Quantity and quality of natural organic matter influence the ecotoxicity of titanium dioxide nanoparticles.

Abstract Nanoparticles’ fate is amongst other parameters determined by the quantity and quality of natural organic matter (NOM). Consequently, the ecotoxicity of nanoparticles is modified, while only little information is available on the NOM characteristics triggering this interplay. This study systematically examined how NOM quantity and quality influences the acute ecotoxicity of titanium dioxide nanoparticles (nTiO2) towards Daphnia magna. Therefore, two nTiO2 products (A-100 and P25; ∼100 nm) were investigated in combination with seven NOM types of variable quality at four levels each (up to 4.00 mg total organic carbon/L). The results showed that – independent of the applied nTiO2 product and NOM type – nTiO2 ecotoxicity decreased up to a factor of >18 with increasing NOM concentration. More importantly, increasing levels of aromaticity and hydrophobicity of the NOM decreased the magnitude of toxic effects caused by nTiO2, which was again independent of the nTiO2 product tested. In the light of the ubiquitary presence of NOM, the ecotoxicological risk of nTiO2 in surface waters with high NOM loads is likely moderate. However, interactions of nTiO2 and NOM in combination with other natural or chemical stressors are not well-understood but seem to be fundamental for a reliable risk assessment of nanoparticles.


Introduction
Considering the continuously increasing production and use of nanoparticles (Gottschalk et al., 2009;Scheringer, 2008) with varying chemical compositions and shapes as ingredients in food, daily care products or industrial applications, their release into aquatic environments seems inevitable. Nanoparticles' fate is determined by the conditions and properties of their surrounding medium (Schaumann et al., 2015), finally affecting their ecotoxicological potential towards aquatic organisms (Blinova et al., 2012;Hall et al., 2009).
However, the implications of environmental conditions such as variations in the quantity and quality of natural organic matter (NOM) on nanoparticle toxicity, have only rarely been considered yet (Seitz et al., 2015a), although NOM is ubiquitously present in surface water bodies (Erhayem & Sohn, 2014a). In particular, NOM has the potential to interact with nanoparticles decreasing their agglomeration and sedimentation behavior (Domingos et al., 2009;Philippe & Schaumann, 2014;Zhu et al., 2014), which seems more effective with increasing NOM hydrophobicity (Lee et al., 2011) and/or aromaticity (Pakarinen et al., 2013). Additionally, one study suggested that nitrogen (N) and sulfur (S) rich NOMs reduce the ecotoxicity of silver nanoparticles (nAg) for a bacterial model species, most likely by limiting the release of toxic silver ions from the particles into the surrounding medium (Gunsolus et al., 2015). Nonetheless, little systematic information is available on NOM characteristics (¼ quality) determining alterations in the ecotoxicological potential of nanoparticles .
In this context, the present study systematically assessed the implications of NOM quantity and quality on the nanoparticleinduced ecotoxicity involving the immobility of the water flea Daphnia magna as response variable. Considering the massive variety of nanomaterials with variable chemical composition (including different surface coatings) and shape (e.g. wires, rods, single-crystals), the present study selected two nanomaterials with different crystalline structure composition, which are produced at high volumes. In this light, nanoparticulate titanium dioxide (nTiO 2 ) -a nanoparticle with relatively high predicted environmental concentrations, namely (up to 110 mg/L in sewage treatment plant effluents; 85% percentile; Sun et al., 2014)served as model. In detail, two nTiO 2 products, i.e. A-100 and P25 (differing in their crystalline structure composition), with an initial particle size of $100 nm, were investigated separately with NOMs of varying quality (including N and S). These NOMs were assessed at four levels of total organic carbon (TOC) representing environmentally relevant concentrations (Ryan et al., 2009;up to 4.00 mg/L). As representatives for NOMs, commercially available humic acids (two different types), seaweed extract, natural organic matter, fulvic acid, Leonardite and Pahokee peat were chosen and further characterized by size-exclusion chromatography -organic carbon detection -organic nitrogen detection (LC-OCD-OND; Huber et al., 2011).

