 2002221 2002221 (8 of 18) © 2020 The Authors. Published by Wiley-VCH GmbH www.advancedsciencenews.com www.advancedscience.com (peroxisome proliferator-activated receptor) and LXR/RXR (liver contribute to the effects observed for ammonium modified Au X receptor/retinoid X receptor).[18] We decided to test this pro- NPs. posal in the present study. To this end, canonical pathway analy- sis was performed on the cytokine data obtained for the full set of ENMs (Figure 2b). As shown in Figure 6a, the ENMs segre- 3. Discussion gated into two groups, with activation or deactivation of PPAR In this study, we performed a multiparametric evaluation of a and LXR/RXR. Furthermore, canonical pathway analysis was per- large panel of ENMs using human in vitro models represent- formed on the transcriptomics data obtained for the metal ENMs. ing the immune system.[31] Our overarching conclusion is that Interestingly, as shown in Figure 6b, the ENMs were found to in- the surface properties are a key determinant of the biological re- activate PPAR and LXR/RXR along with DNA damage and cell sponses toward EMNs, thus supporting our initial hypothesis. cycle related pathways (p < 0.001), suggesting that: i) nuclear re- However, core chemistry also plays a role, and this was evidenced ceptor activation/deactivation in monocyte-like THP-1 cells is a both when studying cytotoxicity and gene expression changes. valid paradigm for ENMs, and ii) cytokine profiling and transcrip- Furthermore, ENMs that do not display cytotoxicity may still elicit tomics results for the studied metal ENMs show good concor- major effects on cells. Indeed, certain PEGylated metal NPs were dance at the level of pathway activation. found to be highly inflammogenic, and the transcriptional re- To further validate the microarray results, we turned our atten- sponses were reminiscent of those induced by viruses, as we dis- tion to the most downregulated biological processes, i.e., DNA cuss below. replication/repair, and cell cycle transition (Table S2, Support- Previous toxicity profiling studies have focused mainly on a ing Information). First, we applied the conventional DNA con- single class of ENMs. For instance, Cai et al.[32] studied a panel tent/cell cycle assay,[9] using THP-1 cells exposed for 24 h at the of Fe2 O3 ENMs by applying metabolomics and proteomics analy- EC10 dose. Etoposide (10 × 10−6 m) was included as a positive ses (at relatively high doses, i.e., 100 µg mL−1 ). Shaw et al.[7] per- control. Figure 7a shows the results for all the metal ENMs tested formed profiling studies of iron oxide ENMs (Fe2 O3 and Fe3 O4 ) (i.e., Ag, Au 5 nm, Au 20 nm, and CdTe, with three different sur- plus QDs (CdSe) using conventional assays. Zhang et al.[33] ex- face modifications). Notably, the ammonium modified Ag and Au plored the biological effects of a panel of metal oxides and Ag (5 and 20 nm) ENMs all triggered a strong cell cycle arrest com- ENMs. Breznan et al.[34] focused on a panel of mesoporous silica parable to that seen for etoposide while this was not the case for particles, while Wu et al.[35] studied a large set of surface modi- their carboxyl modified or PEGylated counterparts. Next, we eval- fied MWCNTs. Taken together, these studies have provided valu- uated DNA damage in cells by using a specific phospho-histone able insights with respect to nano–bio interactions insofar as they H2A.X antibody. Etoposide (10 × 10−6 m) was used as a positive showed that different physicochemical properties, such as sur- control. Both Au-5-NR3+ and Ag-NR3+ , the two ENMs that were face reactivity, or size, or shape/aspect ratio, or a combination shown to be the most closely related according to hierarchical of several properties, may dictate cellular outcomes. Cell type- clustering of DEGs (Figure 4a), triggered a significant degree of specific differences were also noted in several studies. What does DNA damage (Figure 7b) comparable to that seen for etoposide the present study add to this picture? First, using a broad panel treated cells. The carboxyl modified or PEGylated ENMs did not of ENMs encompassing metal, metal oxide, and carbon-based elicit DNA damage. Thus, the transcriptomics results were vali- ENMs, we showed that surface properties are an important driver dated with respect to DNA damage and cell cycle arrest. of the cytotoxicity of ENMs. This was verified in a commonly Previous work has shown that NPs with a high affinity for used human cell line as well as in primary human macrophages. DNA strongly inhibited DNA replication, whereas those with low However, the relation between physicochemical properties and affinity had no or minimal effects.[30] We therefore decided to biological responses is complex and the core chemistry also mat- study the interaction of Au NPs of different diameters and vary- ters. Interestingly, we found that while certain surface modifica- ing surface modifications with respect to their binding to RNA tions, i.e., amino/ammonium termination, may convert ENMs and DNA. To this end, we isolated total cellular RNA and DNA that are inherently inert (such as Au) into a strongly cytotoxic from THP-1 cells using standard protocols. Then, ENMs were in- material, the same surface modification had no effect on other cubated with 25 ng of RNA or 50 ng of DNA for 15 min (on ice). ENMs (such as TiO2 or NDs) that remained biologically inert no To monitor binding, we performed DLS and zeta potential mea- matter the surface properties (in the current model). For compar- surements (data not shown) and PCR (Figure S11a,b, Supporting ison, Orecchioni et al. studied the impact of pristine graphene Information). These studies showed, not unexpectedly, that the oxide (GO) versus GO functionalized with amino groups (GO– positively charged ammonium modified ENMs interacted more NH2 ) on various subpopulations of peripheral blood mononu- readily with RNA (Figure S11a,b, Supporting Information) and clear cells by using single-cell mass cytometry and supplemented DNA (Figure S11c, Supporting Information). These experiments this with genome-wide transcriptomics analysis of GO and GO– were performed with “naked” DNA, while in the cell, DNA is NH2 exposed Jurkat cells and THP-1 cells.