Defining the in vivo mechanism of air pollutant toxicity using murine stress response biomarkers

Background Air pollution can cause a wide range of serious human diseases. For the informed instigation of interventions which prevent these outcomes there is an urgent need to develop robust in vivo biomarkers which provide insights into mechanisms of toxicity and relate pollutants to specific adverse outcomes. Objective To exemplify the application of in vivo stress response reporters in establishing mechanisms of air pollution toxicity and the application of this knowledge in epidemiological studies and potentially in disease prevention. Methods Murine stress-reporter models (oxidative stress/inflammation, DNA damage and Ah receptor -AhR-activity) and primary mouse and primary human nasal cells were exposed to chemicals present in diesel exhaust emissions, particulate matter (PM) standards (PM2.5-SRM2975, PM10-SRM1648b) or fresh roadside PM10. Stress reporter activity was analysed by luminescence assays and histochemical approaches in a panel of murine tissues. Biochemical, genetic and pharmacological approaches were used to establish the mechanism of the stress responses observed. Pneumococcal adhesion was assessed in exposed primary human nasal epithelial cells (HPNEpC). Results Nitro-PAHs induced Hmox1 and CYP1a1 reporters in a time- and dose-dependent, cell- and tissue-specific manner. NRF2 pathway mediated this Hmox1-reporter induction. SRM1658b, but not SRM2975, was a potent inducer of NRF2-dependent Hmox1 reporter activity in lung macrophages. Combined use of HPNEpC and in vivo reporters demonstrated that London roadside PM10 particles induced pneumococcal infection in HPNEpC mediated by oxidative stress responses. Discussion The combined use of in vivo reporter models with HPNEpC provides a robust approach to define the relationship between air pollutant exposure and health risks. These models can be used to hazard ranking environmental pollutants by considering the complexity of mechanisms of toxicity. These data will facilitate the relationship between toxic potential and the level of pollutant exposure in populations to be established and potentially extremely valuable tools for intervention studies.

body at a particular timepoint and lack the cellular and tissue resolution required 150 to draw meaningful interpretations of the responses observed. 151 152 To overcome these knowledge and methodological gaps, our lab has validated 153 a panel of reporter models to detect in vivo the activation of major stress 154 pathways linked to cellular toxicity, including oxidative stress/inflammation 155 (HO1 triple transgenic reporter, HOTT), DNA damage (p21 reporter) or the AhR 156 receptor activation (CYP1a1 reporter). These reporters provide easily 157 measurable readouts of in vivo cellular responses to toxic insults, with high-158 resolution, and in a tissue-and cell-specific manner [23][24][25]. In these reporters, 159 a short viral DNA sequence, known as a T2A sequence, is exploited to provide 160 multiple reporter molecules to be expressed (separately) off the endogenous 161 gene promoter. Recently, we have demonstrated the utility of these models as 162 biomarkers for inorganic arsenic toxicity, a widespread water contaminant 163 classified as a class 1 carcinogen [26]. We hypothesized that our stress 164 reporters could be used as a complementary assay in environmental 165 epidemiological studies to gain in vivo mechanistic insights into the 166 associations observed.

168
Materials and Methods.

172
Marylebone Road (London, UK) PM10 (MR-PM10) was collected using a high-173 volume cyclone [1]. The cyclone was placed kerbside and within 1 metre of 174 traffic, on various days throughout summer 2019 for 5 -8 hours per day. 175 Collections were pooled and stored at room temperature in a sterile glass 176 container. Aliquots of MR-PM10 were diluted in Dulbecco's phosphate-buffered 177 saline (DPBS) to a final concentration of 1 mg/mL and stored as master stock 178 at -20°C. 179 180 Animals.

