Applications of the comet assay in particle toxicology: air pollution and engineered nanomaterials exposure

Transcription

1 Mutagenesis, 2015, 30, doi: /mutage/geu035 Review Review Applications of the comet assay in particle toxicology: air pollution and engineered nanomaterials exposure Peter Møller*, Jette Gjerke Hemmingsen, Ditte Marie Jensen, Pernille Høgh Danielsen, Dorina Gabriela Karottki, Kim Jantzen, Martin Roursgaard, Yi Cao, Ali Kermanizadeh, Henrik Klingberg, Daniel Vest Christophersen, Lars-Georg Hersoug and Steffen Loft Department of Public Health, Section of Environmental Health, University of Copenhagen, Øster Farimagsgade 5A, DK-1014 Copenhagen K, Denmark *To whom correspondence should be addressed. Tel: ; Fax: ; Received March ; Revised May ; Accpeted July Abstract Exposure to ambient air particles is associated with elevated levels of DNA strand breaks (SBs) and endonuclease III, formamidopyrimidine DNA glycosylase (FPG) and oxoguanine DNA glycosylasesensitive sites in cell cultures, animals and humans. In both animals and cell cultures, increases in SB and in oxidatively damaged DNA are seen after exposure to a range of engineered nanomaterials (ENMs), including carbon black, carbon nanotubes, fullerene C 60, ZnO, silver and gold. Exposure to TiO 2 has generated mixed data with regard to SB and oxidatively damaged DNA in cell cultures. Nanosilica does not seem to be associated with generation of FPG-sensitive sites in cell cultures, while large differences in SB generation between studies have been noted. Single-dose airway exposure to nanosized carbon black and multi-walled carbon nanotubes in animal models seems to be associated with elevated DNA damage levels in lung tissue in comparison to similar exposure totio 2 and fullerene C 60. Oral exposure has been associated with augmented DNA damage levels in cells of internal organs, although the doses have been typically very high. Intraveneous and intraperitoneal injection of ENMs have shown contradictory results dependent on the type of ENM and dose in each set of experiments. In conclusion, the exposure to both combustion-derived particles and ENMs is associated with increased levels of DNA damage in the comet assay. Particle size, composition and crystal structure of ENM are considered important determinants of toxicity, whereas their combined contributions to genotoxicity in the comet assay are yet to be thoroughly investigated. Introduction The International Agency for Research on Cancer (IARC) has classified exposure to particulate matter (PM) in air pollution as a human carcinogen (1). This classification was based on convincing evidence from epidemiological studies, in addition to strong evidence from mechanistic studies (1). Numerous studies on cytogenetic changes, mutations, bulky DNA adducts and comet assay endpoints were assessed by the IARC workgroup in pursuit to describe mechanisms of carcinogenicity of air pollution particles. Overall, this sets the blueprint for the comet assay as a widely acknowledged tool in molecular epidemiology and genetic toxicology. This work started ~20 ago with two pioneer studies from the Czech Republic and Mexico, where the comet assay was used alongside other methodologies to assess toxicity in leukocytes from humans in areas with different air pollution exposure (2,3). Since then the comet assay has become increasingly popular in genetic toxicology on ambient air PM and engineered nanomaterials (ENMs). This review focuses on the use of the comet assay for the assessment of genotoxicity following exposure to ambient air PM and ENMs. The Author Published by Oxford University Press on behalf of the UK Environmental Mutagen Society. All rights reserved. For permissions, please journals.permissions@oup.com. 67

2 68 P. Møller et al., 2015, Vol. 30, No. 1 Comet assay endpoints The standard comet assay measures DNA strand breaks (SBs) by using single-cell gel electrophoresis. The cells are embedded in agarose and lysed with detergent in high salt solution to form nucleoids. In the alkaline version of the comet assay, nucleoids are subsequently exposed to electrophoresis at high ph, which promotes migration of nicked DNA towards the anode, resulting in structures that look like comets under fluorescence microscopy (4). This version of the comet assay protocol is typically referred to as measuring DNA SBs or DNA damage, although the specific type of lesion is unknown. In addition, the comet assay can also be modified to measure oxidatively damaged DNA by incubation of nucleoids with DNA repair enzymes. The most widely used enzymes are formamidopyrimidine DNA glycosylase (FPG) or endonuclease III (ENDOIII) from Escherichia coli, while human derived oxoguanine DNA glycosylase (hogg1) has been used less extensively. FPG-sensitive sites include 8-oxo-7,8-dihydro-guanine (8-oxoGua), 4,6-diamino-5-formamidopyrimidine (FapyGua) and 2,6-diamino-4-hydroxy-5-formamidopyrimidine (FapyAde). The incubation with hogg1 has been advocated to be specific for 8-oxoGua rather than FPG (5). ENDOIII detects oxidised pyrimidine lesions, including uracil glycol, thymine glycol, 5-hydroxycytosine and 5-hydroxyuracil (6). The incubation of nucleoids with DNA repair enzymes in the comet assay produces extra SBs (equivalent with longer comets in fluorescence microscopy) because the SBs that have been generated by DNA repair enzymes are added to the basal level of DNA SBs. Unfortunately, there is inconsistency in the manner in which results on oxidatively damaged DNA data are reported. It has been common practice to report the data as enzyme-sensitive sites, which implies extra lesions are generated by the enzyme treatment. Enzyme-sensitive sites are obtained as the difference between the enzyme treatment and the basal level of SB in the same sample. However, there are publications that have reported the total level of DNA damage after the enzymatic treatment (i.e. basal level of SB plus enzyme-generated SB). A number of publications describe that the DNA lesions have been measured by the modified comet assay. Presumably, this is to distinguish any measurements from the standard comet assay, which only measures SB. Nevertheless, this means that certain publications contain data on SB from one experiment and total level of DNA damage in a different experiment. The