Introduction

Nanosilica, also known as the nanoform (<100 nm) of silicon dioxide or silica nanoparticles (SiNPs), possesses distinct physico-chemical characteristics compared to its bulk form; smaller size materials have an increased surface-to-volume ratio and a higher surface reactivity (Oberdörster 2010; Napierska et al. 2010). Due to their appealing properties, SiNPs are now extensively used in agriculture, food, and consumer products including cosmetics (Napierska et al. 2010; Khot et al. 2012; Kasaai 2015; Brinch et al. 2016). Until 2012, nearly 1.5 million tons of SiNPs had already been placed in the global market (Liljenström et al. 2013) and SiNPs became one of the three most produced nanomaterials (NMs) worldwide in 2013. Among the 846 nano-based products listed in a consumer products inventory, approximately 100 claim to contain SiNPs (Vance et al. 2015). Moreover, amorphous SiNPs are being synthesized with highly tunable biocompatibility and stability, and considered as a very promising candidate for various bio-medical applications such as gene carrier, drug delivery, and molecular imaging (Tang and Cheng 2013; Bitar et al. 2012).

In recent years, large-scale industrial production and global commercialization of SiNPs have resulted in increased risk of human exposures at workplaces (Kim et al. 2014a; Oh et al. 2014). Food additive silica (E551) is also in the nano size range (Dekkers et al. 2011), indicating that the general population is probably more exposed than initially anticipated. Moreover, in view of the efforts to use NM in medical applications, SiNPs could also be intentionally introduced into the human body for disease diagnosis and treatments (Croissant et al. 2017). Such growing potentials for exposure raised a global concern regarding the safety and potential adverse health effects of SiNPs.

Human health effects associated with silica exposure, especially crystalline silica (0.5–10 µm), have widely been studied. Occupational exposure to crystalline silica induces silicosis in workers (a fibrotic lung disease) and is also associated with lung cancer, emphysema, and pulmonary tuberculosis (Leung et al. 2012). Conversely, natural amorphous silica is generally considered as less harmful, since the toxicological potential of silica has so far been linked to its crystallinity. Recent studies have revealed that amorphous SiNPs can be as reactive as crystalline particles (Turci et al. 2016). In vivo, amorphous SiNPs are, however, cleared more rapidly from the lung, which may contribute to explain their lower pathogenic potential (Arts et al. 2007). The human health effects of nanosilica remain to be clarified and toxicologists believe that exposure to SiNPs, due to their small size, may bring different adverse effects compared to micron-sized silica (Napierska et al. 2010).

The comprehensive review of Napierska et al. (2010) suggested that exposure to SiNPs (1–100 nm) induced toxic effects in vitro (immortalized mammalian cell lines) and in vivo (rats and mice). Physico-chemical properties such as size, surface area, and surface features were found to play a key role in the toxicity of SiNPs. Importantly, Napierska concluded that physico-chemical properties of SiNPs differ based on their production method and, therefore, may cause different biological effects. However, no definite conclusions were made due to insufficient or no data available for,

  • Detailed physico-chemical characterization of different types of SiNPs;

  • Comparison of the toxicity of different types of SiNPs (based on their production process);

  • Comparison of the toxicity mechanisms of amorphous SiNPs and crystalline silica;

  • Exposure via different routes and adverse effects of chronic exposure in vivo;

  • Correlation of in vitro and in vivo studies and

  • Physico-chemical properties for the safer design of SiNPs.

Therefore, the aim of this review is to summarize the toxicity studies of SiNPs published after the Napierska review (2010), critically discuss the outcomes, and to evaluate how these data gaps have been addressed (Fig. 1).

Fig. 1
figure 1

Systematic selection of studies

Full size image

Methodology

The selection criteria was similar to the method described in Vriens et al. (2017), which was used to construct the MOD-ENP-TOX nanotoxicity database.

Exclusion criteria

To specifically focus on adverse health effects of SiNPs, papers reporting on other interventions such as ecotoxicity, synergistic effects, SiNPs doped with other materials, and therapy-based outcomes were excluded.

Inclusion criteria

Papers reporting,

  1. 1.

    physico-chemical characteristics such as primary size, shape, composition, and crystallinity;

  2. 2.

    toxicological endpoints such as cytotoxicity, apoptosis/necrosis, genotoxicity, oxidative stress, immunotoxicity, and autophagy using immortalized cell lines or primary cells (experimental in vitro studies); and

  3. 3.

    toxic effects in laboratory-animals, more specifically in vivo experiments using rats and mice.

Literature search

We searched two databases for papers published from June 2010 (after Napierska et al. 2010) to December 2016. In the PubMed “http://www.ncbi.nlm.nih.gov/PubMed” and EMBASE “https://www.embase.com” databases, the following keyword combinations were used: “silica nanoparticle” OR “silica NP” AND “toxicity” NOT “review”. We retrieved a list of 859 and 405 articles in English, respectively. In a second step, duplicates were removed and the titles were screened to identify studies that best matched with our search terms, leaving 611 relevant papers. In a third step, we excluded papers that met the exclusion criteria and left 128 most relevant papers. Finally, 82 papers reporting a minimum set of physico-chemical characterization and toxic effects were selected for the main content of the review. Notably, only one study was found on the toxicity of crystalline nanosilica (Chu et al. 2012), but it was finally excluded due to insufficient data on the size of particles. As a result, the review is dealing only with amorphous SiNPs.

Induction of oxidative stress is considered as the major mechanism involved in SiNPs toxicity (Wang et al. 2009; Ye et al. 2010a, b) and, therefore, the in vitro section of the review was structured according to toxic endpoints and its association with oxidative stress. For in vivo, studies were sorted based on exposure route and modalities, since they can significantly influence the toxicokinetics of SiNPs. Throughout the review, the following abbreviations were used to indicate the different types of silica nanoparticles: SiNPs when it is not clear which type was used, C-SiNPs for colloidal silica; S-SiNP for stöber silica; M-SiNPs for mesoporous silica; Pr-SiNPs for precipitated silica; and Py-SiNPs for pyrogenic silica.

The table summarizing in vitro studies (Table 1) was sorted according to the type of SiNPs (colloidal, stöber, mesoporous, pyrogenic, precipitated, and not specified) and cell types. Table 2 (in vivo studies) was organized according to the type of SiNPs and exposure routes.

Table 1 In vitro studies on SiNPs toxicity
Full size table
Table 2 In vivo studies on SiNPs toxicity
Full size table

In vitro studies

Cytotoxicity

Oxidative stress (over production of reactive oxygen species, i.e., ROS) induced by NPs could damage the cellular components and lead to cell death via apoptosis (Fu et al. 2013). Therefore, studies reporting on cytotoxicity and oxidative stress were summarized in this section.

Cytotoxicity associated with oxidative stress

Duan et al. (2013a) showed that S-SiNPs (62 nm) induced time- (6, 12, and 24 h) and dose-dependent (25–100 µg/ml) reduction in cell viability (assessed by 3-(4,5-dimethylthiazol-2-Yl)-2,5-diphenyltetrazolium bromide, i.e., MTT), loss of membrane integrity (lactate dehydrogenase (LDH) release) and apoptosis (Annexin V/PI staining) in human umbilical vein endothelial cells (HUVECs). Apoptosis was also induced in lung (A549) and skin epithelial cells (A431) treated with Pr-SiNPs (15 nm). A dose-dependent increase (25–200 µg/ml for 72 h) in cytotoxicity (MTT and LDH), ROS production (assessed by dichlorodihydrofluorescein assay, i.e., DCFH-DA), lipid peroxidation (measurement of malondialdehyde, i.e., MDA), and apoptosis (caspase 3 and 9 activity) was observed in both cell lines. The lung cells showed, in general, a slightly higher toxic response compared to skin cells (Ahamed 2013).

