Development of Adverse Outcome Pathway for PPARÎ³

This is an open access article published under an ACS AuthorChoice License, which permits copying and redistribution of the article or any adaptations for non-commercial purposes.

Article Cite This: Chem. Res. Toxicol. 2019, 32, 1212−1222

pubs.acs.org/crt

Development of Adverse Outcome Pathway for PPARγ Antagonism Leading to Pulmonary Fibrosis and Chemical Selection for Its Validation: ToxCast Database and a Deep Learning Artificial Neural Network Model-Based Approach Jaeseong Jeong,† Natalia Garcia-Reyero,‡ Lyle Burgoon,‡ Edward Perkins,‡ Taehyun Park,† Changheon Kim,§ Ji-Yeon Roh,∥ and Jinhee Choi*,† Downloaded via 46.161.57.55 on August 28, 2019 at 13:23:09 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

†

School of Environmental Engineering, University of Seoul, 163 Seoulsiripdae-ro, Dongdaemun-gu, Seoul 02504, Republic of Korea United States Army Engineer Research and Development Center (ERDC) Environmental Laboratory, 3909 Halls Ferry Road, Vicksburg, Mississippi 39180, United States § Department of Computer Science, University of Seoul, 163 Seoulsiripdae-ro, Dongdaemun-gu, Seoul 02504, Republic of Korea ∥ Knoell Korea, 37 Gukjegeumyung-ro 2-gil, Yeongdeungpo-gu, Seoul 07327, Republic of Korea ‡

S Supporting Information *

ABSTRACT: Exposure to certain chemicals such as disinfectants through inhalation is suspected to be involved in the development of pulmonary fibrosis, a lung disease in which lung tissue becomes damaged and scarred. Pulmonary fibrosis is known to be regulated by transforming growth factor β (TGF-β) and peroxisome proliferator-activated receptor gamma (PPARγ). Here, we developed an adverse outcome pathway (AOP) to better define the linkage of PPARγ antagonism to the adverse outcome of pulmonary fibrosis. We then conducted a systematic analysis to identify potential chemicals involved in this AOP, using the ToxCast database and deep learning artificial neural network models. We identified chemicals bearing a potential inhalation hazard and exposure hazards from the database that could be related to this AOP. For chemicals that were not present in the ToxCast database, multilayer perceptron models were developed based on the ToxCast assays related to the AOP. The reactivity of ToxCast untested chemicals was then predicted using these deep learning models. Both approaches identified a set of chemicals that could be used to validate the AOP. This study suggests that chemicals categorized using an existing database such as ToxCast can be used to validate an AOP and that deep learning approaches can be used to characterize a range of potential active chemicals for an AOP of interest.

■

formalized way for regulatory purposes.4 Nevertheless, even though the framework has been around since 2010, only a limited number of AOPs have been proposed for inhalation toxicity screening. Inhalation exposure to environmental chemicals can lead to various adverse health effects, including pulmonary fibrosis, one of the most common chronic pulmonary diseases.5,6 Pulmonary fibrosis is a respiratory and potentially fatal disease characterized by the accumulation of scar tissue in the lung insterstitium, which results in loss of alveolar function, destruction of normal lung architecture, and respiratory distress.7 At the cellular level, fibrosis is characterized by the proliferation and accumulation of fibroblasts and scar-forming myofibroblasts with increased synthesis and deposition of

INTRODUCTION Humans are exposed to several environmental chemicals via inhalation through outdoor, indoor, or occupational exposure. Inhalation exposure of some chemicals can lead to pulmonary toxicity and lung diseases. Inhalation chambers are highly costly facilities used to test in vivo the potential inhalation toxicity of chemicals. These tests are not only very expensive but also technically difficult and time consuming. Thus, information on inhalation toxicity is available only for a limited number of chemicals.1,2 In order to efficiently and rapidly screen the potential inhalation toxicity of a large number of chemicals, the development of new approach methodologies (NAMs) has been proposed. The adverse outcome pathway (AOP) framework organizes existing knowledge across biological levels of organization to inform human and ecological risk assessment.3 AOPs are toxicological constructs that connect mechanistic information to apical endpoints in a © 2019 American Chemical Society

Received: January 29, 2019 Published: May 10, 2019 1212

DOI: 10.1021/acs.chemrestox.9b00040 Chem. Res. Toxicol. 2019, 32, 1212−1222

Article

Chemical Research in Toxicology extracellular matrix proteins including collagen and fibronectin.7 Fibroblasts differentiate to myofibroblasts after the appropriate stimuli, including the involvement of TGF-β activation, and those have some of the characteristics of smooth muscle cells. The differentiation of fibroblasts to myofibroblasts leads to the development of fibroblastic foci, similar to the early stages of normal wound healing and scar tissue formation. Scar formation, the accumulation of excess fibrous connective tissue, leads to thickening of the walls and causes reduced oxygen supply in the blood. As a consequence, patients suffer from perpetual shortness of breath. Fibrosis is generally progressive and leads to the destruction of normal lung architecture and has the potential to develop in other organs, such as skin, liver, kidney, and pancreas, where the mechanism is expected to be similar. Here, we developed an AOP for inhalation toxicity screening using pulmonary fibrosis as the adverse outcome (AO). Pulmonary fibrosis has been known to be regulated by transforming growth factor β (TGF-β) and peroxisome proliferator-activated receptor gamma (PPARγ).8−12 As the AOP is a conceptual framework, it needs to be validated by using appropriate stressors. Therefore, in the second part of this study, we conducted a systematic analysis for identifying potential chemicals involved in this AOP. To maximize applicability of this AOP to environmental inhalation toxicants, we selected inhalation-related chemicals for validation. To do this, we conducted a systematic categorization of potential inhalation toxicants using various databases that include environmental chemicals. By using a categorization of the information obtained from the databases, we identified a set of chemicals with potential to cause inhalation hazard. In the AOP framework, chemicals physically interact with a molecular initiating event (MIE), which in turn, activates a set of key events (KEs). However, there is a possibility that the AO may be manifested in other pathways through the same MIE. Therefore, to identify potential chemicals showing activity for each of the MIE and KEs of our proposed AOP, we explored the U.S. Environmental Protection Agency Toxicity Forecaster (ToxCast) database. Positive chemicals on appropriate ToxCast assays related to the AOP were identified. Then deep learning models were developed for the potential inhalation toxicants not present in the ToxCast database. Their reactivity was predicted using deep learning artificial neural network models developed based on the related ToxCast assays. Using both approaches, a set of potential chemicals for the validation of the AOP was proposed.

