Exposure Medium: Key in Identifying Free Ag+ as the Exclusive Species of Silver Nanoparticles with Acute Toxicity to Daphnia magna

It is still not very clear what roles the various Ag species play in the toxicity of silver nanoparticles (AgNPs). In this study, we found that traditional exposure media result in uncontrollable but consistent physicochemical transformation of AgNPs, causing artifacts in determination of median lethal concentration (LC50) and hindering the identification of Ag species responsible for the acute toxicity of AgNPs to Daphnia magna. This obstacle was overcome by using 8 h exposure in 0.1 mmol L−1 NaNO3 medium, in which we measured the 8-h LC50 of seven AgNPs with different sizes and coatings, and determined the concentrations of various Ag species. The LC50 as free Ag+ of the seven AgNPs (0.37–0.44 μg L−1) agreed very well with that of AgNO3 (0.40 μg L−1), and showed the lowest value compared to that as total Ag, total Ag+, and dissolved Ag, demonstrating free Ag+ is exclusively responsible for the acute toxicity of AgNPs to D. magna, while other Ag species in AgNPs have no contribution to the acute toxicity. Our results demonstrated the great importance of developing appropriate exposure media for evaluating risk of nanomaterials.

importance to develop exposure media that are able to preserve the initial physical and chemical properties of AgNPs during the exposure, as well as to accurately determine the various Ag species in AgNPs 30 .
The objective of this work is to evaluate the contribution of various Ag species in AgNP exposure mixtures to the toxicity towards D. magna. To exclude the artifacts from the exposure media, we at first sought an exposure medium that was able to minimize the aggregation of AgNPs, eliminate the precipitation of Ag 1 , and ensure the living of D. magna. Then, the acute toxicity of AgNPs with varied coating and sizes were assessed using the optimized medium, and the median lethal concentration (LC50) in terms of total Ag, free Ag 1 , total Ag 1 , dissolved Ag, and nano Ag were determined with different analytical techniques. Finally, the concentration of various Ag species in exposure mixtures of different AgNPs were compared to LC50 of Ag 1 as AgNO 3 , to identify which species of Ag is responsible for the AgNP acute toxicity towards D. magna.

Results
Characterization of AgNPs. A series of AgNPs varied in their primary size and surface coating were studied herein. Fig. S1 (of Supporting Information (SI)) shows that both the synthesized polyvinylpyrrolidones (PVP) coated AgNPs (herein referred to as AgNP PVP ) and commercially obtained sodium-citrate coated AgNPs (herein referred to as AgNP CIT ) were in spherical or nearspherical structures in ultrapure water. No visible coatings were evident in AgNP PVP transmission electron microscopy (TEM) images, indicating the excess PVP in suspensions were effectively removed in wash procedures. Counted from over 133 particles in TEM images, the respective primary particle sizes of the two synthesized AgNP PVP were 9.9 6 2.2, and 28.1 6 4.6 nm (herein referred to as AgNP PVP10 and AgNP PVP28 ), while that of the 5 commercial AgNP CIT (herein referred to as AgNP CIT10 , AgNP CIT20 , AgNP CIT40 , AgNP CIT60 and AgNP CIT100 ) were 9.2 6 1.7, 20.7 6 2.8, 40.0 6 3.9, 56.3 6 5.9, 95.5 6 7.8 nm, respectively, agreed very well with their commercially reported sizes. All the studied AgNPs had a negative f-potential in ultrapure water, due to that PVP or citrate were used as stabilizing and coating material.