182
All animals used in this study were supplied from the Medical School Resource 183 Unit, University of Dundee, on a C57BL/6N background. Animals were 184 subjected to the following husbandry conditions: mice were housed in 185 temperature-controlled rooms at 21°C, with 45-65% relative humidity and 186 12h/12h light/dark cycle. Mice had ad libitum access to food (RM1 for stock 187 mice; RM3 for mating females; Special Diet Services, 801010 and 801700 188 respectively) and water. Animals were regularly subjected to health and welfare 189 monitoring as standard (twice daily). All cages had sawdust substrate and 190 sizzle-nest material provided. Environmental enrichment was provided for all 191 animals. 192 193 All animal work described was approved by the Welfare and Ethical treatment 194 of Animals Committee of the University of Dundee. Those carrying out this work 195 did so with Personal and Project Licenses granted by the UK Home Office 196 under the Animals (Scientific Procedures) Act 1986, as amended by EU 197 Directive 2010/63/EU. Animals in study plans were inspected regularly by staff 198 trained and experienced in small animal husbandry, with 24-hour access to 199 veterinary advice.

201
Animal numbers were guided by power calculations (G*Power;202 www.gpower.hhu.de), pilot experiments, and previous experience, and 203 experimental design was undertaken in line the 3Rs principles of replacement, 204 reduction, and refinement (www.nc3rs.org.uk).

206
Animal studies design. 207 208 Data in this paper was obtained using male mice, unless otherwise stated. Mice 209 were 12-15wo littermates or age-matched to within 3 weeks of each other. All 210 animals used in this study were heterozygous for the CYP1a1_KI_Cre, HOTT 211 or p21 reporter allele unless otherwise specified. Animals were randomly 212 assigned to control or treatment groups; analysts were not blinded to the 213 identity of biological samples. At the end of studies, all animals were sacrificed 214 according to Schedule 1 of the Animals (Scientific Procedures) Act 1986, by 215 exposure to a rising concentration of CO2 and death confirmed by 216 exsanguination, except animals for primary cells studies (cervical dislocation 217 and confirmation of permanent cessation of the circulation MCF7_AREc32 cells were cultured as described before (Wang et al., 2006). 352 Briefly, cells were maintained in growing media (DMEM + Glutamax, 10% FBS, 353 pen/strep 5ml) supplemented with 0.8mg/ml G418 (Roche, G418-RO) for 354 maintenance or removed for exposure to chemicals. 96-well plates at 5000 355 cells/well.

357
To generate heterozygous HOTT reporter mouse embryonic fibroblasts 358 (HOTThet MEFs) crosses between homozygous HOTT male mice and C57wt 359 females were set up. MEFs were isolated from littermate mouse embryos at 360 day E12.5. The heads and internal organs were removed from the embryos. 361 The remaining tissues were then cut into small pieces, and incubated in trypsin 362 (5 min, 37oC). Resulting individual cells were then plated in 96well plates and 363 cultured in Dulbecco modified Eagle medium (DMEM) containing 10% serum, 364 2 mM l-glutamine, 1% penicillin/streptomycin and used up to passages 3 to 5.

366
Bone marrow precursor cells were extracted from mouse femurs under sterile 367 conditions. Bone marrows were flushed with PBS, cells washed with PBS and 368 resuspended in 10 plates/mouse containing 8ml of BMDM media (DMEM, 10% 369 FBS, heat inactivated, 1mM Pyruvate, 1x GlutaMAX, Pen/Strep and 370 supplemented with 10ng/ml murine M-CSF (Peprotech, 315-02) for plating onto 371 non-tissue culture treated petri dishes. After 4 days, 5ml of media was added. 372 After 7 days adherent macrophages were scraped from the petri dishes with 373 versene (Gibco, 15040) and then plated into tissue culture grade plastic 96 well 374 dishes and left for 24 h to re-attach before exposure to chemical compounds.

398
Cell based assays. 399 400 HOTT or ARE_luc luciferase reporter activity was measured using the 401 Luciferase Assay System (Promega, E1500), according to the manufacturer's 402 instructions and luminescence quantified using the Orion II Microplate 403 Luminometer (Berthold Detection Systems).

405
The CellTiter 96 AQueous One Solution Cell Proliferation Assay (Promega, 406 G3582) was used to determine cell viability, as described by the manufacturer. 407 Briefly, 20ul of the reagent was incubated with 100ul of the cell culture media 408 (96 wells) and after 4h the sample absorbance was read at 490nm.