nTiO 2 preparation and characterization
The nTiO 2 products A-100 (100% anatase) and P25 ($70% anatase and $30% rutile) were provided by Crenox GmbH (Krefeld, Germany) and Evonik (Essen, Germany), respectively. The products had an advertised primary particle size of 6 and 21 nm, and a specific surface area of 230 and 50 m 2 /g. On the basis of these nTiO 2 powders, two dispersant-free, size-homogenized stable dispersions of $100 nm average particle size were obtained by stirred media milling (PML 2, Bühler AG, Uzwil, Switzerland). Subsequent centrifugation ensured the removal of larger particles. Prior to each application, the stock dispersions (2.00 g nTiO 2 /L) were analyzed for their initial particle size distribution (intensity based) using dynamic light scattering (DLS, Delsa Nano C, Beckman Coulter, Krefeld, Germany). Additionally, their average particle size in the test medium was monitored at intervals of 24 h until termination of all bioassays (96 h; Tables S1 and S2). For this purpose, samples of 3 mL were taken 2 cm beneath the water surface from the center of one randomly selected replicate containing 1.00 and 4.00 mg nTiO 2 /L. This procedure was carried out for each NOM and all TOC levels. Moreover, the zeta potential of each nTiO 2 product was determined for each NOM type applied. Therefore, nTiO 2 samples (4.00 mg/L) were taken and analyzed at nominal NOM concentrations of 0.00, 0.04, 0.40 and 4.00 mg TOC/L by using electrophoretic light scattering (Table S3). An overview of all nTiO 2 characteristics is provided in Table S4.

NOM preparation and characterization
Seven different NOM types were used in this study: Sigma Aldrich Humic Acid (¼ SA HA), Seaweed extract (¼ SW; Marinure, Glenside, Scottland), Suwannee River Natural Organic Matter (¼ SR NOM), Suwannee River Humic Acid (¼ SR HA), Suwannee River Fulvic Acid (¼ SR FA), Leonardite (¼ LEO) and Pahokee Peat (¼ PP, the latter five were purchased from the International Humic Substances Society, St. Paul, MN). Prior to the start of each acute toxicity test the respective organic material was dissolved and serially diluted using the actual test medium (for details see sections Test organism and Toxicity testing), finally resulting in four nominal TOC levels: 0.00, 0.04, 0.40 and 4.00 mg TOC/L. Additionally, 50 mL samples of each NOM (8.00 mg TOC/L; nTiO 2 free) were subjected to a dissolved organic carbon (DOC) analysis (LC-OCD-NOD; DOC Labor, Karlsruhe, Germany) allowing for a detailed characterization of the material (Huber et al., 2011; Table 1). Further, information on the NOM N-and Scontent (Table 1) were either obtained from the IHSS webpage (International Humic Substances Society, 2016) or by own CHNSanalyses [vario MICRO cube (elementar), Hanau, Germany].

Test organism
The test organism D. magna (Eurofins-GAB GmbH, Pforzheim, Germany) was kept within climate controlled chambers (Weiss Environmental Technology Inc., Stuttgart, Germany) using a 16:8 h (light:dark) photoperiod at 20 ± 1 C. The animals were cultured in groups of 25 adult organisms using 1.5 L reconstituted hard freshwater (¼ ASTM-M; ASTM International, 2007) with additions of selenium, vitamins (thiamine hydrochloride, cyanocobalamine, biotine) and SW (Seitz et al., 2013) as culture medium. Three times per week all animals were transferred to new culture medium, while feeding took place daily, applying the green algae Desmodesmus sp. with an equivalent of 200 mg C per organism.