[36] Using pathway bound to histones, forming nucleosomes. Nevertheless, the re- analysis, the authors found that the perturbations induced by GO sults suggest that direct nucleic acid binding could potentially were reflective of cytotoxic mechanisms, while the transcriptional Figure 4. Unsupervised clustering of microarray data. A) Hierarchical clustering of differentially expressed genes (DEGs) significantly regulated in THP-1 cells after 24 h exposure to the 31 ENMs at doses inducing a maximum of 10–15% of cell death (EC10 ). The heatmap was colored based on activation z scores of the normalized expression levels for the DEGs; refer to Figure S9 (Supporting Information) for the hierarchical clustering results prior to averaging of the replicates. B) Venn analysis of significant DEGs (FC > 1.5 and adjusted p-value < 0.05) in THP-1 cells exposed to different Au (5 nm) ENMs and Au (20 nm) ENMs. Adv. Sci. 2020, 7, 2002221 2002221 (9 of 18) © 2020 The Authors. Published by Wiley-VCH GmbH www.advancedsciencenews.com www.advancedscience.com Adv. Sci. 2020, 7, 2002221 2002221 (10 of 18) © 2020 The Authors. Published by Wiley-VCH GmbH www.advancedsciencenews.com www.advancedscience.com changes induced by GO–NH2 were consistent with immune regard was that the small, ammonium modified Ag and Au ENMs activation.[36] Thus, the impact of amino/ammonium function- which were found to be strongly cytotoxic in our model were also alization may potentially depend on how the functional groups clustered together based on significant changes in gene expres- are displayed (i.e., on a flat surface as opposed to a surface with sion (at EC10 ). Furthermore, with regard to the six different metal high curvature). We noted that while PEGylation of ENMs, a (Ag, Au, CdTe) ENMs that were found to be the most transcrip- commonly used strategy in nanomedicine,[37] may passivate cer- tionally “active” among the 31 ENMs, we found a remarkably con- tain ENMs (such as CuO) the same surface modification may sistent pattern in terms of affected cellular pathways. Hence, the promote the inflammogenic properties of other ENMs (such as downregulated pathways were related to DNA replication/DNA Au) and may even enhance toxicity of certain ENMs (as seen for repair and cell cycle transition, while the upregulated pathways MWCNTs). Hence, this further supports the view that the man- were associated with cytokine-dependent signaling and inflam- ner in which the functional groups are presented also matters. mation. Importantly, these findings were validated by using con- The stability or lability of surface ligands may also be a factor, ventional cell-based assays. It is noteworthy that not all cytotoxic as suggested in a previous microarray study of surface modified ENMs displayed the same pattern of gene expression changes. Au NPs.[38] Overall, carboxyl functionalization appeared to be the For instance, cytotoxic CuO ENMs were found to activate metal most biocompatible modification in the present study. However, detoxification pathways, likely due to the release of toxic ions, as it is important to qualify this statement in at least two ways. First, shown in several previous studies.[45,46] On the other hand, our macromolecules (such as PEG) may acquire different conforma- proteomics analysis did not distinguish specific effects for any tions depending on a number of factors including the type of of the ENMs and pointed to a general activation of EIF2 signal- bond anchoring the ligand to the surface, the curvature of the par- ing (i.e., protein translation). This shows that gene expression ticle, and the ionic strength and pH of the surrounding media.[39] profiling is a more sensitive method able to discern low-dose ef- Second, differences in surface coating may lead to the adsorption fects of ENMs. However, we reported in a previous study on a of different proteins or other biomolecules which, in turn, may set of Au ENMs that proteomics can be applied to detect changes impact on cellular uptake of ENMs.[40,41] Notwithstanding, our at doses corresponding to EC50 and found a good concordance survey of a large panel of ENMs using two cell models has shown between transcriptomics and proteomics approaches in THP-1 that surface functionalization (coating) is an important determi- cells.[12] In line with the latter findings, a recent proteomics study nant of the cellular effects of ENMs. of metal oxide ENMs (SiO2 and TiO2 ) and GO showed no effects The present study takes as its starting point the adminis- in a lung epithelial cell line at low doses (≤ 1 µg cm−2 ) while doses tered dose (the amount added to the cell cultures) as opposed of 10 µg cm−2 yielded significant changes.[47] to the internalized dose, although the concentration of ENMs It is interesting to note that three ENMs (Au-20-PEG, QD– internalized by cells may, in principle, vary depending upon COOH, and QD–PEG) were noncytotoxic as shown by conven- the surface chemistry (and, potentially, the composition of the tional toxicity assays, yet they were found to elicit significant so-called corona).[42] However, we have studied the internaliza- changes in gene expression in THP-1 cells. Our bioinformat- tion of spherical TiO2 ENMs (pristine and surface-modified) ics analyses revealed that these ENMs activated NF-𝜅B, TLR, by using TEM and found that these ENMs were avidly taken and NLR signaling pathways. Furthermore, the gene expression up by HMDMs regardless of surface functionalization (unpub- profiles matched those of several infectious diseases including lished observations). On the other hand, as shown in the present influenza virus infection, and upstream regulator analysis pre- study, none of these ENMs displayed any cytotoxicity toward dicted both TLR7 and TLR9 as positive regulators and BTK as a macrophages (or THP-1 cells). These results thus dispel the no- negative regulator of gene expression. TLR7 and TLR9 are endo- tion that cellular uptake equals toxicity, at least for TiO2 ENMs. somal pattern recognition receptors that function as sensors of Furthermore, it is of interest to note that the cytotoxicity results single-stranded RNA and double-stranded DNA, respectively.