Informing Selection of Nanomaterial Concentrations for ToxCast in Vitro Testing Based on Occupational Exposure Potential

Background: Little justification is generally provided for selection of in vitro assay testing concentrations for engineered nanomaterials (ENMs). Selection of concentration levels for hazard evaluation based on real-world exposure scenarios is desirable. Objectives: Our goal was to use estimates of lung deposition after occupational exposure to nanomaterials to recommend in vitro testing concentrations for the U.S. Environmental Protection Agency’s ToxCast™ program. Here, we provide testing concentrations for carbon nanotubes (CNTs) and titanium dioxide (TiO2) and silver (Ag) nanoparticles (NPs). Methods: We reviewed published ENM concentrations measured in air in manufacturing and R&D (research and development) laboratories to identify input levels for estimating ENM mass retained in the human lung using the multiple-path particle dosimetry (MPPD) model. Model input parameters were individually varied to estimate alveolar mass retained for different particle sizes (5–1,000 nm), aerosol concentrations (0.1 and 1 mg/m3), aspect ratios (2, 4, 10, and 167), and exposure durations (24 hr and a working lifetime). The calculated lung surface concentrations were then converted to in vitro solution concentrations. Results: Modeled alveolar mass retained after 24 hr is most affected by activity level and aerosol concentration. Alveolar retention for Ag and TiO2 NPs and CNTs for a working-lifetime (45 years) exposure duration is similar to high-end concentrations (~ 30–400 μg/mL) typical of in vitro testing reported in the literature. Conclusions: Analyses performed are generally applicable for providing ENM testing concentrations for in vitro hazard screening studies, although further research is needed to improve the approach. Understanding the relationship between potential real-world exposures and in vitro test concentrations will facilitate interpretation of toxicological results.


Review
Researchers evaluating toxicity and human exposure potential of engineered nano materials (ENMs) are challenged by rapid develop ment of novel materials for new applications as the nano technology industry drives forward. These materials can add sig nificant value to industrial or consumer prod ucts. ENMs have one or more components with at least one dimension in the range of 1-1,000 nm. Components can include nano particles (NPs), nano fibers and nano tubes, nano dots, nano structured surfaces, or nano composites. Carbon nano tubes (CNTs) and metal oxide NPs (two material types having the highest industrial production volumes) are used in plastics, catalysts, battery and fuel cell electrodes, solar cells, paints, coatings, etc. (Klaine et al. 2008). Nanoparticulate silver (Ag) has the greatest number of consumer product applications. Novel nano material (NM) types continue to be synthesized based on the value they may add, often without evaluation of implications for human health, toxicity, environ mental impact, or longterm sustainability. NMs, especially the ones made of metals, semi conductors, and various inor ganic compounds, have the potential for post use risks to humans and the environment (National Nanotechnology Initiative 2008). These concerns need to be examined and addressed before the widespread adoption of nano technologies (Oberdörster et al. 2005).
The U.S. Environmental Protection Agency (U.S. EPA) is beginning to evalu ate exposure and hazard potential of NMs and priori tize them for further animalbased toxicological testing. Prioritization of NM classes and types for targeted testing is impor tant in the early stages of NM development. Currently, only a small portion of the thou sands of commonly used chemicals in the Toxic Substances Control Act (1976) inven tory (U.S. EPA 2004) have under gone ani mal testing because of the high cost (millions of dollars) and long time frame (2-3 years) required per chemical (Judson et al. 2009). Of the unique chemicals (~ 10,000) the U.S. EPA is most concerned with, only a fraction have been evaluated for specific classes of toxicity (Judson et al. 2009). The ToxCast research program of the U.S. EPA was started in 2007 and seeks to predict the potential toxicity of environmental chemicals based on in vitro bio activity profiling at minimal cost compared with fullscale animal testing (Dix et al. 2007). An initial set of approximately 300 chemicals (primarily pesticides) was tested in phase I of ToxCast in 467 highthroughput screening (HTS) biochemical and cellbased assays across nine technologies (Judson et al. 2010). A study has been initiated to evaluate the poten tial of ToxCast methods for screening NMs. A subset of ToxCast in vitro HTS cellbased assays will be run on NMs to produce similar bioactivity profiles and toxicity predictions. Most of the cellbased assays have an exposure time of 24 hr. Initial NM types to be evalu ated include singlewalled carbon nano tubes (SWCNTs) and multi walled carbon nano tubes (MWCNTs), along with Ag, titanium dioxide (TiO 2 ), and gold (Au) NPs.