SiNPs (20 and 80 nm) induced P53-mediated apoptosis in human fetal lung fibroblasts (HFL-1). At the dose of 500 µg/ml, 20 nm SiNPs induced a threefold increase in DCF fluorescence compared to 80 nm. In addition, increased expression of P53, upregulation of cytochrome C (CytC) and caspase 9, and downregulation of anti-apoptotic protein B-cell lymphoma 2 (bcl2) was observed in cells treated with 1000 µg/ml for 48 h (Xu et al. 2012). Another study with lung fibroblasts also showed that SiNPs (20 nm) could reduce cell viability (MTT) by inducing apoptotic cell death (fluorescence microscopy) in a dose-dependent manner (250–1000 µg/ml for 48 h) (Zhang et al. 2011). Athinarayanan et al. (2014) isolated SiNPs (10–50 nm) from commercial food products processed with food additive silica (E551) and exposed human lung fibroblasts (WI-38 cell line) with increasing doses (25–400 µg/ml). After 24 h, they observed cytotoxicity (MTT) in a dose-dependent manner and ROS production (DCFH-DA) at 50 µg/ml.

Cytotoxicity not associated with oxidative stress

Py-SiNPs (12 and 40 nm) induced a significant size and dose- (31.3, 93.8, and 156.3 µg/cm2 culture well) dependent cytotoxicity (LDH, Sulphorodhamine B assay (SRB) and water-soluble tetrazolium-1(WST-1)) in human colon carcinoma cell line (HT29), while no induction of ROS (DCFH-DA) was observed (Gehrke et al. 2013). In the study by Napierska et al. (2012a), 50 μg/ml (24 h) of 16 nm iron-doped S-SiNPs and pure S-SiNPs induced strong cytotoxicity (MTT and LDH) in a human endothelial cell line (EA.hy926), but a significant increase in oxidative stress markers [GSH depletion, malondialdehyde (MDA formation), induction of heme oxygenase-1, glutathione reductase, and NADPH oxidase-1] was observed only for iron-doped SiNPs.

Conclusion: cytotoxicity

Cytotoxicity of SiNPs was investigated using different cell lines and incubation times, making the comparison between studies difficult. However, from Table 3, it is clear that all types of SiNPs induced cytotoxicity. Significant (compared to untreated cells) cytotoxic effects were observed only at or above the concentration of 25 µg/ml. Furthermore, it can be clearly seen that SiNPs induced oxidative stress and mediated apoptosis mainly via the intrinsic or mitochondrial pathway (caspase-dependent pathway) in a size- and dose-dependent manner. ROS-mediated toxicity is believed to be an important mechanism of NP toxicity including SiNPs (Manke et al. 2013). Nevertheless, Py- and S-SiNPs caused cytotoxicity without measurable levels of ROS production. It was demonstrated that the disturbance of membrane integrity due to direct cell-membrane interaction might be another possible mechanism of NP cytotoxicity (Fröhlich et al. 2009; Thomassen et al. 2011). However, neither of these studies did substantiate these observations and, therefore, SiNPs cytotoxic effects in the absence of oxidative stress remain poorly understood.

Table 3 Comparison of toxic effects induced by different types of SiNPs (in vitro)
Full size table

Furthermore, some authors used very high concentrations that may cause “overloading” of cells and modify the nature of NP–cell interactions (Wittmaack 2011). In these cases, it is difficult to evaluate whether the observed effects are physiologically relevant. Although it is challenging, we consider a dose of 384 µg/cm2 or higher as irrelevant to human inhalation exposure for amorphous silica, based on the estimation that can be derived from the occupational exposure levels (OELs) (Fig. 2).

Fig. 2
figure 2

In vitro dose estimation from human tolerable levels (OELs)

Full size image

Genotoxicity

In this section, we presented studies on genotoxic effects of SiNPs as it is used as another major endpoint to characterize hazard of NMs. Direct interaction with DNA, oxidative DNA damage, depletion of anti-oxidants, cell cycle arrest, and abnormal expression of genes have been identified as potential mechanisms of NP mediated (geno)toxicity (Donaldson et al. 2010).

DNA damage associated with oxidative stress

Exposure to SiNPs (15, 30, and 100 nm) resulted in a size- and dose- (2.5–10 µg/ml for 24 h) dependent increase in 8-hydroxy-2′-deoxyguanosine levels (8-OH-dG), phosphorylation of histone on serine-139 (ɣH2AX), and DNA strand breaks (comet) in human keratinocytes (HaCaT) (Gong et al. 2012). Nabeshi et al. (2011a) also demostrated that exposure to SiNPs (70 nm; 10–90 µg/ml for 24 h) resulted in the increase of oxidative DNA damage (8-OH-dG levels) in HaCaT cells. SiNPs were taken up via actin-mediated endocytosis. Micron-sized particles used in these studies showed no or little effects.

The viability of human Caucasian colon adenocarcinoma (Caco-2) cells dropped to 40% when exposed to 15 nm C-SiNPs (64 µg/ml for 24 h), and, at this same concentration, nearly a threefold increase in micronuclei formation, fivefold increase in histone phosphorylation (ɣH2AX), and a significant increase in DCF fluorescence were observed. The particles were localized within lysosomes and endocytic compartments, but not in the nucleus. 55 nm C-SiNPs did not induce any of these effects at the same concentration (Tarantini et al. 2015b).

A non-significant increase in % tail DNA (comet assay) and no chromosomal aberrations were induced by 17 nm SiNPs in human peripheral lymphocytes treated with 100 µg/ml, while a dose-dependent (50–100 µg/ml for 24 h) ROS production (DCFH-DA) and GSH depletion were observed (Rajiv et al. 2015).

Cell cycle arrest associated with oxidative stress

S-SiNPs (62 nm) induced increase in DCF fluorescence and decrease in superoxide dismutase (SOD) and glutathione peroxidase (GSH-Px) in HUVECs in a dose-dependent manner (25–100 µg/ml for 24 h). Oxidative stress was linked to cell cycle arrest at G2/M checkpoint (upregulation of chk 1 and downregulation of Cdc25c, Cyclin B1, and Cdc2) and increase in apoptosis (Duan et al. 2013a). In the study by Li et al. (2011), a size-dependent (19, 43, and 68 nm) increase in oxidative stress (DCF fluorescence) and cell cycle arrest in S and G2/M was observed in HepG2 cells exposed to 100 µg/ml of C-SiNPs. Cell cycle arrest in G2/M phase along with the increase in ROS was also noticed in human hepatic cell line (LC-02) treated with S-SiNPs (50 nm) in a dose-dependent manner (50–200 µg/ml for 24 h) (Wang et al. 2013).

