Environmental Toxicology and Chemistry, Vol. 34, No. 5, pp. 1015–1022, 2015 # 2015 SETAC Printed in the USA
A PRACTICAL APPROACH TO DETERMINE DOSE METRICS FOR NANOMATERIALS CHRISTIAAN J.E. DELMAAR,y WILLIE J.G.M. PEIJNENBURG,*yz AGNES G. OOMEN,y JINGWEN CHEN,x WIM H. DE JONG,y ADRIËNNE J.A.M. SIPS,y ZHUANG WANG,x and MARGRIET V.D.Z. PARKy yNational Institute for Public Health and the Environment, Bilthoven, The Netherlands zInstitute of Environmental Sciences, Faculty of Science, University Leiden, Leiden, The Netherlands xDalian University of Technology, School of Environmental Science and Technology, Dailian, China (Submitted 24 June 2014; Returned for Revision 14 December 2014; Accepted 31 December 2014) Abstract: Traditionally, administered mass is used to describe doses of conventional chemical substances in toxicity studies. For deriving toxic doses of nanomaterials, mass and chemical composition alone may not adequately describe the dose, because particles with the same chemical composition can have completely different toxic mass doses depending on properties such as particle size. Other dose metrics such as particle number, volume, or surface area have been suggested, but consensus is lacking. The discussion regarding the most adequate dose metric for nanomaterials clearly needs a systematic, unbiased approach to determine the most appropriate dose metric for nanomaterials. In the present study, the authors propose such an approach and apply it to results from in vitro and in vivo experiments with silver and silica nanomaterials. The proposed approach is shown to provide a convenient tool to systematically investigate and interpret dose metrics of nanomaterials. Recommendations for study designs aimed at investigating dose metrics are provided. Environ Toxicol Chem 2015;34:1015–1022. # 2015 SETAC Keywords: Nanoparticles
Dose metrics
Risk assessment
Nanomaterials
Dose–response modeling
complex materials such as NPs is inappropriate, but rather should include intrinsic parameters such as surface reactivity. The purpose of the present study is to introduce an approach to assess systematically what dose metric should be used in (eco) toxicological studies with NPs. The approach can be applied to both in vitro and in vivo data, provided these data meet certain criteria as outlined below in General method description. To illustrate, the approach is applied to published data from in vitro experiments with silica (SiO2) and silver (Ag) NPs of different sizes in cell systems [1,7] and to data from experiments with AgNPs and aquatic organisms of 3 trophic levels (algae, zooplankton, and fish) [8]. Based on approaches that are common in mixture toxicity assessment, the contribution of silver ions to the toxicity of the AgNPs suspensions tested was quantified to properly quantify the toxicity of the NPs tested [8]. We will also discuss how this method can be applied to study the use of alternative or additional nanomaterial properties, such as surface reactivity, in the dose metric, at the same time recognizing the possibility of dose metrics being dependent on endpoint tested and on NPs investigated.
INTRODUCTION
In standardized toxicity studies using conventional chemical substances, increasing doses are administered to groups of organisms to determine at what dose the substance under consideration causes adverse effects. For most conventional chemical substances, the dose metric in these studies is expressed as the total administered mass of the substance, because the mass of a substance is related uniquely to the number of molecules of that substance. Consequently, risk assessors generally use the mass of a substance as the dose metric to express estimated, safe, or maximum exposure levels. Generally, a mass concentration alone is sufficient to describe the toxic or safe dose of chemically similar substances in a given exposure scenario. With nanotechnology creating novel, complex structures of different sizes, an entirely new array of materials has emerged. For insoluble or partly soluble nanoparticles (NPs), the mass is not uniquely related to the number of NPs. Furthermore, factors other than chemical composition alone, such as particle size, have been shown to affect toxicity profiles in experimental studies. As a simple example, at equal mass concentrations, silver nanoparticles of 20 nm were at least 10 times more cytotoxic to peritoneal macrophages than 113 nm silver nanoparticles, despite their similar chemical composition [1]. It is now well recognized that for NPs, the administered mass of the chemical substance alone is not enough to describe the dose of NPs. Alternative dose metrics to replace mass have been suggested, including volume, surface area, or particle number [2–6], but consensus among researchers and risk assessors on a universal yet simple dose metric for NPs is lacking. There is also a real possibility that using such a simple dose metric for
METHODS
General method description
The method to derive an appropriate dose metric requires dose–response data from a range of solid, mono-sized NPs of similar chemical composition. These NPs must vary in diameter and be tested in a single experimental system, preferably from multiple and independent experiments. Dose–response data from these experiments must subsequently be analyzed to determine the number of NPs per unit of test medium volume needed to induce a predefined response, R1, for each of the investigated NPs. We refer to these doses as “equi-response doses.” For example, for silver NPs, the number of particles needed to induce a 20% decrease in cell metabolism (R1) is N1 particles/mL for particles with diameter d1 and N2 particles/mL
* Address correspondence to
[email protected]. Published online 7 January 2015 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/etc.2878 1015
1016
Environ Toxicol Chem 34, 2015
W.J.G.M. Peijnenburg et al.