410
The cell glutathione content was determined using the GGSH/GSSH-Glo assay 411 (Promega; V6611) according to manufacturer instructions and luminescence 412 quantified as described above.

494
The nitro-PAHs (1-NP and 3-NFA), but not other tested PAHs, increased NRF2 495 protein levels ( Figure 1D) and the expression of a bona-fide NRF2 target 496 protein, AKR1C1/2. These results indicated that nitro-PAH compounds trigger 497 a NRF2 response. To gain further insight into the mechanism of activation we 498 modulated intracellular redox status in MCF7-AREc32 cells by pre-incubation 499 with the glutathione depleting agent, buthionine sulphoximine (BSO) prior to 500 exposure to 3-NFA. BSO alone did not increase reporter activity ( Figure 1E) 501 [30]. However, the combination of BSO with 3-NFA increased the induction 502 from 4 to a 20-fold. Under these experimental conditions, BSO caused an 80% 503 decrease in the GSH/GSSG ratio ( Figure 1F). 50 µM 3-NFA alone for 24h only 504 had a negligible effect in the GSH/GSSG ratio ( Figure 1G). 505 506 We then established whether primary cell cultures derived from the HOTT 507 reporter mice could be used to identify potentially toxic components of air 508 pollution mixtures. In primary mouse embryonic fibroblasts (MEFs), luciferase 509 activity was increased 3-fold when exposed to 20µM tBHQ for 24h (Figure 2A). 510 When these cells were exposed to different DEP compounds, only 3-NFA (50 511 µM) caused a 2-fold induction of luciferase activity. Only oxy-PAHs were highly 512 cytotoxic to MEF cells at concentrations >1µM ( Figure S1B) and in the absence 513 of reporter activity ( Figure S1C).

515
We next studied the ability of DEP components to activate the HOTT reporter 516 in immune cells. Primary bone marrow-derived macrophages (BMDMs) were 517 exposed to DEP compounds and after 24h luciferase activity was measured 518 ( Figure 2B). The control compound tBHQ induced reporter activity by 5-fold and 519 interestingly, there was a 2-fold increase in activity by a range of compounds, 520 including both PAHs (fluoranthene, pyrene and phenanthrene) and nitro-PAHs 521 (1-NP and 3-NFA) ( Figure 2B). Carbon black did not induce the reporter 522 expression in these cells.

524
We also tested the capacity of DEP compounds to activate the HOTT reporter 525 in primary hepatocytes, a cell type with a high xenobiotic metabolising activity 526 ( Figure 2C). Cells exposed to 20 µM of the pro-oxidant increased luciferase 527 activity on an average to 4-fold. When cells were exposed to 3-NFA (50 µM) a 528 7-fold increase in activity was measured. In contrast, exposure to 50 µM 1-NP 529 only resulted in a 1.5 -2-fold increase ( Figure S1D). 530 531 In addition to Nrf2, Hmox1 is regulated by a number of other transcription 532 factors [31]. To define the involvement of the NRF2 pathway in response to 533 nitro-PAHs in primary hepatocytes we investigated reporter activation in 534 hepatocytes from NRF2-KO_HOTThet mice ( Figure 2E) [26]. In cells derived 535 from NRF2-KO_HOTThet mice there was no increase of luciferase activity on 536 exposure to nitro-PAHs. In addition, exposure of Nrf2 positive primary 537 hepatocytes to 50 µM 3-NFA resulted in an increase in NRF2 protein at 6h and 538 peaking at 16h. Also, the expression of NRF2 inducible proteins (HO-1 and 539 NQO1) was also increased ( Figure 2E). These data demonstrated a key role 540 for Nrf2 in the responses observed. 541 542 2.-Defining DEP toxic mechanisms in vivo using stress reporters. 543 544 Based on the data obtained in our in vitro studies, we investigated the capacity 545 of nitro-PAHs to activate in vivo stress responses [23,24]. Mice carrying the 546 HOTT reporter were treated with a single or 5 consecutive doses (at 24h 547 intervals) of 3-NFA (50mg/kg). Twenty-four hours after the final dose tissues 548 were harvested for histochemical and biochemical analysis (Figure 3 and 549 Figure S2A). As previously described [23], there was a basal ß-galactosidase 550 staining in the brain (hippocampus and cerebellum), kidney (tubular cells), and 551 lung (bronchiole, respiratory epithelium). In mice treated with a single dose of 552 3-NFA a marked increase in LacZ staining was measured in liver (hepatocytes) 553 and kidney (tubular cells) but not in any other tissue examined ( Figure S2A). In 554 mice exposed to repetitive 3-NFA doses, in addition to HOTT reporter 555 expression in liver (hepatocytes and Kupffer cells) and kidney (tubular cells; 556 Figure 3A), a striking increase in activity was observed in the heart 557 (cardiomyocytes) and lungs (epithelial, muscle and bronchioles; Figure 3). 558 There was no staining in any of the other tissues examined, including large 559 intestine and brain ( Figure S2A). Repetitive doses of 3-NFA did not induce 560 significant changes in general toxicity markers, including plasma ALS/AST, 561 H&E staining or mucopolysaccharides secretion in bronchioles (PAS staining; 562 ( Figure S3B, S3C and S3D respectively). NQO1 Western blot analysis 563 confirmed an induction of NQO1 expression, but only in mice exposed to 564 repetitive 3-NFA doses (Figure 3b).