modified comet assay has thus been carried out with a procedure that has an additional incubation period with enzyme for 30 to 45 min at 37 C. It is important that this additional step has a control for non-enzymically generated DNA incisions (i.e. slides incubated with buffer only). The DNA damage in the buffertreated slides is subtracted from that in the enzyme-treated slides to obtain the enzyme-sensitive sites. For hazard identification, it might be relevant to report the total number of lesions, whereas it is a problem in the study of PM-mediated oxidatively damaged DNA because PM may generate SB by non-oxidative mechanisms as well as enzyme-sensitive sites by oxidative mechanisms. In this review, we distinguish between enzyme-sensitive sites and enzyme total sites. The latter refers to the total level of DNA damage that is measured by the enzymic treatment. This measure is regarded as having less significance than enzyme-sensitive sites in the overall evaluation of the associations between exposure to particles and oxidative stressgenerated DNA lesions. Validation of the comet assay The measurement of FPG-sensitive sites in cell cultures and peripheral blood mononuclear cells (PBMCs) has been validated by the European Standards Committee on Oxidative DNA damage (ESCODD) (7,8) and the European Comet Assay Validation Group (ECVAG) (9 13). The collective analysis of the data from ECVAG indicates that laboratories nowadays can detect SB and FPG-sensitive sites in coded samples in a concentration-dependent manner. The ESCODD trials were conducted in the early 2000s and demonstrated that only a few of the participating laboratories could detect concentration-dependent increases in the levels of FPG-sensitive sites in coded samples. Studies from mainly pharmaceutical and chemical toxicologists have focused on validation of the alkaline comet assay in genotoxicity testing to meet regulatory requirements (14 16). In addition, there are published recommendations on the comet assay on cultured cells and animal and humans cells (biomonitoring), although these do not contain a recommendation for analysis of oxidatively damaged DNA (17 20). The collective array of validation studies and recommendations emphasise the confidence gained in the comet assay as reliable tool for genotoxicity measurements. Variation in air pollution particles and ENMs The PM fraction of air pollution is a complex mixture, consisting of particles with different size and chemical composition. The size fractions comprise PM with aerodynamic diameter <10 µm (PM 10 ), 2.5 µm (PM 2.5 ) or 100 nm [ultrafine particles (UFPs) or nanoparticles], while coarse particles have an aerodynamic diameter between 2.5 and 10 µm (PM ). The nature of air pollution particles is dependent on the source of emission as well as the atmospheric conditions in the geographical area where they have been collected. There can be difference in the magnitude of day-to-day exposure and composition of PM from the same location, which may have an impact on the level of genotoxicity (21). By use of model particles researchers can independently investigate the same sample and Standard Reference Materials (SRMs) from the National Institute of Standards and Technology have been widely used for this purpose. SRM1648 and SRM1649 were collected from urban areas in the late 1970s in Washington, DC, and St Louis, MO, USA, respectively. SRM1650 and SRM2975 are samples of diesel exhaust particles (DEPs) collected from the engine of a heat exchanger or the filtering of an industrial dieselpowered forklift, respectively. Despite large international research efforts, including OECD coordinated activities on safety of ENMs, there are yet to appear systematically published studies on genotoxicity of ENM as same reference materials from a central source, although a number of publications have used samples with the same labelling for titanium dioxide and carbon black (e.g. P25 and Printex 90 from Degussa, respectively). The most common fibrous ENMs studied are singlewalled carbon nanotubes (SWCNTs) and multi-walled carbon nanotubes (MWCNTs). Other commonly investigated ENMs are fullerene C 60, silicates [including silicon dioxide (SiO 2 )], zinc oxide (ZnO), nanosilver (AgNP) and nanogold (AuNP). Some of these are available commercially, whereas others have been manufactured specifically for the purpose of research. Measurement of effect size in studies of particlegenerated DNA damage Studies of DNA damage following exposure to air pollution PM or ENMs utilise statistical models to distinguish for differences between samples from exposed and unexposed cells, animals or humans. This is crucial as unexposed cells also contain DNA lesions. The most informative data would be the net increase in DNA lesions per unit of concentration or dose (e.g. lesions/10 6 bp per mass concentration

3 Applications of the comet assay in particle toxicology, 2015, Vol. 30, No of particles) and preferably with a description of the complete dose/ exposure response relationship. However, the primary endpoint in any comet assay is DNA migration in agarose gels, which in turn is dependent on the assay conditions. It is possible to transform the primary comet assay to lesions per 10 6 bp by using ionising radiation as a reference standard (22). However, this is rarely done, with the vast majority of publications reporting DNA damage in primary comet assay units such as tail length, % DNA in the comet tail or tail moment. Moreover, the complete exposure response relationships are rarely reported. In the present review, we have used a standardised mean difference (SMD). This is the difference between exposed and unexposed groups divided by the common standard deviation (SD). The data from the highest exposure group have been used here for the assessment of SMD if the study contained more than one group of exposure. As such, the SMD provides information on the magnitude of effect in SD units. For instance, SMD = 1 means that the difference in DNA damage between two groups was the same as the SD. The SMD is calculated in selected studies in order to compare the effect size between ambient air pollution and ENMs. It is assumed that the interquartile range is similar to the SD in studies that have reported the data as median and quartiles. A high SMD