■

528/2012 (EU-BPR) biocidal active substances (https://echa.europa. eu/information-on-chemicals/biocidal-active-substances) and U.S. Occupational Safety and Health Administration (OSHA) limits for air contaminants (https://www.osha.gov/laws-regs/federalregister/ 1993-06-30-0). Inhalation hazardous chemicals were collected from the U.S. Environmental Protection Agency Integrated Risk Information System’s (EPA IRIS, https://www.epa.gov/iris) inhalation category and OECD Test Guideline 403 (acute inhalation toxicity) test results from the OECD eChemPortal (https://www. echemportal.org/echemportal/index.action). Among the collected chemicals, only chemicals with occupational inhalation exposure information, or for which the consumer product is likely to be inhalation exposure (spray, aerosol, etc.) were selected for analysis. ToxCast Assays Relevant to AOPs. We used the U.S. EPA iCSS ToxCast Dashboard (https://actor.epa.gov/dashboard) and ToxCast and Tox21 summary files (https://www.epa.gov/chemical-research/ exploring-toxcast-data-downloadable-data) to prioritize chemicals for validation of the AOP 206. ToxCast assays with the intended target corresponding to the MIE and KEs (PPARγ, TGF-β, NF-κB, IL-1α, IL-6, TNF-α, MMP-2 and MMP-9) of the AOP 206 were identified, and activity (Hitcall) data for the tested chemicals were collected. Deep Learning Models. We built 25 deep learning artificial neural network models for the AOP 206 using 25 ToxCast in vitro assays data. The input data used for model development consisted of chemical names, the simplified molecular-input line-entry system (SMILES) code for each chemical (denotes structure), and whether or not the chemical was an active one. The chemical structure was transformed into Morgan Fingerprints with a radius of two bonds using an RDKit (http://www.rdkit.org/). We used the synthetic minority oversampling technique (SMOTE) to oversample the class of chemicals.18 The multilayer perceptron (MLP) has an input layer, a dense layer of the same size as the input, with a RELU activator, and a dense single neuron layer as the output with a sigmoidal activator. Model performance was assessed using a stratified 5-fold crossvalidation, and all codes were developed in Python3. Keras and TensorFlow were used for the deep learning, and Sci-kit Learn was used to perform a stratified 5-fold cross-validation. The model building and analysis was performed in a Jupyter notebook. The model project can be downloaded from GitHub at https://github. com/UOSEST/AOP.

■

RESULTS AND DISCUSSION Description of the Adverse Outcome Pathway. Based on literature review, here, we describe an AOP by simplifying the pathways, where antagonism of PPARγ leads to fibrosis

MATERIALS AND METHODS

Development of the Adverse Outcome Pathway. The AOP was developed using information from the literature related to the mechanism of pulmonary fibrosis according to the AOP development guidelines.13−16 We also performed a systematic assessment of confidence of the AOP using tailored Bradford-Hill considerations including (1) biological plausibility for key event relationships (KERs), (2) essentiality of KEs, and (3) empirical support for KERs.17 Although various pathways and feed-back loops can trigger the AO, pulmonary fibrosis, we have constructed a simple AOP having one MIE and one AO for ease of future quantitative experimental validation of the AOP (AOP 206: Peroxisome proliferator-activated receptors γ inactivation leading to lung fibrosis, https://aopwiki.org/aops/206). Chemical Categorization. We identified chemicals having potential to cause inhalation toxicity via the EU Biocides Regulation

Figure 1. AOP for PPARγ inactivation leading to fibrosis. MIE is an inactivation of PPARγ signaling. KE1 is an activate TGF-β signaling. KE2 is an increase inflammatory response. KE3 is an increase EMT. KE4 is an increase in the amount of collagen. The AO or adverse outcome as the result of progression through the MIE and KEs is pulmonary fibrosis. 1213


Article

Chemical Research in Toxicology

Figure 2. Overview of key steps in the interaction between TGF-β and PPARγ signaling and pulmonary fibrosis. TGFBRI: Transforming growth factor-beta type I receptor; TGFBRII: Transforming growth factor-beta type II receptor; Smad 2/3: Mothers against decapentaplegic homologue 2/3; SMAD4: Mothers against decapentaplegic homologue 4; CBP: CREB-binding protein; MLK1: Mitogen-activated protein kinase kinase mlk-1; ROCK1: Rho-associated protein kinase 1; SRF: Serum response factor; TRPV4: Transient receptor potential cation channel subfamily V member 4; COL1A2: Collagen type I alpha 2 chain. (A) Phosphorylation of Smad 2 or 3 by TGF-β receptor. (B) Formation of the Smad2/3/4, p300, CPB complex in nucleus. (C) Ligand activation of PPARγ. (D) Formation of PPARγ, p300, CPB complex for PPARγ-dependent transcription. (E) Transcription of Smad-regulated genes. (F) Globular actin binding and inhibition of MLK1 nuclear transport. (G) Activation of Rho-ROCK signaling. (H) Formation of actin fibers, release of MLK1, and transport of MLK1 into the nucleus. (I) MLK1-Smad3 regulated transcription important in the differentiation of fibroblasts to myofibroblasts. (J) MLK1-SRF regulated transcription important in the production of extra cellular matrix materials leading to fibrosis.