Toxicity testing
For each NOM type, a series of acute toxicity tests with daphnids was conducted. Therefore, ASTM-M (without previous additions of SW) was enriched with 0.00, 0.04, 0.40 or 4.00 mg TOC/L of the respective NOM and either A-100 or P25 at nTiO 2 concentrations ranging from 0.00 mg/L to 16.00 mg/L. Each acute toxicity test was carried out as described in detail in the OECD guideline 202 (OECD, 2004) with an elongated study duration of 96 h, as recommended for nanoparticle testing (Dabrunz et al., 2011). During all tests, groups (n ¼ 4) of five daphnids (524 h) were exposed to the respective NOM and nTiO 2 level and checked for immobility every 24 h. The temperature and light conditions met those described in the section Test organism. The pH during all ecotoxicity tests was within the range of 8.2 ± 0.1.

Statistical analysis
Immobilization data of the acute toxicity tests (for A-100 and P25) were analyzed for their respective 96-h EC 50 (¼ effect Table 1. Characterization of the dissolved organic carbon properties as well as N and S information for the seven different NOM qualities (i.e. Sigma Aldrich Humic Acid (SA HA), Seaweed Extract (SW), Suwannee River (SR) Natural Organic Matter (SR NOM), SR Humic Acid (SR HA), SR Fulvic Acid (SR FA), Leonardite (LEO) and Pahokee Peat (PP)) determined via LC-OCD-NOD or gained by either CHNS-analyses or provider's information (International Humic Substances Society, 2016). concentration at which 50% of the tested population is immobilized) by using Abbott's formula to correct for control mortality (if necessary) and fitting proper dose-response models (based on Akaike's information criterion and expert judgment). Afterwards, EC 50 values were compared for statistical significant differences considering the different NOMs and their respective levels. Therefore, confidence interval (CI) testing was accomplished (Wheeler et al., 2006). In order to evaluate the role of NOM characteristics for the nTiO 2 toxicity, data of the DOC analyses (e.g. aromaticity and molar weight of the respective humic substance; see Table 1) were separately analyzed for relationships with 96-h EC 50 values of nTiO 2 obtained in the presence of the medium NOM level (i.e. 0.40 mg TOC/L) using linear correlation models. To assess the potential of the NOM characteristics to jointly explain the variation in nTiO 2 toxicity, linear regressions with multiple predictors were performed for each nTiO 2 product separately. Thereby, optimal models were determined by using an ANOVA based stepwise comparison approach. All statistical analyses and figures are based on the statistics program R version 2.15.3 (R Core Team, 2013) as well as the required extension packages (Lemon, 2010;Ritz & Streibig, 2005).

Quantity of NOM
In the absence of NOM the 96-h EC 50 values for A-100 and P25 were as low as 0.85 (95% CI: 0.55-1.14) and 2.98 (95% CI: 2.30-3.66) mg nTiO 2 /L, respectively (Table 2). However, irrespective of the applied nTiO 2 product and NOM type, the toxicity of nTiO 2 decreased with increasing NOM concentration (Table 2; Figure  S1 A-C and S2 A-C). The presence of SR NOM, for instance, reduced the toxicity of A-100 by a factor of approximately 1.3, 3.4 and418.8 at 0.04, 0.40 and 4.00 mg TOC/L, respectively, when compared to its absence (Table 2; Figure S1 A-C). Also the particle size distribution and the zeta potential meaningfully changed for different levels of NOM. In detail, higher concentrations of NOM displayed lower average particle sizes (up to approximately 15-fold; Tables S1 and S2) and higher zeta potentials (up to approximately 3.5-fold; Table S3), independent of the nTiO 2 product applied.