implies a large effect, but it can also be obtained by a low SD. The SMD is not dependent on the number of observations because this does not affect the level of genotoxicity and SD. However, the SMD does provide information about the statistical power in the study. For instance, a study design with four groups will require 28 (SMD = 1), 13 (SMD = 1.5), 8 (SMD = 2), 4 (SMD = 3) or 3 (SMD = 3.75) observations in each group to obtain statistical power to reject the null hypothesis with α = 0.05 and β = A typical cell culture study could have 3 4 independent experiments, whereas the group sizes in studies on animals/humans are usually higher. It should be noted that the statistical outcome in the original publications may differ from the estimate of SMD here due to transformation of the raw data to achieve the normal distribution or analysis by non-parametric tests. In this review, the assessment of SMD for the highest exposure implies that the level of genotoxicity between studies may differ due to the differences in maximal exposure. Nevertheless, there is a tendency that the SD increases proportionally with the level of DNA damage in the comet assay. The calculation is therefore somewhat robust to the differences in exposure levels between studies. Measurement of DNA damage in human cells after air pollution exposure DNA damage levels in human air pollution studies have typically been evaluated in leukocytes, lymphocytes, PBMCs or airway epithelial cells. This genotoxicity may occur as the leukocytes enter a pro-oxidant and pro-inflammatory environment in the pulmonary vascular bed of PM-exposed subjects. These investigations have used different designs that can be segregated into controlled exposures, panel or cross-sectional studies. The controlled exposure studies have well-defined measurement of air pollution components at personal level, whereas they typically have limited numbers of subjects. The panel and cross-sectional studies are useful for the assessment of associations between air pollution exposure and genotoxicity because the exposure contrast is achieved by variation over time or location, respectively. Irrespective of the study design, the studies on air pollution exposure in humans have assessed the exposure differently with a number of studies mainly reporting the mass concentration of total suspended particles (TSPs), PM 10 or PM 2.5 or the particle number concentration of UFP. Polycyclic aromatic hydrocarbons (PAHs) or benzene metabolites have been used as personal proxymeasures of air pollution exposure in panel and cross-sectional studies. The PAH metabolites include 1-hydroxypyrene (1-HOP), whereas the benzene metabolites include S-phenylmercapturic acid (S-PMA) and trans, trans-muconic acid (tt-ma). The supplementary Table S1, available at Mutagenesis Online, outlines the studies on the associations between air pollution exposure and levels of DNA damage in human cells, while Figure 1 depicts blood cell effects from humans in studies where it is possible to obtain information on means and SD. A study from Copenhagen, Denmark, used a design with controlled personal exposure of cyclists to traffic-related UFP for ~90 min in the laboratory or in traffic-intense streets. This study demonstrated an association between personal UFP exposure and levels of FPG-sensitive sites in PBMC, whereas there was no association with levels of SB (23). Additionally, the same team using a controlled exposure set-up in a chamber with air from a traffic-intense street demonstrated a correlation between the level of particles (median size mode of 57 nm, representing carbonaceous soot) and levels of SB and FPG-sensitive sites in PBMC, whereas a size mode of 23 nm (representing semi-volatile organic compounds of diesel exhaust) was only associated with elevated levels of FPG-sensitive sites (24). However, the use of a panel study design showed no association between personal exposure to PM 2.5 and levels of SB and FPG-sensitive sites in lymphocytes (25). Likewise, a cross-sectional study exhibited no association between urinary excretion of S-PMA and levels of SB, ENDOIII- and FPG-sensitive sites in lymphocytes (26). These panel and cross-sectional studies had relatively small exposure contrasts, which could intensify the need for a controlled exposure design to avoid effects of confounding factors in panel and cross-sectional studies. Studies from Florence, Italy, demonstrated a positive correlation between ozone concentrations and levels of SB in leukocytes from healthy subjects, whereas no effect on FPG-sensitive sites was observed (27,28). There was also a positive association between ambient air concentrations of ozone and SB levels in nasal epithelial cells, with Florence subjects exhibited higher SB levels compared to subjects from a town with low ozone level (29). Policemen from Rome, Italy, had similar levels of SB in PBMC as compared to a control group of office workers (30). Non-smoking subjects from Athens, Greece, had higher levels of SB in lymphocytes as compared to subjects in a rural area, whereas there was no difference related to air pollution in smokers (31). Subjects from an industrialised region of Flanders, Belgium, had higher SB levels in leukocytes as compared to subjects in low polluted areas, with the SB levels correlating with ozone levels as well as urinary excretion of 1-HOP and tt-ma (32,33). Another set of experiments from Belgium showed that subjects living in air-polluted regions had high levels of SB in leukocytes, whereas there was no correlation between exposure markers (tt-ma and 1-HOP) and SB levels (34). A series of cross-sectional studies from the Czech Republic have assessed DNA damage in subjects exposed in occupational settings with different levels of air pollution. Using a combined panel and cross-sectional study design, it was demonstrated that policemen had higher levels of SB and FPG/ENDOIII sites in lymphocytes in the season with high levels of air pollution as compared to the season with low air pollution (35). Bus drivers from Prague had higher levels of SB in lymphocytes than the control group of office workers, whereas there were no associations between levels of air pollution and ENDOIII/FPG-sensitive sites (36). Other studies from air pollution exposed subjects in the Czech Republic have shown that personal exposure to particles correlated with levels of SB in lymphocytes (2), whereas there was no association between air pollution