Design and conduct of ToxCast screen ing of NMs requires selection of testing con centrations, characterization of materials, and analysis of resulting HTS data. Selection of concentrations used for in vitro toxicity stud ies of NMs often lacks scientific justification, and concentrations are often chosen to be very high to ascertain a toxicological end point without consideration of realworld exposure (Oberdörster et al. 2005). Some research ers have used particle concentrations caus ing "overload" (Warheit et al. 2009), a dose where pulmonary clearance becomes severely impaired (Morrow 1988). Although high testing concentrations may be considered to ensure that NMs show bio activity across the spectrum of assays evaluated, there is also a need for biologically relevant human expo sure information to facilitate interpretation of volume 119 | number 11 | November 2011 • Environmental Health Perspectives assay results (Cohen Hubal 2009). Authors of Toxicity Testing in the 21st Century: A Vision and a Strategy (National Research Council 2007) noted that human exposure information is required to select doses for toxicity testing, facilitating development of environmentally relevant hazard information.
Recognizing the critical need for expo sure information to inform chemical design, evalua tion, and health risk management, the U.S. EPA ExpoCast TM program was initiated in 2010 to meet challenges posed by new tox icity testing approaches (Cohen Hubal et al. 2010). The goal of ExpoCast is to advance characterization of exposure required to trans late findings in computational toxicology to information that can be directly used to sup port exposure and risk assessment. Combining information from ToxCast with information from ExpoCast will help the U.S. EPA priori tize NMs and chemi cals for further evaluation based on potential risk to human health.
Human exposures to ENMs are likely to be higher for workers in occupational settings than for the general population, including consumers (Bergamaschi 2009), and may thus provide upper bounding estimates of exposure potentials. For consumers, the greatest expo sure to ENMs likely comes from products that are ingested or that come into intimate con tact with the body (Kessler 2011). Although ingestion and dermal exposures must also be considered during the product life cycle (manufacturing, usage, and disposal of EMNs) (Oberdörster et al. 2005; U.S. EPA 2010), inhalation may be the key route of human exposure in nano technology manufacturing and R&D (research and development) facili ties (Bergamaschi 2009;Hoet et al. 2004). Many studies have focused on the inhalation exposure route for ENMs and have considered potential airborne releases of NMs from facili ties. After intake of NPcontaining aerosols, high deposition fractions in the alveo lar region (for particles < 100 nm in size) and the head region (diameter < 5 nm) are predicted by the multiplepath particle dosimetry (MPPD) and International Commission on Radiological Protection (ICRP) models (U.S. EPA 2009). NM exposure of the lung parenchyma is of concern because of longterm retention in this lower region and potential for particles to cause cytotoxicity and translocate.

General Approach
The aim of this study was to use information on potential ENM exposure in the occupa tional setting to recommend in vitro testing levels for bioprofiling in U.S. EPA ToxCast program. Our general approach (Figure 1) was to assume that the inhalation exposure route for NMs is of primary concern for humans in occupational settings. Occupational aerosol levels of NMs reported in the litera ture were reviewed and used as inputs for lung dosimetry modeling. We assumed that these reported NM concentrations from manufacturing and R&D laboratory facilities would provide a highend potential for realworld NM exposure to the general population, higher than expo sures that may result from consumer products (Bergamaschi 2009).
We used the maximum reported NM aero sol concentrations (mass per cubic meter of air) as an input for the MPPD model to estimate deposition, clearance, and mass retained in the alveo lar region of the human lung. A sensitiv ity analysis was performed to evaluate MPPD input parameters that most affect NM alveolar retention after 24 hr of exposure. Two expo sure scenarios were considered for further mod eling: exposure over the course of 24 hr (based on the standard assay exposure duration) and 45 years (a full occupational lifetime). For each scenario, we varied the significant parameters to estimate the mass of particles retained in the alveolar region per surface area. Model results of lung surface mass concentrations were then converted (using the reported wellbottom sur face area and volume delivered) to suggest test ing solution mass concentrations for in vitro screening. All of the applied material was assumed to deposit on the bottom of the well. The results suggest upper and lower bounding HTS assay testing concentrations based on potential for realworld NM exposures at short and long durations via the inhalation route in an occupational setting. The concentrations were subsequently compared with in vitro con centrations found in recent literature. Although we have chosen here to consider aerosol mass concentration, we recognize that other lung deposition metrics (based on particle num ber or particle surface area) are also potentially important for understanding health risk.
A small fraction of NPs deposited in the alveolar region may be cleared into the blood stream by absorption. Particles that deposit in the respiratory tract can also be cleared to the gastro intestinal (GI) tract via the pharynx or to the regional lymph nodes (LN) via lym phatic channels. Only lung surface cells would receive the same concentration of NPs as esti mated here for inhalation. Modeling expo sure to other cell types is beyond the scope of this article, but the concentrations from these exposures would likely be significantly lower than those calculated for lung cells.

NM air concentrations.