DNA damage not associated with oxidative stress

Genotoxicity of Py-SNP (20 and 25–70 nm), Pr-SNP (20 nm), and C-SNP (15 and 40–80 nm) SiNPs were studied in Chinese hamster lung fibroblasts. Py-SiNPs (20 nm) induced a significant increase in DNA strand breaks at 66 µg/ml (24 h), while C-SiNPs (15 nm) showed a similar effect only at 252 µg/ml. Neither of these SiNPs did induce ROS. SiNPs in the size range 25–80 nm exerted no or little genotoxicity (Guichard et al. 2016).

Genotoxicity reports without the assessment of oxidative stress

M-SiNPs (100 nm) induced a significant increase in phosphorylated-ɣH2AX-foci in HT-29 cells treated with a dose of 10 µg/ml for 24 h (Sergent et al. 2012). In the human embryonic kidney cell line (HEK293), 579 genes were upregulated and 1263 genes were downregulated after 24 h of exposure (100 µg/ml) to 100 nm M-SiNPs (Zhang et al. 2015). In another study, 15-nm C-SiNPs induced a significant increase in DNA strand breaks (comet assay) in chinese hamster cells (V79) and A549 cells at 100 µg/ml (24 h), but, for 55 nm, this effect was observed only in A549 cells (Maser et al. 2015). A significant increase in DNA tail length (comet assay) was observed in HaCaT cells treated with 30 µg/ml (24 h) of 70 nm SiNPs (Nabeshi et al. 2011b).

Exposure to C-SiNPs (~7 nm) resulted in positive genotoxic effects (Lymphoma assay) in mouse lymphoma cells treated with 100 and 150 µg/ml for 4 h (Demir and Castranova 2016). SiNPs (12 nm) induced DNA strand breaks in RAW 264.7 at 200 and 400 µg/ml, but the induction of micronuclei was noticed only at 400 µg/ml. The particles were internalized in vesicles and in the nucleus (Hashimoto and Imazato 2015).

At any tested concentrations (1–100 µg/ml for 72 h), SiNPs (10–25, 5–30, 35, 15, 80, and 90 nm) neither induced cytotoxicity nor micronuclei in immortalized Balb/3T3 fibroblasts (Uboldi et al. 2012). In another study, Pr-SiNPs (NM-200 and NM-201) and Py-SiNPs (NM-202 and NM-203) with primary size between 14.5 and 16 nm did not induce any micronuclei (cytokinesis block micronucleus assay) in human peripheral lymphocytes exposed to different concentrations (200–1250 µg/ml) over 24 h (Tavares et al. 2014). It is also worthy to note that, in the latter study, the positive control used did not differ from control conditions.

Conclusion: genotoxicity

C-SiNPs and S-SiNPs induced genotoxicity in human tumor cell lines (lung, kidney, skin, and gastro-intestinal systems) and the amplitude of the effect negatively correlated with the size of the NPs. DNA strand breaks were observed at low concentrations (2.5–10 µg/ml), particularly in skin-derived cell lines. The genotoxic effects of C- and S-SiNPs were mainly associated with the induction of oxidative stress, while such information is very limited for other types (Py- and Pr-SiNPs). One study indicated that Py-SiNPs induce DNA damage without the generation of ROS, suggesting that other mechanisms such as direct DNA damage might be involved (Magdolenova et al. 2014). However, it is very difficult to judge whether such genotoxic effect is direct or indirect, since the cellular uptake and subcellular localization of SiNPs are not often reported. Furthermore, several factors such as SiNP properties, cell type, and exposure scenarios (such as concentrations, assays, and endpoints) may influence the outcomes (Magdolenova et al. 2014), making the comparison difficult between studies and indicating an urgent need for the standardization of genotoxicity studies.

Immunotoxicity

NPs entering the body will most probably interact with immune cells, as they are the first line of defence in human body. In this section, we presented the immune responses induced by SiNPs in different cell lines.

Immunotoxicity associated with oxidative stress

Hara et al. (2014) exposed THP-1-derived macrophages to 100 µg/ml of SiNPs (30 nm) for 6 h and found a significant increase in interleukin-1-beta (IL-1β), ROS production, and SiNP uptake via phagocytosis. In the study of Choi et al. (2010), larger sized SiNPs (150–200 nm) were effectively phagocytosed by primary rat microglial cells after 24 h of exposure to different concentrations (0.0728–7.28 µg/ml). A significant increase in ROS, reactive nitrogen species (RNS) and IL-1β was detected at all concentrations.

Immunotoxicity not associated with oxidative stress

At 10 and 20 µg/ml, Di Cristo et al. (2016) found that Py-SiNPs (~14 nm) induced a stronger increase of tumor necrosis factor-alpha (TNF-α), interleukin(IL)-6, and IL-1β in RAW.264.7 macrophages compared to similar sized Pr-SiNPs; Notably, no SiNPs induced ROS in RAW.264.7 macrophages.

Immunotoxicity reports without the assessment of oxidative stress

A significant increase in TNF-α, IL-6, and IL-1α, mitogen activated protein kinases (MAPKs), and nuclear factor (NF)-κB were observed only for C-SiNPs (100 nm) in J774A.1 macrophages exposed to 100 µg/ml of same sized (100 nm) C-SiNPs or M-SiNPs (Lee et al. 2011).

Uemura et al. (2016) showed that SiNPs (10 and 50 nm) caused dose-dependent (6.25–100 µg/ml) increase in the production of TNF-α and decrease of IL-6 in RAW.264.7 macrophages, while their amine surface-modified counterparts did not. Furthermore, 300 and 1000 nm micron-sized particles (both bare and amine modified) also showed a dose-dependent decrease of IL-6. Notably, the effects were stronger for 50 nm compared to other particles. The same cell line was utilized by Yu et al. (2011) to investigate phagocytosis using inductively coupled plasma mass spectroscopy (ICP-MS), and they found that S-SiNPs (25 nm) were phagocytosed at least ten times more than M-SiNPs of same size and high aspect ratio SiNPs (AR 2, 4, and 8). In the study by Napierska et al. (2012b), THP-1 cells dosed with 5 µg/cm2 S-SiNPs (2 nm) showed a significant increase of IL-8, TNF-α, and macrophage inflammatory protein (MIP)-1α, while only a non-significant increase in MIP-1α expression was observed for 16 and 104 nm S-SiNPs.

Conclusion: immunotoxicity

The main cells used to study immune responses to SiNPs were ‘innate’ cells such as monocytes and macrophages. Therefore, the identified in vitro studies only address a very limited part of the immune system, essentially pro-inflammatory responses and potential phagocytosis. Furthermore, the data on immune responses and oxidative stress are very limited and, therefore, no firm conclusions can be made. SiNPs, not only induced stronger pro-inflammatory responses compared to sub-micron and micron sized particles but also size-specific effects within the nano-range in immune cells are observed. Besides size, shape and porosity seem to influence the phagocytosis of SiNPs.

Autophagy

Recently, a growing body of evidence identified autophagy as a cellular defence mechanism against NP toxicity, since it plays a key role in removing misfolded proteins and clearing damaged organelles (Glick et al. 2010). Hence, we present here studies that show induction of autophagy upon exposure to SiNPs.