Figure 3. Graphical representation of the workflow of identification of potential chemicals for validation of AOP 206 using ToxCast and deep learning model.

(Figure 1). The MIE is defined as antagonism of PPARγ, which increases the profibrotic effect of TGF-β/Smad3 signaling19 (KE1). Then TGF-β signaling pathway and oxidative stress pathway lead to increased inflammatory cytokine production (KE2). An increased inflammatory response drives EMT (KE3), which results in deposition of an interwoven network of fibrillar and nonfibrillar collagens (KE4).20 Increasing amounts of collagen lead to increased tissue stiffness, which in turn signals for increased production of extracellular matrix components (ECM) by mechanotransduction, through the rho-ROCK-actin pathways, leading to

Table 1. Number of Chemicals from Each Category for AOP Validation category inhalation exposure chemicals inhalation hazard chemicals

database

number

EU-BPR (vapor, gas, dust and mist) OSHA (vapor, gas, dust and mist) EPA IRIS (inhalation category) OECD eChemPortal (OECD TG403)

136 554 70 72

1214


DNA binding

cytokine

NF-κB (3)

inflammation (10)

protease

growth factor

TGF-β (5)

EMT (3)

nuclear receptor

target family

PPARγ (4)

AOP 206 (25)

1215

matrix metalloprote inase

inflammatory factor

interleukins

NF-κ B

transforming growth factor β

nonsteroidal

target subfamily

187 188 261 262 281 282 209 210 295 296 491 497 267

271 272 1458 94 1346

BSK_KF3CT_TGFb1_down BSK_KF3CT_TGFb1_up ATG_NF_kB_CIS_dn ATG_NF_kB_CIS_up TOX21_NFkB_BLA_agonist_ratio BSK_BE3C_IL1a_down BSK_BE3C_IL1a_up BSK_KF3CT_IL1a_down BSK_KF3CT_IL1a_up BSK_LPS_IL1a_down BSK_LPS_IL1a_up BSK_CASM3C_IL6_down BSK_CASM3C_IL6_up BSK_LPS_TNFa_down BSK_LPS_TNFa_up NVS_ENZ_hMMP2 NVS_ENZ_hMMP9 BSK_KF3CT_MMP9_down

1497 112 199

802 1127 719

TOX21_PPARg_BLA_Agonist_ratio TOX21_PPARg_BLA_antagonist_ratio NVS_NR_hPPARg ATG_TGFb_CIS_dn ATG_TGFb_CIS_up BSK_BE3C_TGFb1_down

134

AEID

ATG_PPARg_TRANS_up

name

assay

Table 2. Brief Description of ToxCast Assays Related to AOP 206

1536-well plate 96-well plate

ME-180

96-well plate

coronary artery smooth muscle cells


cell-free keratinocytes and foreskin fibroblasts

96-well plate

96-well plate

umbilical vein endothelium and peripheral blood mononuclear cells

umbilical vein endothelium and peripheral blood mononuclear cells

96-well plate

keratinocytes and foreskin fibroblasts

bronchial epithelial cells

24-well plate




format

HepG2

keratinocytes and foreskin fibroblasts

bronchial epithelial cells

HepG2

HEK293T HEK293 NA

HepG2

model

24 h

1h

24 h

24 h

24 h

24 h

24 h

24 h

24 h

24 h

24 h

24 h

2h

24 h

24 h

time point

signaling

enzymatic activity

signaling

signaling

signaling

signaling

reporter gene signaling

reporter gene

signaling

signaling

reporter gene

binding

reporter gene reporter gene

function

rosiglitazone GW9662 ciglitazone

GW7647

positive control

ELISA

Fluorescence

ELISA

ELISA

ELISA

ELISA

TNFa

GAL4 β-lactamase reporter gene ELISA

colchicine

galardin (GM6001)

colchicine

colchicine

colchicine

colchicine

colchicine

NA

colchicine

colchicine

RT-PCR and capillary electrophoresis

ELISA

ELISA

Fluorescence Polarization RT-PCR and Capillary NA electrophoresis

RT-PCR and capillary electrophoresis GAL4 β-lactamase reporter gene

detection

readout

Chemical Research in Toxicology Article


Article

Chemical Research in Toxicology Table 3. Accuracy of Deep Learning Model Based on ToxCast Assays Related to AOP 206 AOP 206 PPARγ