Development of Acute Toxicity Testing Medium for Daphnia magna. The EPA and OECD standard exposure media with diverse ions were established with primary concerns for the living maintenance of D. magna. These media have long been applied for testing the acute toxicity of traditional chemicals, as they do not cause physicochemical changes to the tested chemicals. Given the various ions in EPA and OECD media could alter the state and speciation of AgNPs/Ag 1 for their special properties, NaNO 3 was selected to develop a medium that is able to maintain both the state of AgNPs/Ag 1 and the living of D. magna, even though a medium containing single type of ion would not be favourable for the survival of D. magna as the EPA and OECD media did. The 48-h mortality experiments showed that with 0.1 mmol L 21 NaNO 3 medium (pH 7.8 6 0.2 adjusted by NaOH), the D. magna all lived well and showed normal behavior in at least 14 h without feeding, while with NaNO 3 media at other studied concentrations the death of D. magna occurred within 8-h exposure, indicating the media with either higher or lower concentrations of NaNO 3 were more toxic relative to 0.1 mmol L 21 NaNO 3 upon D. magna (Fig. 1). The higher toxicity of media with higher concentration of NaNO 3 would probably due to the toxicity of NO 3 2 to D. magna, whereas the more toxic of media with lower concentration of NaNO 3 could be attributed to the insufficient ionic strength of the media to maintain the lives of D. magna. Thus, test of 8-h acute toxicity to D. magna in 0.1 mmol L 21 NaNO 3 (pH 7.8 6 0.2 adjusted by NaOH) medium was adopted as optimum. We acknowledge that the 8-h exposure was relatively short, but it was a compelled choice due to concerns over the survival assurance of D. magna in this specific exposure medium. It should also be noted that, as the aim of this study was to identify the relative importance of various Ag species in AgNP exposure mixtures to the acute toxicity towards D. magna by using a reliable exposure medium, the relative short term exposure would actually help understand the rapid response of D. magna to certain Ag species, thus reveal the Ag species with relatively high acute toxicity towards D. magna. In addition, it was found that some previous studies also chose short term exposure (#8 h) in toxicity study to D. magna to meet special concerns 48,49 . Figure 2 shows the characterization of a representative AgNP, AgNP CIT20 , at 8 h after being spiked in four exposure media including OECD simplified M4 (SM4) medium, EPA hard water, 0.1 mmol L 21 NaNO 3 , and ultrapure water, respectively. While the OECD and EPA media induced the aggregations of AgNPs (hydrodynamic sizes and TEM results shown in Fig. 2A-D), suppressed the absorbance at 400 nm (Fig. 2E), and turned the suspension color to blue (photograph) relative to that in ultrapure water, the 0.1 mmol L 21 NaNO 3 medium resulted in negligible change in comparison to that in ultrapure water. Fig. 2F also shows the total Ag 1 released from these media during 8-h exposure. At the initial of exposure, the total Ag 1 content accounted for only 1.3% of the total Ag in AgNP CIT20 (4 mg L 21 as total Ag). In OECD and EPA media, proportions of total Ag 1 increased to 5.0% and 2.7%, respectively. Whereas, the total Ag 1 elevated only a little to 1.6% in NaNO 3 medium and to 1.7% in ultrapure water at the end of 8-h exposure. The analysis of variance (ANOVA) results further indicated that the total Ag 1 release from the later two media had no significant differences either during  (p 5 0.922) or at the end of (p 5 0.116) 8-h exposure. These evidence suggested that, in the NaNO 3 medium, although slight increase of Ag 1 content appeared as the exposure proceeded from 0 to 8 h due to the inevitable reactions with dissolved oxygen 16 , the Ag 1 content of AgNP suspension during acute toxicity exposure test was stabilized to the greatest extent as in ultrapure water and showed no significant changes during 8-h exposure (p 5 0.06). These combined evidences in Fig. 2 indicate that the 0.1 mmol L 21 NaNO 3 medium could significantly retard the physicochemical transformation of AgNPs relative to the OECD and EPA media recommended for D. magna acute toxicity exposure.   Table 1, we determined the concentrations of total Ag, free Ag 1 , total Ag 1 and dissolved Ag in solutions/suspensions equivalent to LC50 8-h of AgNO 3 and AgNPs (Table S2 of SI), and further calculated the nano Ag concentration and the proportion of Ag species in total Ag (Table S3 of SI). The measured total Ag at LC50 8-h were 4.68-126.51 mg L 21 for AgNPs and was 0.44 for AgNO 3 (Table S2 of SI), which were generally lower than their corresponding nominal total Ag concentration (Table S1 of SI). This was ascribed to the adsorption of a fringe of Ag species to glass-beakers during the exposure. The LC50 8-h as measured total Ag would be more persuasive than that as nominal total Ag. Hereafter, the total Ag referred to as the measured total Ag in suspension equivalent to LC50 8-h condition, unless otherwise stated. For AgNPs with similar size but different coatings, the measured total Ag at LC50 8-h for AgNP PVP appeared to be relevantly lower than that for AgNP CIT (e.g. 14.01 mg L 21 for AgNP PVP28 vs 36.25 mg L 21 for AgNP CIT20 ). In addition, as the size of AgNPs (with same coating) increased, the total Ag increased as well (Table S2 of SI). For instance, converting the AgNP CIT10 to AgNP CIT100 , total Ag concentrations increased from 5.70 to 126.51 mg L 21 .

