Engineered nanomaterials in the environment: Are they safe?

Abstract

Engineered nanomaterials (NMs) are increasingly released into the environment with the rapid development of nanotechnology. In this review, we present the current understanding on the safety of NMs in the environment from the existing literature. For the four typical NMs selected (i.e., TiO₂, Ag, CuO, graphene), available data on both environmental concentrations and ecotoxicity (e.g., predicted no effect concentrations [PNECs]) were collected from the published literature. All the four typical NMs did not pose a high risk by comparing environmental concentrations and PNECs. Toxicity of NMs as affected by environmental factors was further discussed. It can be concluded that nanotoxicity could be further lowered under environmental conditions for most scenarios. Generally, the environmental risk of NMs is of low concern currently, while the risk in specific scenarios such as surface water and soil near the point sources (e.g., wastewater effluent, sewage sludge) should be paid more attention. Moreover, the NM-induced toxicity could be changed by various environmental factors such as sunlight irradiation, natural organic matter, and mineral particles. In view of the uncertainties on environmental concentrations and ecotoxicity, the challenges on comprehensive understanding on the environmental risks of NMs, and the future research perspectives were addressed and presented.

Graphical abstract

Keywords:

• Environmental risk
• environmentally relevant concentration
• nanotoxicity
• natural organic matter
• sp-ICP-MS
• transformation

1. Introduction

As nanotechnology rapidly develops, the categories and numbers of nano-enabled products are drastically increasing in the last two decades. Based on the data in “The Nanodatabase” in 2019, there are 3111 nano-enabled products, which could be divided into seven categories including “Health and Fitness,” “Home and Garden,” “Automotive,” “Food and Beverage,” “Electronics and Computers,” “Appliances,” and “Goods For Children” (Figure 1a) (The Nanodatabase, 2019). “Nanotechnology Products Database” includes 8902 nano-enabled products in 2019, and the top eight categories were “Cosmetics,” “Textile,” “Construction,” “Medicine,” “Automotive,” “Environment,” “Renewable Energies,” and “Petroleum” (Statnano, 2019). Among these categories, Ag, Ti, and C-based nanomaterials (NMs) are the most commonly-used NMs for these products (Figure 1b). As a result, NMs are releasing into air, water, and soil during the production, consumption, and disposal of nano-enabled products (Giese et al., 2018; Sun et al., 2016). It is of importance to assess the environmental risk of NMs, and possible threats to the health of ecosystems.

Figure 1. Nano-enabled product categories in 2019. (a) Main application fields of nano-enabled products; (b) number of nano-enabled products with different materials. Data from “The Nanodatabase” (December 1, 2019).

Concerns over the toxicity of NMs have been raised since the early 2000s. At the first decade, the toxicological researches mainly focused on (1) the toxicity of different NMs to both model and unique native organisms; (2) the acquisition of various toxicity parameters, such as no observable effect concentration (NOEC), the lowest observed effect concentration (LOEC), median effect concentration (EC₅₀), and lethal concentration (LC₅₀); (3) the effects of texture, size, surface properties on NM toxicity; and (4) the toxicity of NMs at organismal and cellular levels (Auffan et al., 2009; Mahmoudi et al., 2011; Moore, 2006; Nel et al., 2006). At the second decade, the following aspects are further addressed: (1) prediction and measurement of the concentrations of NMs in different environmental compartments; (2) toxicity of NMs as affected by environmental factors; and (3) mechanistic investigations of NM (especially the novel NMs) toxicity at molecular levels. To date, there are already numerous of works on toxicological effects of NMs. However, the understanding of the environmental risks of NMs still lags behind the release of new nano-enabled products and the developments of new applications (Giese et al., 2018). There is still no consensus on the ecotoxicity of NMs, and a number of data and knowledge gaps remain apparent. The environmental scientists attempted to answer the question: are the NMs safe to use? Two aspects: (1) the actual environmental concentrations, and (2) the toxicity of NMs in realistic environments, are essential to answer this question. But there are still many different views and opinions, including the validation of both the prediction and measurement of environmental concentrations in different environmental compartments, the accuracy on the toxicological investigation approaches; and the reliability for the toxicity of NMs in environments (Arvidsson, 2018; Musee, 2018). Therefore, this work will focus on the above two aspects to improve the understanding on the environmental risks of NMs.

In this review, four typical NMs including TiO₂ (high production and environmental concentrations), Ag (high production and toxicity), CuO (high toxicity), and graphene (rapidly increasing production) are mainly considered. For environmental concentrations, the predicted and measured concentrations of NMs in different environmental compartments will be discussed and summarized; For NM toxicity, the toxicological data will be mainly from the newly-released publications in the past five years because there are several important critical reviews that summarized the ecotoxicity of NMs by 2014 (Bour et al., 2015).The toxicity mechanism of NMs will not be discussed in detail in the present work because it has been recently well-reviewed (Buchman et al., 2019). Instead, the effects of environmental relevant conditions on NM toxicity will be emphasized.

2. Environmental concentrations of NMs

Environmental concentrations of NMs could be obtained by both prediction and measurement (Wimmer et al., 2019). This section summarizes the predicted environmental concentrations (PECs) and measured environmental concentrations (MECs) of NMs. PECs and MECs of four typical NMs in a same environmental compartment are compared.

2.1. Model prediction of NMs concentration in water, sediment, and soil

Mathematical modeling is currently one of the most important tools to obtain the environmental concentrations of NMs in different environmental compartments. PEC of NMs was first calculated a decade ago (Gottschalk et al., 2009; Mueller & Nowack, 2008). Afterwards, PECs of various types of NMs in different regions/countries were reported. In this review, we summarized reported PECs of TiO₂, Ag, CuO, and carbon based NMs (e.g., CNTs) in water, sediment, and soil (Table S1), and the PEC results are presented as a box and whiskers plot (Figure 2).

Figure 2. Box and Whisker plots for modeled environmental concentrations of TiO₂, Ag, CuO, and carbon based NMs in water (a), sediment (b), and soil (c). The diagrams were generated based on the modeled concentration values of NMs published in previous studies (Table S1).

For PECs in waters, TiO₂, Ag, CuO, and CNT NMs lie in the ranges at 0–16,000, 0–619, 0.02–6, and 2 × 10⁻⁵–1.82 ng/L, respectively (Figure 2a, Table S1). The ranges are quite wide, which are due to (1) the fitting data from different models; (2) the data from the surface waters in different regions/countries; and (3) different distribution properties of NMs between freshwater and seawater (Table S1). Generally, TiO₂ NMs had higher PECs than the other three NMs, probably due to their highest production volume and widespread applications in the world (Arvidsson et al., 2018; Holden et al., 2014). CNTs showed the lowest PECs in surface water, which could be explained by the lowest production volume among the four NMs (Arvidsson et al., 2018). The high surface hydrophobicity of CNTs could induce strong aggregation and sediments (Zhao et al., 2012), which may be another reason for the low PECs in surface water. In addition, all the PEC data are from European countries, US and South Africa, and there is no information from the surface waters in Asian countries (Figure 2a). Considering that the nano-product volumes in Asian countries (e.g., China) are rapidly increasing (The Nanodatabase, 2019), more efforts are needed to predict the environmental concentrations in the rivers and lakes in these countries. Moreover, the PEC information from seawaters is more limited than freshwater (Figure 2a). NMs in seawater are mainly transported from terrestrial and coastal areas (Matranga & Corsi, 2012). The PECs of NMs in seawater are generally lower than those in freshwater, which could be attributed to (1) high ionic strength-induced aggregation and sedimentation of NMs in seawater; and (2) high mass transport and diffusion of NMs from coastal seawater to the ocean.