Figure 1. The equi-response doses for 3 nanoparticles of different diameter.
for particles with diameter d2 (Figure 1). Then, the doses “N1 particles/mL of diameter d1” and “N2 particles/mL of diameter d2” are equi-response doses. The relationship between these equi-response doses can be used to identify the appropriate dose metric. For this, we impose the requirement that equi-response doses should be equal across NPs meeting the specific requirements given above, provided the doses are expressed in the appropriate dose metric. To illustrate this graphically, a dose of NPs can be represented in a plot as a point in the plane spanned by the dose parameters “number of particles N” and “diameter d” as its axes (Figure 2). The equi-response doses of the tested range of NPs can be plotted in this plane and can be connected to construct a curve on which the response R1 is equal (equiresponse curve). Along this curve, the doses associated with this response, R1, should be equal when expressed in the appropriate dose metric. This dose metric can be found by observing that the curve can be described mathematically as S(N,d) ¼ constant. Thus, S(N,d) is the required dose metric, expressed as a function of the parameters N and d. Frequently suggested dose metrics are administered surface area, volume, or number of particles. If any of these dose metrics are adequate dose descriptors, the equi-response curves will take on specific shapes, as specified in the next paragraphs.
If administered surface area is an adequate dose metric, a large number of small spheres will give the same response as a smaller number of larger spheres with the same total surface area (Figure 3). In other words, the equi-response curve in a (N, d) plot describes a constant surface area in the (N,d) plot. The surface area s of a single sphere is given by the equation s ¼ p d2
ð1Þ
where d is the diameter of the sphere. It follows that the curve for constant total surface area S of N particles is S ¼ N p d2 ¼ constant
ð2Þ
thus, N¼
c1 d2
ð3Þ
where c1 is a particular constant that in equi-response curves depends on the response level of the test. Taking the base-10 logarithm of both sides of the equation allows for a more convenient graphical analysis of the curve expressed as either logðNÞ ¼ logðc1 Þ 2 logðdÞ
ð4Þ
logðNÞ ¼ c2 2 logðdÞ
ð5Þ
or
It follows that for NPs for which surface area is an adequate dose metric in a specific test system, on the logarithmic scale, the equi-response curve has a slope equal to –2 and an offset that reflects the intrinsic toxicity of the NPs tested. An equi-response curve can be obtained in a similar way for the case in which volume is an adequate dose metric. The volume v of a sphere is equal to v¼p Figure 2. Plotting equi-response doses in a Log(N) versus Log(d) plot.
d3 6
ð6Þ
Dose metrics for nanomaterials
Environ Toxicol Chem 34, 2015
1 x1011
number of particles
8 x1010
S1
N1
6 x1010
4 x1010 V1 2 x1010
0 0
20
40
60 diameter [nm]
80
100
The constant total volume, V, of N particles is given by d3 ¼ constant 6
ð7Þ
thus, logðNÞ ¼ c4 3 logðdÞ
log(d) plot derived from experimental data equals –2, then surface area adequately describes the dose. If, on the other hand, the slope equals –3, volume will adequately describe the dose. And, if it equals 0, the adequate dose description is the total number of particles. As a matter of course, the possibility exists that the slope of the equi-response curve may not equal –2, –3, or 0 at all. In this case, the function of the curve may still be used to derive an adequate dose metric (see Identifying the dose metric). The method described was applied to 8 dose–response data sets obtained from various in vitro test systems with SiO2 and AgNPs [1,7] and from test systems with AgNPs and 3 different aquatic organisms (algae, zooplankton, and fish) [8]. Because the NPs were mono-sized, spherical, and had the same surface characteristics within a series of a particular NP, the general approach described in the present study can be used to determine the adequate dose metric and verify if any of the dose metrics’ total number of particles, total particle surface area, or total particle volume is appropriate. Experimental data
Figure 3. Curves for constant surface area, volume, and number of particles based on a specific dose of N particles of diameter d. For example, curve S1 shows that 1 1011 particles with a diameter of 50 nm and 4 1010 particles with a diameter of 80 nm have the same surface area, S1. Similarly, the same volume V1 can be obtained with approximately 1 1011 particles with a diameter of 25 nm and with approximately 2 1010 particles with a diameter of 40 nm. The constant particle number N1 (in this case, 6 1010 particles), however, is independent of the diameter of the particles. The existence of a simplified dose metric-like administered surface area S, administered volume V and total number of particles administered N for nanoparticles would lead to equi-response curves of constant S, V, or N. Note that if such equi-response curves exist, this would reduce the number of parameters required to describe the dose from 2 (N and d) to 1 (S, V, or just N).