4 70 P. Møller et al., 2015, Vol. 30, No. 1 Figure 1. Effect size of DNA SBs and FPG-sensitive sites in blood cells from humans. The bars represent the SMD. There was a negative value for the study in Czech Republic to the far right (SMD = 2.18), which is depicted as zero in the upper graph. Values of SMDs in single studies are shown in supplementary Table S1, available at Mutagenesis Online. exposure and levels of SB in lymphocytes from policemen in Prague (37) as well as leukocytes from mothers or children in the Teplice and Prachatice areas (38). A panel study of subjects, immigrating to Mexico City from rural settings, showed increased numbers of nasal cells with elevated levels of SB during the first 2 weeks of arrival (3,39). Subjects also had higher level of SB in nasal epithelial cells in the season with high air pollution level (40). A number of cross-sectional studies have displayed associations between air pollution (mainly ozone) exposure in Mexico and levels of SB in nasal epithelial cells, exfoliated tear duct cells and circulating leukocytes (3,39,41 43). While most human studies have been carried out in Europe and Mexico, the association between exposure to ambient air pollution and DNA damage was also investigated in a cross-sectional study in the Republic of Benin. This showed associations between urinary excretion of S-PMA and levels of SB and FPG-sensitive sites in PBMCs (44). Interestingly, the samples were analysed in Denmark and the difference in SMD resembles that of the particle number concentration in the two locations. The exposure contrasts were ~ and particles/cm 3 in the studies from Benin and Denmark, respectively. A study of policemen in Bangkok, Thailand, showed a correlation between exposure levels of 1,3-butadiene and SB levels in leukocytes (45). In addition, school children exposed to air pollution in Bangkok had higher levels of SB in leukocytes as compared to children in a provincial area (46 49). Studies from Brazil have demonstrated contradictory data with one study showing high levels of SB in lymphocytes from subjects living near an oil refining plant as compared to subjects from a city with little traffic and industry (50), while another study demonstrated no difference in SB levels in lymphocytes between subjects from urban industrialised and non-industrialised areas (51). The majority of data generated from human air pollution studies have demonstrated positive association between exposure to ambient air particles and levels of SB, ENDOIII- and FPG-sensitive sites. However, the magnitude of these effects differs substantially between the studies (Figure 1). The controlled exposure studies provide compelling evidence for the association between air pollution particle exposure and DNA damage. This is further supported by a number of panel and cross-sectional studies, although it should be acknowledged that these studies have compared exposures in residential or occupational settings, and they may have risk of confounding and exposure misclassification because of mainly area-based exposure assessment. DNA damage in animal tissue after exposure to air pollution particles There are relatively few studies on associations between air pollution exposures and comet assay endpoints in animal models (supplementary Table S2, available at Mutagenesis Online). Three studies observed increased levels of SB in rat lung tissue after intra-tracheal (i.t.) instillation of relatively high doses of air pollution particles from urban areas in China (7.5 and 37 mg/kg) (52 54). Another study using i.t. instillation of SRM1649, administered twice during 26 h (total dose = 1 mg/kg), showed unaltered levels of SB and FPGsensitive sites from lung tissue in mice (55). A number of studies have investigated the genotoxicity of DEP rather than authentic air

5 Applications of the comet assay in particle toxicology, 2015, Vol. 30, No pollution particles. I.t. instillation of SRM1650 (1.6 or 4.7 mg/kg in Guinea pigs) was associated with increased levels of SB in lung tissue (56). However, inhalation of SRM1650 in mice did not result in increased level of SB in bronchoalveolar lavage (BAL) cells after either a single expose to 20 or 80 mg/m 3 for 90 min or the same total dose administered on four consecutive days (57). Repeated inhalation exposure to SRM2975 was associated with increased levels of SB in BAL cells of Tnf knockout mice, whereas no effect was noted in wild-type mice (58). Nevertheless, another study employing the same dose and exposure period of SRM2975 resulted in no effect on SB and FPG-sensitive sites in BAL cells, lung or liver of Ogg1 knockout and wild-type mice (59). A study from Brazil showed that levels of SB in leukocytes of native rodents (Ctenomys minutus) correlated with environmental exposure to automobile emissions (60). However, free-living dogs from different regions of Sao Paulo, Brazil, had the same level of SB in cells from the olfactory or respiratory epithelium (61). Oral exposure of Big Blue rats to SRM1650 ( mg/kg feed) resulted in increased levels of SB in colon epithelial cells, liver and lungs (62,63). The same study showed no effect on ENDOIIIsensitive sites in the liver and colon epithelial cells, whereas the levels of both ENDOIII- and FPG-sensitive sites were increased in the lung (63). The increased level of SB in the liver was observed at a dose of 8 mg/kg feed, corresponding to ~7 mg/kg body weight, although there was no statistical significance in the levels of SB at the lowest dose (1.6-fold). The lack of statistical significance at the lowest dose could be attributed to the lack of sufficient power because there were only six rats in each group. However, oral exposure to SRM2975 (0.8 or 8 mg/kg feed, corresponding to 1.7 mg/kg body weight of the highest dose) did not change the levels of SB, ENDOIIIor FPG-sensitive sites in colon epithelial cells, liver and lung of Ogg1 knockout and wild-type mice (64). The mechanism of genotoxicity in the liver or lungs after oral exposure to DEP is not clarified, but it might be directly related to translocated DEP or indirectly to altered uptake of nutrients or altered microbiome-related metabolism of