In sediments, TiO₂, Ag, CuO, and CNT NMs are ranged at 0–186, 0–0.47, 3.5 × 10⁻³–2.1, and 1 × 10⁻⁴–2.66 × 10⁻²mg/kg, respectively (Figure 2b, Table S1). For all the four NMs, PECs in sediments are higher than those in surface waters. The reason is that NMs released into waters could quickly adsorb to the suspended minerals, finally depositing into the sediment environments (Wimmer et al., 2019). Actually, sediment is the compartment in which most NMs entering surface water end up (Sun et al., 2016). In addition, PECs in seawater sediments are lower than those in freshwater sediments, which is likely due to lower PECs in seawater than freshwater (Figure 2a).

In soils, PECs for all the NMs are 2–4 orders of magnitude lower than that in sediments (Figure 2c). TiO₂ NMs generally show the highest PECs in soils among the NMs, which is consistent with those in sediments and surface water. However, a high PEC for CuO NMs was detected in soil in Denmark, with the concentration at 0.039–0.13 mg/kg (Figure 2c). The high PECs of CuO NMs were explained by the wide usage of CuO and CuCO₃ NMs as wood preservatives which could be released into the soils (Gottschalk et al., 2015). It is noted that the available PEC data in soil are relatively fewer than the other two compartments. All the PECs in soils are obtained from Europe and US (Figure 2c).

For all the environmental compartments (surface water, sediment, and soil), there is no PEC information on graphene NMs. Further modeling investigations on graphene-based NMs including few-layer graphene, graphene oxide (GO), reduced GO are suggested. It is reported that the total input production volume of graphene is 6–7 times higher than CNTs (Arvidsson et al., 2018). Considering that the surface properties, applications and environmental fates of graphene are similar to CNTs, PECs of graphene NMs are speculated to be at a similar concentration level with (or even higher than) CNTs.

Sun et al. (2016) predicted the NM concentrations in both waste streams and environmental compartments in the EU in 2014. It was found that the highest PECs (“mg/kg” level) were in wastewater treatment plant (WWTP) sludge, followed by solid waste, bottom ash of waste incineration plant (WIP), and WIP fly ash. For the soil compartments, WWTP sludge-treated soil had much higher PECs than untreated soil. For TiO₂ NMs, PECs of “WIP bottom ash,” “WIP fly ash,” and “sludge treated soil” were higher than that those in sediments. For Ag NMs, PECs of “solid waste to landfill,” “WIP bottom ash,” “WIP fly ash” were higher than that in sediments. In another report, for all the test NMs (TiO₂, Ag and CNTs), PECs in WWTP effluent were higher than that in surface water (Gottschalk et al., 2010). The same finding was also observed in another case (Gottschalk et al., 2009). PECs in WWTP sludge were very high, with the concentrations at 211 mg/kg for TiO₂, 0.069 mg/kg for CNT, and 1.88 mg/kg for Ag NMs (Gottschalk et al., 2010), which are higher than those in sediments as shown in Figure 2b. TiO₂ NMs showed higher PECs in WWTP effluent and WWTP sludge than Ag and CNT NMs (Gottschalk et al., 2010), and this is consistent with the order of NMs in environmental compartments (surface water, soil, and sediments) (Figure 2). These findings suggest that more attention should be paid to the potential environmental risk of NMs in WWTP effluents and WWTP sludges, as well as the surface waters near the discharge of WWTP effluents and the soils after sludge treatment.

2.2. Measurements of NMs concentration in surface water, sediment, and soil

The released NMs in the environment were first detected by transmission electron microscopy/energy dispersive X-ray analysis (TEM/EDX) and inductively coupled plasma-mass spectrometry (ICP-MS) (Kägi et al., 2008). Currently, electrothermal atomic absorption spectrometry (ET-AAS) (Li et al., 2016), asymmetric-flow field flow fractionation with inductively coupled plasma–mass spectrometry (AF4-ICP-MS) (Hoque et al., 2012), and especially, single particle inductively coupled plasma-mass spectrometry (sp-ICP-MS) (Loula et al., 2019; Peters et al., 2018; Wu et al., 2020; Xiao et al., 2019) have been developed for characterization and quantification of NMs in natural environments. Figure 3 summarized the MECs of NMs in surface water, and sediment based on the published data (Table S2).

Figure 3. Box and Whisker plots for measured concentrations of nanoparticulate Ti, Ag, and Cu in surface waters (a) and sediments (b). The diagrams were generated based on the measured concentration values of NMs published in previous studies (Table S2).

In surface waters, MECs of Ti-, Ag-, and Cu-based NMs are ranged at 64–10,200, 0.3–3200, and 15–136 ng/L (tap water), respectively (Figure 3a). Ti-based NMs had the highest MECs among the three NMs, consistent with the PECs (Figure 2a). MECs of Ti- and Ag-based NMs are at the upper level of their PEC ranges. It is noted that there is only one report having the measured mass concentrations of Cu-based NMs in water (tap water), and the obtained MEC exceeded the maximum value of the PEC range (Venkatesan et al., 2018). The collected MEC data in surface water are fewer than PECs for all the NMs. The waters were sampled from rivers, lakes, swimming pool, and tap water (Table S2). There is only one report focusing on the MECs of NMs in seawater, and nanoparticulate Ti and Ag concentrations in seawater were 0.65 × 10⁷ and 0.54 × 10⁷ particles/mL (Xu et al., 2020). In addition, the MECs are not available for CNT, graphene or other carbon NMs. This is because (1) it is hard to distinguish specific carbon-based NMs from natural carbonaceous materials, and (2) there is no detection technique that could satisfy the accuracy and detection limits of carbon-based NMs at environmentally relevant concentrations (ERCs) (Goodwin et al., 2018). In addition, the MECs data are from European countries, North America, and China (Table S2), while no information is found from African countries and Australia (Figure 3a).

Few data for MECs in sediments were reported (Figure 3b). Ti- and Ag-based NMs were at 24.7, and 0.0033–0.2 mg/kg, respectively (Metcalfe et al., 2018; Xiao et al., 2019). MECs of Ti- and Ag-based NMs are in the range of PECs, but tend to be at the upper level of the ranges, which could be attributed to the strong backgrounds of Ti and Ag, and the presence of naturally occurring nanoparticles (Garner et al., 2017). To date, there are very few data on the MECs in soils due to the limitations on NM extraction and separation methods: (1) the extraction process and combined separation techniques (e.g., filtration, centrifugation) could change the physicochemical properties (e.g., dissolution, aggregation) of NMs; (2) the interactions of NMs with soil particles limit the tracking and detecting of low-concentration NMs; and (3) soil organic matter, particularly at high content, could highly impact the feasibility of the NM extraction approach for sp-ICP-MS analysis (Li et al., 2019; Mahdi et al., 2017). Mahdi et al. (2017) extracted the spiking Ag NMs from soils by using the combined methods of prewetting and aqueous extraction, and the recovery of Ag NMs (60 nm) was only 32% for sandy soil and 46% for clayey soil. Li et al. (2019) optimized the extraction approach, and analyzed Ag NMs in soils (Changsha, China) using sp-ICP-MS. It is found that the recovery for Ag NMs was 59.6%–87.4%, and the MECs of Ag NMs (>30 nm) were in the range of (1–3) ×10⁷ particles/g soil.