V ¼Np
1017
ð8Þ
It follows that the equation describing the curve on a logarithmic scale for constant volume has a slope of –3. Finally, the curve of constant particle number N is independent of diameter d and is given by N ¼ constant
ð9Þ
logðNÞ ¼ c5
ð10Þ
or
Hence, the equation describing the curve for constant particle number has a slope of 0 in a log(N) versus log(d) plot; in other words, it is a horizontal line. To summarize, the equi-response curve, constructed from dose–response data, can be used to determine the appropriate dose metric. If the equi-response curve is used to test whether any of the simple dose metrics, administered surface area, volume, or number of particles is appropriate, the equirespsonse curve on a double log scale should be a straight line. If the slope of the equi-response curve in the log(N) versus
Experimental dose–response data sets were obtained from previously reported test systems with SiO2 and AgNPs and cell lines [1,7] and from test systems with AgNPs and aquatic organisms of 3 trophic levels (algae, daphnid, and fish) [8] (Table 1). For the experiments in cell lines, 4 kinds of freshly prepared SiO2 NPs (primary size 11 nm, 248 nm, and 2–34 nm particles differing in zeta potential) and 3 kinds of phosphate buffered Ag NPs (primary size 20 nm, 80 nm, and 110 nm provided by NanoComposix) were used. The manufacturing process, transmission electron microscopy images, zeta potential measurements, and hydrodynamic diameters in water and cell culture medium have been reported in Park et al. [1,7]. For the metabolic activity and membrane integrity assays, mouse peritoneal macrophages (RAW 264.7) or embryonic fibroblasts (L929) were exposed to freshly prepared suspensions of particles in cell culture medium for 24 h. Metabolic activity of the cells was determined using the WST-1 assay, whereas membrane integrity was measured by determining the amount of lactate dehydrogenase leaking out of the cells [1,7]. For the induction of reactive oxygen species (ROS), RAW 264.7 macrophages were exposed to particles for 4 h and generating ROS was determined using a dichlorofluorescein probe as reported in Park et al. [1,7]. Inhibition of stem cell differentiation was measured using the standard embryonic stem cell test, during which mouse embryonic stem cells were exposed to particles for 10 consecutive days [1,7]. Four kinds of AgNPs with different original sizes were selected for the experiments with aquatic organisms and were either provided by the Joint Research Centre of the European Union (EU) as dispersions of polyoxyethylene glycerol trioleate and polyoxyethylene sorbitan mono-laurat (15 nm AgNPs). Alternatively, they were purchased as powders (20 nm, 35 nm, and 80 nm) from Nanostructured and Amorphous Materials. Stocks of AgNPs were prepared fresh before each batch test as Bradford et al. [9] described. The NPs were prepared in Pyrex bottles protected from light using 0.01 M MOPS (3-morpholino propane sulfonic acid, buffer at pH 7.0) in a reconstituted medium consisting of 1.36 mM Ca(NO3)2, 0.73 mM Mg(NO3)2, 1.19 mM NaNO3, and 0.20 mM KNO3. For each batch of toxicity tests with aquatic organisms, a single control for each test concentration was included. For toxicity testing of Raphidocelis subcapitata, the exposure time was 4.5 h. The inhibition of the photosynthetic efficiency was
1018
Environ Toxicol Chem 34, 2015
W.J.G.M. Peijnenburg et al.