Autophagy associated with oxidative stress

The same S-SiNPs (62 nm) were used in three studies to investigate the induction of autophagy. Along with the dose-dependent (25–100 µg/ml) increase in ROS production, increase in autophagy bio-marker-microtubule-associated protein 1A/1B-light chain 3 (LC3) and monodansylcadaverine (MDC) labelled autophagic vacuoles were detected in HepG2 cells treated with 62 nm S-SiNPs. In addition, transmission electron microscopy (TEM) images revealed that autophagosomes and autolysosomes induced in the presence of SiNPs (Yu et al. 2014). The same S-SiNPs (62 nm) induced increase of LC3-II/LC3-I ratio and decrease of p-mTOR/mTOR, p-P13 K/P13 K and p-Akt/Akt in HUVEC cells in a dose-dependent (25–100 µg/ml) manner (Duan et al. 2013b, 2014a, b). The results of Guo et al. (2016) also suggest that 50 µg/ml of S-SiNPs (58 nm) could induce autophagy via MAPK/Bcl-2 and PI3K/Akt/mTOR signaling in HUVECs.

After 4 h of exposure to 200 µg/ml of S-SiNPs (50 nm), autophagosomes and ROS production was observed in HaCaT cells. The TEM images revealed that SiNPs were in the cytoplasm and lysosomes, but not in the nucleus. (Liang et al. 2014).

SiNPs (4–13 nm, 62.5 µg/ml) induced a time-dependent (24, 48, and 72 h) reduction in cell viability and increase in oxidative stress (DCF fluorescence and GSH depletion) in the lung fibroblast cell line MRC-5. Compared to control, a significant increase of autophagic vacuoles and LC-3 II/LC3-I ratio was also observed in a time-dependent manner (Voicu et al. 2015). A549 cells, when exposed to 100 and 1000 µg/ml of 20-nm SiNPs, showed threefold and fivefold increase in MDC fluorescence, respectively. In addition, autophagy genes such as ATG-12 and BECN were significantly upregulated (30- and 50-fold, respectively) along with increased production of ROS in cells dosed with 1000 µg/ml (Nowak et al. 2014).

Conclusion: autophagy

SiNPs, particularly S-SiNPs induced autophagy mainly via oxidative stress-mediated upregulation of autophagy-related genes and differential regulation of Akt/mTOR signaling. Similar to cytotoxicity, 25 µg/ml appeared to be the lowest exposure concentration at which SiNPs exhibited significant effects. Furthermore, induced autophagy is correlated to cytotoxicity, suggesting that exposure to SiNPs caused irreversible (serious) cellular damage and resulted in autophagic cell death. Besides autophagy induction, lysosomal and autophagy dysfunction could be a potential mechanism of NPs toxicity (Stern et al. 2012), which has, however, not been investigated for SiNPs.

Toxic effects on blood cells and endothelial dysfunction

Several studies suggest that NPs, when inhaled or ingested, can translocate across barriers (such as air–blood) of the body, enter the circulation, and interact with the cardiovascular system. In this section, we summarized the studies that report the effects of SiNPs on blood and endothelial cells.

Toxic effects on blood cells

Nemmar et al. (2015) showed that mouse blood platelets could aggregate after 3 min of incubation with 5 or 25 µg/ml of 50-nm C-SiNPs, while in the study of Jose Corbalan et al. (2012), such aggregation was observed in 15 min (10 µg/ml of 10 nm C-SiNPs). The latter study also showed fourfold reduction of the nitric oxide (NO)/peroxynitrite (ONOO−) ratio compared to non-treated platelets.

Maurer-jones et al. (2010) investigated the role of porosity of SiNPs on blood cell toxicity. M-SiNPs (25 nm) reduced the cell viability of red blood cells (RBCs) to 50% at the concentration of 270 µg/ml, while non-porous S-SiNPs of similar size required only 20 µg/ml to reach this level of cytotoxicity. In another study, 10% hemolysis (LC10) of RBCs was observed at 36 µg/ml for S-SiNPs (115 nm), while M-SiNPs required 154 µg/ml to induce the same effects. For amine-coated counterparts, LC10 were 97 and 30 µg/ml for S- and M-SiNPs, respectively (Yu et al. 2011).

Endothelium dysfunction

Exposure to different concentrations (12.5–100 µg/ml) of S-SiNPs (58 nm) resulted in a dose-dependent increase in inflammatory mediators such as IL-1β, IL-8, and TNFα, intercellular adhesion molecule-1 (ICAM-1), vascular cell adhesion molecule (VCAM-1), and monocyte chemoattractant protein-1 (MCP-1) in HUVEC cells (Guo et al. 2015). In the study of Corbalan et al. (2011), 10 µg/ml of 10-nm C-SiNPs induced extremely low NO/NOO− ratio (~0.1) in HUVEC cells. Furthermore, free radical production, pro-inflammatory cytokines (IL-6 and IL-8), and NF-κB-binding activity were significantly increased in treated cells. In these two studies, increase in ROS was observed at all tested concentrations.

S-SiNPs (62 nm) induced an imbalance in the ratio NO/nitric oxide synthase (NOS) enzyme in HUVEC cells and such imbalance resulted in a significant increase of pro-inflammatory response (c-reactive protein CRP, IL-1β, IL-6, and TNFα) in a dose- (50–100 µg/ml) dependent manner (Duan et al. 2014b).

Conclusion: Toxic effects on blood cells and endothelial dysfunction

Endothelial cells and platelets together play a key role in maintaining the vascular homeostasis (Rajendran et al. 2013). C-SiNPs induced oxidative stress and disturbed NO/NOO− ratio, which resulted in the aggregation of platelets and endothelial dysfunction. This information is not available for other types of SiNPs. Furthermore, C-SiNPs mediated endothelial dysfunction resulted in pro-inflammatory signals via the secretion of cytokines and adhesion molecules. Together, these results suggest the potential of SiNPs to cause vascular thrombosis and atherosclerosis (Radomski et al. 2005). Furthermore, SiNPs caused hemolysis of RBCs in a size-, charge-, and porosity-dependent manner.

Neurotoxicity

NPs of very small size are capable of translocating across the blood–brain barrier (Hu and Gao 2010). Therefore, studies investigating effects on cells relevant for neurotoxicity are presented here.

Rat medulla tumor cells (PC12 cell line) incubated with the supernatant of 20 nm SiNPs-treated-microglial cells (250 µg/ml and 500 µg/ml for 24 h) did not show any effects compared to the control. Earlier in this study, no secretion of bio-mediators was observed in SiNPs-treated-microglial cells (Xue et al. 2012). In contrast to this study, PC12 cells exposed directly to SiNPs (25 nm; 25–200 µg/ml for 24 h) showed increased uptake and a dose-dependent increase in the induction of autophagy (increase in LC-II and Beclin 1) and inhibition of PI3 K-Akt-mTOR signaling (Xie and Wu 2016). Yang et al. (2014a) showed that exposure to SiNPs (15 nm; 10 µg/ml for 24 h) induced pathological signs of Alzheimer’s disease such as altered expression of amyloid precursor protein (APP) and neprilysin, enhanced phosphorylation of tau at Ser262 and Ser396, and activation of glycogen syntheses kinase (GSK)-3β in human SK-N-SH and mouse neuro2a neuroblastoma cells.

Conclusion: neurotoxicity

In vitro studies used cell lines of CNS-based cells, mainly neuron like cells. Although the data on the neurotoxic effects of SiNPs are very limited, studies above suggest that SiNPs can induce adverse effects including the markers of Alzheimer’s disease, when in direct contact with neuroblastoma cells.