[48] for the entire panel of ENMs were largely concordant between BTK, in turn, is involved in the sensing of multiple microbes, in the two cell models, despite the fact that primary macrophages part, via TLRs.[49] Taken together, these observations imply that are more proficient in terms of phagocytosis when compared to Au-20-PEG, QD–COOH, and QD–PEG are “sensed” as microbes. cell lines. This may not be surprising if one considers that viruses are nano- The second major point of interest is that we have shown, us- sized particles (e.g., influenza virus particles are between 80 and ing a low-dose (equipotent) exposure regimen (to avoid excessive 120 nm) that display repetitive molecular motifs on their surface; cell death as this may confound the results), that transcriptional these molecular patterns are, in turn, recognized by extracellular responses to ENMs are related to surface properties. The cur- or intracellular pattern recognition receptors, such as TLRs.[50] rent dataset spanning 31 different ENMs is the largest transcrip- Moreover, viruses were recently shown to display a “corona” tomics dataset thus far, and our results provide a useful resource of surface adsorbed proteins,[51] further blurring the distinction for in silico exploration of nano–bio interactions, as exemplified between synthetic and natural nanoparticles. Previous studies in previous publications.[43,44] The most striking finding in this have provided evidence that QD–COOH and QD–PEG activate Figure 5. Functional analysis of the transcriptomics data. A) Functional annotation of the microarray data obtained for 23 out of 31 ENMs (i.e., ENMs for which significant DEGs were identified) was performed using the KEGG database[29] focusing on “human disease” categories. The results were visualized by using the FunMappOne graphical tool.[63] The activation z scores are shown (red: activation, green: deactivation). The cutoff for p-values was p < 0.001. B) Upstream regulator analysis of the transcriptomics results was performed,[64,65] and TLR7/TLR9 were identified as putative regulators on the basis of z scores. Refer to Figure S10 (Supporting Information) for upstream regulator results with a negative activation z score and the legend for the upstream regulator analysis shown in panel B. Adv. Sci. 2020, 7, 2002221 2002221 (11 of 18) © 2020 The Authors. Published by Wiley-VCH GmbH www.advancedsciencenews.com www.advancedscience.com Adv. Sci. 2020, 7, 2002221 2002221 (12 of 18) © 2020 The Authors. Published by Wiley-VCH GmbH www.advancedsciencenews.com www.advancedscience.com TLR signaling pathways in murine RAW264.7 macrophages,[52,53] of the type XRSH (X = COOH, N(CH3 )3 , and CH3 ) were used in though this does not prove a direct interaction of QDs with order to replace the citrate ligands on the nanoparticle surface. TLRs. Moreover, Au NPs were shown to modulate TLR9 signaling Poly(ethylene glycol) methyl ether thiol (average 550 g mol−1 ) was within macrophage lysosomes in a size-dependent manner.[54] the bifunctional ligand used in the synthesis of the Au-20-PEG. Luo et al. recently demonstrated that PEGylated GO stimulates CuO ENMs were synthesized by the precipitate decomposition cytokine responses in peritoneal macrophages, and comparable method. A precursor, basic copper carbonate (Cu2 (OH)2 CO3 ), results were obtained for the unrelated 2D material, MoS2 .[55] was first prepared by precipitation reaction from an aqueous Therefore, the widely held assumption that PEGylation serves to solution composed of 1 m copper nitrate and sodium carbon- passivate ENMs may not be correct. Instead, it appears that PE- ate. The obtained precursor was dewatered, dried, and milled. Gylation promotes inflammogenic responses, at least for certain Nanocrystalline CuO particles were obtained by thermal decom- ENMs (such as Au). QDs (with or without PEG) may be especially position of the precursor at 300 °C for 2 h. For surface modifi- prone to trigger “virus-like” responses by virtue of their small size cation of the CuO ENMs, the thiol-ligands, HSHCH2 CH2 NH2 , and spherical shape. HSCH2 COOH, and HSCH2 COOPEGOCH3 , were used due to In sum, we report multiparametric assessment of a large panel their strong ability to bind to copper oxide in order to obtain of ENMs and show that surface properties are key determinants CuO–NH2 , CuO–COOH, and CuO–PEG ENMs, respectively. Ag of cellular (immune) responses to ENMs. However, the core ENMs were prepared by reduction of Ag+ to Ag0 with excess ice- chemistry (composition) of the ENMs is also important. Our find- cold NaBH4 solution. The bifunctional ligands of the type XRSH ings also demonstrate that ENMs can be discriminated on the (X = COOH, N(CH3 )3 , and CH3 ) were added and also provided basis of transcriptional profiling. an efficient stabilization. Ag-PEG was obtained by using the [poly (ethylene glycol) methyl ether thiol, average 550 g mol−1 ] ligand. CdTe QDs were synthesized by PlasmaChem GmbH (Berlin) 4. Experimental Section in the frame of FP7-NANOSOLUTIONS. Briefly, cadmium and ENM Synthesis and Functionalization: Spherical TiO2 ENMs tellurium precursors were mixed together with a thiocarboxylic (TiO2 s-core) were prepared at room temperature by hydrolysis of acid, which serves as a stabilizing ligand. After introduction of the titanium tetrachloride solution with a further condensation of the reducing atmosphere, the solution was refluxed until a desirable reaction products, and stabilization by nitric acid. TiO2 nanorods fluorescence emission wavelength is achieved. The carboxylated (TiO2 r-core) were obtained by forced hydrolysis in acidic condi- QDs (CdTe–COOH) had an emission wavelength of 590 nm. tions at normal pressure, leading to high yield and desired diam- Modification of CdTe–COOH with amino and PEG groups was eter and aspect ratio. TiO2 -spheres and TiO2 -rods were surface performed through the ligand exchange procedure with the modified using a multistep procedure. First, the particle cores respective thiols. Briefly, CdTe–COOH QDs were extensively were modified with an NH2 -terminated silane (APTMS), yielding washed with a solvent mixture containing an excess of a target lig- to amino-modified TiO2 . Then, to produce COOH and PEG mod- and (HS- n R-(PEG) or HS-R-NH2 ) followed by incubation of the ified particles, succinic anhydride