ingested compounds. Intraperitoneal (i.p.) injection of SRM2975 ( mg/kg) increased the level of FPG-sensitive sites in the liver of mice at the lowest dose, whereas no effect was noted in terms of SB or ENDOIII-sensitive sites (65). In general, there seems to be a tendency that doses above 1 mg/ kg have been associated with statistically significant increased levels of DNA damage in lung tissue after airway exposure. This should not be regarded as a threshold of effect, but rather that PM from air pollution or DEP generates a sufficient number of lesions to be statistically significant with the commonly used group size of animals. This is exemplified by increased levels of SB observed at a dose of 7 mg/kg in the liver after oral exposure to DEP, whereas increased levels of FPG-sensitive sites could be detected at a dose as low as 0.05 mg/kg (62,65). DNA damage in animal tissues after exposure to ENMs The pulmonary exposure to ENMs is principally via i.t. instillation or inhalation routes (supplementary Table S3, available at Mutagenesis Online). I.t. instillations of 0.2 mg/mouse of carbon black (Printex 90) or fullerene C 60 were associated with increased levels of SB in lung tissue of mice (66). Exposure to Printex 90 ( µg/mouse by i.t. instillation) was also associated with increased levels of FPG-sensitive sites in lung tissue at Day 1 and 3 post-instillation (67). These levels were similar to the controls at Day 28 after the exposure (67). Similarly, i.t. instillation of fullerene C 60 (maximal dose = 2.5 mg/kg) or nanosized TiO 2 (maximal dose = 5 mg/kg) was not associated with an increased level of SB in lung tissue of rats after either single or repeated dose exposure (68,69). SWCNTs did not increase the level of SB or FPG-sensitive sites in mice after administration by i.t. instillation of 0.5 mg/kg at both 26 and 2 h before sacrifice (total dose = 1 mg/kg) (55). However, i.t. instillation of MWCNT (0.05 or 0.2 mg/kg) was associated with an increased level of SB in mouse lung tissue at 3 h after the administration (70). Inhalation exposure to TiO 2 (maximal exposure = 28.5 mg/m 3, 4 h/ day for 5 days) did not increase SB levels in alveolar type II and Clara cells from the lung of mice (71). There were also unaltered levels of SB and FPG total sites in lung cells from mice and rats after inhalation exposure to nanosized carbon particles ( µg/m 3 for 4 h, or 3 4 h on three consecutive days) in mice and rats (72). I.t. instillation of AuNP did not increase the level of SB in lung tissues of rats after a relatively low dose (18 µg/rat) at 72 h after the exposure (73). Figure 2 depicts the SMD of SB or FPG-sensitive sites in lung tissue after i.t. instillation of ENMs or air pollution particles, integrated from multiple studies. There appears to be a dose-dependent increase in levels of SB after exposure to carbon nanotubes (CNTs). There are increased levels of SB in lung tissue after exposure to carbon black, although the dose response relationship is less obvious as compared to that of CNTs. Figure 2 also suggests that i.t. instillation of fullerene C 60 or nanosized TiO 2 generates fewer SB in lung tissue than carbon black and CNTs at the same dose. There are relatively few studies focusing on FPG-sensitive sites subsequent to i.t. instillation, while the dose response relationship can only be assessed for nanosized carbon black. As an overview of ENMs airway exposure only i.t. instilled carbon black and MWCNT demonstrated increased levels of SB and FPG-sensitive sites in lung tissue. A number of studies have assessed levels of SB in BAL cells after airway exposure to ENMs. It was shown that inhalation of Printex 90 (20 mg/m 3 for 90 min on four consecutive days) was associated with increased levels of SB in BAL cells (58). Another study indicated that the levels of SB in BAL cells were decreased after repeated i.t. instillations (268 µg/mouse as total dose after four exposures) of Printex 90 (74). However, a single i.t. instillation of Printex 90 ( µg/mouse) was associated with increased levels of SB in BAL cells (67). It has also been shown that there were increased levels of SB in BAL cells after i.t. instillation of 54 µg/mouse of Printex 90, whereas AuNPs and fullerene C 60 did not increase the level of SB (75). The same protocol was used to study SB levels in BAL cells after i.t. instillation of pure TiO 2, SiO 2 and Printex 90 as well as sanding dust generated from boards coated with paints or lacquers with conventional pigments or ENMs (76,77). There were no differences in SB levels in BAL cells from mice exposed to dust from paints and lacquers with or without ENMs, whereas some of the pure nano and fine TiO 2 samples and sanding dust from SiO 2 -containing lacquers generated SB in BAL cells (76,77). These studies are relevant in nanotoxicology because they involve both pure ENMs and the ENM-containing products in the test model. Nevertheless, the composition of BAL cells differs in exposed and unexposed animals because of influx of inflammatory cells into the airways. Therefore, it is uncertain whether the increased levels of SB in BAL cells are because of particle-generated genotoxicity or differences in DNA damage between the subgroups of BAL cells. I.t. instillation of TiO 2 (20 mg/kg twice weekly for 12 weeks) in mice was associated with increased levels of SB in circulating PMBCs (78). It was also shown that inhalation exposure (42 mg/m 3 for 1 h/ day on gestation days 8 18) in mice resulted in increased SB levels in