Quality of NOM
Also, the identity (Table 1) of the applied NOM meaningfully influenced the toxicity of nTiO 2 for Daphnia. For instance, SA HA, LEO and PP had the highest potential to reduce -by a factor of approximately 10 for A-100 -the toxicity of nTiO 2 at concentrations equal to or above 0.40 mg TOC/L (Table 2; Figures S1 and S2). These patterns were subsequently linked to the quality or properties of the NOM: SUVA measurements, for example, indicated a partly statistically significant positive linear relationship with the respective 96-h EC 50 values of both A-100 and P25 explaining around 60% of the variability ( Figure 1A and Table 3). A similar relationship was uncovered for the content of hydrophobic organic carbon (HOC) and low-molecular weight neutral compounds, which also increased with the 96-h EC 50 values for D. magna. This was statistically significantly different from the null-model only for A-100 (Table 3). The aromaticity of the NOM increased the 96-h EC 50 values and thus reduced the ecotoxicity of nTiO 2 with increasing quantities; the observation was statistically significant for both nTiO 2 products explaining between 66% and 88% of the variability in the toxicity data ( Table  3). The strength of the relationships between EC 50 values and NOM qualities (here defined as the slope of the linear relationship) deviated strongly between P25 and A-100 ( Figure 1A-J and Table 3). Higher slopes and thus more frequent statistically significant linear correlations were identified for A-100 relative to P25 (Table 3). Further, multiple linear regression models revealed for A-100 that HOC -as the only NOM characteristic -explains approximately 94% of the observed variability in nTiO 2 -induced toxicity (Table 3). For P25, in contrast, the best fitting model included  besides HOC also aromaticity, explaining approximately 92% of the observed variability, whereas the interaction term of both factors was not statistically significant (p value: 0.15). Finally, the statistical analysis suggested that the remaining NOM characteristics such as the quantity of hydrophilic DOC, biopolymers, building blocks, low-molecular acids, N-and S-content affected nTiO 2 -related ecotoxicity only marginally for both nTiO 2 -products (Table 3; Figure S3).

Discussion
The EC 50 values in absence of NOM are in accordance with previous studies conducted in our laboratory using the same test species and identical nTiO 2 products (Bundschuh et al., 2012;Dabrunz et al., 2011;Seitz et al., 2014). Similarly, A-100 showed a higher toxicity than P25, which had in previous studies been associated with (i) a higher surface reactivity of anatase compared to rutile and (ii) different crystalline structure properties, e.g. inducing a higher production of harmful reactive oxygen species . Moreover, the reduced toxicity of nTiO 2 with increasing levels of NOM can be assigned to steric repulsion forces among nanoparticles, which become more important at higher NOM concentrations (Erhayem & Sohn, 2014b). These forces kept nTiO 2 in the water phase but presumably also decreased their bioavailability (Erhayem & Sohn, 2014a;Rosenfeldt et al., 2014; Tables S1 and S2). This contradiction may be explained by the potential of NOM to cover both the surface of nTiO 2 as well as Daphnia. Thereby, NOM prevents a particle attachment on the organism's surface (¼ biological surface coating) (Lin et al., 2012), which is proposed as one mode of toxic action for nTiO 2 (Dabrunz et al., 2011). Hence, NOM has the potential to widely eliminate associated adverse acute effects on Daphnia's health in terms of mobility (Table 2; Figures S1 A-C and S2 A-C) and molting behavior (Dabrunz et al., 2011;Noss et al., 2013). Moreover, with increasing NOM concentrations larger quantities of NOM adsorb to the nanoparticle surface (Erhayem & Sohn, 2014a), as indirectly displayed by the zeta potential measurements of the present study (Table S3). This in turn may explain the dose-dependent reduction of nTiO 2induced toxicity by NOM ( Figure S1 A-C and S2 A-C). The efficiency of NOM to mitigate nTiO 2 -induced toxicity (Table 2, Figure S1 A-C and S2 A-C) and to influence the nTiO 2 particle size distribution (Tables S1 and S2) seems to be driven not only by its quantity but also by its quality (Erhayem & Sohn, 2014a). In this context, Lee et al. (2011) uncovered differences in the toxicity of quantum dots in the presence of NOMs of varying quality for Daphnia, which was positively correlated with the hydrophobicity of NOMs when using specific ultraviolet absorbance (SUVA) as proxy. During the present study, SUVA, HOC and aromaticity measurements, indicate a partly statistically significant positive linear relationship with respective 96-h EC 50 values of both A-100 and P25. The remaining NOM qualities, such as hydrophilic DOC, biopolymers, humic substance, lowmolecular weight neutrals or low-molecular weight acids but also N-and S-content did not affect the nTiO 2 -related ecotoxicity ( Figure S3). Whereas for the most of these NOM qualities, no further information regarding their influence on the ecotoxicity of nTiO 2 is available, the lack of explanatory power of N-and Scontent is contrary to Gunsolus et al. (2015). These authors observed a reduced toxicity with increasing N-and S-content of the NOM for nAg, which may be explained by the sulfidation of nAg surfaces, reducing the release of toxic Ag-ions and thus the toxicity towards bacteria -a pattern, however, irrelevant for the present investigation. The data of the present study, hence, suggest that mainly two NOM characteristics (i.e. HOC together with the aromaticity) determine the adsorption potential of NOM onto nTiO 2 inducing a higher concentration of coated particles in the water phase . In detail, organic materials exhibiting high levels of aromaticity and phenolic content have a greater affinity to the nTiO 2 surface (Erhayem & Sohn, 2014b) and thus potentially induce a stronger (long lasting) stabilization effect. Further, higher aromatic contents increase the hydrophobicity (and hence the HOC) of NOM resulting in larger amounts of NOM adsorbed to the nTiO 2 surfaces (Erhayem & Sohn, 2014b). This results in steric repulsion forces among coated particles/organisms, that subsequently reduce the nanoparticlerelated acute ecotoxicity as observed in the present study.
However, the strength of the relationships between EC 50 values and NOM qualities differed strongly between P25 and A-100 ( Figure 1A-J and Table 3), whereas higher slopes were identified for A-100 relative to P25 (Table 3). This indicates a higher efficacy of NOM to mitigate A-100-induced toxicity. Although both nTiO 2 -products had comparable initial particle sizes ($100 nm), the observed phenomenon may be attributed to the generally lower ecotoxicity of P25 and a higher surface area of A-100 compared to P25. This can be explained by a higher surface roughness of A-100 , potentially enabling larger amounts of NOM to adsorb to the surface, inducing increased electrostatic repulsion forces when compared to P25.