We reviewed occu pational exposure studies that meas ured air borne levels of ENMs. The instruments used to obtain particle number concentrations were typically the condensation particle coun ter (CPC), scanning mobility particle sizer (SMPS), and the fast mobility particle sizer (FMPS). In some cases, personal air samplers collected NPs on filters from the breathing zone of workers during the work day. The SMPS and FMPS instruments provide real time temporal changes in particle size. The data give particle number concentrations (particles per cubic centimeter of air) versus particle diameter across the size distribution. The CPC provides particle number concentra tion (particles per cubic centimeter of air) for particles in the range of 2.5 to > 1,000 nm. General approach for recommending in vitro testing levels, considering exposure to NMs from occupational-setting indoor air via the inhalation route resulting in respiratory tract uptake. Estimated exposure potential is converted to levels for NM testing in HTS cellular assays. The instruments can also report the change in particle number concentration versus time.
We searched for the highest aerosol parti cle number concentrations for TiO 2 and Ag NPs and for CNTs (including MWCNTs) in manufacturing and R&D settings (Table 1). Background particle number concentrations were subtracted from the maxi mum particle number concentrations if they were reported. Typically, particle counts per volume (cubic centimeters) of air are reported, whereas exposure limits are set as mass concentrations (milli grams per cubic meter). To convert from reported particle count concentration to mass concentration, TiO 2 and Ag NPs were assumed to be spherical, and the reported size (taken to be geometric particle diameter) was used to calculate a particle volume. The CNTs were assumed to be cylindrical, and reported diameter and length were used to obtain particle volume. We assumed a CNT length of 0.5 μm if it was not reported. A density of 4, 10, and 2 g/cm 3 was assumed for TiO 2 NPs, Ag NPs, and CNT, respec tively, based on specifications of similar mate rials from supplier web sites (Nanostructured & Amorphous Materials Inc. 2010; Sigma Aldrich 2010). The highend reported particle counts were approximated to mass concentra tions by multiplying particle volume by the assumed density (Table 1). The calculated mass concentrations were typically less than approximately 0.1 mg NM per volume (cubic meters) of air (Table 1). One study on MWCNTs reported a higher mass concentration of 0.3208 mg/m 3 ). However, this value was from personal sampler filters with typical sampling durations of 183-409 min. The mass concen tration would be lower if calculated over the time duration. Data normalized over expo sure characterization duration from a liquid phase production facility of Ag NPs yielded a mass concentration of 0.46 mg/m 3 for 1 min (Park et al. 2009). In that study, both change in particle number concentration versus time and total number of particles (with diame ters between 10 and 250 nm) counted over a range of time were reported. A conserva tive aerosol concentration of 1 mg/m 3 was taken to be an upper exposure limit. Although the U.S. National Institute for Occupational Safety and Health (NIOSH) does not have a recom mended exposure limit (REL) for TiO 2 NPs, a draft NIOSH bulletin (NIOSH 2005) recommended "0.1 mg/m 3 for ultra fine TiO 2 , as timeweighted average concen trations (TWA) for up to 10 hr/day during a 40hour work week," where "ultrafine" is defined as the fraction of respirable particles with primary particle diameter < 100 nm. A recent draft NIOSH bulletin (NIOSH 2010) proposed a REL of 0.007 mg/m 3 for CNTs and carbon nano fibers. Using a differ ent approach, an occupational exposure limit of 0.05 mg/m 3 was derived for Baytubes, a more flexible MWCNT type (Pauluhn 2010). There is no limit set for Ag NPs in the United States. However, the Occupational Safety and Health Administration (OSHA) estab lished a permissible exposure limit (PEL) of 0.01 mg/m 3 (which is the same as the REL set by NIOSH) for all forms of airborne Ag (Miller et al. 2010). The American Conference of Governmental Industrial Hygienists (ACGIH) set a threshold limit value (TLV) of 0.1 mg/m 3 for metallic Ag and 0.01 mg/m 3 for soluble Ag compounds (Miller et al. 2010). In the present study, the mass concentrations derived based on meas ured aerosol levels were taken as a basis and used as inputs to model the mass of NPs that could deposit and be retained deep in human lungs.