6 72 P. Møller et al., 2015, Vol. 30, No. 1 Figure 2. Dose response relationship of DNA SBs or FPG-sensitive sites in lung tissue after i.t. instillation of particles in animals. The symbols represent the SMD. Values of SMDs in single studies are shown in supplementary Tables S2 and S3, available at Mutagenesis Online. the liver of dams and offspring, whereas there were unaltered levels of FPG-sensitive sites in the liver of the offspring (74). A single oral exposure to high doses of Fe 2 O 3 ( mg/ kg) was not associated with increased levels of SB in leukocytes from albino Wistar rats at 6, 24, 48 or 72 h after the exposure (79). Similarly, oral exposure to nanosized ZnO (50 mg/kg for 1 h) did not alter the level of SB in leukocytes, bone marrow or spleen of mice, whereas it ameliorated SB generation by ionising radiation in these cells (80). A study of repeated oral exposure to ZnO ( mg/ kg administered three times during 45 h) resulted in no effect on SB levels in the stomach or liver tissue of rats (81). However, oral exposure to ZnO (600 or 1000 mg/day for five consecutive days) was associated with an increased level of SB in cardiac tissue of rats (82). It was also reported that oral exposure to ZnO (50 or 300 mg/ kg for 14 days) increased the level of FPG total sites in the liver of rats, whereas there was no effect in the kidney (83). Ingestion of TiO 2 in the drinking water (500 mg/kg for 5 days) was associated with increased levels of SB in blood cells of mice (84). In addition, oral gavage of TiO 2 ( mg/kg for 7 days) in mice was associated with elevated SB levels in the liver and bone marrow, whereas there were unaltered SB levels in brain tissue (85). It has also been shown that a 14-day oral exposure to TiO 2 was associated with elevated SB levels in the liver of mice at doses of 10, 50 and 100 mg/kg and increased levels of FPG total sites at doses of 50 and 100 mg/kg (86). A 28-day repeated oral exposure study to MnO 2 ( mg/kg) also showed increased levels of SB in leukocytes of rats (87). Overall, the studies on oral exposure tend to demonstrate increased levels of DNA damage, although the doses of ENMs have been relatively high. A number of studies have assessed the effect of intravenous (i.v.) injection of ENMs in animal models. I.v. injection of AgNPs (5 or 10 mg/kg) was associated with an increased level of SB in epididymal sperm cells at 24 h after the exposure (88). However, i.v. injection of AgNPs or TiO 2 (5 10 mg/kg) was not associated with increased levels of SB in bone marrow cells (89). A single exposure to AuNPs (36 µg/rat) was associated with increased levels of SB in the kidney at 1 h after administration, whereas there were unaltered levels in the liver and spleen in rats (90). Repeated exposure to AgNPs (4 40 mg/ kg every 5 day for 32 days) increased the levels of SB in leukocytes (91). The exposure to AgNPs (25 mg/kg for three consecutive days) was associated with genotoxicity in the liver of mice, measured by the ENDOIII- and hogg1-modified comet assay, whereas there was no difference in terms of SB (92). I.v. injection of AuNP ( mg/ kg administered at 48, 24 and 4 h before sacrifice) did not increase the level of SB in the liver, spleen and blood cells, whereas there was a modest increase in SB levels in the liver of rats after exposure of nanosized silica particles (93). I.p. injection of silica nanocrystals (5 or 50 mg/kg) in mice was associated with increased levels of SB in bone marrow cells at 24 h, whereas there were no effect at 3 h or 7 days after the exposure (94). One study exposed mice by i.p. injection to AgNP and observed increased levels of SB in the spleen, although this was attributed to the presence of an anionic surfactant in the dispersion (95). I.p. injection of Al 2 O 3 (34 74 mg/kg, corresponding to 30 60% of the LD 50 value) was associated with increased levels of SB in brain tissue of rats (96). The exposure to MWCNT by i.p injection ( mg/ kg once a day for 5 days) increased the level of SB in leukocytes in a dose-dependent manner (97).

7 Applications of the comet assay in particle toxicology, 2015, Vol. 30, No The exposure to ENMs in animals has generated mixed results with regard to generation of SB and oxidatively damaged DNA. So far, only with carbon-based ENMs has increased generation of SB in lung tissue after airway exposure been demonstrated, which is similar to the dose response relationships of DEP or air pollution particles (also show a high content of elemental carbon) (Figure 2). Importantly, the studies on i.t. instillation indicate that particles have different DNA damage potential. CNTs and carbon black seem to be the most potent on mass concentration basis in comparison to fullerene C 60 and nanosized TiO 2. The studies on oral exposure have shown increased SB level in other tissues than the gastrointestinal tract. This suggests uptake of ENMs or transmission of effect by other mechanisms than direct ENM-mediated DNA damage. However, it should be acknowledged that the exposures in these studies have been very high (above 30 mg/kg, which is higher than the DEP exposure studies maximal 7 mg/kg). I.v. exposure studies have used doses above 5 mg/kg as single or repeated doses. The doses by i.p. injection are relatively high as well and it is a non-physiological route of exposure. The relevance of these very high exposure doses in risk assessment is not clear. With regard to airways exposure, it is important to note that high doses may be associated with pulmonary overload, which should be considered in the interpretation of any data on DNA damage. On a similar note, the administration of high doses systemically should be justified because they may not be relevant if humans are exposed from external sources. The studies do not yet allow assessment of specific ENM characteristics such as size, shape, suspension medium/protocol or agglomeration state that determine the genotoxic potential in animals. DNA damage in cell cultures after exposure to air pollution particles Many studies have investigated levels of DNA damage in cells exposed to authentic air pollution particles and reference materials (supplementary Table S4, available at Mutagenesis Online). The exposure of A549 or THP-1 cells to aqueous suspensions of PM from traffic-intense streets was associated with increased levels of SB and FPG-sensitive sites (98 100). Suspensions of urban PM 2.5 were more potent compared to PM 10 on mass basis for SB generation in BEAS-2B cells, whereas there was no effect on FPG total sites (101). The levels of SB and FPG total sites in A549 cells were increased after exposure to aqueous extracts of PM from an industrial site as compared to particles from a highway site (102). There were higher levels of SB and FPG-sensitive sites in both A549 and THP-1 cells after exposure to PM from a town with many wood stoves as compared to samples from a rural setting in Denmark (103). There are numerous studies that have only assessed the effects of aqueous suspensions of ambient air PM on the level of SB in various cell types. Air pollution particles generated SB in A549 cells, leukocytes and