Table 1. Data from various experimental test systems (cell systems and aquatic organisms) used to derive equi-response curvesa ECx [Number of particles 1010/mL] (Confidence interval)
Particle diameter [nm] Reduction of metabolic activity of RAW 264.7 macrophages by SiO2 NP (EC20) 11 nm 34 nmb 34 nmc 248 nm Reduction of metabolic activity of L929 fibroblasts by AgNP (EC20) 20 nm 80 nm 113 nm Induction of reactive oxygen species in RAW 264.7 macrophages by AgNP (EC20) 20 nm 80 nm 113 nm Inhibition of embryonic stem cell (D3) differentiation by AgNP (EC20) 20 nm 80 nm 113 nm Decrease of membrane integrity in L929 fibroblasts by AgNP (EC20) 20 nm 80 nm 113 nm Inhibition of photosynthetic efficiency of Raphidocelis subcapitata by AgNP (EC40) 15 nm 20 nm 80 nm Immobility of Chydorus sphaericus by AgNP (EC40) 15 nm 20 nm 35 nm 80 nm Teratogenicity in Danio rerio by AgNP (EC40) 15 nm 20 nm 35 nm 80 nm
1500 mL (1174–1957 mL) 16 mL (14–19 mL) 27 mL (24–29 mL) 0.13 mL (0.11–0.17 mL) 6.4 mL (6.1–6.6 mL) 0.60 mL (0.53–0.64 mL) 0.15 mL (0.14–0.16 mL) 57 mL (50–64 mL) 3.6 mL (3.2–4.1 mL) 1.5 mL (1.3–1.8 mL) 27 mL (23–30 mL) 1.1 mL (0.9–1.3 mL) 0.37 mL (0.33–0.40 mL) 0.46 mL (0.30–0.64 mL) 0.19 mL (0.082–0.28 mL) 0.030 mL (0.020–0.039 mL) 1.91 mL (1.62–2.25 mL) 1.30 mL (1.06–1.46 mL) 0.38 mL (0.31–0.43 mL) 0.248 mL (0.248–0.248 mL) 0.177 mL (0.173–0.181 mL) 0.094 mL (0.073–0.111 mL) 0.015 mL (0.05–0.016 mL) 0.714 mL 0.790 mL 0.269 mL 0.064 mL
(0.629–0.724 mL) (0.773–0.805 mL) (0.219–0.290 mL) (0.056–0.070 mL)
a Each equi-response level (based on effective concentrations [EC] inducing a 20% and 40% response compared to control for cell systems and for aquatic organisms, respectively [EC20, EC40]) and its confidence interval was obtained from dose–response models fit to data from multiple experiments. b Data from a duplicate experiment using particles of the same diameter. c Data from a duplicate experiment using particles of the same diameter.
determined using a pulse-amplitude modulation fluorometer (Walz; measuring head WATER-PAM/F) and quantified by recoding the changes in the fluorescence signal. The test was replicated 4 times. The Chydotox testing for the freshwater zooplankton, Chydorus sphaericus, was used as a promising alternative for the existing Daphnia sp. 48 h acute immobility test [10,11]. Twenty newborn neonates (< 24 h old) were divided into 4 batches each in vials at room temperature. After 48 h, the vials were placed under a reverse dissecting microscope, and immobility was determined by activating the animals by slightly shaking the vial and monitoring them for 30 s. The zebrafish (Danio rerio) embryo test was performed according to a procedure Lammer et al. [12] described. The control and the Ag nanoparticles were tested with 10 eggs/ embryos (4 cell stage–64 cell stage) selected from 20 healthy appearing, fertilized eggs. The selected embryos were transferred into 24-well cell culture plates to incubate for 96 h (26 18C). The embryos were scored for severity of toxicity as follows: 0 ¼ normal, no toxic response; 1 ¼ minor, 1 or 2 toxic endpoints; 2 ¼ moderate, 2 or 3 toxic endpoints; 3 ¼ severe, more than 3 toxic endpoints; 4 ¼ dead. The development status of the embryos and larvae was observed microscopically every day, and sublethal developmental morphology and teratogenicity endpoints were scored after 24 h, 48 h, 72 h, and 96 h. The embryo test was performed in 3 independent experiments.