and PEG-COOH were bound obtained QDs in the suspension of a pure target ligand. The pro- to amino group through amide bonds, respectively. For the syn- cess of ligand exchange was monitored by the dispersibility of the thesis of the two Au ENMs (Au-5 and Au-20) the strategy con- QDs in water and change of the optical properties of QDs. NDs sisted in growing the metallic NPs with the simultaneous at- were prepared by detonation in two stages: first, synthesis of nan- tachment of self-assembled thiol monolayers on the growing nu- odiamond soot by explosion of mixed trinitrotoluene-hexogen ex- clei in order to allow the surface reaction to take place during plosive charges (TNT–RDX 60/40) in a large-sized chamber (50 metal nucleation and growth. Briefly, Au-5 were prepared us- m3 ) at a raised pressure of a cooling gas, and second, oxida- ing reduction of Au3+ to Au0 with NaBH4 in the presence of bi- tion with a HNO3 –HClO4 mixture of metals and nondiamond functional ligands of the type XRSH (X = COOH, N(CH3 )3 , or carbon and isolation of the carboxylated NDs (ND–COOH). CH3 ) so that the surface is terminated with these functionalities. The carboxylic groups were used as reactive sites for the The ligands bind to the gold surface through their thiolate (RS ligands –C(O)–NH–(CH2 )2 NH–(CH2 )2 NH–(CH2 )2 –NH3 Cl and –C(O)O–CH2 CH2 –(O–CH2 –CH2 )11–O–CH3 in order to obtain end). Au-5-PEG were obtained by direct synthesis with controlled ratio of [HAuCl4 ]/[poly(ethylene glycol) methyl ether thiol aver- ND–NH2 and ND–PEG. MWCNT-core, MWCNT–NH2 and age 550 g mol−1 ]. The Au-20 were synthesized by reduction of MWCNT–COOH were produced by Nanocyl SA (Sambreville) Au3+ to Au0 with Na3 citrate which yielded Au ENMs of core size in the frame of FP7-NANOSOLUTIONS via catalytic carbon > 10 nm. The citrate reduction process involves hot gold chlo- vapor deposition (CCVD) (NC7000, NC3152, and NC3151, re- ride and sodium citrate as reactants. In this reaction, the citrate spectively). For PEGylated MWCNTs, 1,2-distearoyl-sn-glycero- molecules act as both reducing and stabilizing agents, allowing 3-phosphoethanolamine-N-[methoxy(poly ethylene glycol)-2000] for the formation of the colloidal gold. The bifunctional ligands (DSPE-PEG 2000) was used for a noncovalent functionalization Figure 6. Pathway analysis of immune responses. A) Pathway analysis was performed by using IPA software,[22] by integrating the normalized cytokine profiling results of THP-1 cells exposed to the 31 ENMs. The results shown highlight the grouping of ENMs into activators and deactivators of nuclear receptor (PPAR and LXR/RXR) pathways, respectively. The heatmaps show the significant canonical pathways with activation score > 2, and adjusted p-value < 0.05 for at least one activated pathway per treatment. B) Pathway analysis was also performed on transcriptomics results obtained for the metal ENMs. The heatmap based on activation z scores depicts the activation (red) or inactivation (blue) of the most significant canonical pathways associated with the various conditions. The significance values for the canonical pathways were calculated by Fisher’s exact test right-tailed and indicates the probability of association of DEGs with the respective pathway. The cutoff for the p-value was p < 0.001 for at least one of the conditions. Adv. Sci. 2020, 7, 2002221 2002221 (13 of 18) © 2020 The Authors. Published by Wiley-VCH GmbH www.advancedsciencenews.com www.advancedscience.com Figure 7. Validation of microarray results. A) Quantification of cell cycle analysis results for THP-1 cells exposed for 24 h at the EC10 dose of the indicated metal ENMs, i.e., Ag, Au (5 nm), and CdTe. N.C., negative control (medium alone); P.C., positive control (etoposide, 10 × 10−6 m). Data analyzed by one- way ANOVA followed by Dunnett test: *p < 0.05, ***p < 0.001, ****p < 0.0001. B) DNA damage was determined by staining cells with an anti-histone H2A.X (Ser139) antibody with quantification by flow cytometry. Cells were exposed as described in (A). N.C., negative control; P.C., positive control (etoposide). Data analyzed by one-way ANOVA followed by Dunnett’s test for comparison to N.C. or by Bonferroni’s test for multiple comparisons: *p < 0.05, ***p < 0.001, ****p < 0.0001, ### p < 0.001. process. 125 mg of DSPE-PEG 2000 was first dissolved in 20 mL processing. UV–vis measurements were performed on a UV–vis of water, then added to the pristine MWCNT (25 mg) at a 1:5 spectrometer Lambda 750 (Perkin Elmer). The parameters for re- [MWCNT:DSPE-PEG 2000] weight ratio. The mixture was then fractive index were set according to the ENM core type measured, sonicated at room temperature for 1 h and diluted to 250 mL with the range of acquisition covered 200 to 800 nm and the interval water before filtration through a Millipore 100 kDa cut-off filter 5 was set at 0.5 or 1 nm. The software UVWinLab was used for times to remove any unbound DSPE-PEG 2000. The final prod- data processing. Photoluminescence analysis of QDs was done uct of MWCNT–PEG was then resuspended in water to give the on a LS-55B Luminescence spectrometer (Perkin Elmer). The ex- final product of MWCNT–PEG. citation wavelength was set at 550 nm and emission was collected ENM Characterization: DLS and zeta potential measure- from 560 to 700 nm. The software FLWinLab was used for data ments were performed on freshly prepared ENM suspensions processing. Raman spectroscopy of MWCNTs was performed us- using a Malvern Zetasizer Nano ZS (model ZEN3600) operating ing a WITec alpha 300 RAS instrument, coupled to a 532 nm with a 632.8 nm laser wavelength. The angle of detection was set wavelength laser operating with the WITec project FOUR soft- at 173° and the temperature was maintained at 25 °C through- ware. One single acquisition represents the average of 10 spectra out the analysis. The Zetasizer v.6.32 software was used for data each integrated over 0.5 s. The presence of impurities (Al, Si, Mn, Adv. Sci. 2020, 7, 2002221 2002221 (14 of 18) © 2020 The Authors. Published by Wiley-VCH GmbH www.advancedsciencenews.com www.advancedscience.com Fe, Co, Ni, Cu, Cr, Ga, Fe, Br, Sr, As, Mo, Ag, Sn, Sb, Te, W, Au, Pb) solution was added to the cells. After 24 h, 100 µL of the super- was