Li et al. (2016) examined the MECs of Ag-based NMs in wastewater influent and effluent from a WWTP in Germany, in which 10.1–357, and 0.7–11.1 ng/L of Ag-based NMs were detected, demonstrating that most of Ag-based NMs could be removed in the WWTP. However, the effluent of Ag-based NMs still resulted in a high concentration range (2.0–8.6 ng/L) in the river near the WWTP, which was higher than the nearby uncontaminated surface waters (Li et al., 2016). In another report, Ag-based NMs from the water samples near WWTP were determined to be 2.47–69.08 ng/L, much higher than the level of the river water (1–2 ng/L) (Wimmer et al., 2019). The above findings suggested that WWTPs could act as point sources for the NMs in surface water and sediments. Tou et al. (2017) sampled the sewage sludge from Shanghai (China) for sp-ICP-MS analysis, and found that the metal-based NMs (Ti, Fe, Zn, Sn, and Pb) were at the range of 10⁷–10¹¹ particles/g, and Ti-based NMs were at 1.51 × 10⁹ particles/g. Although there is no available information on the concentrations of NMs in sludge-treated soils, the MECs are speculated to be much higher than the pristine soils.

3. Environmental risk of NMs

3.1. Risk assessment of NMs in different environmental compartments

To better understand the environmental risk, a classic risk assessment by comparing PECs with predicted no effect concentrations (PNECs) was commonly used (Arvidsson, 2018; Coll et al., 2016). We thus collected NOEC or LOEC values of TiO₂, Ag, CuO and graphene NMs from the latest ecotoxicological literature (2014–2019) (Tables S3–S6), and obtained the PNECs from the species sensitivity distribution (SSD). Based on the SSD analysis, the calculated PNECs for TiO₂, Ag, CuO, and graphene-based NMs in aquatic environments were 1.92 × 10⁻¹, 4.85 × 10⁻⁵, 9.06 × 10⁻³, and 1.15 × 10⁻²mg/L, respectively (Figure 4), clearly suggesting the lowest toxicity for TiO₂ NMs, and the highest toxicity for Ag NMs in aquatic environments. Coll et al. (2016) reported the PNECs from the most sensitive to least sensitive in the order of Ag (1.7 × 10⁻⁵mg/L) ≪ TiO₂ (15.7 × 10⁻³mg/L) ≪ (CNT 55.6 × 10⁻³mg/L), and Chen et al. (2018) reported the PNECs by following an order: Ag (0.0036 mg/L) < CuO (0.049 mg/L) < TiO₂ (3.1 mg/L), which are consistent with this work (Figure 4). The reported PNECs of these four NMs in WWTP effluent, sediment and soil compartments were also collected from the published literature (Table S7), and these PNECs were in a wide range for each NMs, indicating the high uncertainties during ecotoxicological investigations. In addition, most toxicological works focused on PNECs in aqueous and soil environments. The PNEC information in WWTP effluent and sediment environments is scarce.

Figure 4. Predicted no effect concentrations (PNECs) of TiO₂ (a), Ag (b), CuO (c), and graphene (d) NMs in aquatic environments. Based on probabilistic species sensitivity distributions (PSSD) analysis, the calculated PNECs for TiO₂, Ag, CuO, and graphene NMs are 1.92 (0.32–6.41) × 10⁻¹, 4.85 (0.13–53.87) × 10⁻⁵, 9.06 (0.39–65.40) × 10⁻³, 1.15 (0.03–10.6) × 10⁻²mg/L, respectively.

Risk characterization ratios (RCRs), i.e., the ratio of PEC to PNEC, are shown in Table S8. For most scenarios in surface water, RCR values were lower than 1, indicating low risk in aqueous environments. Only one report observed that the RCRs of Ag NMs in the surface waters of Europe were a bit higher than 1 (i.e., 1.1, 1.03) (Gottschalk et al., 2009). For soil compartments, RCRs for all the four NMs were much lower than 1, demonstrating that NMs would pose very low or no risk in soils. Only a few works focused on the environmental risk in sediments, in which the RCRs were all much lower than 1 (Table S8). Considering that RCR between 0.1 and 1 could be ranked as “medium risk,” Table S8 shows that most of RCRs were lower than 0.01, suggesting low risk in these environmental compartments. However, NMs in WWTP effluents may pose significant risk. For example, in some scenarios, the RCR values were higher than 1 in wastewater (effluent) for TiO₂, Ag, and CuO NMs (Gottschalk et al., 2009; Kjølholt et al., 2015).

In addition to assessing the environmental risks of NMs using RCRs, other approaches were also employed. For example, besides NOEC at HC₅ (5% of species will be harmed), LOEC, NOEC, LC₅₀, EC₅₀, and NOEC for single species are also able to be used for the risk assessment. Garner et al. (2017) reported that under all release scenarios in fresh water, CuO NMs did not exceed the NOEC HC₅, and TiO₂ occasionally exceeded the NOEC HC₅ in all scenarios (especially for the highest release scenario). In agricultural and urban soils, neither TiO₂ nor CuO would exhibit any toxicity even under the most extreme release scenarios (Garner et al., 2017). Moreover, Arvidsson et al. (2018) assessed seven NMs using the two proxy measures: production volume and short-term aquatic ecotoxicity. It was found that Ag NMs ended up in the high ecotoxicity-low production volume quadrant; graphene, TiO₂ and SiO₂ NMs ended up in the low ecotoxicity-high production volume quadrant; CeO₂, ZnO, and CNT ended up in the low ecotoxicity-low production volume. There are no NMs (mentioned above) located in the high ecotoxicity-low production volume quadrant, suggesting that none of NMs are of high environmental concern currently.

Sewage sludges also contained NMs, with much higher concentrations than those in the environmental compartments (Tou et al., 2017). The application of sewage sludges in agricultural soils could pose NM-related risk. Schlich et al. (2013) assessed the risk of Ag NMs in sewage sludge-treated soil, and found that the PNEC (0.05 mg/kg soil) would be exceeded by applying Ag NM-containing sewage sludge in soil for 50 years. In another study, long-term (20 years) field applications of sewage sludge to soil resulted in an increase of Ag content to 51 mg/kg. Ag contents in the vegetable crops (red beet, carrot, and sugar beet) were generally lower than 0.7 mg/kg after soil-plant transfer when grown in sludge-applied soil, much lower than the safety threshold for the use by animals and humans, suggesting that environmental risk of Ag transferred to food chain is low (Wang et al., 2018). Although low risk to food chain, Ag (51 mg/kg) in the sewage sludge-treated soils may still have adverse effect to soil organisms (especially soil bacteria) according to the PNEC value of Ag NMs (0.05 mg/kg soil) (Schlich et al., 2013). No report on the risk assessment of other NMs was found in sewage sludge-treated soils.