ToxCast AEID

number of chemicals

active chemicals

802

7947

372

1127

7202

565

ATG_PPARg_TRANS_up

134

3422

939

NVS_NR_hPPARg

719

782

180

ATG_TGFb_CIS_dn

1497

3422

124

ATG_TGFb_CIS_up

112

3422

138

BSK_BE3C_TGFb1_down

199

1445

230

BSK_KF3CT_TGFb1_down

271

1445

294

BSK_KF3CT_TGFb1_up

272

1456

12

ATG_NF_kB_CIS_dn

1458

3422

281

ATG_NF_kB_CIS_up

94

3422

141

1346

7202

243

BSK_BE3C_IL1a_down

187

1445

271

BSK_BE3C_IL1a_up

188

1456

18

BSK_KF3CT_IL1a_down

261

1445

298

BSK_KF3CT_IL1a_up

262

1456

28

BSK_LPS_IL1a_down

281

1445

311

BSK_LPS_IL1a_up

282

1445

22

BSK_CASM3C_IL6_down

209

1445

210

BSK_CASM3C_IL6_up

210

1445

49

BSK_LPS_TNFa_down BSK_LPS_TNFa_up

295 296

1445 1445

294 77

NVS_ENZ_hMMP2

491

1113

42

NVS_ENZ_hMMP9

497

2007

125

BSK_KF3CT_MMP9_down

267

1445

392

assay TOX21_PPARg_BLA_Agonist_ratio TOX21_PPARg_BLA_antagonist_ratio

TGF-β

NF-κB

TOX21_NFkB_BLA_agonist_ratio inflammation

EMT

tissue damage and scarring or fibrosis, the AO. This AOP can be found in the AOP Wiki as AOP 206 (https://aopwiki.org/ aops/206). MIE (Inactivation of PPARγ) and KER with KE1 (Activation of TGF-β Signaling). PPARs are ligand-activated transcription factors that belong to the nuclear hormone receptor family and regulate a wide range of physiological activities. Three different isoforms have been identified: PPARα, PPARβ/δ, and PPARγ.21 PPARγ is expressed in many types of lung cells, including fibroblasts, and has anti-inflammatory properties.22 The biological plausibility is high, as the anti-inflammatory effects of PPARγ ligands are well-described,7 and a number of studies have explored the effects of PPARγ ligands as potential antifibrotic agents.23−26

average model accuracy (min−max)

true positive rate

false positive rate

true negative rate

false negative rate

98.35 (97.82− 98.81) 96.73 (96.01− 98.00) 87.83 (84.99− 89.41) 90.28 (88.79− 91.25) 98.33 (98.02− 98.63) 98.44 (97.71− 99.16) 94.35 (92.38− 96.29) 99.34 (98.26− 99.65) 92.43 (91.3− 94.34) 96.46 (95.38− 97.45) 98.44 (98.32− 98.62) 98.85 (98.56− 99.06) 92.75 (89.97− 93.81) 99.82 (99.65− 100) 91.92 (90.39− 93.44) 99.29 (98.94− 99.47) 92.01 (90.28− 94.48) 99.64 (99.29− 100) 95.62 (92.9− 97.97) 99.06 (98.74− 99.64) 92.08 (90−93.91) 98.02 (96.34− 98.9) 98.73 (97.65− 99.76) 98.37 (97.99− 98.93) 86.73 (84.28− 89.07)

1.00

0.02

0.98

0.00

0.99

0.04

0.96

0.01

0.94

0.16

0.84

0.06

0.96

0.13

0.87

0.04

0.99

0.02

0.98

0.01

0.99

0.02

0.98

0.01

0.98

0.13

0.87

0.02

1.00

0.01

0.99

0.00

0.95

0.13

0.87

0.05

0.99

0.06

0.94

0.01

1.00

0.02

0.98

0.00

1.00

0.05

0.95

0.00

0.98

0.15

0.85

0.02

1.00

0.00

1.00

0.00

0.96

0.15

0.85

0.04

1.00

0.01

0.99

0.00

0.94

0.11

0.89

0.06

0.99

0.01

0.99

0.01

0.98

0.05

0.95

0.02

1.00

0.01

0.99

0.00

0.95 0.99

0.10 0.04

0.90 0.96

0.05 0.01

1.00

0.03

0.97

0.00

1.00

0.03

0.97

0.00

0.91

0.17

0.83

0.09

The essentiality of MIE is also high. TGF-β drives the differentiation of lung fibroblasts to myofibroblasts, a key step in fibrosis formation. On the other hand, PPARγ ligands differentiate fibroblasts to fat-storing adipocytes, suggesting that PPARγ agonists may oppose the fibrogenic effects of TGF-β.7,11 Therefore, blocking PPARγ function would remove this opposing regulatory pathway, increasing the profibrotic effects of TGF-β.19,27 However, the empirical support of KER is moderate. In several studies, the relationship between PPARγ and TGF-β was confirmed by experiments using PPARγ ligands, and the PPARγ agonists inhibit the ability of TGF-β1 to induce myofibroblast differentiation and collagen secretion.10,12,25,28,29 For the quantitative relationship, there were studies of PPARγ ligand decreased intrinsic expression of TGF-β1 and TGF-β1216


Article

Chemical Research in Toxicology Table 4. Number of Active and Inactive Chemicals Based on ToxCast and Deep Learning Model ToxCast AOP 206 PPARγ

TGF-β

NF-κB

inflammation

EMT

assay TOX21_PPARg_BLA_Agonist_ratio TOX21_PPARg_BLA_antagonist_ratio ATG_PPARg_TRANS_up NVS_NR_hPPARg PPARγ total ATG_TGFb_CIS_dn ATG_TGFb_CIS_up BSK_BE3C_TGFb1_down BSK_KF3CT_TGFb1_down BSK_KF3CT_TGFb1_up TGF-β total ATG_NF_kB_CIS_dn ATG_NF_kB_CIS_up TOX21_NFkB_BLA_agonist_ratio NF-κB total BSK_BE3C_IL1a_down BSK_BE3C_IL1a_up BSK_KF3CT_IL1a_down BSK_KF3CT_IL1a_up BSK_LPS_IL1a_down BSK_LPS_IL1a_up BSK_CASM3C_IL6_down BSK_CASM3C_IL6_up BSK_LPS_TNFa_down BSK_LPS_TNFa_up inflammation total NVS_ENZ_hMMP2 NVS_ENZ_hMMP9 BSK_KF3CT_MMP9_down EMT total

deep learning

ToxCast AEID

number of chemicals

active chemicals (%)

number of chemicals

active chemicals (%)

inactive chemicals

802 1127 134 719

295 290 208 75 298 208 208 130 130 130 208 208 208 290 294 130 130 130 130 130 130 130 130 130 130 130 122 166 130 168