investigated by inductively coupled plasma-mass spectrome- natants were collected into 96-well plates and stored at −80 °C for try. The presence of functional groups was confirmed with XPS, subsequent analysis by multiplex immunoassay (see below). The FTIR, and thermogravimetric analysis (TGA). Complete datasets cytotoxic effects were evaluated by incubating cells with the Ala- can be retrieved from the NANOSOLUTIONS repository (refer mar Blue reagent (ThermoFischer Scientific) for 4 h at 37 °C. The to Table S1 for a summary, Supporting Information). resulting fluorescence was determined at 540/590 nm (ex/em), Endotoxin Assessment: The procedure for the detection of using an Infinite 200 Tecan microplate reader. The statistical gram-negative bacterial endotoxin followed the QCL-1000 End- analysis was performed by one-way ANOVA analysis using Prism point Chromogenic LAL assay (Lonza) protocol. The enzy- 5.02 (GraphPad), on results obtained from three independent ex- matic reaction was performed on ENMs alone, ENMs in pres- periments (three individual donors in case of HMDMs), each per- ence of LPS or ENMs in presence of LPS and polymyxin B formed in triplicate. (PolyB). Briefly, the stock suspensions of ENMs were diluted at Cytokine Measurements: For the multiplex immunoassay, 100 µg mL−1 in endotoxin-free culture medium and 50 µL of each HMDMs and THP-1 cells were exposed to ENM doses inducing suspension were dispensed into a 96-well plate. LPS and PolyB 10–15% cell death (EC10 ). The procedure followed the Bio-Plex were added at 1–3 EU mL−1 or 5–10 × 10−6 m, respectively, fol- Pro Human Cytokine Immunoassay 27-plex, Group I (BioRad lowing by 1 h incubation at room temperature. The reaction was LUMINEX) protocol, as described.[57] Briefly, the beads were first initiated by first adding the proenzyme and then the substrate. dispensed into a 96-well plate and incubated with antigen stan- Then, a stop solution of acetic acid 25% v/v in dH2O was added dards or samples for 1 h at room temperature. The beads were and the absorbance was determined at 405 nm, using an Infi- then washed, followed by incubation with the biotinylated detec- nite 200 Tecan microplate reader operating with Magellan v7.2 tion antibodies for 30 min at room temperature. After washing software. For ENMs for which assay interference was shown, the the unbound biotinylated antibodies, beads were incubated with macrophage-based TET assay was deployed.[17] Briefly, HMDMs the reporter SA-PE conjugate for 10 min at room temperature. were obtained from human monocytes seeded into a 96-well plate Following removal of the excess of SA-PE, the beads were passed and differentiated with recombinant human M-CSF (PeproTech) through the Bio-Plex 200 array reader to detect the 27 cytokines, for 3 days. ENMs suspensions were diluted to 5 µg mL−1 in cell chemokines, and growth factors (IL-1𝛽, IL-1ra, IL-2, IL-4, IL-5, IL- culture medium, supplemented or not with PolyB (10 × 10−6 m). 6, IL-7, IL-8, IL-9, IL-10, IL-12, IL-13, IL-15, IL-17, Eotaxin, FGF HMDMs were then exposed for 24 h to the ENM suspensions, basic, G-CSF, GM-CSF, IFN-𝛾, IP-10, MCP-1, MIP-1𝛼, PDGF, or culture medium alone, or LPS at 100 ng mL−1 (positive con- MIP-1𝛽, RANTES, TNF-𝛼, and VEGF) by measuring the fluores- trol for TNF-𝛼 production). Following incubation, cell culture su- cence of the bound SA-PE. Data processing was performed with pernatants were collected and the levels of TNF-𝛼 were moni- the Bio-Plex Manager v6.0 software. The calibration curves were tored using a specific ELISA (Mabtech) according to the manu- determined by adjusting logistic functions to the standard points facturer’s protocol. assessed and limits of detection were established. Data retrieved Cell Line and Primary Cells: HMDMs are obtained from pe- from the multiplex assay were analyzed using hierarchical clus- ripheral blood mononuclear cells as described.[56] Briefly, cells tering analysis.[18] Pathway analysis was performed on normal- were isolated from buffy coats of healthy adult donors by den- ized cytokine expression data using the Ingenuity Pathway Anal- sity gradient centrifugation followed by positive selection of ysis (IPA) software (Ingenuity Systems, Redwood City, CA). CD14+ monocytes using CD14 MACS magnetic beads (Mil- DNA Damage/Cell Cycle Analysis: THP-1 cells were exposed tenyi Biotech). Cells were then cultured in RPMI-1640 medium for 24 h to the indicated ENMs at EC10 concentrations. Etopo- supplemented with 10% heat-inactivated bovine serum (FBS), side (10 × 10−6 m) was used as a positive control. Detection of 2 × 10−3 m glutamine, 100 U mL−1 penicillin, and 100 µg mL−1 cellular 𝛾H2AX and simultaneous cell cycle analysis were car- streptomycin and M-CSF (50 ng mL−1 ) was added for 72 h to in- ried out by using the FlowCellect Cell Cycle Checkpoint H2A.X duce macrophage differentiation. The human monocytic THP-1 DNA Damage Kit (Millipore) according to the manufacturer’s cell line was purchased from ATCC (UK). The cells were cultured instructions. Briefly, cells were washed twice with phosphate- in RPMI-1640 medium supplemented with 10% FBS, 2 × 10−3 m buffered saline (PBS), fixed, and permeabilized with the fixation glutamine, 100 U mL−1 penicillin, 100 µg mL−1 streptomycin, and permeabilization buffers, respectively, and thereafter stained and 0.05 × 10−3 m 𝛽-mercaptoethanol. The cell density was with an anti-phospho-Histone H2A.X (Ser139)-AlexaFluor 488 strictly maintained at 0.1–2.0 × 106 cells mL−1 . conjugated-antibody. DNA staining was performed by using a Cytotoxicity Assessment: HMDMs and undifferentiated THP- PI/RNAse solution. Histone H2A.X resides downstream of the 1 cells were seeded into 96-well plates at a density of 106 DNA damage kinase signaling cascade and phosphorylation of cells mL−1 (100 µL per well). ENMs were freshly prepared and H2A.X at serine 139 is an important indicator of DNA damage.