Equi-response concentrations were derived by dose– response analysis on data from the experimental systems described. The entire dose–response curve was used to determine specific effect concentrations and their 95% confidence intervals. In the present examples, 20% and 40% change in the endpoint compared with the control for both the cell systems and the aquatic organisms was defined as the response of interest for each test system (Table 1). For practical reasons, experiments with NPs generally are performed using mass as an initial dose metric. Because both SiO2 and AgNPs were spherical, with known diameter d, mass density r, and a relatively narrow primary size distribution (standard deviation < 10%), the nominal massbased ECx equi-response concentrations, specified in the experiments as mg/mL, can be converted to a number-based concentration CN as CN ¼ CM
6 prd3
ð11Þ
To construct equi-response curves for each test system, the logarithm of the number-based equi-response concentration was plotted against the logarithm of the particle’s diameter for each NP tested. To test whether a simple dose metric such as
Dose metrics for nanomaterials
Environ Toxicol Chem 34, 2015
log(particle number)
14 13 12 11 10 y =-2.9x + 16.0 R2 = 0.98
9 8 0.5
1
1.5 2 log(diameter) [nm]
2.5
Figure 4. Equi-response curve for SiO2 nanoparticles (11 nm, 34 nm, and 248 nm) representing a 20% reduction in metabolic activity of RAW264.7 macrophages compared with control.
administered surface area or volume adequately describes the data, the linear equation logðNÞ ¼ a logðdÞ þ b
ð12Þ
was fit to the data using a least squares minimization principle. A bootstrap procedure was used to estimate the uncertainty in the slope of the equi-response curves (i.e., the parameter a). Particle number, surface area, and volume were considered adequate dose metrics if the 95% confidence interval around the slope a of the equi-response curve included 0, –2, and –3, respectively.
1019
particles, of similar primary sizes, they have a constant density. This implies that the dose in this case may also be expressed as total administered mass. In contrast, both the dose–response relation for AgNPs in the test systems for metabolic activity in L929 fibroblasts (a ¼–2.05 [ 0.05]) and induction of ROS in RAW 264.7 macrophages (a ¼–2.1 [ 0.1]) are best described by specifying dose in terms of administered surface area (Figure 5A and 5B, respectively). For the aquatic organisms, the dose of AgNPs affecting the photosynthesis efficiency of algae (Raphidocelis subcapitata) was also best described by surface area, whereas for both teratogenic effects in zebrafish (Danio rerio) and immobilization of the zooplankton (Chydorus sphaericus), volume (or mass) was the most appropriate dose metric (Figure 6). The slope a of –2.4 ( 0.1) of the dose–response relation of AgNPs for inhibiting differentiation in embryonic stem cells indicated that the dose can be described by a parameter related to the diameter of the particles, but not administered surface area or volume (Figure 7). To illustrate what this type of dose metric means in practice when deriving toxic doses, we consider the response of AgNPs inhibiting embryonic stem cell differentiation by 20% compared with control. The regression line to the logarithmic 20% response curve in the log(d), log(N) plot, derived from the embryonic stem cell test experiments with the 20 nm, 80 nm, and 113 nm particles had any of the equations y ¼ 2:4x þ 14:6
ð13Þ
logðNÞ ¼ 2:4 logðdÞ þ 14:6
ð14Þ
N ¼ d2:4 1014:6
ð15Þ
RESULTS
Identifying the dose metric
For all but 1 test system, regression analysis suggested a linear relation between log(N) and log(d) (R2 of > 0.95; Figures 4–8). The slope of the linear regression lines (parameter a) for the experimental systems gives an indication on what the appropriate dose metric may be (Table 2). The observation of the slope a being equal to –2.9 ( 0.1) in the dose–response relation for the test system for metabolic activity in RAW 264.7 macrophages suggests that the dose of SiO2 NPs is best expressed as administered volume (Figure 4). Because the administered NPs consisted of solid, nonporous spherical
In words, for NPs consisting of similar Ag particles with any distribution of sizes, a suitable dose metric for inhibiting stem cell differentiation is N d2.4. Based on the size distribution of the NP of interest, we can convert every fraction of N particles with diameter d to the dose metric N d2.4 and determine its sum, that is, the dose of NPs in N d2.4. The dose that will then inhibit embryonic stem cell differentiation by 20% is 1014.6, that is [N d2.4]. For the test system measuring membrane integrity of L929 fibroblasts, the data did not seem to fit a linear regression line very well (R2 of 0.75). This may indicate that the dose metrics of administered surface area, volume, or number of particles are 12
12
b log(particle number)
log(particle number)
a 11
10 y = -2.0x + 13.5 R2 = 0.97
9
8
11
10
y = -2.1x + 14.5 R2 = 1.00
9
8 1
1.5
2
log(diameter) [nm]
2.5
1
1.5 2 log(diameter) [nm]
2.5
Figure 5. Equi-response curves for Ag nanoparticles (20 nm, 80 nm, and 113 nm) for metabolic activity in L929 fibroblasts (A) and induction of reactive oxygen species in RAW264.7 macrophages (B). The equi-response curves represent an effect level of 20% compared with control.