lymphocytes (104,105). Moreover, SRM1649 that had been treated with solvents to remove organic molecules generated SB in human fibroblasts (106). However, washed particles of SRM1648 had less SB generation in THP-1 and A549 cells as compared to untreated SRM1648 (107). Nevertheless, there is persuasive evidence indicating that aqueous suspensions of air pollution particles generate SB in cell cultures. This is further supported by observations of associations between increased levels of SB and exposure to air pollution particles collected from ambient air in urban areas of east Asia (53,108,109), Europe (106, ), Mexico (104,119,120), USA (121) and Argentina (122). A number of studies have shown increased levels of DNA damage in cell cultures after exposure to DEP. It was shown that DEP from conventional or biodiesel increased the level of SB and FPG-sensitive sites in A549 cells (123). In addition, exposure to SRM1650 and SRM2975 have been associated with elevated levels of SB, FPG- and hogg1-sensitive sites in A549, THP-1, HepG2 and endothelial cells (57,98,99, ). Figure 3 depicts the effect size in studies that have investigated the genotoxic effect of SRM1650 or SRM2975 in cultured cells. Interestingly, these studies differ greatly in SMD of SB levels (range ) compared with the SMD of FPG-sensitive sites (range ). DNA damage in cell cultures after exposure to ENMs There are numerous ENMs that have been investigated in cell cultures for their potential genotoxicity, exceeding the number of those utilising animal models. Supplementary Table S5, available at Mutagenesis Online, illustrates comet assay endpoints from those studies on ENMs in cultured cells that have also been investigated in animal models. These include fullerene C 60, SWCNT, MWCNT, AuNP, AgNP, TiO 2, SiO 2, ZnO and nanosized carbon black as depicted in Figure 4. The studies on TiO 2 and silicates have shown mixed results, whereas the studies on other type of ENMs seem to show predominantly positive associations between exposure and increased levels of SB in cultured cells. The studies differ substantially in design and choice of statistical analysis. In addition, there is a tendency that reports on increased levels of genotoxicity have become more common after the establishment of nanotoxicology as a branch of toxicology. The investigations on TiO 2 are useful in this respect because they span the period before the comet assay became a standard test in nanotoxicology to the present day. The first of such investigations observed photogenotoxicity of TiO 2 in fibroblasts or lymphoma cells, albeit at very high concentrations (800 and 3200 µg/ml), as well as generation of SB in lymphoma cells in the absence of ultraviolet light (128,129). These observations were later supported by studies of photogenotoxicity in keratinocytes, whereas there were unaltered levels of SB when the cells were incubated with either anatase or rutile form of TiO 2 in the dark (130). Another study showed higher levels of SB in TiO 2 exposed Caco-2 cells when the samples had been processed in the comet assay under normal laboratory light, whereas the genotoxicity level was lower for samples that were processed in the dark (131). Figure 5 depicts the accumulation of publications on nanosized TiO 2 in cell cultures that have investigated levels of SB. There are approximately the same number of studies showing increased level of SB in cell cultures after exposure to TiO 2 ( ) as there are studies that have shown no or mixed effects of TiO 2 ( , ). There seems to be no simple explanation for the discrepancy between the studies with regard to the type of cultured cells, concentrations of TiO 2 or incubation period. The observation that laboratory light has an influence on the generation of SB by adventitious photogenotoxicity cannot be assessed in all publications because there is insufficient information about such assay procedures. In addition, it has been speculated that the quantity of serum or protein in the dispersion solution is an important factor for genotoxicity (144). Figure 6 shows the effect sizes of SB levels in cell cultures after TiO 2 exposure with different primary particle size and crystal structure. There seems to be a tendency that anatase nanosized TiO 2 (SMD = 3.45) is more potent than fine particles (SMD = 2.06). Most of these studies have used either 3 6 h incubation time (SMD = 2.17, n = 15) or 24 h (SMD = 2.95, n = 20) for determination of SB levels (P = 0.58, Student s t-test with unequal variance between groups). Direct

8 74 P. Møller et al., 2015, Vol. 30, No. 1 Figure 3. Effect size of DNA SBs and FPG-sensitive sites in cell cultures that have been exposed to DEPs. The effect size (SMD) for each publication is reported in supplementary Table S4, available at Mutagenesis Online. Each column represents one study. 35 Increased SB level Unaltered SB level Number of publications TiO2 CNT ZnO AgNP Silica CB AuNP C60 Figure 4. Distribution of studies showing either positive or null effect on levels of SB after exposure of cultured cells to ENMs. comparisons have shown that TiO 2 samples of different size (nano or fine size) and crystal structure (anatase or rutile) all generated SB in A549 cells; however, there was no clear relationship between the particle characteristics and SB level (143). A similar study showed that nanosize and fine anatase TiO 2 generated SB in Syrian hamster embryo cells, whereas rutile forms tended to generate lower levels of SB than anatase forms (146). There appears to be an increasing number of publications on TiO 2 have reported elevated levels of DNA damage after incubation with ENDOIII, FPG or hogg1 (131,132,135,136, ,148,149,154,156) as compared to those showing no effect (145,151,161,162,165). Only two of these studies reported the results as FPG-sensitive sites (135,144), whereas the other studies reported the levels of genotoxicity as FPG total sites. Nevertheless, these studies would probably have detected a higher level of FPG-sensitive sites, although they might not have been statistically significant. We have recalculated the original results to FPG-sensitive sites in Figure 7. All studies show elevated levels of FPG-sensitive sites in TiO 2 -exposed cells as compared to controls. However, this effect does not appear to depend on the cell type.