[58] the following working concentrations were applied: 0.78, 1.56, Samples were analyzed on a BD LSRFortessa and data were pro- 3.13, 6.25, 12.5, 25, 50, 100 µg mL−1 in 100 µL of complete RPMI cessed using FCS Express DeNovo software. cell medium supplemented with 10% FBS. Dimethyl sulfoxide Nucleic Acid Binding Studies: Qiagen AllPrep (DMSO) 5%, LPS (100 ng mL−1 ), and culture media alone were DNA/RNA/miRNA Universal kit (Qiagen GmbH) was used used as positive controls for cell death, immune activation, and as to extract RNA and DNA according to the manufacturer’s pro- a negative control, respectively. After differentiation of HMDMs, tocol. Briefly, cells were quickly washed once with ice-cold PBS, the cell medium was removed, and cells were exposed directly to and lysed directly into the well with 350 µL RTL plus buffer 100 µL of the treatment solutions prepared at the final working from the kit supplemented with 𝛽-mercaptoethanol. Samples concentration. For THP-1 cells, 100 µL of a twice concentrated were stored at −80 °C before RNA/DNA extraction. RNA was Adv. Sci. 2020, 7, 2002221 2002221 (15 of 18) © 2020 The Authors. Published by Wiley-VCH GmbH www.advancedsciencenews.com www.advancedscience.com eluted with 30 µL of RNAse free water and DNA was eluted with licate. After exposure, the cells were washed twice with PBS by 50 µL of elution buffer (EB). Au NPs were dispersed by sonica- centrifugation 300 g, 5 min at room temperature. Proteins were tion for 5 min in a bath sonicator at 47 kHz before they were extracted, digested, and analyzed using a Q Exactive Plus Hy- diluted to working suspensions (5.0, 0.5, 0.05, 0.005, 0.0005, and brid Quadrupole-Orbitrap mass spectrometer. Data processing 0.00 005 µg mL−1 ). For RNA binding, 25 ng of total RNA from was done using Raw2MGF and ClusterMGF from the Quanti THP-1 cells was incubated with each of the six EMNs in 100 µL workflow 18. The data were searched against the human com- total volume using PBS as buffer. The incubation was performed plete proteome database (Uniprot) using the Mascot search en- on ice for 15 or 30 min. Then, the suspension was centrifuged gine and quantified by Quant, an in-house developed software with 16 000 × g for 30 min at 4 °C and the supernatant was for label-free quantification. Student’s t-test, adjusted for multi- collected. ENMs were then washed four times with 1 mL of ple testing false discovery rate, where used to compare the differ- ice-cold PBS, centrifuged after each wash with 16 000 × g for ent treatments. Heatmaps were generated by calculating: i) the 30 min at 4 °C and the supernatants were saved. The washed median of each sample, ii) the variance from the median, and iii) ENMs were resuspended into 100 µL of PBS. During this time, unsupervised clustering of the top 50 or 100 proteins identified 25 ng total RNA in a volume of 100 µL was kept on ice to be for each of the tested ENMs. used as unbound control. The experiments were repeated three Microarray Analysis: For transcriptomics, 106 cells per well times. For DNA binding, 50 ng of DNA from THP-1 cells was were seeded in a 12-well plate one day prior to exposure. Cells incubated with each of the six ENMs in a volume of 10 µL diluted were exposed for 24 h to the respective EC10 doses of freshly in water for 15 min on ice; 50 ng DNA in a volume of 10 µL resuspended ENMs and to a fixed concentration for the noncy- without EMNs was used as unbound control. totoxic ENMs. Cell culture medium without ENMs was used as Quantitative Real-Time PCR: cDNA was synthesized with negative control. Each condition was performed in triplicate. Af- Maxima First Strand cDNA Synthesis Kit (K1642; Thermo Sci- ter exposure, the cells were washed twice with ice-cold PBS by entific) according to manufacturer’s instructions. 12 µL of tem- centrifugation at 300 g, 5 min at 4 °C. RNA extraction was per- plate (control RNA, first supernatant, wash supernatants, fi- formed using the AllPrep DNA/RNA/miRNA Universal Kit, ac- nal suspension of ENMs, or water as no-template control) cording to the manufacturer’s instruction. RNA samples were la- was used in each reaction. Quantitative real-time PCR (qRT- beled with Cy3 or Cy5 (Quick Amp Gene Expression Labelling kit, PCR) was analyzed with undiluted cDNA using SYBR Green Agilent). Labeled samples were randomly hybridized to Agilent primers for human 𝛽-actin and GAPDH (0.2 × 10−6 m) and SurePrint G3Human GE 8 × 60K DNA microarrays, according the FastStart Universal SYBR Green Master Mix (Roche Di- to the manufacturer’s protocol (Agilent). Hybridized slides were agnostics). The following SYBR Green primers were used: 𝛽- scanned using the Agilent microarray scanner (Model G2505C) actin forward: 5’-GACGACATGGAGAAAATCTG-3; 𝛽-actin re- and raw data were extracted using Agilent feature extraction soft- verse: ´5’-ATGATCTGGGTCATCTTCTC-3´, GAPDH forward: ware (version 12.0.1.1). Downstream analyses of the normal- 5´-ACGGATTTGGTCGTATTGGG-3; and GAPDH reverse: ´5´- ized data[59] were performed using the IPA software (Ingenuity TGATTTTGGAGGGATCTCGC-3´. Human placenta cDNA was Systems, Redwood City, CA, http://www.ingenuity.com—version used as positive control and water as negative control. Reac- 33 559 992) to examine the canonical signaling pathways associ- tions were amplified by using the 7500 Fast Real-Time PCR sys- ated with the observed changes.[22] Venn diagrams of the differ- tem (Applied Biosystems). Agarose gel electrophoresis. PCR was entially expressed genes (DEGs) were plotted with the web-based run in the presence of 1 × PCR buffer, 1.5 × 10−3 m MgCl2 , tool, Venny 2.1.0.[60] GO enrichment analyses were performed by 0.4 × 10−3 m dNTPs, 1 × 10−6 m of each primer, and 8.5 units Hot- using Enrichr,[61,62] and the GO database version 2017b. KEGG Star Taq Plus DNA Polymerase (Qiagen GmbH). The whole DNA- (Kyoto Encyclopedia of Genes and Genomes) pathways were ana- ENM incubation reaction (10 µL) was used as a template for PCR. lyzed and visualized by using the FunMappOne graphical tool.