3.2. Toxicity of NMs at ERCs

For risk assessment of NMs, both PECs and PNECs are modeled/calculated based on many assumptions which often lack of direct evidence for the toxicological information at ERCs. Currently, the research on the toxicity of NMs is still increasing, with more than 30,000 articles in the past two decades (Figure 5a). However, we only found around 30 articles claiming that their toxicological investigations toward TiO₂, Ag, CuO, and graphene-based NMs are at ERCs (Table 1). The following discussion in this section is based on these toxicological investigations, and the ERCs used for nanotoxicological studies are shown in Figure 5b. The ERCs of TiO₂, Ag, CuO, and graphene-based NMs used for toxicological investigations (Figure 5b) are within or slightly higher the ranges of PECs and MECs (Figures 2 and 3). Most of these investigations focused on aquatic environments, followed by wastewater, soils, and sediments. All the four NMs were investigated at ERCs. For graphene-based NMs, only one report was found, in which GO at 1–100 μg/L was exposed to zebrafish embryos (Zhang et al., 2017).

Figure 5. The number of published papers on the toxicity of NMs from 2000 to 2019 (a), and current toxicological investigations that considered environmentally relevant concentrations of NMs (b). Data in panel (a) were collected from “Web of Science.” The environmental concentration data in panel (b) were collected from all the available publications (Table 1).

Table 1. Current investigations on the impact of nanomaterials (NMs) at environmentally relevant concentrations.

NMs	Environmental compartments	Environmentally relevant concentrations	Organisms	Toxicity and mechanisms	References
TiO2	Water	700 µg/L	Microbial communities	TiO2 NMs at 700 μg/L had no significant effect on the structure and distribution of the microbial communities.	Londono et al. (2017)
	Water	Direct exposure: 7–120 μg/L; exposure with food: 4–830 μg/L	Zebra mussels (Dreissena polymorpha)	The bioaccumulation of NMs in mussels was observed after exposure for 1 h (Toxicity was not investigated).	Bourgeault et al. (2015)
	Water	50 μg/L	Juvenile rainbow trout (Oncorhynchus mykiss)	TiO2 NMs did not lead to significant accumulation of Ti in internal tissues; no obvious toxicity was observed.	Boyle et al. (2013)
	Water	0.0005–50 μg/L	Nematode (Caenorhabditis elegans)	TiO2 NMs at 0.0005–10 μg/L did not obviously affect the survival of nematodes. In contrast, TiO2 NMs at 50 μg/L increased the mortality of nematodes. TiO2 NMs at 0.05–50 μg/L significantly decreased both body bends and head thrashes of nematodes, and significantly induced the ROS production.	Wu et al. (2013)
	Water	0.001 and 0.01 mg/L	Nematode (Caenorhabditis elegans)	Big size (60, 90nm) of TiO2 NMs at 0.001 and 0.01 mg/L did not obviously influence the body bends and head thrashes of nematodes. In contrast, small size (4 and 10 nm) of TiO2 NMs significantly decreased the body bends and head thrashes of nematodes.	Li et al. (2012)
	Water	1 μg/L	Wild nematodes	TiO2 NMs could induce the significant increase in intestinal ROS production, and decrease in locomotion behavior.	Dong et al. (2018)
	Water	25 and 250 μg/L	Mesocosms (phytoplankton, zooplankton, macroinvertebrates, macrophytes and fish)	TiO2 NMs caused a reduction in available soluble reactive phosphorus in the mesocosms by 15 and 23%, but not in the amount of total phosphorus. The biomass of Cladocera, Copepoda, phytoplankton, macrophytes, chironomids and fish was unaffected. The biomass of Rotifera was significantly reduced by 32 and 57% after exposure to TiO2 at 25 μg/L and 250 μg/L, respectively.	Jovanović et al. (2016)
	Water	0.0001, 0.001, 0.01, 0.1, and 1 μg/L	Nematode (Caenorhabditis elegans)	The significant decrease in locomotion behavior as indicated by head thrash and body bend was detected in sod-2, sod-3, mtl-2, and hsp-16.48 mutants.	Wu et al. (2014)
	Water	TiO2 at 0.01 mg/L (pre-mixing with algae)	Daphnia magna	TiO2 NMs induced low mortality, but caused a decrease in growth and reproduction.	Fouqueray et al. (2012)
	Seawater	0.1, 1, and 10 mg/L	Blood clam (Tegillarca granosa)	No mortality was observed. TiO2 NMs were neurotoxic to the blood clam as indicted by increased neurotransmitter concentrations, decreased AChE activity, and the down-regulated expression of neurotransmitter-related genes.	Guan et al. (2018)
	Seawater	1 mg/L	Marine scallop (Azumapecten farreri)	No mortality of scallop was observed during exposure. TiO2 NMs caused oxidative damage on the scallops and neurotoxicity.	Xia et al. (2017)
	Wastewater effluent	0.05 mg/L	Periphyton	No detectable increase on the concentration of Ti in periphyton and no detectable effects on algal cell density, chlorophyll-a, or periphyton mass was observed.	Wright et al. (2018)
	Wastewater in reactor	1 mg/L	N-related microbial community	TiO2 NMs lead to the decrease in nitrogen removal efficiency, anammox rate, and relative abundance of anaerobic ammonia-oxidizing bacteria (AAOB).	Zhang et al. (2018)
	Sediment	2.5, 25, and 250 mg/kg	Freshwater midge (Chironomus tentans)	The mortality and emergence ratio were not affected by TiO2 NMs. The morphological changes (e.g., mentum teeth, mandibles) in larvae were observed.	Savić-Zdravković et al. (2018)
	Soil	1 mg/kg	Microbial communities	TiO2 NMs had no significant effect on C-mineralization potential activity or bacterial abundance.	Simonin et al. (2015)
	Agricultural soils	0.05–500 mg/kg dry-soil	Soil microbial communities related to nitrification	TiO2 NMs had negative impact on the nitrification process in soil.	Simonin et al. (2017)
Ag	Water (Lake Michigan Water)	5–40 µg/L	Bacterium (Escherichia coli)	The presence of Ag NMs induces a stress response in Escherichia coli, as indicated by a decrease in ATP production.	Wilke et al. (2016)
	Bacterial cultures	0.1, 0.5, and 1 mg/L	Bacterium (Pseudomonas putida)	Ag NMs inhibited the growth of the bacteria. Ag NMs caused reactive oxygen species (ROS) production, glutathione depletion and inactivation of the antioxidant enzymes in bacteria.	Khan et al. (2015)
	Seawater	10 μg/L	Mussels (Mytilus galloprovincialis)	Although there was tendency for the up- (CYP4y1 and Cathepsin) and downregulation (CAT, GST, Caspase, HSP70, and EF 1) of specific genes, none of the analyzed gene transcription profiles were significantly altered in the gills after 15-day exposure.	Bebianno et al. (2015)
	Wastewater in constructed wetlands	50, 200 μg/L	N-related microbial community	Ag NMs had negative impacts on the key enzymatic activities and the abundance of nitrogen-related genes, and finally affected the treatment performance of constructed wetlands.	Huang, Cao, et al. (2019)
	Treated wastewater	TiO2 700 μg/L, ZnO 70 μg/L, Ag 200 μg/L	River microbial community	The bacterial community structure was changed after NM exposure, but was independent with NM concentration.	Londono et al. (2019)
	Sediment	4 μg/kg	Microbial community	In highly impacted sediments, bacterial communities were resilient to Ag NM exposure; in less impacted sediments, bacterial communities were more sensitive to Ag NM exposure.	Welz et al. (2018)
	Soil	0.01–1 mg /kg soil	Soil bacterial community	Actino- and alpha-Proteobacteria were statistically unaffected after 1-year Ag NM exposure. The abundances of Acidobacteria, Bacteroidetes and beta-Proteobacteria were significantly diminished after exposure. Even at low concentrations, Ag NMs may cause negative effect for the nitrification, organic carbon transformation and chitin degradation.	Grün and Emmerling (2018)
Cu or CuO	Water	20 µg/L Cu NMs	Juvenile (Epinephelus coioides)	No mortality was observed. The growth parameters were significantly lower after exposure. Liver damage was caused as indicated from the increased apoptosis index.	Wang et al. (2014)
	Seawater	10 μg Cu/L of CuO	Mussels (Mytilus galloprovincialis)	No mortality was detected during the whole exposure period. CuO NMs induced oxidative stress in mussels by overwhelming gills antioxidant defense system exposure. Neurotoxic effects were only detected at the end of the exposure period	Gomes et al. (2011)
	Seawater	10 μg Cu/L of CuO	Marine invertebrates Worms (Hediste diversicolor) and bivalves (Scrobicularia plana)	Several biomarkers are activated by CuO NMs: GST, CAT and SOD in Scrobicularia plana; CAT and GST in Hediste diversicolor. No significant effects were shown for the markers of neurotoxicity (ChE) or oxidative damage (TBARS).	Buffet et al. (2011)
	Seawater	10 μg Cu/L of CuO	Endobenthic species (Scrobicularia plana and Hediste diversicolor)	No mortality was observed for the endobenthic species after exposure for 21 days. For Scrobicularia plana clams, the burrowing kinetics was significantly affected after 21-day exposure. In Hediste diversicolor, no significant difference was observed after 7 days, whereas the feeding and burrowing had significantly decreased after 14-day exposure.	Buffet et al. (2013)
	Wastewater in reactor	1 mg/L	N-related microbial community	CuO NMs decreased the nitrogen removal efficiency, anammox rate, and relative abundance of anaerobic ammonia-oxidizing bacteria (AAOB).	Zhang et al. (2018)
	Wastewater	1 mg/L	N-related microbial community	CuO NMs significantly altered the bacterial community composition and decreased the abundances of Anaerolineaceae, Acidobacteria, Aminicenantes, and Anaerolinea. Bacterial genera depleted by CuO NMs were related to carbohydrate and glycan biosynthesis and metabolism, and biosynthesis of other secondary metabolites.	Miao et al. (2019)
	Wastewater	1 mg/L	N-related microbial community	CuO NMs at 1 mg/L did not significantly influence the LDH release or ROS production of bacteria; EPS content and composition of the biofilm was not changed.	Miao et al. (2017)
GO	Water	1–100 μg/L	Zebrafish embryos	The development of zebrafish embryos exposed to GO at 1–100 μg/L was impaired due to DNA modification, protein carbonylation and oxidative stress.	Zhang et al. (2017)