Median
The LC50 8-h as free Ag 1 concentration for AgNO 3 was 0.40 6 0.05 mg L 21 . As the measured total Ag should be equivalent to its free Ag 1 for AgNO 3 , we performed the ANOVA to test on them. The result showed no significant difference between the two measured values (p 5 0.27, Table S4 of SI), thus partially proved the credibility of free Ag 1 results determined by ISE method. LC50 8-h as free Ag 1 concentrations in seven AgNPs were almost constant with values in range of 0.37-0.44 mg L 21 (Table S2 of SI). The ANOVA results showed that the free Ag 1 contents in AgNP suspensions had no significant differences against coatings or sizes of AgNPs (p 5 0.91, .0.05, Table S4 of SI). However, they accounted for wide-range proportions of 0.32% to 9.31% of the measured total Ag, and the proportion increased significantly as the size of AgNP increased (Table S3 of SI).
Concentrations of total Ag 1 and dissolved Ag were relatively higher than that of free Ag 1 in identical AgNP suspension, with concentration ranges of 1.05-2.62 mg L 21 and 0.91-1.58 mg L 21 , respectively. Although LC50 8-h as total Ag 1 and dissolved Ag both differed significantly among AgNP suspensions (p = 0.05, Table S4 of SI), neither of them showed size-dependent changes among AgNPs.
Compared with the above Ag species, the nano Ag was the primary portion in AgNP suspension. For AgNPs with the same coating, the proportion of nano Ag generally increased with the size of AgNPs. Take AgNP CIT for instance, the proportion of nano Ag increased from 81.57% to 98.30% as the size of AgNP CIT elevated from 10 nm up to 100 nm.
Relevance of Different Silver Species to the Toxicity towards Daphnia magna. With the above measured concentration of various Ag species in AgNP suspensions equivalent to LC50 8-h , we can present the LC50 8-h values in terms of different Ag species. As shown in Fig. 4, for the seven tested AgNPs, while the LC50 8-h as nano Ag spanned over an order of magnitude, the LC50 8-h as different dissolved Ag species were relatively close to that as Ag 1 in AgNO 3 . To identify the most responsible Ag species to the toxicity of AgNP suspension towards D. magna, ANOVA was performed to test the significances of concentrations of different dissolved Ag species among AgNPs with that from the LC50 8-h values for AgNO 3 (Table S4 of SI). First, the free Ag 1 concentrations in AgNP suspensions at their LC50 8-h concentration levels had no significant differences from the measured free Ag 1 LC50 8-h for AgNO 3 (0.40 mg L 21 as Ag 1 ) either (p 5 0.66, .0.05). For concentrations of total Ag 1 and dissolved Ag, however, the ANOVA results showed that the two Ag species had significant difference from the measured free Ag 1 LC50 8-h for AgNO 3 (p = 0.05).

Discussion
In traditional exposure media, a critical obstacle in the toxicity study of AgNPs was the occurrence of uncontrollable but consistent physical and chemical transformation (e.g. aggregation and dissolution) of AgNPs, due to the high chemical reactivity of Ag 0 with dissolved O 2 , and Ag 1 with anions like Cl 2 . In addition, the relatively high ionic strength of the exposure media also lead to physical changes of AgNPs, which in turn affects the dissolution of AgNPs to release Ag 1 . Therefore, the measured concentrations of Ag species in traditional media at an exact time point cannot profile the concentration during the entire exposure period (as shown in Fig. 2F). However, currently most of the reported LC50 of different Ag species to D. magna or other aquatic organisms in OECD/EPA media were related to concentrations of corresponding Ag species at the initial or end point of 24-h or 48-h exposure, which inevitably causes large artifacts in the measured LC50 values due to the significant changes of state and speciation of AgNPs/Ag 1 , regardless they were presented as total Ag,   6,7,9 , others show free Ag 1 predominant toxicity 10,12 and particle-specific toxicity 13,14 .