[63] The primers used were LDLR forward: CACCTGGCTGTTTC- Statistical Analysis: Experiments were performed in at least CTTGAT; and LDLR reverse: CGGTCAGGGGATATGAGTCT. three biological replicates (for primary cells, three individual Reactions were amplified by using the T100 Thermal Cycler (Bio- donors) and samples were analyzed using three technical tripli- Rad) and a program of: 95 °C for 5 min; 35 rounds of 95 °C for cates. Data are average values ± S.D. Statistical analysis was per- 30 s, 62 °C for 30 s, and 72 °C for 1 min; finishing with 5 min of formed by one-way ANOVA using Prism 5.02 (GraphPad Soft- elongation at 72 °C. The products were visualized on agarose gel ware, Inc.), assuming equal variances with p < 0.05. For pro- (1%) electrophoresis together with Thermo Scientific GeneRuler teomics, transcriptomics, and multiplex immunoassay data anal- 100 bp Plus DNA Ladder (Thermo Scientific). DLS and zeta po- ysis, see above. tential measurements were performed on freshly prepared ENM suspensions as described above. Measurements were done for Supporting Information ENMs alone versus ENMs (5 µg) incubated with RNA (25 ng) or DNA (50 ng). Supporting Information is available from the Wiley Online Library or from Proteomics Analysis: For proteomics samples, 106 cells per the author. well were seeded in a 12-well plate one day prior to exposure. Cells were exposed for 24 h to the respective EC10 doses of freshly Acknowledgements resuspended ENMs. Staurosporine (at the EC10 dose equivalent), Supported by the European Commission through the FP7- LPS (100 ng mL−1 ), and cell culture medium were used as pos- NANOSOLUTIONS project (Grant Agreement No. 309329). Burcu itive controls for cell death, immune activation, and as a nega- Açikgöz, Dokuz Eylül University, Turkey, is acknowledged for assistance tive control, respectively. Each condition was performed in trip- with nucleic acid corona experiments through the Erasmus program. Adv. Sci. 2020, 7, 2002221 2002221 (16 of 18) © 2020 The Authors. Published by Wiley-VCH GmbH www.advancedsciencenews.com www.advancedscience.com Conflict of Interest [11] C. Pisani, J. C. Gaillard, V. Nouvel, M. Odorico, J. Armengaud, O. Prat, BMC Genomics 2015, 16, 315. The authors declare no conflict of interest. [12] A. Gallud, K. Klöditz, J. Ytterberg, N. Östberg, S. Katayama, T. Skoog, V. Gogvadze, Y. Z. Chen, D. Xue, S. Moya, J. Ruiz, D. Astruc, R. Zubarev, J. Kere, B. Fadeel, Sci. Rep. 2019, 9, 4366. Author Contributions [13] M. Faria, M. Björnmalm, K. J. Thurecht, S. J. Kent, R. G. Parton, M. Kavallaris, A. P. R. Johnston, J. J. Gooding, S. R. Corrie, B. J. Boyd, A.G. performed experiments and analyzed data; M.D. performed valida- P. Thordarson, A. K. Whittaker, M. M. Stevens, C. A. Prestidge, C. tion experiments and analyzed data; P.K. performed microarray analyses J. H. Porter, W. J. Parak, T. P. Davis, E. J. Crampin, F. Caruso, Nat. and analyzed data; V.M. and V.F. performed statistical analyses of the tran- Nanotechnol. 2018, 13, 777. scriptomics data; J.Y. and R.Z. performed proteomics analyses, T.S. per- [14] H. S. Leong, K. S. Butler, C. J. Brinker, M. Azzawi, S. Conlan, C. Dufés, formed nucleic acid binding studies; J.K. coordinated the omics studies, M.C., K.L., Z.A., J.R., M.M., R.H., and S.M. synthesized and character- A. Owen, S. Rannard, C. Scott, C. Chen, M. A. Dobrovolskaia, S. V. Ko- ized ENMs; D.A. and K.K. supervised the material synthesis; K.S. coor- zlov, A. Prina-Mello, R. Schmid, P. Wick, F. Caputo, P. Boisseau, R. M. dinated the project; H.A. supervised the microarray work; D.G. supervised Crist, S. E. McNeil, B. Fadeel, L. Tran, S. F. Hansen, N. B. Hartmann, the bioinformatics analyses; B.F. supervised the study, analyzed data, and L. P. W. Clausen, L. M. Skjolding, A. Baun, M. Ågerstrand, Z. Gu, D. wrote the paper; all authors approved the final version. A. Lamprou, C. Hoskins, et al., Nat. Nanotechnol. 2019, 14, 629. [15] L. Farcal, F. Torres Andón, L. Di Cristo, B. M. Rotoli, O. Bussolati, E. Bergamaschi, A. Mech, N. B. Hartmann, K. Rasmussen, J. Riego- Data Availability Statement Sintes, J. Ponti, A. Kinsner-Ovaskainen, F. Rossi, A. Oomen, P. Bos, R. Chen, R. Bai, C. Chen, L. Rocks, N. Fulton, B. Ross, G. Hutchi- The transcriptomics data have been deposited in the NCBI Gene Expres- son, L. Tran, S. Mues, R. Ossig, J. Schnekenburger, L. Campagnolo, sion Omnibus (GEO) database (accession no. GSE148705), while the pro- L. Vecchione, A. Pietroiusti, B. Fadeel, PLoS One 2015, 10, e0127174. teomics data have been deposited to the ProteomeXchange consortium [16] S. Smulders, J. P. Kaiser, S. Zuin, K. L. Van Landuyt, L. Golanski, J. via the PRIDE repository with the dataset identifier: PXD019267. Other Vanoirbeek, P. Wick, P. H. Hoet, Part. Fibre Toxicol. 2012, 9, 41. relevant source data are available through the data repository of FP7- [17] S. P. Mukherjee, N. Lozano, M. Kucki, A. E. Del Rio-Castillo, L. New- NANOSOLUTIONS (http://nanosolutions.cs.unisa.it/login.cgi). man, E. Vázquez, K. Kostarelos, P. Wick, B. Fadeel, PLoS One 2016, 11, e0166816. [18] K. Bhattacharya, G. Kiliç, P. M. Costa, B. Fadeel, Nanotoxicology 2017, Keywords 11, 809. [19] S. Tuomela, R. Autio, T. Buerki-Thurnherr, O. Arslan, A. Kunzmann, immunotoxicity, nanomaterials, nanotoxicology, proteomics, transcrip- B. Andersson-Willman, P. Wick, S. Mathur, A. Scheynius, H. F. Krug, tomics B. Fadeel, R. Lahesmaa, PLoS One 2013, 8, e68415. [20] S. P. Mukherjee, K. Kostarelos, B. Fadeel, Adv. Healthcare Mater. 2018, Received: June 12, 2020 7, 1700815. Revised: July 29, 2020 [21] N. K. Tarasova, A. Gallud, A. J. Ytterberg, A. Chernobrovkin, J. R. Aran- Published online: October 11, 2020 zaes, D. Astruc, A. Antipov, Y. Fedutik, B. Fadeel, R. A. Zubarev, J. Proteome Res. 2017, 16, 689. [22] A. Krämer, J. Green, J. Pollard, S. Tugendreich, Bioinformatics 2014, 30, 523. [1] B. Fadeel, L. Farcal, B. Hardy, S. Vázquez-Campos, D. Hristozov, A. [23] R. J. Jackson, C. U. Hellen, T. V. Pestova, Nat. Rev. Mol. Cell Biol. 2010, Marcomini, I. Lynch, E. Valsami-Jones, H. Alenius, K. Savolainen, 11, 113. Nat. Nanotechnol. 2018, 13, 537. [24] P. Kinaret, V. Marwah, V. Fortino, M. Ilves, H. Wolff, L. Ruokolainen, [2] K. Paunovska, D. Loughrey, C. D. Sago, R. Langer, J. E. Dahlman, Adv. P. Auvinen, K. Savolainen, H. Alenius, D. Greco, ACS Nano 2017, 11, Mater. 2019, 31, 1902798. 3786. [3] Y. Li, J. Wang, F. Zhao, B. Bai, G. Nie, A. E. Nel, Y. Zhao, Natl. Sci. Rev. [25] The Gene Ontology Consortium, Nucleic Acids Res. 2019, 47, D330. 2018, 5, 365. [26] W. A. Schroder, T. T. Le, L. Major, S. Street, J. Gardner, E. Lambley, [4] X. Bai, F. Liu, Y. Liu, C. Li, S. Wang, H. Zhou, W. Wang, H. Zhu, D. A. K. Markey, K. P. MacDonald, R. J. Fish, R. Thomas, A. Suhrbier, J. Winkler, B. Yan, Toxicol. Appl. Pharmacol. 2017, 323, 66. Immunol. 