566
To establish the role of NRF2 in the 3-NFA-induced activation of the HOTT 567 reporter we treated reporter mice nulled at the Nrf2 locus with repetitive doses 568 of 3-NFA ( Figure 4A). Contrary to the robust 3-NFA induced reporter activity in 569 the liver, kidney, lung, spleen and heart of NRF2wt-HOTT mice, reporter 570 expression was largely abrogated in lung, heart and kidneys of Nrf2 null 571 animals. Interestingly, modest reporter activation persisted in hepatocytes and 572 Kupffer cells in these mice and Hmox1 reporter expression was activated in 573 white pulp macrophages of the spleen by 3-NFA in both the NRF2 wt and null 574 lines. To test the possibility that an inflammatory response could contribute to 575 the reporter activation by prolonged 3-NFA exposure, we measured the 576 expression of IL-6 and IL-1 genes in a 3-NFA responsive tissue (heart). 3-NFA 577 did not increased the expression of these cytokines ( Figure 4B), although there 578 was an upregulation in the basal levels of these inflammatory markers (IL6 and 579 IL1b) in NRF2-KO_HOTT reporter mice. This is in alignment with our previous 580 observations (43) and reflects the interplay of NRF2 signalling and the immune 581 responses.

583
We then extended these studies to another DEP compound, 1-NP. NRF2wt-584 HOTThet or NRF2-KO-HOTThet reporter mice were exposed to vehicle, single 585 or repetitive doses of 1-NP at 50mg/kg ( Figure 5A). Contrary to 3-NFA, a single 586 dose of 1-NP did not activate the reporter in any of the tissues examined. 587 Repetitive treatment only activated reporter expression in liver and sparsely in 588 lung epithelial cells. No reporter expression by 1-NP was observed in NRF2-589 KO_HOTT reporter mice. Consistent with this finding, hepatic NQO1 protein 590 was increased NRF2wt-HOTT mice after 5 consecutive doses but absent in 591 NRF2-KO tissue ( Figure 5B). 592 593 Finally, we dosed HOTT reporter mice with fluoranthene following a repetitive 594 doses protocol ( Figure S3A). No activation of the HOTT reporter was observed 595 in any of the tissues examined. AhR activation however was observed as 596 hepatic CYP1a1 protein levels were increased ( Figure S3B). 597 598

Studies on AhR and DNA damage-mediated responses 599 600
Nitro-PAHs can activate AhR as well as DNA damage responses [32][33][34]. We 601 therefore examined the ability of 3-NFA to activate these pathways using a 602 further set of in vivo reporters. AhR reporter mice (CYP1a1_KI_Cre-Tom rep ) 603 were treated with 5 daily doses of 50mg/kg 3-NFA. Basal activity was seen in 604 some tissues, including liver, heart, kidney ( Figure 6A), lung and spleen (not 605 shown). On exposure to 3-NFA activation of the reporter was observed in 606 hepatocytes only. In this experiment, AhR activation was confirmed by showing 607 increased expression CYP1a1 and NQO-1 by western blot after the 3-NFA 608 treatment ( Figure 6C).