1020
Environ Toxicol Chem 34, 2015
W.J.G.M. Peijnenburg et al.
12
10.5 log(particle number)
log(particle number)
11
10 9.5 y = -2.1x + 13.1 R 2 = 0.99
9
11
10
y = -2.4x + 14.6 R2 = 1.00
9
8.5 1
1.5
2 8
log(diameter) [nm]
1
9
1.5 2 log(diameter) [nm]
2.5
Figure 7. Equi-response curve for Ag nanoparticles (20 nm, 80 nm, and 113 nm) representing a 20% inhibition of stem cells compared with control.
log(particle number)
8.5 8 7.5 7
y = -2.9x + 12.2 R2 = 0.98
6.5 6 1
1.5
2
log(diameter) [nm]
10 log(particle number)
9.5 9 8.5 8
y = -3.1x + 13.5 R2 = 0.96
7.5
risk assessments. To date, a systematic evaluation of what such a dose metric may be is lacking [6]. Research on dose metrics has been performed for isolated cases only and has used ad hoc methods. Consequently, conclusions cannot be compared or generalized to other materials and other experimental systems easily. The approach to identify appropriate dose metrics introduced in the present paper provides a method to study dose metrics in a systematic way. It allows an unbiased assessment of whether specific scenarios (i.e., “the response in this experimental system is determined by administered surface area/ volume/number of particles”) are to be accepted or rejected, quantified by the confidence interval of the slope of a regression line. Moreover, it can identify alternative adequate dose metrics as well; that is, dose metrics other than surface area, volume, or particle number. This offers a great advantage compared with the more generally used method of placing different dose metrics on the axes of dose–response plots and visually determining which dose metric causes the dose–response curves to overlap. The approach was illustrated by applying it to data from in vitro and in vivo experimental systems with spherical SiO2 and AgNPs. In 7 of 8 cases, the observations indicated the existence
7
12
6.5 1.5
2
log(diameter) [nm]
Figure 6. Equi-response curves for Ag nanoparticles (15 nm, 20 nm, 35 nm, and 80 nm) for photosynthesis by Raphidocelis subcapitata (A), the immobilization of Chydorus sphaericus (B), and morphology and teratogenicity effects of Danio rerio (C). The equi-response curves represent an effect level of 40% compared with control.
not adequate for this test system (Figure 8). To draw more robust conclusions, more data (i.e., measurements on additional NPs of similar chemical composition and morphology) are needed. DISCUSSION
Identifying an appropriate dose metric for NPs is of utmost importance to interpret results from toxicity experiments and for
log(particle number)
1
11
10
9 y = -1.3x + 11.4 R2 = 0.75 8 1
1.5 2 log(diameter) [nm]
2.5
Figure 8. Equi-response curve for Ag nanoparticles (20 nm, 80 nm, and 113 nm) representing a 20% decrease in membrane integrity of L929 fibroblasts compared with control.