In this study, by using 8-h exposure in 0.1 mmol L 21 NaNO 3 medium, the uncertainty of physical changes of AgNPs and their influences on the dissolution of AgNPs during exposure were significantly excluded. The ANOVA results on Ag 1 contents in the representative AgNP CIT20 suspension further showed that, the Ag 1 contents during the 8-h exposure in the 0.1 mmol L 21 NaNO 3 medium had no significant changes (p 5 0.06, .0.05), suggesting the optimized exposure medium in the present study could significantly retard the dissolution and oxidation of AgNPs to Ag 1 during the 8-h exposure procedure. Therefore, this more reliable exposure medium ensured the quantification of Ag species in solutions equivalent to the LC50 8-h conditions, and the identification of the relative importance of various Ag species in AgNP exposure mixtures to the toxicity towards Daphnia magna. This excluded the artifacts in evaluation of LC50, thus unlike the wide span of LC50 value as free Ag 1 (0.57-1.1 mg L 21 ) reported in a previous study 12 , almost the same LC50 value (0.37-0.44 mg L 21 ) as free Ag 1 was obtained for 7 different AgNPs in this study.
Based on the LC50 values (Fig. 4) that was accurately measured with the proposed exposure procedure, it is possible to weight the actual contribution of different Ag species to the toxicity of AgNPs, and therefore distinguish the intrinsic Ag species responsible for the acute toxicity of AgNP suspension to D. magna.
For all the tested seven AgNPs with different coatings and sizes, the values of LC50 8-h as free Ag 1 were very close to that of AgNO 3 , and were lower than that as total Ag 1 and as dissolved Ag. Given that the total Ag 1 covered the free Ag 1 , Ag 1 complexes and absorbed Ag 1 on AgNP surfaces, and the dissolved Ag contained free Ag 1 , Ag 1 complexes and tiny AgNPs, these results thus indicated that the other parts except for free Ag 1 showed no contribution to the acute toxicity, namely free Ag 1 is the exclusive contributor of acute toxicity to D. magna. If the measured LC50 as total Ag 1 or dissolved Ag had been lower than that as free Ag 1 , it would have indicated that the AgNP adsorbed Ag 1 , ligand complexed Ag 1 , and tiny AgNPs contributed to the observed toxicity.
To aquatic organisms, it is generally believed that free metal ions are bioavailable, while other metal species such as metal ions com-plexed with organic ligands, and associated to dissolved organic matter or particles are not bioavailable [53][54][55] . Since AgNPs were reported to be directly available, it is controversial if AgNPs showed particle effects to the acute toxicity to organisms. Our results suggest that particulate AgNPs, even tiny AgNPs (,2 nm) that was found to present in the dissolved Ag (Fig. S2 of SI), showed negligible contribution to the acute toxicity to D. magna. Although the 8-h acute toxicity exposure performed in the present study was relatively shorter than that in traditional experiments, it in fact proved that the free Ag 1 showed highly acute toxicity towards D. magna at a very short term relative to any other Ag species. It is noteworthy that this is the first study elucidating free Ag 1 is exclusively responsible for the acute toxicity of AgNPs to D. magna, though previous studies reported that free Ag 1 dominate the toxicity of AgNPs to bacteria 34,56 .