Miscellaneous issues

Influence of cell lines on SiNPs cytotoxicity

A dose- (80–640 µg/ml) dependent decrease in the viability and increase of apoptosis were observed in HepG2 cells in presence of SiNPs (7 and 20 nm), but a significant reduction was observed in normal human liver cells (LC-02) only at the unrealistic dose level of 640 µg/ml (Lu et al. 2011). In another study, SiNPs (10–50 nm) induced a dose-dependent (100–600 µg/ml) increase in LDH release in Caco-2 cells, but a significant release of LDH was observed only at the unrealistic dose of 600 µg/ml in human gastric epithelial cells (GES-1). Furthermore, exposure to these SiNPs (200 µg/ml for 48 h) also induced cell cycle arrest in S phase for GES cells and G2/M in Caco-2 cells (Yang et al. 2014b). In a porcine kidney cell line (LLC PK1) exposed to 20-nm SiNPs, a dose- (5–50 µg/ml) dependent increase in DCF fluorescence and MDA formation was observed, whereas human kidney cells (HK-2) showed little effects at 50 µg/ml (Passagne et al. 2012).

Conclusion: miscellaneous issues

No firm conclusions can be drawn from these cases; however, the cytotoxicity of SiNPs appears to vary with species and cell line.

Physiologically relevant cultures

Lung co-culture models

Co-cultures of lung cells are usually made with epithelial cells on the apical and endothelial cells on the basal compartment of a transwell membrane, with or without monocytes on top of the epithelial cells. In a co-culture (A 549 at the apical and ISO-HAS-1 at the basolateral compartment) exposed to 100-µg/ml 30-nm C-SiNPs (coated with or without surfactant), nearly a fivefold increase of IL-8 release for both forms of SiNPs was observed in both compartments (Kasper et al. 2015). When other epithelial cells were used—H441 cells—at the apical together with ISO-HAS-1 cells at the basolateral compartment, these C-SiNPs induced IL-8 were expressed in both compartments, while SiCAM-1 (6–600 µg/ml) and IL-6 (at 6 and 60 µg/ml) were observed only in the apical part (Kasper et al. 2011). The same co-culture model was exposed to 100 µg/ml of S-SiNPs (15, 35, and 80 nm) and the authors noticed an increase of IL-8, TNF-α, and surfactant protein (SP-A1 and SP-A2) expression compared to the control. In addition, less IL-8 and surfactant protein expression, and more TNF-α were observed in the co-culture added with THP-1, notably the effect was the highest for 35 nm (Farcal et al. 2012).

Napierska et al. (2012b) tested SiNPs with primary size 2, 16, 60, and 104 nm (dosed at 10 µg/cm2) in a co-culture (A549 at the apical and EA. hy926 at the basolateral compartment) and observed increase in cytokines such as IL-6, IL-8, TNF-α, and MIP-1α only for 60 and 2 nm (except IL-8). When THP-1 were added to the co-culture, a significant increase in IL-8 and decrease in TNF-α were observed only for 2 nm. The expression of cytokines was also differentially regulated for 16 and 104 nm before and after THP-1 added, but the effects were stronger for 60 and 2 nm, particularly 60-nm NPs.

Air–liquid interface

At the air–liquid interface (ALI), aerosolized and deposited 12-nm Py-SiNPs (52 µg/cm2) and 50-nm S-SiNPs (117 µg/cm2) induced significantly less biological effects (LDH leakage, IL-8 release, COX-2 expression, and p38 phosphorylation) in A549 cells compared to A549 exposed to 15.6 µg/cm2 under sub-merged conditions (Panas et al. 2014).

Conclusion: physiologically relevant cultures

Sub-merged (co) cultures and ALI systems (Lenz et al. 2013; Panas et al. 2014) have been claimed to more closely mimicking the in vivo exposure scenarios compared to monocultures. In these systems, the toxicity and pro-inflammatory responses are significantly modulated by SiNPs, which represent the complexity of in vivo systems and need for the establishment of physiologically relevant in vitro cultures. However, at this moment, it is difficult to know whether these biological responses were influenced by SiNPs physico-chemical properties.

Chronic in vitro studies

In vitro chronic Py-SiNPs (12 nm) exposure of intestinal epithelial cell line (C2BBe1) was examined in a recent study. The cells were exposed to 10-µg/cm2 SiNPs for 24 h. After 24 h, the medium was replaced (without NPs) and cells were allowed to grow for 4–6 days. At the end of incubation, cells were passaged and again exposed for 24 h and grown for 4–6 days; this cycle was repeated for 29 passages (total life span). The cells and supernatants were collected at the end of each passage for analysis. Though the particles were internalized (only in a fraction of cells), no significant induction of necrosis, apoptosis, and LDH release and decrease in cell viability was observed for any of these conditions (McCracken et al. 2013).

In vivo studies

Ingestion exposure

Single exposure

Lee et al. (2014) investigated the tissue distribution and excretion profiles in Sprague–Dawley (SD) rats orally administered with a very high dose of 500 or 1000-mg/kg bw of C-SiNPs (20 and 100 nm). Silicon (Si) levels were significantly elevated in liver, kidney, lung, and spleen at 6-h post-administration, while no such increase was noticed in brain, ovaries or testes, esophagus, stomach, and intestine even after 7 days of post-administration. Nearly 75–80% of administered SiNPs were excreted via urine and 7–8% via feces. The author also noticed that 20 nm SiNPs excreted faster than the 100 nm SiNPs. In another study, C-SiNPs (12 nm) did not induce any toxicity after 14 days in SD-rats administered with a single oral dose (1959 or 2061 mg/kg bw) (Yun et al. 2015).

In the study of Li et al. (2015), M-SiNPs with different aspect ratios (AR, spherical 83 nm with AR 1, short rods with AR 1.75 and long rods with AR 5) were administered in mice (40 mg/kg bw via gavage). After 7 days, a very high amount of Si was detected only in the liver of spherical SiNPs exposed mice. Urinary excretion, intestinal absorption, and organ distribution of SiNPs were decreased with increasing aspect ratio. Furthermore, increase in renal damage such as hemorrhage, vascular congestion, and renal tubular necrosis with increasing aspect ratio was observed after 14 days of administration.

Kim et al. (2014b) exposed C57BL/6 mice (oral gavage; 750 mg/kg bw) to C-SiNPs (20 and 100 nm) that had been modified with or without l-arginine to determine the influence of surface charge on immunotoxicity in vivo. After 14 days, the author noticed a size-dependent decrease in WBCs cell count and cytokines (in blood), and reduced proliferation of B cells and T cells (from spleen) only for uncoated SiNPs.

Repeated exposure

C-SiNPs (12 nm) neither induce abnormal changes in blood biochemical and hematological parameters nor accumulated in any organs of the acutely (489.8, 979.5, or 1959 mg/kg bw/day, during 14 days) or sub-chronically (244.9, 489.8, or 975.9 mg/kg bw/day, during 13 weeks) exposed (via gavage) SD rats (Yun et al. 2015). In the study by Kim et al. (2014c), C-SiNPs (20 and 100 nm) did not induce any significant changes (compared to control) in clinical signs, blood biochemical, hematological, and histopathological analysis in sub-chronically exposed Crl: CD (SD) specific pathogen-free rats (500, 1000, or 2000 mg/kg bw/day for 90 days).