2010, 184, 2663. [5] B. Fadeel, Front Toxicol. 2019, https://doi.org/10.3389/ftox.2019. [27] W. A. Schroder, T. D. Hirata, T. T. Le, J. Gardner, G. M. Boyle, J. Ellis, 00001. [Epub ahead of print]. E. Nakayama, D. Pathirana, H. I. Nakaya, A. Suhrbier, Sci. Rep. 2019, [6] I. Hansjosten, J. Rapp, L. Reiner, R. Vatter, S. Fritsch-Decker, R. Per- 9, 12421. avali, T. Palosaari, E. Joossens, K. Gerloff, P. Macko, M. Whelan, D. [28] C. F. A. Vogel, T. Haarmann-Stemmann, Curr. Opin. Toxicol. 2017, 2, Gilliland, I. Ojea-Jimenez, M. P. Monopoli, L. Rocks, D. Garry, K. Daw- 109. son, P. J. F. Röttgermann, A. Murschhauser, J. O. Rädler, S. V. Y. Tang, [29] M. Kanehisa, Y. Sato, M. Furumichi, K. Morishima, M. Tanabe, Nucleic P. Gooden, M. A. Belinga-Desaunay, A. O. Khan, S. Briffa, E. Guggen- Acids Res. 2019, 47, D590. heim, A. Papadiamantis, I. Lynch, E. Valsami-Jones, S. Diabaté, C. [30] K. Li, X. Zhao, B. K Hammer, S. Du, Y. Chen, ACS Nano 2013, 7, 9664. Weiss, Arch Toxicol. 2018, 92, 633. [31] B. Fadeel, Front Immunol. 2019, 10, 133. [7] S. Y. Shaw, E. C. Westly, M. J. Pittet, A. Subramanian, S. L. Schreiber, [32] X. Cai, J. Dong, J. Liu, H. Zheng, C. Kaweeteerawat, F. Wang, Z. Ji, R. R. Weissleder, Proc. Natl. Acad. Sci. USA 2008, 105, 7387. Li, Nat. Commun. 2018, 9, 4416. [8] V. Kodali, M. H. Littke, S. C. Tilton, J. G. Teeguarden, L. Shi, C. W. [33] J. Zhang, S. Wang, M. Gao, R. Li, S. Liu, ACS Sustainable Chem. Eng. Frevert, W. Wang, J. G. Pounds, B. D. Thrall, ACS Nano 2013, 7, 6997. 2018, 6, 10374. [9] N. Feliu, P. Kohonen, J. Ji, Y. Zhang, H. L. Karlsson, L. Palmberg, A. [34] D. Breznan, D. D. Das, C. MacKinnon-Roy, S. Bernatchez, A. Sayari, Nyström, B. Fadeel, ACS Nano 2015, 9, 146. M. Hill, R. Vincent, P. Kumarathasan, ACS Nano 2018, 12, 12062. [10] P. Falagan-Lotsch, E. M. Grzincic, C. J. Murphy, Proc. Natl. Acad. Sci. [35] L. Wu, Y. Zhang, C. Zhang, X. Cui, S. Zhai, Y. Liu, C. Li, H. Zhu, G. Qu, USA 2016, 113, 13318. G. Jiang, B. Yan, ACS Nano 2014, 8, 2087. Adv. Sci. 2020, 7, 2002221 2002221 (17 of 18) © 2020 The Authors. Published by Wiley-VCH GmbH www.advancedsciencenews.com www.advancedscience.com [36] M. Orecchioni, D. Bedognetti, L. Newman, C. Fuoco, F. Spada, W. [51] K. Ezzat, M. Pernemalm, S. Pålsson, T. C. Roberts, P. Järver, A. Don- Hendrickx, F. M. Marincola, F. Sgarrella, A. F. Rodrigues, C. Ménard- dalska, B. Bestas, M. J. Sobkowiak, B. Levänen, M. Sköld, E. A. Moyon, G. Cesareni, K. Kostarelos, A. Bianco, L. G. Delogu, Nat. Com- Thompson, O. Saher, O. K. Kari, T. Lajunen, E. Sverremark Ekström, mun. 2017, 8, 1109. C. Nilsson, Y. Ishchenko, T. Malm, M. J. A. Wood, U. F. Power, S. Ma- [37] J. S. Suk, Q. Xu, N. Kim, J. Hanes, L. M. Ensign, Adv. Drug Delivery sich, A. Lindén, J. K. Sandberg, J. Lehtiö, A. L. Spetz, S. El Andaloussi, Rev. 2016, 99, 28. Nat. Commun. 2019, 10, 2331. [38] E. M. Grzincic, J. A. Yang, J. Drnevich, P. Falagan-Lotsch, C. J. Murphy, [52] C. C. Ho, Y. H. Luo, T. H. Chuang, C. S. Yang, Y. C. Ling, P. Lin, Toxi- Nanoscale 2015, 7, 1349. cology 2013, 308, 1. [39] S. M. Louie, R. D. Tilton, G. V. Lowry, Environ. Sci. Nano 2016, 3, [53] C. C. Ho, Y. H. Luo, T. H. Chuang, P. Lin, Toxicology 2016, 344–346, 283. 61. [40] C. D. Walkey, J. B. Olsen, H. Guo, A. Emili, W. C. Chan, J. Am. Chem. [54] C. Y. Tsai, S. L. Lu, C. W. Hu, C. S. Yeh, G. B. Lee, H. Y. Lei, J. Immunol. Soc. 2012, 134, 2139. 2012, 188, 68. [41] C. Vogt, M. Pernemalm, P. Kohonen, S. Laurent, K. Hultenby, M. [55] N. Luo, J. K. Weber, S. Wang, B. Luan, H. Yue, X. Xi, J. Du, Z. Yang, W. Vahter, J. Lehtiö, M. S. Toprak, B. Fadeel, PLoS One 2015, 10, Wei, R. Zhou, G. Ma, Nat. Commun. 2017, 8, 14537. e0129008. [56] N. Feliu, M. V. Walter, M. I. Montanez, A. Kunzmann, A. Hult, A. Nys- [42] C. D. Walkey, W. C. Chan, Chem. Soc. Rev. 2012, 41, 2780. tröm, M. Malkoch, B. Fadeel, Biomaterials 2012, 33, 1970. [43] A. Serra, I. Letunic, V. Fortino, R. D. Handy, B. Fadeel, R. Tagliaferri, [57] A. Gallud, O. Bondarenko, N. Feliu, N. Kupferschmidt, R. Atluri, A. D. Greco, Sci. Rep. 2019, 9, 179. Garcia-Bennett, B. Fadeel, Biomaterials 2017, 121, 28. [44] V. Zhernovkov, T. Santra, H. Cassidy, O. Rukhlenko, D. Matallanas, [58] E. P. Rogakou, D. R. Pilch, A. H. Orr, V. S. Ivanova, W. M. Bonner, J. A. Krstic, W. Kolch, V. Lobaskin, B. N. Kholodenko, Toxicol. Sci. 2019, Biol. Chem. 1998, 273, 5858. 171, 303. [59] B. M. Bolstad, R. A. Irizarry, M. Astrand, T. P. Speed, Bioinformatics [45] S. Triboulet, C. Aude-Garcia, L. Armand, V. Collin-Faure, M. Cheval- 2003, 19, 185. let, H. Diemer, A. Gerdil, F. Proamer, J. M. Strub, A. Habert, N. Her- [60] J. C. Oliveros, http://bioinfogp.cnb.csic.es/tools/venny/index.html lin, A. Van Dorsselaer, M. Carrière, T. Rabilloud, PLoS One 2015, 10, (accessed: 2007). e0124496. [61] E. Y. Chen, C. M. Tan, Y. Kou, Q. Duan, Z. Wang, G. V. Meirelles, N. [46] M. Ilves, P. A. S. Kinaret, J. Ndika, P. Karisola, V. Marwah, V. Fortino, Y. R. Clark, A. Ma’ayan, BMC Bioinformatics 2013, 14, 128. Fedutik, M. Correia, N. Ehrlich, K. Loeschner, A. Besinis, J. Vassallo, [62] M. V. Kuleshov, M. R. Jones, A. D. Rouillard, N. F. Fernandez, Q. R. D. Handy, H. Wolff, K. Savolainen, D. Greco, H. Alenius, Part Fibre Duan, Z. Wang, S. Koplev, S. L. Jenkins, K. M. Jagodnik, A. Lachmann, Toxicol. 2019, 16, 28. M. G. McDermott, C. D. Monteiro, G. W. Gundersen, A. Ma’ayan, Nu- [47] I. Karkossa, A. Bannuscher, B. Hellack, A. Bahl, S. Buhs, P. Nollau, A. cleic Acids Res. 2016, 44, W90. Luch, K. Schubert, M. von Bergen, A. Haase, Part Fibre Toxicol. 2019, [63] G. Scala, A. Serra, V. S. Marwah, L. A. Saarimäki, D. Greco, BMC Bioin- 16, 38. formatics 2019, 20, 79. [48] G. M. Barton, J. C. Kagan, Nat. Rev. Immunol. 2009, 9, 535. [64] S. P. Mukherjee, G. Gupta, K. Klöditz, J. Wang, A. F. Rodrigues, K. [49] A. N. R. Weber, Z. Bittner, X. Liu, T. M. Dang, M. P. Radsak, C. Brunner, Kostarelos, B. Fadeel, Small 2020, 16, 1907686. Front Immunol. 2017, 8, 1454. [65] A. F. Rodrigues, L. Newman, D. Jasim, S. P. Mukherjee, J. Wang, I. [50] N. J. Gay, M. F. Symmons, M. Gangloff, C. E. Bryant, Nat. Rev. Im- A. Vacchi, C. Ménard-Moyon, A. Bianco, B. Fadeel, K. Kostarelos, C. munol. 2014, 14, 546. Bussy, Adv. Sci. 2020, 7, 1903200. Adv. Sci. 2020, 7, 2002221 2002221 (18 of 18) © 2020 The Authors. Published by Wiley-VCH GmbH