Numerous studies have reported the influence of size and coating on the toxicity of AgNPs. Difference of 48-h LC50 values among three AgNPs with different coatings 12 and varied sizes 7 were reported. The present study showed that the acute toxicity of AgNPs to D. magna represented as total Ag decreased in the order of AgNO 3 ? AgNP PVP . AgNP CIT , and the toxicity of AgNP CIT decreased with the increase of size. Specifically, the LC50 8-h as total Ag among AgNPs with different sizes and surface coatings varied by a factor of 27. Apparently, this was probably due to the changes of dissolution of different types of AgNPs as influenced by nanoparticle size and surface-coating. For AgNPs with the same coating, the proportion of total Ag 1 generally decreased with the increasing size of AgNPs. Take AgNP CIT for instance, the proportion of total Ag 1 decreased from 18.43% to 1.70% as the size of AgNP CIT elevated from 10 nm to 100 nm. This was in accordance with expectation since the increase of AgNP particle size reduced the relative surface area, thus retarded the release of dissolved Ag species, and exhibit lower toxic responses 20 . For AgNP PVP and AgNP CIT with similar sizes, the proportion of total Ag 1 for AgNP PVP was generally higher than that of AgNP CIT , which could be attributed to the reducibility of sodium-citrate relatively retarded the dissolution of Ag 0 to Ag 1 17 . However, the LC50 8-h as free Ag 1 showed no significant difference among AgNPs and agreed very well with that of AgNO 3 . Together, our results indicate free Ag 1 was the intrinsic and ultimate reason that governs the acute toxicity of AgNPs, while size and surface coating are apparent factors that influence the toxicity through affecting the free Ag 1 concentration of AgNPs.
In summary, uncontrollable but consistent physical and chemical transformation of AgNPs occurred in traditional OECD and EPA exposure media, which hinders the identification of Ag species of AgNPs responsible for the acute toxicity to D. magna. By using 8 h exposure in 0.1 mmol L 21 NaNO 3 medium, the artifacts in determination of LC50 were excluded, and the LC50 8-h as free Ag 1 of all the seven AgNPs were found to be almost the same and agree with that of AgNO 3 . More importantly, with this developed exposure procedure, we found that free Ag 1 is exclusively responsible for the acute toxicity of AgNPs to D. magna, while other Ag species in AgNPs like Ag 1 complexes, Ag 1 absorbed on AgNP surfaces, and nanoparticulate Ag showed no contribution to the acute toxicity. Our results also showed that free Ag 1 was the intrinsic and ultimate factor that governs the acute toxicity of AgNPs, while size and surface coating are apparent factors that influence the toxicity through affecting the free Ag 1 concentration of AgNPs.  PVP with molecular weight of 10,000 Da (PVP10,000) and 58,000 Da (PVP58,000) were purchased from Sigma-Aldrich and Aladdin Chemistry (Shanghai, China), respectively. The FL-70 was obtained from Fisher Scientific (Fair Lawn, NJ). Silver nitrate (AgNO 3 ), sodium thiosulfate (Na 2 S 2 O 3 ), sodium borohydride (NaBH 4 ), and other salts were analytical-reagent grade or above and obtained from Sinopharm Chemical Reagent Beijing (Beijing, China). Ethylene glycol (99 1 % extra pure) was purchased from Acros Organics (Geel, Belgium). Nitric acid (65%) was obtained from Merck (Darmstadt, Germany). Amicon Ultra-15 centrifugal filter (with nominal molecular weight cut-off of 30 kDa) were obtained from Millipore (Darmstadt, Germany). All reagents were used without further purification. Ultrapure water (18 MV cm) produced from a Millipore Milli-Q Gradient system (Billerica, MA) were used to prepare all solutions throughout experiments.

Organisms
Synthesis and Characterization of AgNPs. Besides the 5 commercial AgNP CIT , another two AgNP PVP (AgNP PVP10 and AgNP PVP28 ) were used in the present study. The two AgNP PVP were synthesized following methods from published works 57,58 , and details were shown in SI.
Characterization and Quantification of AgNPs. The TEM observation was preformed on JEM-2100F (JEOL) (see details in SI). The concentrations of AgNPs were quantified by ICP-MS (Agilent 7700, Santa Clara, CA), after room temperature HNO 3 digestion as described in literature 46 . All the AgNP stock suspensions were kept in the dark at 4uC.
Optimization of Acute Toxicity Testing Medium. To exclude the potential aggregation, precipitation and complexation of AgNPs deriving from the artifact of exposure media, NaNO 3 was selected as a candidate for acute toxicity exposure media for AgNPs. To optimize the concentration of NaNO 3 for acute toxicity experiment, the mortality of seven D. magna (6-24 h) in 50 mL media with different NaNO 3 concentrations (0.05, 0.1, 0.5, 1, 5, 10 mmol L 21 ) was monitored at several time intervals during 48 h. The D. magna were not fed in the duration. All the prepared media were aerated for at least 24 h before exposure to D. magna, and the final pH of the media was all adjusted to 7.8 6 0.2 using NaOH. The additional ionic strength by NaOH was no more than 0.05 mmol L 21 , thus was assumed to be no additive effect on D. magna and AgNPs. Mortality experiments at each concentration point were repeated six times, and mortalities in each sample were assessed at time intervals of 4, 8, 14, 24, 36 and 48 h. Finally, the 0.1 mmol L 21 NaNO 3 medium (pH 7.8 6 0.2 adjusted by NaOH) was selected for a 8-h acute toxicity of AgNPs to D. magna.