SD rats were administered (via oral gavage) with low doses (5, 10 or 20 mg/kgbw) of Pr-SiNPs (NM 200 and 201) and Py-SiNPs (NM 202 and 203) at 0, 24 and 45 h and sacrificed at 3 h after the last administration. No significant genotoxicity was observed in cells extracted from different organs (duodenum, colon, blood, kidney, liver, and spleen), and no pathological conditions were recorded in any of these organs (Tarantini et al. 2015a).

Hofmann et al. (2015) investigated the prenatal toxicity by exposing pregnant Wistar rats to Pr-SiNPs (NM 200) from gestation day 6–19. At doses of 100, 300, or 1000 mg/kgbw/day, administered SiNPs did not induce malformations in fetuses or death of the rats.

Zande et al. (2014) investigated the sub-acute and sub-chronic toxicity of Py-SiNPs (7 nm) and NM-202 (10–25 nm). In the sub-acute part, male SD rats were orally (via food) exposed to 100, 1000, or 2500 mg Py-SiNPs/kg bw/day for 28 days, while in the sub-chronic part, the rats were exposed only to the highest dose (2500 mg/kg bw/day) repeatedly for 84 days. ICP-MS analysis of target organs showed that NM-202 was significantly distributed in the lung, kidney, and spleen. Although no Si was detected in the liver, histopathological analysis and gene expression studies revealed that fibrosis was induced in the liver after 84 days of exposure. Si was found in the spleen of Py-SiNPs exposed rats only at the end of chronic exposure. The examined parameters such as blood biochemistry, antibody levels (IgG and IgM), and cytokines indicated no systemic toxicity in any of the Py- or NM-202 treated rats.

Yoshida et al. (2014) found an increased intestinal absorption of carboxyl (twofold) and amine-coated SiNPs (1.5-fold) compared to uncoated SiNPs (70 nm) in orally exposed BALB/c mice (2.5 mg/mouse/day for 28 days). In this study, the whole small intestine was processed using the everted gut sac method and a significant amount of Si was detected by ICP-MS. However, no signs of systemic toxicity were observed for any of these SiNPs.

In male Wistar rats that exposed to 10–15 nm SiNPs (oral gavage; 333.3 mg/kg bw/day for 5 days), histopathological analysis revealed gross tissue damage in kidney (cell swelling and necrosis), lung (interstitial pneumonia and bronchopneumonia), and in the testis (congestion, reduction of spermatogenesis and edema). In addition, blood biochemical parameters such as albumin, cholesterol, triglycerides, total proteins, urea, high-density lipoprotein (HDL), and low-density lipoprotein (LDL), as well as alkaline phosphatase (ALP) and aspartate aminotransferase (AST) activities were significantly increased in treated mice (Hassankhani et al. 2014).

Inhalation exposure

Single exposure

In the study of Morris et al. (2016), C57BL/6 mice were intratracheally instilled with 4- or 20-mg S-SiNPs/kg bw and significant effects were observed only at the dose of 20 mg/kg. Twenty hours after instillation, approximately 20- and 10-fold higher cell number was observed in the bronchoalveolar lavage (BAL) of mice treated with the bare and amine-coated SiNPs, respectively, compared to control mice; neutrophils were also increased about 30- and 20-fold, respectively.

In another study, 15- and 55-nm C-SiNPs neither induced DNA damage nor micronuclei in lung and bone marrow (erythrocytes) cells of Wistar rats at 72 h of post-instillation (360 µg) (Maser et al. 2015). However, the distribution of administered SiNPs to these organs was not investigated.

Repeated exposure

A size- and dose-dependent increase in the distribution of SiNPs was observed in the serum and heart of male Wistar rats intratracheally instilled (2, 5, or 10 mg/kg bw/day for 16 days) with SiNPs (30, 60, 90, and 300 nm). Blood parameters (WBCs and platelets), inflammatory bio-markers (TNF-α, IL-1β, and IL-6), and oxidative stress bio-markers (ROS and MDA formation) were significant increased, while NO, NOS, and eNOS were significantly decreased in the serum of the treated mice (Du et al. 2013).

A study compared the toxicity and biodistribution of pristine SiNPs (19 nm) and aged paints containing SiNPs in BALB/c mice. The suspensions (20 µg/aspiration) were oropharyngeally aspirated once a week for 5 weeks, and mice were sacrificed either at 2- or at 28-day post-final aspiration treatment. Pristine SiNPs were significantly distributed in the lungs and liver, while only a low amount of Si was detected in the liver of paint-exposed mice. No signs of toxicity were observed, except a slight inflammation (a slight increase in macrophages and neutrophils together with an increase of IL-1β) with pristine SiNPs (Smulders et al. 2014).

Male Wistar rats were intranasally instilled with 150 µg SiNPs (10 and 80 nm) repeatedly for 30 days. After 30 days, a size-dependent increase in the levels of hydrogen peroxide (H2O2), MDA formation, TNF-α, IL-1β, MCP-1, and NF-ĸβ was observed in the frontal cortex, corpus striatum, and hippocampus of the brain. A similar quantity of Si was detected in all these three regions. In addition, a significant decrease in GSH levels was observed in these tissues (Parveen et al. 2015). The latter study shows the potential translocation of (mainly small) SiNP from the nose to the brain, but it has to be noted that the mice received a high dose during 30 days.

Intratracheally administered (2 mg/kg bw; 15 times, once every 3 days) S-SiNPs (58 nm) resulted in reproductive toxicity in C57 mice via enhanced ROS production and cell cycle arrest (G0/G1 phase) in the testicular tissues, and by decreasing the number of mature sperms and primary spermatocytes (Zhang et al. 2016). In another study, repeated intracheal instillation of S-SiNPs (58 nm; 0, 7, 21, and 35 mg/kg bw; 5 times, once every 3 days) induced local (macrophage activation in lung, liver, and spleen) and systemic inflammation (Increase in serum IL-8, TNF-α, and IL-6) in BALB/c mice. In addition, SiNPs were also detected in the lysosomes of macrophages in lung, liver, and heart tissues (Yang et al. 2016).

Dermal exposure

Repeated exposure

The skin of SD rats was repeatedly exposed to different doses (500, 1000 mg, or 2000 mg/kg bw) of l-arginine coated C-SiNPs (20 nm) for 90 days (6-h exposure/day). The repeated exposure neither induced gross changes in the skin nor in any organs. In addition, hematological and blood biochemical parameters did not change in SiNPs-treated mice compared to controls (Ryu et al. 2014). In another study, Shim et al. (2014) showed that C-SINPS (20 and 100 nm) did not induce toxicity in the right and left brain or distribution in the cerebellum, hippocampus, or striatum of the dermally exposed rats (1000 or 2000 mg/kgbw daily for 90 days). In contrast to these studies, a significant increase in apoptosis (TUNEL positive cells) was observed in the skin of the BALB/c mice topically exposed to 70-nm SiNPs (250 mg/ear/day for 28 days). The author reported that the SiNPs were found not only in the skin but also in the regional lymph nodes, cerebral cortex, hippocampus, and in the liver (Nabeshi et al. 2011b).