6 (2.0) 11 (3.8) 34 (16.3) 15 (20.0) 52 (17.4) 5 (2.4) 6 (2.9) 17 (13.1) 23 (17.7) 0 (0) 36 (17.3) 17 (8.2) 2 (1.0) 9 (3.1) 24 (8.2) 17 (1.31) 1 (0.8) 22 (16.9) 1 (0.8) 25 (19.2) 0 (0) 17 (13.1) 5 (3.8) 25 (19.2) 11 (8.5) 50 (38.5) 1 (0.8) 5 (3.0) 25 (19.2) 29 (17.3)

314 319 401 526 312 401 401 473 473 473 401 401 401 319 316 473 473 473 474 473 473 473 473 473 473 474 481 440 473 438

15 (4.8) 8 (2.5) 38 (9.5) 25 (4.8) 53 (17.0) 15 (3.7) 38 (9.5) 49 (10.4) 25 (5.3) 6 (1.3) 85 (21.2) 25 (6.2) 7 (1.7) 4 (1.3) 26 (8.2) 38 (8) 2 (0.4) 47 (9.9) 7 (1.5) 56 (11.8) 5 (1.1) 37 (7.8) 2 (0.4) 35 (7.4) 17 (3.6) 134 (28.3) 14 (2.9) 29 (6.6) 37 (7.8) 69 (15.8)

588 590 537 561 505 589 565 537 555 597 488 567 600 596 560 548 600 534 596 522 598 549 596 543 575 420 588 572 541 508

1497 112 199 271 272 1458 94 1346 187 188 261 262 281 282 209 210 295 296 491 497 267

fibrosis in mice,42 reduced fibrogenesis in the lungs of rats with hypoxia-induced pulmonary hypertension,43 and prevented bleomycin-induced lung fibrosis44 and skin fibrosis.45 However, the empirical support between TGF-β and inflammation is moderate. As a result of immunohistochemical staining for lung sections of pulmonary fibrosis patients, increased production of TGF-β has been associated with chronic inflammatory and fibrotic diseases of humans and rodents.46 There was limited research on quantitative relationships between TGF-β and inflammation-related markers. KE2 (Increase, Inflammation) and KER with KE3 (Induction, Epithelial Mesenchymal Transition). In general, the overproduction of cytokines leads to infiltration of inflammatory cells and proliferation of fibroblast-related interstitial cells.47 The classical proinflammatory cytokines IL-6, IL-1β, and TNF-α released from innate immune cells like monocytes are also profibrotic. The biological plausibility is moderate. Th-2 cytokines (IL-4, IL-5, and IL-13) and proinflammatory cytokines (IL-1β, IL-6, and TNF-α) are linked to MMP and fibrosis.48 The essentiality of KE2 is also high. Experiments in various animal models, such as TNF-α blockers or TNF-receptordeficient mice, have shown that cytokine TNF-α has profibrosis properties. And IL-6 is an essential component of fibrosis, mainly resulting in reduced degradation of the matrix protein.48 Also, the attenuation of cell-related inflammation leads to reduction of the activity of MMP.49

induced phosphorylation of Akt and α-SMA and fibronectin expression as a marker of myofibroblast differentiation, in a dose-dependent manner.10,12,28 KE1 (Activation of TGF-β Signaling) and KER with KE2 (Increase, Inflammation). Lung fibrosis is the end result of a TGF-β-activated signaling cascade in lung fibroblasts (Figure 2).30,31 In this pathway, TGF-β activation leads to the phosphorylation and activation of Smad2 and Smad3, which heterodimerize with Smad4 and recruit the histone acetyltransferase p300 and the CREB binding protein (CBP) at the promoters of Smad3-dependent genes to activate transcription.32−37 Smad3 activation has been shown to be a necessary step in the progression of pulmonary fibrosis and myofibroblast differentiation, with knockout of Smad3 being sufficient to prevent pulmonary fibrosis and myofibroblast differentiation.38 The Smad:p300 complex causes transcription of Smad3-dependent genes, including MLK1, a key regulator of fibroblast differentiation to myofibroblasts. The biological plausibility of the KER is high. The activated TGF-β signaling pathway stimulates the expression of multiple proinflammatory and fibrotic cytokines such as tumor necrosis factor-α (TNF-α), IL-1β, IL-6, and IL-13, promoting the fibrotic response.39 The essentiality of KE1 is high, as an antagonist of TGF-β transduction signaling significantly reduced bleomycin-induced lung fibrosis.40,41 Furthermore, knockout and knockdown experiments with MLK1 were sufficient to prevent cardiac 1217


dye and chemical intermediate

√

√

768-52-5

surfactants

pesticides and flame retardants

n-isopropyl aniline

87-86-5

ACUTE TOX. 2

herbicide

Cetyl pyridinium chloride

123-03-5

√

√

√

poly(oxy-1,2-ethanediyl), αsulfo-ω-(dodecyloxy)-sodium salt graphite

9004-82-4

√

preservatives and Antioxidants

processing aids

Pentachlorophenol

1,2,3,4,5,5-hexachloro-1,3cyclopentadiene

77-47-4

√

use category disinfectant

√

ACUTE TOX. 2

manufacturing and electronic products antimicrobial product

1,2-benzisothiazol-3(2H)one

2634-33-5

IRIS

insecticide

√

√ √

OSHA

BPR

eChme Portal (TG403)

toxicity related source

√

disulfiram

97-77-8

7782-42-5

rotenone

83-79-4

chemical name

glutaraldehyde

CAS no.