Characterization of AgNPs in Different Exposure Media.
To compare the effect of the optimized exposure medium with traditional media on the stability of AgNP suspensions, a representative AgNPs (AgNP CIT20 , 4 mg L 21 as total Ag) were prepared in OECD SM4 medium (which included most components of M4 59 but excluded CoCl 2 , KI, Na 2 SeO 3 , NH 4 VO 3 and Vitamins), EPA hard water 60 , 0.1 mmol L 21 NaNO 3 medium and ultrapure water, respectively. The total Ag 1 release from AgNP CIT20 in these media were monitored during the 8-h exposure. After 8-h exposure, the characterizations of AgNP CIT20 , including UV-vis spectra (UV-3600 Spectrometer, Shimadzu, Japan), TEM imaging (preformed on H-7500 (Hitachi)), hydrodynamic sizes (Dynamic Light Scattering, DLS, Malvern, UK) and photographing of these samples were performed. TEM samples were prepared by loading 10 mL aliquots of suspensions onto ultrathin carbon-coated copper grid, and drying at room temperature under vacuum.
Acute Toxicity Test of AgNPs to Daphnia magna. For acute toxicity test, seven D. magna (6-24 h old) were placed in 50-mL glass beakers containing 50 mL aliquots of the optimized NaNO 3 exposure medium containing different concentrations (as total Ag) of each AgNPs. Triplicates of seven (or more) concentrations and medium-only control were tested for Ag 1 (in AgNO 3 ) and each AgNPs. Exposures were conducted at ,25uC in light for 8 h without feeding D. magna. Mortalities were assessed at the end of 8-h exposure. The LC50 8-h , including 95% confidence intervals, were determined using SPSS 16.0 software applying the probit regression. The calculated LC50 8-h represented the nominal total Ag concentration, and concentration range of lethal effect (1% and 99%) were presented in Table S1 of SI.
Determination of Ag Species in AgNP Suspension. The Ag species measurements, including total Ag, free Ag 1 , total Ag 1 and dissolved Ag, were all conducted in new suspensions equivalent to the LC50 8-h of AgNO 3 and AgNPs (as nominal total Ag), which were all prepared in 0.1 mmol L 21 NaNO 3 medium. Definition and analytical methods used for the Ag species were listed in Table 1. The detailed determination procedures are as follows: Determination of Total Ag by ICP-MS after Digestion. Concentrations of total Ag in suspensions equivalent to LC50 8-h (as nominal total Ag) of AgNO 3 and AgNPs were measured using ICP-MS after room temperature HNO 3 digestion 46,61 . Briefly, 0.5 mL AgNP sample was first digested with 0.5 mL 65% HNO 3 for 15 min, and then diluted and mixed with 10 mL ultrapure water for ICP-MS measurement. The digestionquantification procedure was repeated in triplicates for each sample. The calibration curve (with linear correlation coefficient, R 2 . 0.999) was obtained from a series of standards of 0.1, 0.2, 0.5, 1, 2, 5, 10 and 20 mg L 21 Ag 1 prepared in 5% HNO 3 .