Lung dosimetry modeling. MPPD model application. We estimated particle deposi tion and clearance in human lungs using the recently developed, publicly available MPPD model (version 2.1 for NPs, presently sup ported by Applied Research Associates Inc., Raleigh, NC). The model can be used to esti mate particle dosimetry in both human and rat airways (Anjilvel and Asgharian 1995;Asgharian et al. 2001). It calculates deposi tion and clearance of particles ranging from ultrafine (0.001 μm) to coarse (100 μm) in the respiratory tract, based on userprovided input on airway morphometry, clearance rates, particle properties (density, diameter, and size distribution), and exposure scenario (aerosol concentration, activity breathing pattern, and exposure duration). Three main particle depo sition mechanisms (impaction, sedimentation, and diffusion) are incorporated in the model, and deposition in different regions of the lung are calculated using published analytic formu las (Anjilvel and Asgharian 1995). Clearance from each lung region is treated competitively between absorption into the blood and par ticle transport processes (from the respiratory tract to the GI tract and to lymph nodes, and from one region to another) (ICRP 1994). Retention in the human alveolarinterstitial region is represented by three compartments, which clear at fast, medium, and slow rates to the lymph nodes and the bronchiolar region (ICRP 1994). Although the clearance kinetics in the MPPD model were based on studies of micro sized particles, evidence sug gests efficient surface macrophage uptake and clearance of both microparticles and NPs as well as penetration of both sizes of particles through the human lung epithelium into the interstitial region, from which they are slowly cleared (Geiser and Kreyling 2010). In addi tion, the MPPD model (version 2.1) incor porates improved estimates of particle losses from the airway by diffusion and includes particlespecific axial diffusion and dispersion effects in the transport equation (Asgharian Table 1. NM exposure concentrations in lab and manufacturing sites. We selected an initial baseline set of MPPD inputs (Table 2) based on data from the ICRP report (ICRP 1994), which provided morphological charac teristics and physio logi cal parameters for the human respiratory tract. We organized the MPPD model input parameters into three categories: individual characteristics, exposure scenario, and mate rial properties. For the individual characteris tics input, the airway morphometry selected was the human Yeh/Schum symmetric lung model (Yeh and Schum 1980). Default val ues were selected for the clearance rates and other parameters. For the exposure scenario input, 0.1 mg/m 3 aerosol concentration was selected, and light exercise activity breathing pattern for an adult male was assumed with 20 breaths/min frequency at 1,250 mL tidal volume, (V T ) (ICRP 1994). Oronasalmouth breather was selected for breathing scenario, because humans typically switch to breathing partly through the mouth and through the nose at ventilation rates between light and heavy exercise (ICRP 1994). For the particle properties input, we selected a particle count median diameter of 40 nm, assuming a single mode of lognormal size distribution with size geometric standard deviation (GSD) of 1.25 based on the ICRP report. Inhalability was not considered because it approaches 100% for small (< 5 μm) particles (ICRP 1994). The lengthtodiameter aspect ratio was set to 1.

NM
Sensitivity analysis. Key determinants of MPPD model predictions of mass (milli grams) retained in the alveolar region were determined by systematically altering each input baseline parameter one at a time, while holding the oth ers constant, and rerunning the model based on a 24hr exposure duration with 1 week of total time (Table 3) to allow for clearance. For the individual charac teristic inputs, we evaluated two different size (based on total number of airways) human stochastic lung models because they provide more realistic lung geometry than the symmetric lung model. Calculations were also performed using an agespecific symmetric lung model for a 3yearold child. Although this group is unlikely to be exposed occupa tionally, we wanted to check model results for a vulnerable population group. The alveolar interstitial rate constants for fast, medium, slow, and lymph node human clearance were doubled, halved, increased by an order of magnitude, and decreased by an order of mag nitude. Tracheal mucosal velocity was not con sidered because it affects only tracheo bronchial clearance rates and residence times and will not affect longterm alveolar burden. For the expo sure scenario inputs, the aerosol concentra tion was decreased by one order of magnitude from 0.1 mg/m 3 . As a conservative estimate in case the mass per air volume concentration was much higher than reported, the aerosol mass concentration was also increased by one and two orders of magnitude. Both heavyexercise and resting breathing patterns were evaluated, as well as purely nasal and oral breathing. For the particle properties inputs, we considered a low size diameter of 5 nm, a high diameter of 100 nm, a low GSD of 1 (monodisperse diam eter distribution), and a high GSD of 4 (poly disperse diameter distribution). Additionally, aspect ratios from 4 to 1,000 were evaluated with a length GSD of 1.0 (as a conservative estimate) and a density of 2.
For this sensitivity analysis, if the alveo lar mass retained using the new setting resulted in a percentage change ≥ 10% of the baseline amount, the parameter was considered to be significant and was evaluated further. If the alveolar mass retained using the new setting yielded a negative percentage change com pared with the baseline setting, then the input was not considered, because we are interested in a conservative exposure approach that may over estimate particle deposition and retention deep in the lungs. If alveolar retention output did not change linearly with change in input, additional input changes were considered to better characterize model behavior over the relevant range. The MPPD input parameters determined to be significant were evaluated further to calculate mass retained in the alveo lar region per alveolar surface area, based on two exposure durations: a shortterm exposure duration of 24 hr and a longterm occupa tional lifetime exposure. The longterm sce nario assumed a 45year full working lifetime (Schulte et al. 2010) with 8 hr inhalation per day, 5 days/week, 52 weeks/year. The alveolar surface area (~ 106,350 cm 2 ) was obtained from the MPPD model results report by sum ming the pulmonary surface area for lung gen erations 17 to 24. This alveolar surface area accounts for only surface area of the airways (alveolar ducts) and not the alveolar sacs, and thus is a low estimate of the actual alveolar surface area. The MPPD calculations were performed for different particle sizes (5,10,20,30,40,50,60,70, and 100 nm), aerosol concentrations (0.1 and 1 mg/m 3 ), and expo sure durations. Larger particle sizes (200, 500, and 1,000 nm) were also run because particle aggregation of nano sized particles may occur in air (Maynard et al. 2004;Methner et al. 2010a) or inside the human respiratory tract. For CNTs, an aspect ratio of 167 was selected based on material dimensions (5 μm length, 30 nm diameter) of one sample to be tested in ToxCast. Aspect ratios of 2, 4, and 10 were also run for the different particle sizes. These aspect ratios were chosen based on electron images of SWCNT aggregates from the litera ture (Baron et al. 2008). Searching for realistic airborne CNT aspect ratios was challenging because many exposure studies found no evi dence of carbonbased nano tubes or nano tube bundles in air samples (Bello et al. 2008(Bello et al. , 2009. In one study of seven CNThandling workplaces, transmission electron micrographs reveal clumped structures with aspect ratios of approximately 8-10 and diameters of approxi mately 100 nm ). However, these particle aggregates are mostly metal com ponents rather than CNTs.