111-30-8

exposure related source

1218

0/0

0/0

1/3

0/0

0/0

10/25

12/24

8/25

5/24

6/25

active/ total assay

PPARγ (1127)

PPARγ (802), TGF-β (271), NF-κB (1458), inflammation (261; 295), EMT (267) PPARγ (1127), NF-κB (94), inflammation (281; 295), EMT (267) TGF-β (199; 271), NF-κB (1458), inflammation (187; 261; 281; 295), EMT (267) PPARγ (802; 1127), TGF-β (199; 271), NF-κB (1458; 1346), inflammation (187; 261; 281; 209; 295), EMT (267) PPARγ (1127), TGF-β (199; 271), NF-κB (1458), inflammation (187; 261; 281; 209; 295), EMT (267)

target (AEID)

HTS in vitro data

8/25

7/25

12/22

14/25

12/25

0/0

0/1

0/0

0/1

0/0

active/ total assay

PPARγ (134; 802; 1127; 719), TGF-β (1497; 112; 199; 271), NF-κB (1458), inflammation (187; 261), EMT (267) PPARγ (134), TGF-β (1497; 199; 271; 272), NF-κB (94; 1346), inflammation (187; 261; 281; 209; 295), EMT (497; 267) PPARγ (134), TGF-β (1497; 199; 271; 272), NF-κB (1458), inflammation (187; 261; 281; 209; 295), EMT (267) PPARγ (134; 1127), TGF-β (112; 199), NF-κB (94), inflammation (187), EMT (267) TGF-β (199), NF-κB (1458), inflammation (187; 261; 281; 209; 295), EMT (267)

target (AEID)

predictive result

Table 5. Summary of Selected Chemicals for Validation of AOP 206, Based on BPR, OSHA, IRIS, and OECD eChemPortal Database and ToxCast Assay

Chemical Research in Toxicology Article


Article

Chemical Research in Toxicology

4). Lastly, a set of chemicals for the validation of AOP 206 was proposed (Step 5). Step 1. In the EU-BPR and OSHA chemical inventory, 136 and 554 chemicals with vapor, gas, dust, and mist properties were selected, respectively, as potential inhalation exposure chemicals. Seventy chemicals from the U.S. EPA IRIS (inhalation toxicity category) and 72 chemicals from the OECD eChemPortal (OECD TG 403) were selected as potential compounds to cause an inhalation hazard. A total of 654 unique chemicals were identified, and a summary of the chemical list is presented in Table 1. A full list of chemicals and their exposure and usage information is found in Table S1. Step 2. ToxCast assays relevant to the MIE and KEs of AOP 206 were identified. In total, 25 assays were selected, including 4 PPARγ assays, 5 TGF-β assays, 3 NF-κB assays, 10 inflammation (IL-1α, IL-6, and TNF-α) assays, and 3 EMT (MMP-2 and MMP-9) assays. The PPARγ assays were related to the identification of MIE, whereas other assays were related to pathways involved in KEs. A brief description of each assay is presented in Table 2. Step 3. To predict potential activity of the chemicals for each MIE and KE, we developed MLP models using chemical structure and their activity.59 We obtained the chemical structure information (SMILES) from PubChem (https:// pubchem.ncbi.nlm.nih.gov) and ToxCast in vitro assay data from the iCSS ToxCast Dashboard. We built 25 models for AOP 206 targets (i.e., 4 PPARγ, 5 TGF-β, 3 NF-κB, 10 inflammation, and 3 EMT). A summary of accuracy and the number of active chemicals used for learning each model is presented in Table 3. The active/inactive ratio of the ToxCast assay data set was unbalanced, for example, TOX21_PPARg_BLA_Agonist_ratio (AEID 802) data set consisted of 372 PPARγ active (agonist) and 7575 inactive chemicals. Since there are large differences between the number of active and inactive chemicals, the active chemicals were oversampled to the inactive chemicals using the SMOTE algorithm18 for all data sets. RDKit identified 2048-bit fingerprints across these chemicals, and 100 epochs were performed for each fold in the stratified 5-fold cross validation. The final models had an accuracy of ranging from 86.73% to 99.82% based on 5-fold cross validation, with true positive rates ranging from 0.91 to 1.00 and false positive rates ranging from 0.00 to 0.17. Step 4. We investigated whether or not chemicals selected in Step 1 were tested in the selected ToxCast assays. Active, tested chemicals were identified for each assay, whereas for chemicals not present in ToxCast, their activity status was predicted using the MLP models (Table 4). For PPARγ, 17.4% of the chemicals were found active in ToxCast, and 17.0% were predicted to be active in the MLP models; similar results were shown in TGF-β and EMT. Related to inflammation, the most abundant active chemicals were identified, showing 38.5% in ToxCast and 28.3% in MLP. In contrast, there were the least active chemicals identified in NF-κB, showing 8.2% in both ToxCast and MLP. Step 5. Potential chemicals for the validation of AOP 206 were selected using inhalation hazard and exposure information from the database, experimental results from ToxCast, or predicted results from the deep learning model (Table 5). Among the 654 chemicals, 299 chemicals were identified as active chemicals in at least one event of AOP 206 (i.e., PPARγ, TGF-β, NF-κB, inflammation, and EMT). Fourteen chemicals were identified as affecting all events in AOP 206, and six chemicals (heptachlor, aldrin, n-isopropyl aniline, hexachlor-