610
We then investigated whether exposure to 3-NFA induced a DNA damage 611 response in vivo. using the DNA damage-inducible p21 reporter [24]. Following 612 5 daily doses of 50mg/kg 3-NFA no changes over the basal expression of the 613 reporter was observed in any of the tissues examined, including liver, kidney 614 and heart ( Figure 6B). Treatment with cisplatin, used as a positive control, 615 resulted in marked p21 reporter activation in liver and large intestine ( Figure  616 6D). 617 618 3.-Pharmacological approaches to define toxic mechanisms involved in 619 stress reporter activation by DEP components. 620 621 To further define the mechanism of reporter activation by 3-NFA we employed 622 a pharmacological approach. In the first instance, we tested the ability of the 623 antioxidant NAC to modulate the reporter activation following 3-NFA treatment.

624
We have previously shown that NAC at the concentration used (300mg/kg) 625 does not increase the activation of the HOTT reporter [23,26]. NAC 626 administration completely prevented the induction of Lac-Z staining in the liver, 627 but the reporter activity remained unaffected in the kidney.

629
In order to establish whether the Nrf2-independent reporter activity in the 630 spleen induced by 3-NFA was due to an inflammatory response, NRF2-631 KO_HOTThet reporter mice were treated with either with 5 consecutive doses 632 of 3-NFA alone or in combination with celecoxib (a selective Cox-2 inhibitor; 633 Figure 7B) or dexamethasone (pan-NFKB signalling inhibitor; Figure 7C).

634
Neither of these treatments attenuated reporter activation suggesting an 635 alternative mechanism of induction. CYP1a1 induction in the liver ( Figure 7D, 636 E) and a reduction in plasma IL-1b were used as positive controls ( Figure 7F). 637 638 4.-Metabolic activation of air pollutants assessed in stress reporter mice.

640
Metabolism at nitro-groups is an important mechanism of activation of 641 substituted PAHs to toxic products [35]. Before apply in vivo stress reporters to 642 investigate this mechanism of toxicity, we first exposed the metabolically 643 competent MCF7_AREc32 cells to AhR agonists (TCDD 200ng/ml) 24h before 644 and during exposure to increasing concentrations of selected PAHs and nitro-645 PAHs ( Figure 8A). TCDD alone did not activate the luciferase activity, despite 646 triggering a strong induction of CYP1a1 mRNA. 3-NFA also induced CYP1a1 647 in these cells, but to a lesser extent ( Figure S4A). Treatment with a combination 648 of TCDD and nitro-PAHs resulted in a marked additional 6-fold increase of 649 luciferase activity compared to the activation by nitro-PAHs alone. TCDD did 650 not increase reporter activity when cells were incubated with the PAHs 651 fluoranthene or pyrene. In the light of these results, we investigated whether 652 co-exposure of nitro-PAHs with other DEP components (i.e. PAHs) increased 653 luciferase activity. Cells were treated with low doses of 1-NP or 3-NFA in 654 combination with single PAHs for 24h. PAHs alone did not further induce the 655 activation of the ARE_luc reporter by nitro-PAHs ( Figure S4B).