Determination of Free Ag 1 by ISE. Concentrations of free Ag 1 in AgNPs at LC50 8-h (as nominal total Ag) were measured using a silver/sulfide ISE (Van-London pHoenix, Houston, TX). The standards for ISE measurement, with nominal concentrations in range of 0.2 to 5.0 mg L 21 of Ag 1 (in AgNO 3 ), were prepared in 0.1 mmol L 21 NaNO 3 medium in 50-mL glass beakers. Given the predictable sorption of Ag 1 to glass beakers after the preparation standard solutions would result in deviations of calibration, the Ag 1 concentrations of ISE standards were determined by ICP-MS simultaneously with the ISE standards' measurement ( Fig. S3 of SI). The ICP-MS detected concentrations (in range of 0.18-3.99 mg L 21 ) were plotted against potential differences to prepare the calibration curve of ISE for measuring the free Ag 1 in samples of AgNPs and AgNO 3 . A linear calibration obtained over the whole concentration range with slope of 50.7 mV was represented in Fig. S4 of SI. Free Ag 1 measurements for each AgNP suspension were conducted at least in triplicates. The electrode was carefully washed by ultrapure water after each measurement, and dried off by dust-free paper prior to the next measurement.
Determination of Total Ag 1 by LC-ICP-MS. Concentrations of total Ag 1 in AgNPs at LC50 8-h (as nominal total Ag) were measured using on-line coupled LC (Agilent 1200, Santa Clara, CA) and ICP-MS (Agilent 7700). This LC-ICP-MS was recently established by our group for speciation analysis of nanoparticulate Ag (1-100 nm) from dissoluble Ag 1 , which was defined to includes free Ag 1 , Ag 1 complexed with ligands, Ag 1 adsorbed on the AgNP surface, and originally undissolved Ag(I) salts that can be dissolved in the presence of Na 2 S 2 O 3 46 . For the simple AgNP exposure mixture used in this study, the dissoluble Ag 1 equals to the total Ag 1 . Briefly, the LC separation of total Ag 1 from the AgNP sample (20 mL) was performed with an Venusil Durashell-NH 2 amino column (250 3 4.6 mm I.D., 5 mm particle size, 500 Å pore size, Bonna-Agela Technologies Inc., Tianjin, China) and a mobile phase of 0.1% (v/v) FL-70 (a surfactant) and 2 mmol L 21 Na 2 S 2 O 3 in ultrapure water at a flow rate of 0.7 mL min 21 . The total Ag 1 was quantified by the coupled ICP-MS system. For the measurements of total Ag 1 in AgNPs at LC50 8-h , into the AgNP samples were added 2 mmol L 21 Na 2 S 2 O 3 just before injection into the LC-ICP-MS system, with each sampling conducted in triplicates. The calibration curve prepared by the AgNO 3 standards of 0.1, 0.2, 0.5, 1, 2 and 5 mg L 21 Ag 1 in 2 mmol L 21 Na 2 S 2 O 3 shows good linear correlation of R 2 . 0.997. Given the excellent recovery of 87.4 6 5.2% for total Ag 146 , the total Ag 1 values determined by LC-ICP-MS method in this study were without recovery correction.
Determination of Dissolved Ag by Centrifugal Ultrafiltration and ICP-MS. The concentrations of dissolved Ag in AgNPs at LC50 8-h (as nominal total Ag) were measured by centrifugal ultrafiltration off-line coupled with ICP-MS (UF&ICP-MS). The separation of dissolved Ag from AgNPs followed the method in our previous work 27 , which was modified from literature reported by Li and Lenhart 21 . Briefly, 10 mL aliquot of prepared sample was added to centrifugal ultrafilter devices (Amicon Ultra-15, 30 kDa), and the dissolved Ag was separated from AgNPs after centrifugation at 5000 rpm for 20 min. The dissolved Ag in the filtrate was collected and mixed with 0.2 mL 65% HNO 3 for ICP-MS analysis. The quantification was conducted with calibration curve of Ag 1 ranging from 0.1 to 5 mg L 21 with R 2 . 0.997. The recovery of Ag through the 30-kDa membrane was 66.6 6 14.6%, obtained by testing 5 mL AgNO 3 solution at concentrations of Ag 1 as 1.0 and 5.0 mg L 21 in 0.1 mmol L 21 NaNO 3 with the same sample preparation procedure. The dissolved Ag values presented here were all with recovery correction, and each measurement was conducted in triplicates.
To test the existence of tiny particles in dissolved Ag, 30-kDa centrifugal ultrafiltration off-line coupled with LC-ICP-MS were performed on AgNP PVP10 at appropriate concentration. Fig. S2 of SI shows the presence of tiny particles in the filtrate of AgNP with particle size of ,1 nm, which was estimated according to the retention times of AgNP with different sizes following our previous method 46 .