Determining in vitro concentrations. Based on MPPD model predictions, we determined associated in vitro concentrations by calculat ing mass retained in the alveolar region of the lung per alveolar surface area for each particle size, at two aerosol concentrations (0.1 and 1 mg/m 3 ) and for each exposure scenario (24 hr and 45 years). We assumed that the NM mass retained at the lung surface can be directly correlated to NM mass sedimented on the bottom surface of a well. To convert to mass of NM per volume of solution, we mul tiplied the resulting mass per alveolar surface area concentration (micrograms per square centimeter) by the bottom surface area of a single well in a 12, 96 or 384well plate and divided by the volume of culture medium added to each well (as obtained from the assay contractors). The 12, 96, and 384well plates had a single wellbottom surface area of 3.8, 0.32, and 0.056 cm 2 and volume of 1,000, 200, and 50 μL, respectively. The con verted concentrations were compared with in vitro concentrations tested using human, mouse, and rat cell lines in the litera ture [see Supplemental Material, Tables S1-S3 (http:// dx.doi.org/10.1289/ehp.1103750)]. Although final selection of concentration will include consideration of the MPPD model output and conversion, there will still be a need to test at levels based on where bioactivity has been demon strated in the literature, bounded by concentration levels that can be dispersed with longterm stability in cell culture media.
Another method to determine high range in vitro concentrations to test could be to eval uate a NM steadystate mass in the alveo lar region of the lung. Steady state occurs when the clearance rate equals the rate of deposition and the NM mass retained reaches a constant value. According to Brown et al. (2005), it takes > 10 years to reach a steadystate lung burden for insoluble 1 μmsized particles for a 0.01mg/m 3 aerosol concentration based on resting human breathing pattern.

Results and Discussion
Key MPPD model input parameters. For the MPPD baseline settings used here (Table 2), a steadystate retention dose would take > 80 years to achieve for 40nm particles based on inhalation of an aerosol concentra tion of 0.1 mg/ m 3 for 8 hr/day, 5 days/week. Because of the long time to achieve steady state, we did not use that method. Instead, we focused on modeling potential exposure scenarios and under standing implications of associated model inputs. Model results for the baseline input parameters (Table 2) resulted in 1.22 mg alveolar mass retained.
Results of the sensitivity analysis [alveolar mass retained, percentage change in model output and input, and sensitivity percent age (output percentage change by input per centage change)] are presented in Table 3. Although inter actions between input param eters may occur, we assumed that key param eters could be uncovered by varying one parameter per run. Based on this analysis, aerosol concentration and heavyexercise breathing pattern were the most important MPPD input parameters, as these increased alveolar retention by > 10%. For variations to the inputs for individual characteristics (Table 3), the choice of airway morphometry using the the human stochastic lung model resulted in a lower retention compared with the symmetric lung model. The agespecific symmetric lung model for a 3yearold child resulted in lower mass retained at 0.16 mg because of lower intake (functional residual capacity, upper respiratory tract volume, and V T ) compared with the adult male default baseline condition. Thus, the Yeh/Schum symmetric model provided a conservative esti mate of NM particle dosimetry. Nasal and oral breathing scenario did not significantly affect the results and was set to the baseline of oro nasal breathing. Increasing and decreasing the default alveolarinterstitial rate constants did not significantly affect the result as indi cated by the sensitivity percentage in Table 3. The alveolarinterstitial rate constants were set to the default values and the lung model to symmetric to further calculate alveolar mass retained per alveolar surface area.