657
As a precursor to similar in vivo investigations, we carried out studies with 658 primary hepatocytes derived from the HOTT reporter mice. Mice were 659 administered vehicle or the AhR agonist Aroclor-1254, five days before cell 660 isolation. Cells were then cultured in 96-well plates and exposed to a panel of 661 DEP compounds for 24h ( Figure 8B). AhR activation was confirmed by 662 measuring Cyp1a1 protein ( Figure S4C). Interestingly nitro-PAHs increased 663 luciferase activity in hepatocytes derived from vehicle-treated reporter mice, 664 however, on cells from Aroclor-1254 treated mice a reduction rather than an 665 activation of the HOTT reporter with the compounds studies was observed. In 666 addition, co-incubation of primary HOTTrep hepatocytes from untreated mice 667 with TCDD and DEP compounds did not potentiate the activation of the 668 luciferase reporter ( Figure 8C). We confirmed the induction of Cyp1a1 669 expression by TCDD in whole cell lysates ( Figure S4D).

671
The paradoxical data using primary hepatocytes can be explained by the 672 complex balance between metabolic activation and detoxification as has been 673 already described for certain PAHs [36]. We were therefore intrigued to 674 establish the effects of the DEPs on reporter activity in vivo. Reporter mice were 675 treated with the AhR agonist 3-methylcholanthrene (3-MC), (which does not 676 induce a NRF2 response in vivo) and then to a single dose of 50mg/kg 3-NFA. 677 Increased reporter expression was observed in liver and kidney of 3-MC-678 treated mice over untreated controls but not in any other tissues ( Figures 8D,  679 S4-E and S5). In order to establish whether the use of in vivo reporters in mice identified toxic 686 mechanisms of relevance to epidemiological studies in man, we dosed reporter 687 mice with particulate matter representing different emission scenarios ( Figure  688 9). Initially, HOTThet were exposed by oropharyngeal deposition to a single 689 dose (250µg) or a repetitive dose (5x 50µg, once daily for 5 days) of the 690 standard reference diesel particulate material (SRM 2975). Despite significant 691 particle uptake in the alveolar epithelium, after 24h no reporter activation was 692 seen in the lung or any other tissues examined ( Figure 10A and not shown).

693
We then tested a different pollutant mixture, SRM1648b (road dust PM10, Figure  694 9B) in NRF2wt-HOTThet and NRF2ko-HOTThet mice. Histological analysis 695 showed that particles were mainly accumulated in the bronchioles, particularly 696 in the terminal bronchiolar region and some extension up the bronchiolar tree 697 ( Figure 9B and C). Exposure to these particles induced reporter expression in 698 a number of different cell populations in lung tissues, including bronchiolar cells 699 and epithelial cells in both exposure protocols. Strikingly, these signals were 700 significantly reduced in NRF2-KO mice, indicating that NRF2 plays a major role 701 in the cellular responses associated with PM10 exposure ( Figure 9C). 702 703 Epidemiological studies have described an association of PM10 urban 704 particulate matter exposure with increased vulnerability to bacterial pneumonia.

705
We have previously shown that PM10 (Leicester, Ghana) has the capacity to 706 increase pneumococcal adhesion as well as increase platelet-activating factor 707 receptor (PAFR) expression in pulmonary cell lines. In order to establish 708 whether the reporter mice could be used to evaluate toxicity mechanisms 709 resulting from real time exposures, we collected London roadside PM (MR-710 PM10) using a high-volume cyclone. HOTThet reporter mice exposed to MR-711 PM10 revealed an increase in positive staining cells scattered through the lung 712 parenchyma, particularly aggregated around inhaled particles ( Figure 10A).

713
The morphology of the cells was consistent with being alveolar macrophages 714 and this was confirmed by an immunohistochemical analysis using the F4/80 715 macrophage marker. To define the human relevance of these findings we 716 incubated primary human nasal epithelial cells (HPNEpC) with MR-PM 10 (10 717 µg/mL). A significant increase in both PAFR expression ( Figure 10B, p<0.01) 718 and pneumococcal adhesion ( Figures 10C-D, Yc, p<0.01) was observed. The 719 PAFR receptor blocker CV3988 reduced MR-PM10 pneumococcal adhesion to 720 baseline levels ( Figure 10C, p<0.05). The antioxidant NAC inhibited MR-PM10 721 stimulated adhesion ( Figure 10D, p<0.05) supporting a role for oxidative stress 722 in this process. NAC had no effect pneumococcal adhesion in unstimulated 723 cells.