9 Applications of the comet assay in particle toxicology, 2015, Vol. 30, No Studies on silica materials seem to have reported both negative (131,160, ) and positive results on levels of SB in cultured cells ( ). The null effect studies on SB in cultured cells after exposure to silicates have generally high quality with assessment of genotoxicity in multiple concentrations, time points or cell types as well as use of H 2 O 2 as positive control (160,167,168,170). In addition, these studies have also shown unaltered levels of oxidatively damaged DNA as measured by the FPG treatment ( ). One of the positive studies showed a concentration-dependent increased level in human endothelial cells by exposure to nanosized silica (174). However, other positive studies appear to have based their conclusion on single experiments where the total number of comets per gel has been used in the statistical analysis ( ). The use of single comets rather the cell cultures as experimental unit for the statistical analysis is considered to be wrong because comets from the cells gel have not been randomly assigned to the treatment or comet assay procedure and it greatly inflates the statistical power (175). This may also explain the apparent discrepancy in effect size on SB levels reported in different studies on nanosized silica (Figure 8). The majority of studies on nanosized ZnO have shown positive association between exposure and increased levels of SB (131,135,137,148,149,171, ). The assessment of total sites by treatment with FPG or hogg1 has likewise shown mainly elevated levels after exposure to nanosized ZnO (131,135,148,149, ). The studies on CNTs also show mainly increased levels of SB in cultured cells (124,135,148,149,171, ), although there are also a few null effect studies (192,193). The association between CNT exposure and levels of FPG-sensitive sites has likewise been conflicting (124,135,187,193) as have studies on FPG total sites (148,149,188,190). There are relatively few studies that have investigated the exposure to fullerene C 60 in cell cultures. It has been shown that fullerene C 60 increased the level of SB in lymphocytes (194). Another study showed no effect of exposure to fullerene C 60 on SB levels in lung epithelial cells, whereas there were increased levels of FPG-sensitive sites (193). There were increased levels of both SB and FPG-sensitive sites in HepG2 cells after exposure to fullerene C 60 (124). The studies on nanosized carbon black have mainly shown increased SB levels (107, ,171,173,191,193, ), although there are also null effect studies (115,131,198). The effects on FPG-sensitive sites in carbon black exposed cells are mixed as demonstrated by null effect in Caco-2 cells (131) and Accumulated number of publications Positive Negative/mixed Publication year Figure 5. Accumulation of studies on TiO 2 generated DNA SBs in cultured cells. 100 SMD of DNA strand breaks Nano Anatase Nano P25 Nano Rutile Nano Unknown Fine Anatase Fine Rutile Figure 6. Effect size of DNA SB generation in cultured cells after exposure to TiO 2. The symbols represent the SMD. Values for SMDs in single studies are shown in supplementary Table S5, available at Mutagenesis Online. The median value for each group is indicated by the vertical line.

10 76 P. Møller et al., 2015, Vol. 30, No Lung Liver Skin Liver SMD of FPG sensitive sites Gonade Liver Colon Lymp Lung Colon Skin Skin Kidney Testis Figure 7. Effect size of FPG-sensitive sites in cultured cells that have been exposed to nanosized TiO 2. The results in terms of FPG-sensitive sites have been recalculated from original results. The effect size (SMD) for each publication is reported in supplementary Table S5, available at Mutagenesis Online. Each column represents one study and the cell type is specified. SMD of strand breaks Skin positive associations in HepG2, lung epithelial and endothelial cells (124,196). The majority of studies on AgNP have shown positive results on SB in cultured cells (148,149,169, ), although there also are some studies that have shown null effect (162,205,206). Likewise, there appear to be more studies that have shown elevated levels of lesions that are detected by the FPG enzyme (148,149,199,205) than studies that have shown no effect (162,169). The studies on AuNP have shown increased levels of SB in cultured fibroblasts, small airway epithelial and HepG2 cells ( ). Discussion Fibroblast HUVEC Colon Skin Figure 8. Effect size of DNA SBs in cultured cells that have been exposed to nanosized silica. The effect size (SMD) for each publication is reported in supplementary Table S5, available at Mutagenesis Online. Each column represents one study and the cell type is specified. The studies on airway exposure to air pollution particles have shown increased levels of DNA damage in circulating blood cells, which is supported by observations of genotoxicity in lung cells of exposed animals and cultured cells. The effect size in humans is typically <1 SD unit, indicating that a relatively large group size is required to show significant levels of effects. This is supported by observations that 8-oxodG and FPG-sensitive sites had similar effect size in circulating Lymp Colon Lung Fibroblast Lung blood cells from air pollution exposed subjects (212). There are studies on ENM exposure in animals that have shown increased levels of DNA damage in lung tissue. The exposure to nanosized carbon black (Printex 90) and CNTs seems to generate higher levels of DNA damage in lung tissue after single-dose i.t. instillation than fullerene C 60, TiO 2 and combustion-derived particles on the same mass concentration, although it should be emphasised that it is based on relatively few studies. There are only few studies on inhalation exposure to ENMs, which have shown unaltered levels of DNA damage in lung tissue after exposure to nanosized carbon particles and TiO 2 (71,72). Inhalation exposure is obviously more relevant than i.t. instillation as a way of administration of ENMs and combustion-derived particles. In addition, high bolus doses by i.t. instillation or long-term inhalation may lead to pulmonary overload in rats as reviewed extensively for exposure to poorly soluble particles such as carbon black, TiO2, talc and coal dust ( ). This is typically explained by a mechanism where deposition of high doses of particles in the lungs is associated with impairment of macrophage-mediated lung clearance, accumulation of excessive lung burdens of particles and initiation and propagation of inflammatory responses to which rats are particularly vulnerable. In situations of chronic inflammation, the on-going release of reactive oxygen species from pulmonary macrophages and neutrophils can damage the lung tissue and stimulates tissue repair, increasing the risk of DNA transcription errors as well as directly causing DNA damage. Therefore, elevated levels of DNA damage in high-exposure studies may have been caused by inflammatory reactions that would not be relevant at low-dose exposures that are relevant to the human exposure situation. Still, it has been shown that the net increase in oxidatively damaged DNA in lung tissue increases with dose of particles in even the low-dose area (216,217). Cell culture studies are generally associated with larger effects than animal studies on the same type of particles irrespective of the source (from combustion processes or ENMs). It is interesting that reports on TiO 2 -generated DNA damage seem to coincide with the establishment of nanotoxicology as a research field. This raises some concerns related to publication bias, although this can also be related to the development of better dispersion protocols or selection of more potent nanosized TiO 2 samples. Another odd discrepancy is most notable in the studies of nanosized silica that show either very high effect of SB generation or virtually no effect at all. Certain studies with very high effect of nanosized silica appear to have been based on only one experiment and the