Conclusions
The present study clearly displayed the importance of NOM quantity and quality -particularly aromaticity and HOC -for its potential to reduce acute nTiO 2 -induced toxicity. In this context, it has to be noted that the nTiO 2 triggered effects strongly dependent on its crystalline structure . However, as NOMs are ubiquitously present in surface water bodies at different concentrations (Ryan et al., 2009), the findings of the present study suggest that the ecotoxicological risk associated with nTiO 2 release into the aquatic environment is moderate. Moreover, the fate and toxicity of nTiO 2 in nature may, as a consequence of more complex NOM matrices (e.g. bacterial and algal exudates as well as living cells), but also the additional influence of other environmental parameters such as the pH and ionic strength of the surface water (Cupi et al., 2015;Erhayem & Sohn, 2014b;Gallego-Urrea et al., 2014;Keller et al., 2010;Loosli et al., 2013;Lüderwald et al., 2016;Metreveli et al., 2016;Seitz et al., 2015a) deviate from the observations of the present study. For instance, the toxicity of nAg in presence of NOM is higher at lower pH values (Seitz et al., 2015b). Similarly, lower ionic strengths in presence of NOM during aging of nTiO 2 revealed an increased toxicity to daphnids relative to an aging under high ionic strength conditions (Seitz et al., 2015a). This complexity of interactions between nanoparticles and their surrounding environment calls for further systematic approaches disentangling the importance of the individual parameters and their combination on nanoparticleinduced effects in aquatic systems . Irrespective of these considerations, the increase in the dwelling time of NOM-coated nTiO 2 in the water column elevates the exposure of filter-feeding pelagic organisms, sensitizing subsequent generations (Bundschuh et al., 2012), and consequently raises chronic risk for aquatic species.