727
We have used a panel of stress reporter mice to understand at a mechanistic 728 level how environmental agents induce pathways associated with chemical 729 toxicity. Although in vitro approaches can provide important insights into toxic 730 mechanisms they have significant limitations in defining what actually happens 731 in vivo. These include the physiological complexity of cell-cell interactions, 732 metabolic activation, multicellular inflammatory responses (often considered to 733 be a key factor in defining toxicity), crosstalk between tissues or bona fide 734 replication of specific cell types. However, a major limitation of in vivo studies 735 is the enormous amount of time, effort and cost involved in detailed pathological 736 analysis of any cellular changes observed and also the fact that changes can 737 only often be detected in the presence of overt toxic events. In order to 738 circumvent this major challenge, we have for a number of years been 739 developing a panel of reporter mice which reflect the activation of pathways 740 directly associated with chemical toxicity. The activation of these pathways, 741 including NRF2, p53 or AhR, provides insights into toxic mechanisms. The 742 application of stress reporters to reflect toxic potential is widely accepted as 743 reflected in the ToxCast 21 program or other environmental toxicology studies 744 [37,38].

746
In this paper we have exemplified the power of the in vivo stress reporters to 747 understand toxic mechanisms of air pollution. Our studies provided a hazard 748 ranking of toxic compounds and/or particles that can be used to inform 749 epidemiology studies and identify health interventions aimed at reducing 750 exposure to specific air pollutants. Additionally, through the use of a range of 751 stress responsive reporters (e.g. oxidative stress, inflammation, AhR 752 interactions and DNA damage), we obtained mechanistic insights into the 753 toxicity of air pollutants. While the reporters used in this study represent some 754 of the major mechanisms of toxicity related to the deleterious health effects of 755 environmental exposure to air pollutants [39,40], this approach can be 756 extended to study additional mechanisms of toxicity such as those induced by 757 endocrine disruptors. 758 759 Our approach involved a combination of in vitro and in vivo studies. We used 760 primary cells derived from the reporter mice in in vitro screens to prioritize 761 chemicals with a higher toxic potential for subsequent in vivo studies. We 762 selected a panel of chemicals commonly identified in urban diesel exhaust 763 particles, including PAHs, nitro-PAHs, oxy-PAHs and carbonaceous particles. 764 The advantage of using primary cell assays derived from reporter mice 765 compared to immortalised cell lines is that they do not bear genetic mutations 766 in cell signalling that could compromise the interpretation of results. For 767 example, the human lung cancer-derived alveolar cell line A549 often used in 768 toxicology studies has constitutively high levels of antioxidant proteins as a 769 result of the stabilization of NRF2 as a consequence of mutations in Keap1 [41].

771
Our in vitro studies demonstrated that different compounds found in diesel 772 exhaust particles activate distinctive adaptive response pathways. Consistent 773 with the literature, we found that PAH derivatives (i.e., oxy-and nitro-PAHs) 774 have the highest capacity to induce oxidative stress, AhR responses and/or 775 cytotoxicity. Interestingly, although both oxy-and nitro-PAHs have been linked 776 to oxidative stress [42], the oxy-PAHs, 1,4 naphthoquinone, 9-10 777 phenanthrenequinone, did not activate an adaptive oxidative stress response 778 in our studies. This could be ascribed to the differences in endpoints been 779 measured. For example, 1,4 naphthoquinone increased the oxidation of DCFH-780 DA in melanoma cells at comparable concentrations used in our assays [43]. 781 In contrast to quinones, we found that nitro-PAHs elicited a robust adaptive cell 782 response in both human and murine cell reporter assays. Overall, our in vitro 783 assays supported the conclusion that nitro-PAHs have a distinctive capacity to 784 induce oxidative stress responses triggered by the NRF2 pathway. On this 785 basis, we selected nitro-PAHs as chemical exemplars for our in vivo studies.