For the exposure scenario inputs (Table 3), the correlation between alveolar mass retained and aerosol concentration was linear: The amount retained in the alveolar region changed linearly by one order of magnitude as the expo sure aerosol concentration (and thereby the intake) was increased or decreased by one order of magnitude, yielding a sensitivity of 100%. Using resting or heavyexercise breathing pat tern resulted in a sensitivity of approximately 100%, indicating that alveolar mass retained changed almost linearly with minute ventila tion (breathing frequency by V T ). To further calculate alveolar mass retained per alveolar surface area, the breathing scenario was set to light exercise, based on the assumption that this was the most realistic for a full working lifetime. The aerosol concentration of 1 mg/m 3 was taken as a conservative estimate of poten tial worker exposure. Particle property input changes to par ticle diameter, size GSD, and aspect ratio (length:diameter) did not result in linear changes to output alveolar mass retention, as observed in the sensitivity percent column in Table 3. Particle diameter of 20 nm resulted in maximum alveolar mass retained of 1.51 mg for diameters between 5 and 100 nm and was approximately 24% increase in output com pared with baseline 40 nm size. A GSD value of 1 (mono disperse size distribution) yielded a higher mass amount retained, but it was only 9.08% more than the baselinesize GSD value of 1.25 and did not meet the sensitivity analy sis requirements. All other input changes for diameter and GSD lowered the alveolar mass retained compared with the baseline settings. In a report on TiO 2 particles, Hameri et al. (2009) listed a GSD of 1.66, and in a study of Ag NPs, Park et al. (2009) listed GSD val ues of 4.63-6.3. Although a highersize GSD value is expected for realistic size distribution of NMs, this parameter was set to 1.25 as a conservative estimate that would increase alveolar retention. Changes to the aspect ratio input at constant aerosol concentration and minute ventilation lowered the alveolar mass retained compared with the baseline (aspect ratio 1 in Table 3). Only an aspect ratio of 20 slightly increased the alveolar mass retained. The sensitively percent to the aspect ratio parameter was low.
Concentrations recommended for in vitro testing. In Figure 2, results of the deposition modeling are presented as a function of mate rial charac teris tics for the two exposure sce narios of interest. Because MPPD alveolar mass retention was linearly proportional to the inputted aerosol concentration, we plot ted mass per lung surface area per inputted aerosol concentration versus particle diam eter. The alveolar retention per surface area for Ag and TiO 2 spherical NPs for a full working lifetime was highest for 20nm diam eter particles (48.9 μg/cm 2 ), based on an exposure aerosol concentration of 1 mg/m 3 (Figure 2A). Relative to this peak lung sur face concentration, the amount decreased to 20.3 μg/cm 2 as size was increased to 100 nm, and also decreased to 25.1 μg/cm 2 for 5nm particles. For the 12, 96 and 384well plates used by the different assay contractors, the peak lung surface concentration equates to 186 [i.e., 48.9 μg/cm 2 × (3.8 cm 2 /mL)], 78.2, and 54.8 μg/mL, respectively. These amounts for a full working lifetime lie within the range of the highest in vitro assay concentrations tested in the literature for Ag NPs and TiO 2 NPs on human, rat, and mouse cell lines. The highest amount tested for Ag NPs ranges from 1.6 to 500 μg/mL, whereas for TiO 2 NPs, the highside range is 100-1,000 μg/mL or 20-520 μg/cm 2 [see Supplemental Material, Tables S1and S2 (http://dx.doi.org/10.1289/ ehp.1103750)]. Concentrations for most of the Ag NPs tested fell within 50-400 μg/mL, whereas those for TiO 2 NPs fell within 100-250 μg/mL. Because the MPPD model uses a low estimate of alveolar surface area, a more realistic estimate would result in lower alveolar mass retained per surface area (by approximately one order of magni tude), which would correspond to a lower in vitro concentration for a given exposure duration. For a full working lifetime exposure duration to 0.1 mg/m 3 , the peak lung surface concentration was 4.9 μg/cm 2 (for particles with a diameter of 20 nm) and the range was 2.0-4.9 μg/cm 2 for particle with diameters of 5-100 nm (Figure 2A). Because alveolar retention is directly proportional to aerosol concentration, reducing the input aerosol concentration by a factor of 10 results in a linear reduction of the calculated wellplate concentration (micrograms per milli liter) by a factor of 10. The calculated wellplate con centration for a full working lifetime is similar to the low range (1.6-10.8 μg/mL) of the highest concentrations tested in in vitro assays for Ag NPs, but it is below the range tested for TiO 2 NPs [see Supplemental Material, Tables S1 and S2 (http://dx.doi.org/10.1289/ ehp.1103750)].
The lung surface concentration for a 24hr exposure duration to 1 mg/m 3 aerosol con centration of TiO 2 or Ag NPs ranged from 0.061 to 0.15 μg/cm 2 for particles 5-100 nm in diameter (Figure 2A). This range is more than two orders of magnitude lower than the range for a full working lifetime (Figure 2A). The peak lung surface concentration equates to 0.570, 0.240, and 0.168 μg/mL for the 12, 96 and 384well plates, respectively. In previous studies, the lowest amount tested for Ag NPs ranged from 0.108 to 25 μg/mL [see Supplemental Material, Table S1 (http:// dx.doi.org/10.1289/ehp.1103750)], whereas for TiO 2 NPs the range was 0.002-10 μg/mL or 0.0052-5 μg/cm 2 (see Supplemental Material, Table S2). For 24hr exposure dura tion, the alveolar surface concentrations calcu lated using the MPPD model fell within the range (closer to the lower end) of the lowest in vitro concentrations tested. Thus, the con centrations reported in previous studies (see Supplemental Material, Tables S1 and S2) are similar to the lowerbound assay test concen trations we derived using the estimated lung retention after 24 hr exposure. Each of the studies had a set exposure duration ranging from 1 to 144 hr for Ag NPs and from 5 min to 120 hr for TiO 2 NPs (see Supplemental Material, Tables S1 and S2). Rerunning all the baseline settings for 20nm particles for 24hr exposure duration would require a very high aerosol concentration of approximately 330 mg/m 3 to result in a similar peak alveolar surface concentration (~ 48.9 μg/cm 2 ) for TiO 2 and Ag NPs.