In surface water, TiO₂ NMs at ERCs (700 μg/L) had no significant effect on the structure and distribution of microbial communities, and the eukaryotic communities were not changed either (Londono et al., 2017). In another report, TiO₂ NMs at 250 μg/L were introduced into the mesocosms, and the phytoplankton were unaffected (Jovanović et al., 2016). For other aquatic organisms such as crustaceans (Daphnia magna) (Fouqueray et al., 2012), juvenile rainbow trout (Oncorhynchus mykiss) (Boyle et al., 2013), marine scallop (Azumapecten farreri) (Xia et al., 2017), and blood clam (Tegillarca granosa) (Guan et al., 2018), no mortality was observed during TiO₂ NM exposure at ERCs. But other biological stresses such as oxidative damage, neurotoxicity, and/or decrease in growth and reproduction were observed during exposure. Ag NMs in aquatic phase caused the growth inhibition on and stress responses to bacteria (Khan et al., 2015; Wilke et al., 2016). Although Ag NMs at ERCs induced the up- and down-regulation of specific genes, the analyzed gene transcription profiles were unchanged in the gills of mussels (Mytilus galloprovincialis) after 15-day exposure (Bebianno et al., 2015). For GO, it is reported that the development of zebrafish embryos was impaired due to DNA modification, protein carbonylation and oxidative stress at its ERC (1–100 μg/L) (Zhang et al., 2017). In wastewater, the changes in bacterial community composition and abundances were observed after exposure to TiO₂, Ag, or CuO NMs (Huang, Cao, et al., 2019; Londono et al., 2019; Miao et al., 2019; Zhang et al., 2018). The treatment performance of constructed wetlands, and the nitrogen removal efficiency in the wastewater reactor were thus decreased. However, no significant effect on algal cell density, chlorophyll-a, or periphyton biomass was observed after TiO₂ NM exposure (Wright et al., 2018).

In soil compartment, TiO₂ NMs at ERCs exhibited insignificant effect on bacterial abundance or C-mineralization potential activity (Simonin et al., 2015), while showed negative impact on the nitrification process in soil (Simonin et al., 2017). After Ag NM exposure, actino- and alpha-Proteobacteria in soil were unaffected after 1-year exposure, but the abundances of Acidobacteria, Bacteroidetes and beta-Proteobacteria were significantly decreased after exposure. The nitrification, organic carbon transformation and chitin degradation in soil may be decreased by Ag NMs (Grün & Emmerling, 2018). It should be noted that the ERCs (“mg/kg” level) of NMs used in soil compartments were higher than their PECs or MECs. The actual negative effect of NMs on soil microbial communities may be not obvious. The microbial communities in sewage sludge-treated soil are likely easier to be affected by NMs, which should receive more attention. Currently, related information in sediment compartment is scarce. Savić-Zdravković et al. (2018) reported that TiO₂ NMs at ERCs did not affect the mortality or emergence ratio of freshwater midge (Chironomus tentans), but morphological changes (e.g., mentum teeth, mandibles) in larvae were observed. Welz et al. (2018) found that low content of Ag NMs (4 μg/kg) could modify the bacterial community structure in sediments. Environmental concentrations of NMs in sediments were much higher than soil, the effect of NMs at ERCs in sediments thus deserves further study.