787
The application of multiple stress reporters for each individual compound or 788 mixture in vivo allows a more complete assessment of toxic potential to be 789 established. For example, we did not observe DNA damage responses where 790 oxidative stress responses were observed. Furthermore, the use of multiple 791 reporters allows cause and effect and relative dose and temporal responses to 792 be established. We have shown for example that acute exposure to 3-NFA 793 triggered NRF2-dependent Hmox1 reporter activation in hepatocytes and 794 kidney tubular cells, indicative of an oxidative stress response. However, on 795 chronic exposure activation was additionally observed in cardiomyocytes, lung 796 (epithelial, bronchial and immune cells) and spleen (white pulp macrophages). 797 We also demonstrated that compounds within a related chemical class (i.e. 798 nitro-PAHs) can exhibit profound differences in the induction of toxic responses 799 in vivo. For example, unlike 3-NFA, chronic exposure to 1-NP only resulted in 800 a mild response in hepatocytes. These in vivo results highlight the importance 801 of evaluating individual components in pollutants of concern to predict 802 differential contributions to disease progression. They also identify potential 803 caveats when assuming similar toxic properties of compounds based on 804 physicochemical characteristics. For example, nitro-PAHs are grouped into a 805 similar mode of action by which their bio-activation will induce oxidative DNA 806 damage, formation of DNA adducts and mutagenesis. While this is indeed the 807 case for 3-nitrobenzanthrone or 6-nitrochrysene [44,45], (our in vitro and in 808 vivo studies with 3-NFA do not support this. In fact, despite 3-NFA exhibiting a 809 higher capacity to induce NRF2 responses than 1-NP, we did not observe a 810 DNA damage response for either compound. These results are consistent with 811 studies in laboratory animals, which demonstrated that 3-NFA is only an 812 extremely weak carcinogen (<5% tumours after 100 weeks exposure to 1g of 813 compound) [46]. Our findings provide evidence that understanding toxic 814 potential at a mechanistic level is of key importance in the design of 815 epidemiological studies. For example, in diesel mixtures 1-NP is found at higher 816 concentrations than 3-NFA, 3-NFA metabolites are a much more relevant 817 marker of toxic potential. Toxic potential is inevitably defined by the level of an 818 individual compound, the duration of exposure and its potency as a chemical 819 toxin.

821
A powerful application of in vivo reporter systems is to define through the use 822 of genetic or pharmacological interventions the pathways affected by chemical 823 exposure. For example, Hmox1 can be regulated in vivo by a number of 824 different signalling pathways, including NRF2, heme and inflammation [23,47]. 825 In our study this was exemplified by the finding that induction of reporter activity 826 was lost on a Nrf2 null background. Also pharmacological intervention, using 827 the antioxidant NAC, attenuated our reporter activity induced by 3-NFA in the 828 liver. Additional studies to define the mechanism of NRF2 activation by nitro-829 PAHs are in progress.

831
Metabolic transformation of PAHs, for example P450 mediated metabolic 832 activation or redox cycling of quinones, can play a key role in defining the 833 toxicity of chemicals and chemical mixtures. Currently, in vitro assays often 834 used in epidemiological studies as a proxy for "toxic potential", such as 835 oxidative potential, do not take this key factor in defining chemical toxicity into 836 account [48] [35,45]. In the case of nitro-PAHs, two biochemical pathways are 837 responsible for the metabolic activation. These involve cytochrome P450-838 mediated ring C-oxidation to epoxides, with subsequent rearrangement to nitro-839 pyrenols or hydration to dihydrodiols; and through nitro-reduction by reductases 840 (e.g. P450 reductase, NQO1/AKR1C, and to a lesser extent P450s CYP1A1/2) 841 to nitroso-PAH, N-hydroxy-amino-PAH or amino-PAH [49,50]. In our studies 842 we clearly demonstrated that AhR activation both in vitro and in vivo led to a 843 marked increase in reporter activation after 3-NFA exposure. CYP1A1 844 expression is highly regulated by AhR, and indeed CYP1a1-mediated 845 generation of electrophiles has been intimately linked with the activation of 846