Dose metrics for nanomaterials
Environ Toxicol Chem 34, 2015
1021
Table 2. Summary of the slope a of the equi-response curves in a log(N) versus log(d) plot, plus 95% confidence intervals, determined from the 90% confidence intervals of the respective dose–response models by bootstrapping Nanomaterial SiO2 Ag Ag Ag Ag Ag Ag Ag
Test system
R2
Metabolic activity of RAW264.7 macrophages Metabolic activity of L929 fibroblasts Reactive oxygen species in RAW 264.7 macrophages Embryonic stem cell test Membrane integrity of L929 fibroblasts Photosynthesis in R. subcapitata Immobilization in C. sphaericus Teratogenicity in D. rerio
0.98 0.97 1.00 1.00 0.75 0.99 0.98 0.96
of a simplified dose metric related to particle diameter. In 6 cases, the dose metric required only a single parameter (either surface area or volume) to adequately describe the dose of the NP considered. Even in the relatively simple in vitro test systems, the dose metric differed among all of the NPs and systems considered, indicating that the use of a single dose metric for all NPs and for all relevant toxicological endpoints in all experimental systems is probably not a realistic scenario. In 1 case (inhibiting stem cell differentiation by silver nanoparticles), surface area, particle number, and volume were all inadequate dose metrics. Instead, the dose metric required 2 parameters (number of particles N and diameter d), in the form of N d2.4. Although this dose metric is unusual, we have demonstrated that even this type of dose metric can be used for defining doses of NPs. This type of dose metric may be the result of more than a single NP property determining the effect. Indeed, the doses were calculated based on administered doses rather than the actual cellular exposures. Cellular exposure and uptake may increase with the diameter of the particle, whereas the toxicity of the particle may increase in relation to its surface area. The properties that drive the transfer to the site of action, often referred to as kinetics and fate, may differ from the properties driving the toxicity of the material that is at the site of action. Clearly, this combination of NP properties determining the ultimate effects will be even more complex with in vivo test systems. The method can also be applied using dose–response data based on cellular rather than nominal concentrations; for example, as calculated using the model Hinderliter et al. [13] proposed or bioavailable concentrations in target tissues in the body or in target species in the environment. A dose metric derived based on these local bioavailable concentrations may be entirely different than a dose metric based on actual exposure concentrations and intrinsic and extrinsic particle properties. The principle of comparing the equi-response curve to the curves of constant surface area or volume can also be applied to NPs with shapes other than spheres, such as cylinders and cubes, but the slopes of the constant curves would differ. In addition, it is likely that in many cases additional NP properties other than size are required to adequately describe the dose metric. For these cases, continuous parameters such as surface charge, reactivity, and aspect ratio can be analyzed rather than or in addition to diameter in the graphical method. When the number of NP characteristics required to describe a dose adequately is 3, equi-response surfaces can be constructed using an extra axis in the plot. If more than 3 NP characteristics are required to describe the dose, the principle of constructing equi-response surfaces and analyzing the mathematical relation between doses on these surfaces is still an appropriate method to identify an
a –2.91 –2.05 –2.06 –2.41 –1.34 –2.13 –2.94 –3.11
(–3.03 (–2.11 (–2.19 (–2.53 (–1.71 (–2.50 (–3.00 (–3.25
to to to to to to to to
–2.79) –2.00) –1.93) –2.29) –0.98) –1.83) –2.87) –2.96)
adequate metric for the dose. However, this cannot easily be demonstrated in a graphical way. Unfortunately, far too little systematic data currently exists to illustrate the use of the approach for NP properties other than diameter. Combining data from different studies is problematic, because a slight change in protocol can have a profound impact on the results. Fortunately, many large projects such as those funded by the EU’s Framework Program are aware of this and are making significant efforts to standardize testing protocols. It is likely that the data originating from these projects in the near future can be used to study dose metrics further by applying the proposed approach. Identifying an adequate dose metric, especially in more complex cases, will always benefit from including equiresponse data from a wide range of NPs that vary in one characteristic and equi-response levels with narrow confidence intervals. Equi-response levels in these experimental systems need to be derived using adequate dose–response analysis, which will generally benefit from including more dose levels and experiment replicates. More specifically, studies with at least 4 different variations in an NP characteristic (e.g., size) are preferred to verify whether the equi-response levels fit a straight equi-response curve. Another issue to consider when designing experiments to analyze dose metrics is that in practice, most NPs will consist of particles with a distribution of characteristics (e.g., particle