The empirical support is moderate. TNF-α induces the expression of vimentin and MMP, and inflammatory cytokines were associated with elevated EMT genes, indicating causal relationships.50 For the quantitative relationship, there were studies that the proinflammatory cytokine, IL-17A-induced dose-dependent downregulation of E-cadherin and upregulation of α-SMA.51 And TNF-α only slightly decreases Ecadherin expression in a concentration-dependent manner.52 KE3 (Induction, Epithelial Mesenchymal Transition) and KER with KE4 (Collagen Deposition). The EMT acquires mesenchymal markers, including neural cadherin (N-cadherin), vimentin, integrin, fibronectin, and MMPs, while epithelial cells are gradually transformed into mesenchymal-like cells, losing epithelial functions and characteristics.53 The biological plausibility is moderate. MMPs, the EMT markers, are able to degrade collagen types I and III, key collagens of the irreversible scar in hepatic cirrhosis.54 The essentiality of KE3 is high, whereas EMT is necessary for proper re-epithelialization and extracellular matrix (ECM) deposition, an uncontrolled continued transition from epithelial cells to myofibroblasts can result in fibrosis.53,55 Inhibition of MMP activity leads to accumulation of matrix proteins like collagen in the extracellular space.54 The empirical support is moderate. Bleomycin-induced fibrosis is reduced in transgenic animals.30,56 There was limited research on quantitative relationships between EMT and collagen. KE4 (Collagen Deposition) and KER with AO (Pulmonary Fibrosis). Collagen deposition is a part of the tissue healing process induced by epithelial cell injury. The biological plausibility is high. The total amount of collagen deposited by fibroblasts is a controlled balance between collagen synthesis and catabolism. In the remodeling phase, when this balance is disrupted and collagen deposition increases, scar formation, organs, or peri-implantational fibrosis occur.57 The essentiality of KE4 is moderate. The disruption of the epithelial layer integrity can enhance inflammatory cell infiltration and, in turn, worsen the fibrotic process.58 The empirical support is moderate. Morphological analysis in the bleomycin-treated meprinβ KO mice revealed a decrease in collagen deposition and tissue density in lung fibrosis.58 There is a limited understanding on quantitative relationships between collagen and fibrosis. Identification of Potential Chemicals for Validation of AOP 206. After development of the AOP, we then tried to identify potential chemicals for validation. A total of 10 chemicals to validate AOP 206 were selected, considering both inhalation exposure and hazard using a tiered approach, as shown in Figure 3. First, 654 chemicals having potential for inhalation hazard and exposure were identified using databases including EUBPR, OSHA, EPA IRIS, and OECD eChemPortal (Step 1). Twenty five ToxCast assays related to pathways involved in AOP 206 were then identified (Step 2). For chemicals not tested in selected ToxCast assays, 25 MLP models were developed based on the ToxCast assays to predict each chemical’s activity (Step 3). Among the chemicals having potential inhalation exposure and hazard identified in Step 1, those chemicals which were active when tested in the aforementioned ToxCast assays were identified, and, if not tested, the activity was predicted using the MLP models (Step 1219


Article

Chemical Research in Toxicology obutadiene, disulfiram, and endosulfan) potentially affected only KEs (i.e., TGF-β, NF-κB, inflammation, and EMT). The full results of all 654 chemicals are in Table S2. Finally, we selected five active chemicals from ToxCast and five active chemicals from the MLP model for validation of AOP 206 based on the number of active assays and coverage of AOP 206 and the use category. The five chemicals identified from ToxCast were glutaraldehyde, rotenone, disulfiram, 1,2benzisothiazol-3(2H)-one, and 1,2,3,4,5,5-hexachloro-1,3-cyclopentadiene, and the five chemicals predicted from the MLP models were cetylpyridinium chloride, pentachlorophenol, poly(oxy-1,2-ethanediyl), α-sulfo-ω-(dodecyloxy)-sodium salt, graphite, and n-isopropyl aniline. Experimental validation of the AOP 206 using these prioritized chemicals is ongoing.

Jinhee Choi: 0000-0003-3393-7505 Funding

This work was supported by the Korean Ministry of Environment under the Environmental Health R&D Program (2017001370001). Notes

The authors declare no competing financial interest.