Results of the present study show that the alveolar mass retention per surface area for CNTs (with a lengthtodiameter aspect ratio of 167) for a full working lifetime expo sure to 1 mg/m 3 aerosol concentration ranged from 12.4 to 46.5 μg/cm 2 ( Figure 2B), similar to the range for spherical particles. As CNT diameter decreased from 100 nm, the mass retained per surface area increased to a maxi mum of 46.5 μg/cm 2 for 5nm diameter nano tubes. In previous studies, the highest amount tested in vitro for CNTs ranged from 50 to 1,000 μg/mL [see Supplemental Material, Table S3 (http://dx.doi.org/10.1289/ ehp.1103750)]. Most of the CNT concen trations tested fell within 50-400 μg/mL. For the more realistic aspect ratios of 4 and 10, we observed peak mass per surface area concentration at 40 nm and approximately 25 nm, respectively ( Figure 2B). This peak concentration decreased with increasing diam eter ( Figure 2B). It is possible that the CNTs will form aggregates of larger diameter and lower aspect ratios, as reported by Baron et al. (2008). Using the model, we observed that  particles with aspect ratios of approximately 20 had a maximum deposition fraction in the alveolar region. For the aspect ratios 2, 4, and 10, the mass per surface area retained for diameters > 40 nm followed a similar trend. The lung surface concentration of aspect ratio 2 was similar to the trend for spherical particles (Figure 2A) at the same aerosol con centration and exposure duration. Applications of approach. The approach we took in this study was a simple screening level assessment using the latest quantitative NM aerosol data in occupational settings to determine concentrations that may deposit and be retained deep in the human respiratory tract. The methodology we used and the alveo lar retention results obtained can be generally applied to inform in vitro study designs, which include other NM types. Wherever possible, conservative MPPD input parameters were selected so that results would indicate a higher alveolar retention, although we attempted to choose realistic inputs as well. Our results indi cate that a full lifetime occupational exposure to a concentration of 1 mg/m 3 (one order of magnitude higher than what has typically been reported) (Table 1) is required to reach the highest concentrations currently being tested in vitro in most studies [see Supplemental Material, Tables S1-S3 (http://dx.doi. org/10.1289/ehp.1103750)]. Because in vitro studies use different cell culture containers, in order to convert the lung surface concentra tions provided here, the specific wellbottom surface area and medium volume presented to each well are required.
Note that we are comparing lung surface concentrations to concentrations being tested in a range of cell types and would expect only a small percentage of particles to reach cells in other organs of the body following absorp tion into the bloodstream. Nevertheless, retention of NPs in the deep lung alveolar region is important because these particles potentially can be absorbed quickly into the bloodstream. Such a phase of rapid absorp tion is observed immediately after inhala tion, even with relatively insoluble materials (ICRP 1994). Recently, in a study in rats, Choi et al. (2010) found a) that NPs with hydro dynamic diameters < 6 nm and zwit terionic surface charge can rapidly enter the bloodstream from the lung and then be sub sequently cleared by the kidneys; and b) that NPs < 34 nm with a non cationic surface charge translocate rapidly from the lung to the media stinal lymph nodes. However, for technetiumradiolabeled 100nm, 35nm, and 4-20nm diameter carbon particles, no sig nificant systemic translocation of particles has been observed in humans (Mills et al. 2006;Möller et al. 2008;Wiebert et al. 2006). Gold NPs (5-8 nm) have been found at a low frac tion (0.03-0.06% of lung concentration) in the blood of rats 1-7 days after inhalation (Takenaka et al. 2006). The type and amount of surface charge or coating may be a key fac tor for trans location of particles and should be evaluated. There is also a potential for larger mass amounts of NMs per lung surface area to be deposited in the tracheobronchial region. All airway surfaces may not receive the same amount of deposited particles, and local ized hot spots for deposition in the vicinity of airway bifurcations have been predicted (up to 100-1,000 times higher than the average mass per surface area for particles > 10 nm) using mathematical modeling techniques (Farkas et al. 2006;U.S. EPA. 2009). However, we did not consider mass retained in this region because a large portion of the particles depos ited is assumed to be cleared within 24-48 hr by action of the mucociliary escalator (U.S. EPA 2009). Potential future work will con sider GI tract exposure to NPs cleared from the tracheo bronchial region because it may be significant for aggregated NPs at heavy exercise breathing conditions.