Parenteral exposure

Single exposure

In a 48-h study, exposure to 15-nm SiNPs (dose of 50 mg/kg bw by intravenous injection) resulted in a significant increase of CD68-positive Kupffer cells (KCs), WBCs, lymphocytes, monocytes, neutrophils, and TNF-α in the serum of male SD rats. There was a decrease in GSH activity and elevation in MDA levels in the liver of treated mice. Furthermore, bio-markers of liver dysfunction such as lactate, phosphorylcholine, sn-glycero-3-phosphocholine, tyrosine, phenylalanine, and lysine were increased in blood, while the levels of succinate, glucose, and glycine were significantly decreased in mice exposed to SiNPs (Chen et al. 2013).

In another study, 50 nm C-SiNPs (0.25 mg/kg bw) induced a significant increase in blood parameters such as leukocyte number, creatinine kinase (CK), ALT, AST, and LDH release in male tuck ordinary mice at 24-h post-intraperitoneal injection. Oxidative stress bio-markers such as MDA formation, SOD, and catalase were significantly increased in the lungs, liver, kidney, and brain of SiNPs-treated mice. Cytokines such as IL-6, IL-1β, and TNF-α were significantly increased in the lung and differentially expressed in other organs. Furthermore, all organs showed a significant DNA damage compared to saline treated mice and the extent of damage was in the order heart > kidney > lung > liver > brain (Nemmar et al. 2016).

In the study of Duan et al. (2014a, b), ICR mice were intravenously injected with 29.5, 103.5, or 177.5 mg/kg bw of 62 nm S-SiNPs. After 14 days, induction of autophagic vacuoles and mitochondrial rupturing in the heart tissues was observed using TEM. In addition, increase in LC3 positive staining in the heart tissues was detected at 103.5 and 177.5 mg/kg bw, and the expression of ICAM-1 and VCAM-1 was significantly decreased at 177.5 mg/kg. Huang et al. (2011) injected the mice intravenously with mesoporous rods (pure or PEGylated) at a dose of 20 mg/kg bw and found a dysfunction in biliary excretion and glomerular filtration. Blood analysis showed that there was a significant increase in total bilirubin (TBIL), blood urea nitrogen (BUN), and creatinine (CREA). The mesoporous rods mainly distributed in the lung, liver, and spleen, but PEGylation significantly reduced their distribution in these organs. In addition, short rods cleared via urine and feces more rapidly than long rods.

Yu et al. (2013) estimated the i.v. median lethal dose (LD50) for 64 nm C-SiNPs as 262 mg/kg bw in ICR mice using the Dixon’s Up and Down method. After 14-day post-administration, intravenously injected C-SiNPs (64 nm, 177.5 mg/kg bw) induced a significant increase in liver injury bio-markers (LDH, AST, and ALT). Histopathological analysis of target organs showed that the number of megakaryocytes in the spleen was significantly increased, and pulmonary hyperemia and interstitial thickening were observed in the lungs. Nearly 35% of the injected SiNPs were distributed in the spleen, 12.5% in the liver, and 2% in the lung of SNP-treated mice.

At 24 h of post-intravenous administration (30 mg/kg bw), TEM imaging of the BALB/c mice tissues showed that 70-nm SiNPs were localized in the regional lymph nodes, cerebral cortex, hippocampus, and the cytoplasm, and nucleus of hepatocytes (liver). Using in vivo imaging, the authors also found that the SiNPs were distributed near the liver immediately after the administration (20 min) and moved near the intestinal tract over time (6 h). Micron-sized silica particles (300 and 1000 nm) used in these studies were accumulated mainly around the gall bladder (Nabeshi et al. 2011b).

A significant increase in the levels of kidney injury bio-markers such as BUN and CREA and morphological changes associated with renal interstitial fibrosis was noticed in BALB/c mice intraperitoneally treated with 198-nm M-SiNPs (150, 300, or 600 mg/kg bw for 2 or 12 days). There was also a significant increase in the levels of fibronectin (FN), TGF-β and ICAM-1, and nuclear translocation of p65 in 300- or 600-mg/kg treated mice (Chen et al. 2015).

Zhuravskii et al. (2016) demonstrated that intravenously injected (70 mg/kg bw) Py-SiNPs (13 ± 5 nm) distributed and persisted in the liver of Wistar rats at 60-day post-administration. The authors also indicated that the administered SiNPs induced fibrosis and liver tissue remodeling by noticing the increase in blood ALT, and presence of mast cells, connective tissues, and foreign body-type granulomas in the liver.

Conclusion: in vivo

The in vivo toxicity studies have been carried out using rats and mice, and exposure through various routes of administration. In general, short-term exposure to SiNPs induced adverse effects in the lungs, kidneys, liver, and brain. SiNPs administered mainly via inhalation, ingestion, and intravenous routes were majorly distributed in the liver, lungs, spleen, and kidneys, and in the brain of the intranasally exposed rats. Most of the administered SiNPs were excreted via feces and to a lesser extent via urine, in a size- and shape-dependent manner. However, in most cases, the administered doses were very high compared to relevant human inhalation (Barsan 2007) and ingestion (Winkler et al. 2016; Dekkers et al. 2011) exposures to amorphous silica. Some studies showed accumulation of SiNPs in organs such as the liver, but such accumulation was not associated with any major effects. However, long-term effects of accumulated SiNPs were not studied. It is surprising that in contrast to acute studies, no toxicity (local or systemic) was observed in chronic oral and dermal exposure studies, regardless of the size of SiNPs and the high doses used. Moreover, surface-modified SiNPs showed a significant increase in absorption by the GI tract compared to bare SiNPs. Dosing SiNPs via i.v. showed in all studies some sign of damage/toxicity. No clear type-or size-dependent effect can be identified from the set of reviewed studies. Furthermore, where genotoxicity was studied, it was not clear whether SiNPs were taken up by the target cells or were able to reach the tissue examined, which is obviously needed to exert a direct genotoxic effect.

Discussion

Growing production and use of SiNPs increase the risk of human exposures. Available toxicity studies mainly focused on effects after exposure via inhalation or skin (occupational exposure), ingestion (food additive), or nano-therapeutics (parenteral exposure). In vitro and in vivo studies demonstrated that SiNPs can induce adverse effects, but multiple ‘inconsistencies’ were found in the collected toxicity data set. As small differences in the physico-chemical properties of SiNPs could contribute to significant variation in the toxicity (Napierska et al. 2010), it is critical to discriminate how these variations influence toxicity. It is also worthy to note that the choice of cell type, culture system, assay conditions (Fede et al. 2012; Geys et al. 2010; Hayashi et al. 2017), and exposure route influence the toxic responses to SiNPs. However, the nature of these variations remains unclear, which continues to limit our understanding of SiNPs toxicity and hampers their hazard assessment.

Physico-chemical characterization

In this review, a basic set of physico-chemical characteristics of primary SiNPs such as primary size, shape, crystallinity, and chemical composition (or) purity was set as inclusion criteria. Most studies associated the toxic endpoints to SiNPs size, while only a few related the toxicity to SiNPs porosity, shape, surface charge, and surface chemistry. Furthermore, hydrodynamic diameter is another widely reported property in these studies and the results suggest that, in most cases, SiNPs are often aggregated/agglomerated (AA) in cell culture media. Conversely, AA formation is also very likely in real-world matrices such as in air (Kim et al. 2014a), water, and in commercial products such as food (Dekkers et al. 2011). Such AA formation not only substantially alters the overall characteristics (such as size, shape and surface topology) but also potentially influences the biological outcomes (Luyts et al. 2013; Drescher et al. 2011). Although some progress has been made in the characterization of SiNPs AA in recent years (De Temmerman et al. 2012), their biological effects remain poorly understood.