4. Toxicity of NMs as affected by environmental factors

4.1. Nanotoxicity mechanisms

Nanotoxicity mechanism is a main emphasis for nanotoxicological investigations, which could be divided into three stages: external toxicity, interface interaction and intracellular response (Figure 6). At “external toxicity” stage, NMs could cause both direct and indirect toxicity to cells/organisms (Figure 6). The direct toxicity of NMs can be triggered at external surface. The physical contact could lead to membrane damage, and some of NMs are able to be taken up by cells during interface interaction (e.g., endocytosis) with cell membrane. Indirect toxicity is also important at the “external toxicity” stage, including shading effect, extracellular ROS generation, nutrient depletion and ionic toxicity (Figure 6). Shading effect by graphene-based NMs was reported to cause indirect toxicity to both algae and protozoan (Hu et al., 2015; Zhao et al., 2017). Extracellular ROS generation could be facilitated by the photo-active NMs (e.g., TiO₂). NMs with large surface area are able to adsorb nutrients and cause nutrient depletion to both algae and human cells (Creighton et al., 2013; Zhao et al., 2017). Ionic toxicity is originated from the dissolution of NMs (e.g., Ag, CuO), which is one of the most important toxicity mechanisms. After cellular internalization, the intracellular responses are activated, which are extensively investigated in the past two decades. At cellular levels, the internalized NMs could lead to oxidative stress, mitochondrial dysfunction, and DNA damage. Moreover, intracellular NMs could be further dissolved and/or biodegraded, which also contributes to the above responses and toxicity.

Figure 6. Schematic illustration of the mechanisms for NMs-induced toxicity and related processes as regulated by environmental factors. The four stages include environmental processes, external toxicity, interface interaction, and intracellular responses.

It should be noted that all these toxicity mechanisms (I–III, Figure 6) and the obtained toxicological data are highly influenced by environmental conditions. Environmental factors such as sunlight irradiation, natural organic matter (NOM), and minerals are able to change the toxicity of NMs through regulating the key processes of NMs. The key NM processes include surface transformation, heteroaggregation, adsorption and dissolution (Figure 6), which are closely related to the external toxicity processes of NMs. How are these environmental conditions specifically regulating the processes and toxicity mechanisms of NMs? What are the changes of nanotoxicity under natural conditions? These questions are further discussed in the following sections.

4.2. Sunlight irradiation

TiO₂, Ag, CuO and graphene NMs are all photo-active (Markiewicz et al., 2018; Mitrano et al., 2015). Among them, TiO₂ is the most prominent example for photoactive NMs. Due to high photocatalytic activity, TiO₂ NMs could result in photo-toxicity to bacteria (e.g., Escherichia coli and Aeromonas hydrophila), which was highly related to the production of ROS (e.g., OH·) (Tong et al., 2013). It is noted that anatase is more toxic than rutile TiO₂ NMs under light irradiation due to higher photocatalytic activity. Ma et al. (2012) found that TiO₂ NMs under simulated sunlight irradiation exhibited much higher toxicity than that under ambient laboratory light toward D. magna and Japanese medaka. However, the photoactivity of TiO₂ NMs was not obvious when interacting with some organisms such as algae and plants. For example, no enhanced toxicity to green algae (Raphidocelis subcapitata) was observed during exposure to TiO₂ NMs under visible, UVA, and UVB irradiation (Lee & An, 2013). TiO₂ NMs were reported to facilitate the photosynthesis, stomatal conductance and growth of plants in some cases (Hu et al., 2020; Liu et al., 2019; Tan et al., 2018).

Different from TiO₂, Ag NMs are readily to undergo physical and chemical transformation. Cheng et al. (2011) reported that sunlight irreversibly decreased the dispersion stability of Ag NMs, and the toxicity to a wetland plant (Lolium multiflorum) was highly suppressed. When releasing into natural environments, Ag NMs would not be the sole species. During oxidation and dissolution, Ag⁺ ions are releasing, and the more stable forms such as AgCl and Ag₂S could be formed, which are the persistent forms in natural environments (Li et al., 2017). It is reported that the dissolution of Ag NMs was unchanged by sunlight irradiation, but altered in the presence of Cl⁻ ions. The phototoxicity of Ag NMs toward E. coli was decreased by Cl⁻ at low concentrations (e.g., 0.01 M, 0.1 M), which was highly related to the dissolved Ag⁺ ions (Li et al., 2018). Sulfidation of Ag NMs also reduced the toxicity to zebrafish (Devi et al., 2015). In addition, Ag⁺ ions could be formed to Ag NMs under sunlight irradiation with the assistance of different organic matters such as dissolved organic matter and extracellular polymeric substances (EPS) from bacteria and algae (Kathiraven et al., 2015; Yin et al., 2012; Zhang et al., 2016). Even GO could facilitate the formation of Ag NMs from Ag⁺ ions (Cao et al., 2019). Sunlight can accelerate the transfer of electrons from the EPS to adjacent Ag⁺, and/or excite organic matters to produce strong reducing species for the photoreduction of Ag⁺ ions, thus causing the formation of Ag NMs (Zhang et al., 2016). This photoreduction process could decrease the toxicity and risk of Ag⁺ ions in natural environments because Ag NMs are less toxic in most scenarios (Choi et al., 2018).

Carbon NMs such as graphene and CNTs could undergo photo-transformation due to their conjugated bond system (Markiewicz et al., 2018). This photo-transformation is mainly surface oxidation, but the surface change is minor because of the relative stable graphitic structures. However, GO is very photoactive, and the photoreduction of GO surface has been reported (Han et al., 2019; Matsumoto et al., 2011). For bacteria, the photo-reduction increased the toxicity of GO to E. coli (Hou et al., 2017). In another work, the antibacterial activity of GO was increased after short-time (1–3 h) UV irradiation, and increased after long-time irradiation (12–60 h) (Gao et al., 2019). For algae, the photo-transformation of GO also enhanced the toxicity, mainly due to the production of dissolved organic byproducts (low molecular weight species) (Zhao et al., 2020). However, there are no toxicological data on the long-term photo-transformation of GO. Although GO could be well-dispersed in waters, long-term photoreduction is expected to lower the suspension stability of GO due to enhanced hydrophobicity. Therefore, the toxicity to aquatic organisms is likely to be lower, and more attention should be paid on benthonic organisms. CuO NMs are not very photo-active. The role of sunlight irradiation in the transformation (e.g., dissolution, sulfidation) of CuO NMs is unknown. Possible effect of sunlight irradiation on CuO NM toxicity needs further investigation.

4.3. Natural organic matter

NOM is a key component in natural waters to alter the fate and toxicity of NMs, with the concentrations ranging at 1–100 mg/L in surface water, soil water, and groundwater (Sharma et al., 2019). All the environmental processes (surface transformation, aggregation, adsorption, dissolution) of NMs as shown in Figure 6 could be impacted by NOM through NMs-NOM interaction. There are already many studies focusing on the toxicity of NMs as altered by NOM. NOM could change nanotoxicity through regulating (I) suspension stability of NMs, (II) electrostatic repulsion between NMs and organisms/cells, (III) steric hindrance, (IV) generation of ROS, and (V) bioavailability of dissolved metal ions (Wang et al., 2016). For most scenarios, NOM could decrease the toxicity of NMs to organisms in both aquatic and soil environments. For TiO₂ and graphene-based NMs, they do not dissolve much in water. Therefore, NOM could mitigate their toxicity through Mechanisms I-IV. For example, humic acid (HA) reduced the toxicity of TiO₂ NMs to alga (Chlorella sp.) through preventing the adhesion to algae due to electrostatic and steric repulsions (Lin et al., 2012). HA could also mitigate the toxicity of GO to both alga and D. magna due to the alleviation of oxidative damage and steric hindrance (Zhang et al., 2019; Zhao et al., 2019). For Ag and CuO NMs, Mechanism V (bioavailability of dissolved metal ions) is also important for the nanotoxicity mitigation in addition to Mechanisms I-IV. NOM in aqueous systems could reduce the bioavailable metal ions for organisms through (1) decreasing NM dissolution (Gunsolus et al., 2015), and/or (2) the formation of complexes of NOM-metal ions (Seitz et al., 2015). In addition, soil NOM could also increase Ag retention by the soil solid particles and decrease dissolved Ag levels (Li et al., 2017).