Limitations of approach. There are several limitations in estimating concentrations for in vitro testing using the latest available NP aerosollevel data from occupational settings. The instrumentation technology to measure spherical NPs typically provides non specific particle counts over a broad size range. For example, the SMPS provides particle counts in a size range of 2.5-1,000 nm, whereas the CPC used in several studies (Table 1) measures particles 10-1,000 nm. Particle counts become increasingly insensitive to particle sizes that are nearer to the lower limit of detection (Maynard and Aitken 2007). Measurements for non spherical particles such as CNTs may not be reliable and may need to be corrected because these instruments are designed to count spheri cal particles. Additionally, to compare particle number concentration for the same type of materials across different occupational set tings or manufacturing processes, the FMPS and SMPS data reported need to be normal ized by dividing by the number of channels. Instrument data are not always normalized and thus may be reported as a higher count over a particle size distribution than what actually occurs. It is not currently possible to distin guish between NPs, aggregates of the same compounds, and aggregates of a mixture of par ticles, dust, and other airborne particle types. NPs often can agglomerate in air, which is why we present results for potentially more realistic sizes > 100 nm (Figure 2A,B). Instruments such as the universal NP analyzer (UNPA), which uses a CPC, a differential mobility ana lyzer (DMA), and an NP surface area monitor (NSAM) are being developed to determine the primary particle size and measure the num ber, surface area, and volume distributions of gasborne NP agglomerates (Wang et al. 2010). To distinguish NPs from background particles, both realtime instrumentation meas ure ments and qualitative analysis by electron microscopy are required (OnoOgasawara et al. 2009). In addition, chemical analy sis is necessary for quantitatively assessing exposure to NMs at facilities with high levels of back ground NPs. Models are being developed to predict the change in NP number concentra tion for a defined source and a defined envi ronment based on a given background aerosol concentration (Seipenbusch et al. 2008). NPs do not reach the receptor in their original size as an aerosol, but change their size and number concentration by coagulation either within the same type of materials or by inter action with a background aerosol (Seipenbusch et al. 2008).
Here we provide MPPD results in which we assumed no changes to the original aero sol concentration and performed simulations using the reported size of the particles. If parti cles have a tendency to aggregate and agglom erate above an aggregate size of 100 nm, the amount deposited and retained in different regions of the lung will be less (Figure 2A,B). In the case of SWCNTs, large aggregates > 10 μm in diameter can form by diffusion and van der Waals inter actions between nano tubes in air or in aqueous solutions (Mutlu et al. 2010). Other drawbacks based on the method used include limitations with the MPPD model, such as a low estimate of alveolar surface area. Currently, distinctions between NM types cannot be made based on NM physico chemical characteristics. The only input possible in the latest version of the model is lengthtodiameter aspect ratio for cylindrical particles. The clearance calculations in the model are based on experi mental data for spherical particles, and fibers with elon gated structures may have different clearance kinetics. NMs have unique physico chemical charac teris tics that may affect their deposition, retention, and toxicity. These characteristics include particle shape and shape distribution, large surface area:volume ratio, chemical com position and crystalline form, surface compo sition and coating, and surface charge. There is a need to understand which physico chemical charac teris tics most affect the deposition and alveolar retention of NPs and to further incor porate these key parameters into the model.
Another limitation in the approach may be the conversion of lung surface concentra tions to in vitro test concentrations, assum ing that the NMs will quickly (relative to the duration of the assay) settle onto the cells at the bottom of the well plate. If the particle transport (diffusion, sedi menta tion) time is slower than the in vitro assay testing time (which could possibly be the case for particle agglomerates, depending on their mass, size, and density) (Hinderliter et al. 2010), then the localized NM concentrations near the cells at volume 119 | number 11 | November 2011 • Environmental Health Perspectives the bottom of the well may be lower than we estimated. A recently developed computational model of partico kinetics (sedimentation, dif fusion) and target cell dosimetry for in vitro systems addressed this issue (Hinderliter et al. 2010) and could be used to calculate dose rates and target cell doses to compare with the total assay exposure time. Further, bio activity pro files attained for NMs would need to take into account the localized concentration.

Conclusions
Consideration of potential exposures during design of in vitro toxicity tests would improve interpretation of hazard screening results for use in risk assessment. The methodology described here is a first step toward improving selection of NM concentrations to test in vitro based on realworld inhalation exposure potential. The results obtained can be generally applied to other in vitro study designs and for other NM types. The approach here reveals that current highrange in vitro testing concentrations being used are similar to predicted lung surface area concentrations based on inhalation exposure to NMs of a high aerosol concentration in an occupational setting over the course of a full working lifetime. This methodology can be improved by better measurements of NMs in occupational settings, addition of particle prop erty input parameters to the MPPD model, and considerations of delivered dose to cells.