Toxicity of different types of SiNPs

Synthetic amorphous SiNPs are produced via different methods such as thermal (pyrogenic) or wet route (colloidal, precipitated and gel). The physico-chemical characteristics of SiNPs produced by these methods differ (Fruijtier-Pölloth 2012; Napierska et al. 2010) and may influence the biological outcomes. In this review, nearly 70% of the papers clearly reported the synthesis method. Among those 70%, about 80% reported the toxicity of wet method based SiNPs and only approximately 20% on the other types (such as pyrogenic and mesoporous). A general overview of toxic effects induced by different types of SiNPs is presented in Table 3. Cytotoxicity and genotoxicity induced by S- and C-SiNPs are strongly correlated with the induction of oxidative stress. For Pr-SiNPs, oxidative stress was associated with cytotoxicity but not genotoxicity. Interestingly, Py-SiNPs caused cytotoxicity, mostly without the generation of oxidative stress. In addition, recent studies showed that Py-SiNPs are biologically more reactive than C-SiNPs (Zhang et al. 2012) and Pr-SiNPs (Di Cristo et al. 2016) of same composition and size. It is known that C-, S-, and Pr-SiNPs are hydrophilic in nature, while Py-SiNPs are hydrophobic due to the de-hydroxylation of surface OH groups during the production process (Napierska et al. 2010), and such difference in surface chemistries might contribute to different biological activities. However, more systematic studies are required to verify these differences depending on production processes.

Toxicity mechanisms of amorphous SiNPs and crystalline silica

A fundamental question was raised by Napierska et al. (2010): do amorphous SiNPs induce biological responses similar to crystalline silica? Oxidative DNA and membrane damage have been reported as the major toxic mechanisms involved in the health effects of micron-sized crystalline silica, which is also observed for C- and Pr-SiNPs. The latter two possess silanols on the surface while Py-SiNPs mostly contain siloxanes (Napierska et al. 2010). In addition, recent studies showed that the presence of surface moieties such as silanols is more correlated with crystalline silica toxicity than crystallinity (Zhang et al. 2012; Turci et al. 2016). Thus, silanols appear to be a common surface feature in C-SiNPs, Pr-SiNPs, and crystalline silica, which might contribute to a similar toxic activity. Furthermore, toxicity elicited by Py-SiNPs appears to be oxidative stress-independent, indicating that Py-SiNPs induce adverse effects via other mechanisms (Gehrke et al. 2013; Napierska et al. 2012a). More studies are required to verify these hypotheses.

Influence of exposure routes in vivo

The exposure route obviously influences the in vivo absorption, biodistribution, and toxicity of SiNPs. For instance, after oropharyngeal aspiration or intratracheal instillation, the lungs are clearly the main target (Smulders et al. 2014), while after ingestion (Zande et al. 2014) or iv injection (Yu et al. 2013), the liver and/or the spleen were the targets. SiNPs accumulated in the brain of intranasally exposed rats (Parveen et al. 2015) but did not after oral exposure (Shim et al. 2014). In addition to the exposure route, physico-chemical properties such as size and shape clearly influence the clearance, distribution, and toxicity (Huang et al. 2011; Li et al. 2015).

Adverse effects of chronic exposure in vivo

In repeated (oral and dermal) dose studies (≥28 days), SiNPs did not induce any local or systemic toxicity even in (very) highly dosed rats. A recent study demonstrated that mild and highly dosed rats excreted most of the orally administered SiNPs via the feces (Yun et al. 2015). In addition, Van der Zande et al. (2014) demonstrated the in vitro gelation of SiNPs with increasing concentrations, which might reduce the gastro-intestinal absorption in highly dosed animals and increase excretion via feces. Therefore, the use of (low) realistic exposure doses appears more appropriate, since the toxicokinetics may depend on the level of exposure (Paek et al. 2014). Moreover, information on the physico-chemical properties of ingested and digested SiNPs is lacking, representing a huge knowledge gap in the risk assessment of SiNPs in food (Dekkers et al. 2012).

Correlation between in vitro and in vivo studies

Correlation between in vitro and in vivo effects is an indication that specific cells or tissues are potential targets for SiNPs toxicity. The results of in vitro and in vivo experiments (included within the same study) suggest that exposure to SiNPs could induce Kupffer cell mediated liver injury (Chen et al. 2013), kidney injury via the activation of NF-kB signaling pathways (Chen et al. 2015), and endothelial dysfunction via autophagy (Duan et al. 2014a). Several in vitro studies showed that SiNPs caused DNA double-strand breaks in a wide range of immortalized cell lines at low doses, but no such genotoxic effects were observed, even in animals at high doses. In addition, SiNPs did not induce micronuclei either in vitro or in vivo.

Dosimetry in vitro

Accurate in vitro dosimetry is an important yet complex aspect of nanotoxicology (Lison et al. 2014). In many studies reviewed here, authors often tested high in vitro exposure doses. Considering the exposure dose as nominal dose may be appropriate only for well dispersed and stable SiNP suspensions (Lison et al. 2008). However, Py-SiNPs, for instance, often re-agglomerate in culture media with an effective density lower than the material density (Deloid et al. 2014), thereby potentially affecting the dose reaching the cells (Cohen et al. 2014). Thus, a realistic estimation of the delivered dose is necessary to compare the biological effects between studies and to establish good in vitro and in vivo correlations (Pal et al. 2015).

Physico-chemical properties for the safer design of SiNPs

Immune responses after the administration of SiNPs are very crucial, as they can induce cascades of events by the secretion of cytokines, which may be harmful or beneficial. The data collected in this review show that the smaller the size, the stronger the pro-inflammatory effect. In addition to the size, the surface charge appears to play a role, since less negative charges seem to suppress the immune response. Porosity is another crucial factor influencing blood biocompatibility, i.e., the more porous the SiNPs, the less hemolysis of RBCs (Maurer-jones et al. 2010). Furthermore, SiNPs aspect ratio was shown to closely relate to in vivo organ retention and clearance (Huang et al. 2011). Therefore, the size, surface area, porosity, and geometry (shape) appear to be the key parameters for designing less-toxic and less-inflammagenic SiNPs, for bio-medical applications for instance.

In conclusion, SiNPs generally exhibit acute toxic effects in vitro and in vivo. The data on chronic effects of SiNPs exposure are rather conflicting with the acute effects and are still insufficient to draw firm conclusions. No concrete data were found to conclude whether amorphous SiNPs induce fibrosis like micrometric crystalline silica. Moreover, translation to human health effects is impossible at this moment due to the lack of realistic exposure and epidemiological data. Surface moieties (such as silanols, silanolates, and siloxanes) of SiNPs are found to be production process-specific and seem to be the key determinants of SiNPs toxicity. AA formation for some SiNPs is very dynamic in physiological media, but it is still unclear how it contributes to the hazard characterization. Therefore, the design of safe(r) SiNPs for food, medical, and other applications will only be possible when physico-chemical characteristics can be unambiguously linked to toxicity. Furthermore, detailed investigations on the SNP bioaccumulation/bioavailability and their long-term consequences in vivo are required for a safer use.