There are a few investigations which show the nanotoxicity enhancement in the presence of NOM. Castro et al. (2018) investigated the toxicity of GO to nine organisms including green algae (Raphidocelis subcapitata), aquatic plant (Lemna minor), lettuce (Lactuca sativa), planktonic microcrustacean (D. magna), brine shrimp (Artemia salina), chironomidae (Chironomus sancticaroli), freshwater polyp (Hydra attenuata), and nematodes (Caenorhabditis elegans). HA only increased the toxicity of GO to D. magna and C. elegans, and the reason for the observed toxicity enhancement was unclear. The increased toxicity of TiO₂ NMs to zebrafish (Danio rerio) by HA was also reported (Yang et al., 2013) due to increased oxidative stress. HA was coated on the surface of TiO₂ NMs and increased suspension stability, however, this was not responsible for the increased toxicity due to decreased TiO₂ NM uptake. On the contrary, NOM increased the accumulation of graphene sheets in the bodies of zebrafish, especially for gut and gill (Lu et al., 2017), but the role of NOM in the biological responses was unknown. All these unclear points need to be further investigated to better understand the nanotoxicity enhancement mechanisms induced by NOM.

It is noted that toxicity of NMs could be influenced by NOM types. It is reported that bovine serum albumin (BSA) increased the toxicity of NMs (e.g., Ag) by increasing their accumulation and dispersion. HA reduced the toxicity of most of NMs by scavenging free radicals and reducing the accumulation of NMs. These findings highlighted the significant influence of the NOM types on the toxicity of NMs in aquatic environments (Liang et al., 2020). Sunlight irradiation is another factor that may influence nanotoxicity. NOM molecules such as HA and citric acid serve as free radical scavengers to mitigate the toxicity of NMs (e.g., Ag, CuO, CeO₂) (Gao et al., 2012; Peng et al., 2015; Tella et al., 2014). However, under sunlight irradiation, HA could be as a photosensitizer or precursor for ROS production (Dasari & Hwang, 2010), which may be the reason for the stronger TiO₂ NM toxicity in the presence of HA under sunlight irradiation than that without irradiation (Yang et al., 2013). Moreover, sunlight irradiation is able to degrade NOM (Yang et al., 2013), and the NOM-induced nanotoxicity may be ultimately reduced if the NOM-coated NMs are thoroughly exposed to long-term sunlight irradiation.

4.4. Minerals

Heteroaggregation is a dominant driving factor for NM settling from the water columns (Espinasse et al., 2018). Minerals are important components of soils and sediments, and are widely present in natural waters (Zhao et al., 2015). NMs in these environmental compartments can heteroaggregate with mineral particles. Heteroaggregation could occur between minerals and all the four types of NMs (TiO₂, Ag, CuO, graphene). GO sheets were able to heteroaggregate with positively charged goethite and Al₂O₃ particles, while did not with negatively charged montmorillonite, kaolinite and SiO₂ particles, confirming that the above heteroaggregation was dominated by electrostatic interaction (Zhao et al., 2015). Similar heteroaggregation was found between GO and hematite (Feng et al., 2017). Interestingly, negatively charged Ag NMs could heteroaggregate with negatively charged montmorillonite through the face-edge conformation (Zhou et al., 2012). This is because that the edge of montmorillonite remained positively charged when the solution pH was higher than its zero point of charge (ZPC). All these heteroaggregation processes could destabilize NMs in aquatic environments. Due to heteroaggregation, the toxicity of NMs in the presence of minerals was highly decreased for all the scenarios. It is reported that GO toxicity to protozoan was reduced by the formation of GO-kaolin heteroaggregates (Kryuchkova & Fakhrullin, 2018). Heteroaggregation with hematite decreased antimicrobial activity of Ag NMs by physically preventing the direct contact with bacteria (E. coli) (Huynh et al., 2014). GO toxicity to alga (Auxenochlorella pyrenoidosa) was reduced by Al₂O₃ particles with different shapes due to the formation of GO-Al₂O₃ heteroaggregates and the decrease in direct contact and membrane damage (Zhao et al., 2018). NOM further decreased the toxicity of GO in the presence of Al₂O₃ particles due to the surface coating of GO-Al₂O₃ heteroaggregates by NOM and the enhanced steric hindrance (Zhao et al., 2018). In addition to heteroaggregation, the reduction in bioavailable metal ions released from NMs (e.g., Ag, CuO) was another mechanism for the NM toxicity mitigation. For example, hematite reduced Ag NM toxicity to another alga Chlamydomonas reinhardtii by decreasing the bioavailability of Ag ions through adsorption (Huang, Wei, et al., 2019).

Different types of NMs are co-existing in environmental compartments. Antagonistic effect was observed during the co-exposure of different NMs, which could also be explained by the above two mechanisms (heteroaggregation and reduction in bioavailable metal ions). For example, combined mixtures of different NMs (e.g., CuO, TiO₂, Fe₂O₃) were significantly less toxic in comparison to individual NMs due to increased heteroaggregation (Jośko et al., 2017). The antagonistic effect of negatively charged GO and ZnO NMs to freshwater fish larva (Danio rerio) was also observed in the incubation systems, which was mainly due to the adsorption of the released Zn²⁺ ions by GO (Ye et al., 2018). This mechanism was also responsible for the reduced toxicity of TiO₂ and ZnO NMs to bacteria (Tong et al., 2015). The combined toxicity of NMs may also influenced by sunlight irradiation. Under dark condition, TiO₂ NMs were reported to attenuate Ag NM toxicity due to the adsorption of dissolved Ag⁺ (Wilke et al., 2016). However, under sunlight irradiation, TiO₂ and Ag NMs showed synergistic toxicity due to enhanced oxidative stress (Wilke et al., 2018).

4.5. Aging

NMs undergo aging after releasing into the compartments. The aging processes are the combined effect of sunlight irradiation, NOM, minerals, as well as biomolecules (Figure 6), which include chemical transformation, adsorption, and heteroaggregation. Adsorption of NOM and heteroaggregation of minerals could reduce NM toxicity, which has been discussed above. For chemical transformation, sulfidation of NMs (e.g., CuO, Ag) was a main process in water, sediments, and soil compartments. For CuO NMs, sulfidation is an important transformation in biological and environmental systems, which was reported to reduce the toxicity of CuO NMs to human cells (Wang et al., 2013). For Ag NMs, sulfidation reduced the mortality and reproduction of C. elegans caused by Ag NMs. This is because sulfidation not only decreased dissolution of Ag NMs, but also reduced the bioavailability of intact Ag NMs (Starnes et al., 2015). AgCl is another transformed product in the environment. Cl⁻ could decrease Ag NM toxicity to bacteria by forming AgCl and decreasing bioavailable Ag⁺ ions in aquatic phase (Chambers et al., 2014). The chemical transformation processes are further regulated by NOM. It is reported that Ag NMs were transformed to Ag₂S (main species), AgCl, Ag₃PO₄, and Ag₂O in the flooded soil after incubation for 28 days, and soil NOM further increased transformation and retention of Ag NMs in soils (Li et al., 2017). In another work, aging by NOM (fulvic acid, FA) was found to reduce the toxicity of pristine and sulfurized Ag NMs to soil bacteria in media with organic matter (FA) (Schultz et al., 2018). All these aging processes under laboratory conditions suggest that toxicity of NMs is likely to be reduced in natural environments.