diameter) rather than having a single, well-defined mean value for this characteristic. Preferably, for the purpose of studying dose metrics, distributions of particle characteristics within an NP should be as narrow as possible. As such, NP aggregation and agglomeration in solution will impact the actual available dose for biota and hence the bioavailability of NPs in solution. The issues of particle size distribution and effective bioavailability require extending the methodological framework proposed, which is feasible when more mechanistic information is available on the factors driving N toxicity. From a mathematical perspective, there clearly are no limitations in this respect. In conclusion, identifying adequate dose metrics for NPs is highly relevant for interpreting results from toxicity studies, for risk assessment, and for regulatory purposes. The approach depicted in the present study provides an unbiased way to identify an appropriate dose metric for a particular experimental system and thus provides a means to determine whether a simplified metric such as volume or surface area is appropriate to describe the dose of NPs with in vitro and in vivo studies. Information on the dose metric provides insight on the ability to compare and combine different in vitro or in vivo studies. This is necessary to advance the existing and future knowledge on
1022
Environ Toxicol Chem 34, 2015
hazard and exposure of NPs. It should be noted that the approach presented in the present study needs further and more robust experimental verification. On the other hand, however, the approach provides clear recommendations for future study designs. Acknowledgment—The present study was supported by the Dutch National Institute for Public Health and the Environment (RIVM). The research described was the result of combining the conceptual ideas with experimental data obtained within three RIVM projects, that is, the Dosimetry project, the Nanotechnology: Potential Health Risks project, and the IRAN-project. The Dosimetry project is an assignment of the Dutch Government to RIVM as a contribution to the Working Party on Manufactured Nanomaterials (WPMN) program of the Organisation for Economic Co-operation and Development. The IRAN project and the Nanotechnology: Potential Health Risks projects are part of the strategic research program at RIVM. Data availability—Data, associated metadata, and calculation tools are available publicly. All data has been published, and references are cited within the article text. REFERENCES 1. Park MVDZ, Neigh AM, Vermeulen JP, de la Fonteyne LJJ, Verharen HW, Briede JJ, van Loveren H, de Jong WH. 2011. The effect of particle size on the cytotoxicity, inflammation, developmental toxicity and genotoxicity of silver nanoparticles. Biomaterials 32:9810–9817. 2. Montellier C, Tran L, Macnee W, Faux S, Jones A, Miller B, Donaldson K. 2007. The pro-inflammatory effects of low-toxicity low-solubility particles, nanoparticles and fine particles, on epithelial cells in vitro: The role of surface area. J Occup Environ Med 64:609–615. 3. Pauluhn J. 2011. Poorly soluble particulates: Searching for a unifying denominator of nanoparticles and fine particles for DNEL estimation. Toxicology 279:176–188.
W.J.G.M. Peijnenburg et al. 4. Sager TM, Castranova V. 2009. Surface area of particle administered versus mass in determining the pulmonary toxicity of ultrafine and fine carbon black: Comparison to ultrafine titanium dioxide. Part Fibre Toxicol 6:15. 5. Wittmaack K. 2007. In search of the most relevant parameter for quantifying lung inflammatory response to nanoparticle exposure: Particle number, surface area, or what? Environ Health Perspect 115:187–194. 6. Tinkle SS. 2010. Maximizing safe design of engineered nanomaterials: The NIH and NIEHS research perspective. Wiley Interdiscip Rev: Nanomed Nanobiotechnol 2:88–98. 7. Park MVDZ, Lynch I, Ramírez-García S, Dawson KA, de la Fonteyne L, Gremmer E, Briede JJ, Slob W, Elsaesser A, Howard CV, van Loveren H, de Jong WH. 2011. In vitro evaluation of cytotoxic and inflammation properties of silica nanoparticles of different sizes in murine RAW 264.7 macrophages. J Nanopart Res 13:6775– 6787. 8. Wang Z. 2012. Aquatic toxicity of nanosilver colloids to different trophic organisms: Contributions of particles and free silver ion. Environ Toxicol Chem 31:2408–2413. 9. Bradford A. 2009. Impact of silver nanoparticle contamination on the genetic diversity of natural bacterial assemblages in estuarine sediments. Environ Sci Technol 43:4530–4536. 10. Dekker T. 2006. Development and application of a sediment toxicity test using the benthic cladoceran Chydorus sphaericus. Environ Pollut 140:231–238. 11. Organisation for Economic Development and Co-Operation. 2014. Guidance document for quantitative method for evaluating antibacterial activity of porous and non-porous antibacterial treated materials. Series on Testing and Assessment. No. 202. ENV/JM/MONO(2014) 18. Paris, France. 12. Lammer E. 2009. Is the fish embryo toxicity test (FET) with the zebrafish (Danio rerio) a potential alternative for the fish acute toxicity test? Comp Biochem Physiol Part C Toxicol Pharmcol 149:196–209. 13. Hinderliter PM. 2010. ISDD: A computational model of particle sedimentation, diffusion and target cell dosimetry for in vitro toxicity studies. Part Fibre Toxicol 7:36.