The microcosm and mesocosm approaches could be more realistic to reveal the ecotoxicity of NMs in terrestrial and aquatic ecosystems (Jovanović et al., 2016; Lu et al., 2020; Vijayaraj et al., 2018). A mesocosm experiment (5 weeks) was conducted to investigate the toxicity of TiO₂ NMs to periphytic algae. TiO₂ NMs at 0.05 mg/L (ECRs) did not show significant changes on growth and algal assemblage composition, suggesting that TiO₂ NMs had low environmental risk in natural waters (Wright et al., 2018). However, TiO₂ NMs at 5 mg/L (concentration in wastewater) decreased periphyton biomass and altered algal assemblage structure, indicating that the toxicity of TiO₂ NMs in wastewater should be paid more attention. In another mesocosm study, TiO₂ NMs at 0.025 and 0.25 mg/L reduced the biomass of Rotifera, however, the growth of Cladocera, Copepoda, phytoplankton, macrophytes, chironomids and fish was unaffected (Jovanović et al., 2016). Therefore, TiO₂ NMs at ECRs would not significantly affect the main functions of freshwater ecosystem. For CuO NMs (12.5 µg/L) in a mesocosm, no mortality was observed for the endobenthic species (Scrobicularia plana and Hediste diversicolor) after exposure for 21 days (Buffet et al., 2013), although some biochemical parameters such as behavioral biomarkers, antioxidant defenses and genotoxicity were changed during exposure.

5. Concluding remarks, challenges, and future research perspectives

Environmental risk assessment of NMs requires abundant and accurate data on their environmental concentrations and ecotoxicity. For environmental concentrations, the collected PEC and MEC data in this review show that the concentrations of the four types of NMs (TiO₂, Ag, CuO, graphene) are very low in surface water and soil, while more NMs are deposited in sediments. Wastewater effluents and sewage sludges contain relatively high concentrations of NMs, and the environmental concentrations for the surface waters near the WWTPs and sewage sludge-treated soils are therefore at high levels. Currently, the ecotoxicity of NMs in surface water and soil compartments are of low concern as indicated by much higher PNECs than PECs. In wastewater, NMs showed higher toxicity than natural surface water, which could affect the treatment performance of WWTP. The toxicity of NMs at ERCs demonstrated that individual bacteria and microbial communities are very sensitive to NM exposure. For other organisms, no or low mortality was observed, instead, the biological responses such as antioxidative defense and gene expression could be triggered under NM exposure at ERCs.

It should be noted that the NM-induced toxicity under laboratory conditions could be changed by environmental factors including sunlight irradiation, NOM, and minerals. Generally, sunlight irradiation enhanced the toxicity of TiO₂ and GO among the four types of NMs. NOM reduced the toxicity of NMs for most scenarios, in which the electrostatic and steric hindrance between NMs and organisms/cells induced by surface coating could be the main mechanisms. For minerals, toxicity of all the four types of NMs was mitigated for all the scenarios due to heteroaggregation, and the adsorption of released metal ions (for Ag and CuO) on minerals. The aging (e.g., sulfidation) of NMs in the environment generally reduced nanotoxicity. Taken together, the toxicity of NMs could be reduced under real environmental conditions, and the obtained toxic parameters under laboratory conditions may likely be overestimated.

Therefore, it can be generally concluded that the environmental risks of the four NMs are of low concern in surface water and soil compartments presently. More concerns should be raised on nanotoxicity in sediments, and any environmental compartments contaminated by wastewater, sewage sludges, and/or accidental spills. However, there are still many uncertainties/challenges on both environmental concentrations and ecotoxicity, which are identified as follows:

1. For the modeling of PECs, there are several fate models (e.g., materials flow analysis model, steady-state fate and transport model, dynamic stochastic model, nanoFate model) for predicting the environmental distribution of NMs (Garner et al., 2017; Giese et al., 2018; Musee, 2018). Each model has respective assumptions and limitations. Although the environmental processes of NMs such as dissolution, transformation and aggregation were considered in some fate models, there are many assumptions and uncertainties, which require the confirmation by experimental investigations. For examples, the production volume of NMs in the world or specific countries is largely uncertain, while it is crucial for the outcome of exposure modeling (Giese et al., 2018); the homoaggregation of NMs was commonly assumed to negligible (Garner et al., 2017); NMs in environments could be taken up by organisms, but this factor was not be included in the fate description during modeling. More efforts are needed for more accurate predictions.

2. For MECs, the measurement techniques such as sp-ICP-MS cannot distinguish a type of NMs with different forms (e.g., different surface coatings, crystal forms). It is also hard to distinguish engineered NMs from naturally occurring nanoparticles in environmental samples (Sun et al., 2014). In addition, the quantification of metal-based NMs by sp-ICP-MS is based on the assumptions: (1) all the particles are spherical; and (2) the molecular formula, and chemical compositions are already known (Tou et al., 2017). Moreover, the quantification technology of carbon-based NMs still has many limitations. The development of new analytical methods should be specific for very low concentrations of targeted NMs in different environmental compartments.

3. For ecotoxicity, most of the mechanistic investigations on nanotoxicity are based on the toxicological experiments at high concentrations. Therefore, the provided mechanistic insights are likely not useful for the risk assessments at ECRs (Krug, 2014). In addition, heteroaggregation and aging processes in aqueous environments could induce the settling of NMs into sediments. This process brings further risks to benthonic organisms, which deserves future and focused studies. Moreover, more efforts should be made on the compartments near the NM “point sources” (e.g., production plant outfalls and waste treatment plants). More nanotoxicity studies on field and outdoor mesocosms are also needed.

4. One of the main aims of environmental risk assessment of NMs is to better guide the application of nanotechnology. The environmental concentrations and PNECs of NMs summarized in this review may provide useful information on the selection of suitable and safe NM types and concentrations during nanotechnology applications (e.g., nano-agriculture; nano-remediation) in environments. In addition, the interfacial interactions between NMs and cells are helpful for designing the NMs with controlled shape, size and surface properties to minimize their toxicity and environment-related issues (Amde et al., 2017). For example, the safer NMs could be designed by coating with NOM and forming biomolecule coronas (Markiewicz et al., 2018).

Additional information

Funding

This work was supported by National Natural Science Foundation of China (41822705, 41530642), Taishan Scholars Program of Shandong Province (tsqn201909051), Natural Science Foundation of Shandong Province (JQ201805), USDA-NIFA Hatch program (MAS 00549), and UMass Amherst Conti Fellowship (2019–2020).