■

(1) Rovida, C., and Hartung, T. (2009) Re-Evaluation of Animal Numbers and Costs for in Vivo Tests to Accomplish REACH Legislation Requirements for Chemicals - A Report by the Transatlantic Think Tank for Toxicology (T4). ALTEX. 26 (3), 187−208. (2) Clippinger, A. J., Allen, D., Jarabek, A. M., Corvaro, M., Gaça, M., Gehen, S., Hotchkiss, J. A., Patlewicz, G., Melbourne, J., Hinderliter, P., et al. (2018) Alternative Approaches for Acute Inhalation Toxicity Testing to Address Global Regulatory and NonRegulatory Data Requirements: An International Workshop Report. Toxicol. In Vitro 48, 53−70. (3) Ankley, G. T., Bennett, R. S., Erickson, R. J., Hoff, D. J., Hornung, M. W., Johnson, R. D., Mount, D. R., Nichols, J. W., Russom, C. L., Schmieder, P. K., et al. (2010) Adverse Outcome Pathways: A Conceptual Framework to Support Ecotoxicology Research and Risk Assessment. Environ. Toxicol. Chem. 29 (3), 730−741. (4) Leist, M., Ghallab, A., Graepel, R., Marchan, R., Hassan, R., Bennekou, S. H., Limonciel, A., Vinken, M., Schildknecht, S., Waldmann, T., et al. (2017) Adverse Outcome Pathways: Opportunities, Limitations and Open Questions. Arch. Toxicol. 91 (11), 3477−3505. (5) Taskar, V. S., and Coultas, D. B. (2006) Is Idiopathic Pulmonary Fibrosis an Environmental Disease? Proc. Am. Thorac. Soc. 3 (4), 293−298. (6) Yoshida, T., Ohnuma, A., Horiuchi, H., and Harada, T. (2011) Pulmonary Fibrosis in Response to Environmental Cues and Molecular Targets Involved in Its Pathogenesis. J. Toxicol. Pathol. 24 (1), 9−24. (7) Lakatos, H. F., Thatcher, T. H., Kottmann, R. M., Garcia, T. M., Phipps, R. P., and Sime, P. J. (2007) The Role of PPARs in Lung Fibrosis. PPAR Res. 2007, 71323. (8) Burgess, H. A., Daugherty, L. E., Thatcher, T. H., Lakatos, H. F., Ray, D. M., Redonnet, M., Phipps, R. P., and Sime, P. J. (2005) PPARγ Agonists Inhibit TGF-β Induced Pulmonary Myofibroblast Differentiation and Collagen Production: Implications for Therapy of Lung Fibrosis. Am. J. Physiol. Cell. Mol. Physiol. 288 (6), L1146− L1153. (9) Tan, X., Dagher, H., Hutton, C. A., and Bourke, J. E. (2010) Effects of PPARγ Ligands on TGF-Β1-Induced Epithelial-Mesenchymal Transition in Alveolar Epithelial Cells. Respir. Res. 11, 1−13. (10) Kulkarni, A. A., Thatcher, T. H., Olsen, K. C., Maggirwar, S. B., Phipps, R. P., and Sime, P. J. (2011) PPAR-γ Ligands Repress TGFβInduced Myofibroblast Differentiation by Targeting the PI3K/Akt Pathway: Implications for Therapy of Fibrosis. PLoS One 6 (1), No. e15909. (11) Deng, Y. L., Xiong, X. Z., and Cheng, N. S. (2012) Organ Fibrosis Inhibited by Blocking Transforming Growth Factor-β Signaling via Peroxisome Proliferator-Activated Receptor γ Agonists. Hepatobiliary Pancreatic Dis. Int. 11 (5), 467−478. (12) Nuwormegbe, S. A., Sohn, J. H., and Kim, S. W. (2017) A PPAR-Gamma Agonist Rosiglitazone Suppresses Fibrotic Response in Human Pterygium Fibroblasts by Modulating the P38 MAPK Pathway. Invest. Ophthalmol. Visual Sci. 58 (12), 5217−5226. (13) (2013) Guidance Document on Developing and Assessing Adverse Outcome Pathways, Vol. 184, No. 184, OECD, Paris, France. (14) Villeneuve, D. L., Crump, D., Garcia-Reyero, N., Hecker, M., Hutchinson, T. H., LaLone, C. A., Landesmann, B., Lettieri, T., Munn, S., Nepelska, M., et al. (2014) Adverse Outcome Pathway

■

CONCLUSIONS In this study, we describe an AOP where antagonism of PPARγ leads to pulmonary fibrosis for the development of an inhalation toxicity screening platform. Although there may be a number of pathways leading to pulmonary fibrosis, the MIE target is set to PPARγ, whose activity is determined by ligand binding, thus making various in silico approaches including molecular docking applicable. This allows a quick and simple screening of AO, pulmonary fibrosis, using in silico approaches for chemicals suspected to cause inhalation toxicity. This, of course, requires a quantitative relationship as well as qualitative relationship between the KEs. To date, however, there is limited evidence of quantitative relationships in the literature. Also, the overall biological plausibility and essentiality of each KE are high, but the empirical support and quantitative relationships are moderate or weak. Therefore, it is necessary to qualitatively and quantitatively validate the KERs by selecting chemicals that follow this AOP as a future study. However, finding appropriate chemicals for validating each event of an AOP is not an easy task. Here, we proposed a tiered approach for selecting chemicals by combining inhalation-related database, ToxCast database, and deep learning artificial neural network models. The complexity of the fibrosis formation network, the number of potential interacting MIEs, as well as the number of pathways that could affect it highlight the importance of assays and databases such as ToxCast to explore potential MIEs and KEs. This study also highlights the fact that toxicity prediction based on deep learning models has the potential to accurately classify a range of chemicals for AOP validation.

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.chemrestox.9b00040. Table S1. A full list of chemicals having inhalation exposure and hazardous potential (XLSX) Table S2. A full analysis results of all 654 chemicals using ToxCast database and the deep learning model (XLSX)

■

REFERENCES

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. ORCID

Edward Perkins: 0000-0003-1693-7714 1220


Article

Chemical Research in Toxicology (AOP) Development I: Strategies and Principles. Toxicol. Sci. 142 (2), 312−320. (15) Villeneuve, D. L., Crump, D., Garcia-Reyero, N., Hecker, M., Hutchinson, T. H., LaLone, C. A., Landesmann, B., Lettieri, T., Munn, S., Nepelska, M., et al. (2014) Adverse Outcome Pathway Development II: Best Practices. Toxicol. Sci. 142 (2), 321−330. (16) Perkins, E. J., Gayen, K., Shoemaker, J. E., Antczak, P., Burgoon, L., Falciani, F., Gutsell, S., Hodges, G., Kienzler, A., Knapen, D., et al. (2019) Chemical Hazard Prediction and Hypothesis Testing Using Quantitative Adverse Outcome Pathways. ALTEX. 36 (1), 91−102. (17) Becker, R. A., Ankley, G. T., Edwards, S. W., Kennedy, S. W., Linkov, I., Meek, B., Sachana, M., Segner, H., Van Der Burg, B., Villeneuve, D. L., et al. (2015) Increasing Scientific Confidence in Adverse Outcome Pathways: Application of Tailored Bradford-Hill Considerations for Evaluating Weight of Evidence. Regul. Toxicol. Pharmacol. 72 (3), 514−537. (18) Kass, R. E., and Raftery, A. E. (1995) Bayes Factors. J. Am. Stat. Assoc. 90 (430), 773−795. (19) Ferguson, H. E., Kulkarni, A., Lehmann, G. M., Garcia-Bates, T. M., Thatcher, T. H., Huxlin, K. R., Phipps, R. P., and Sime, P. J. (2009) Electrophilic Peroxisome Proliferator-Activated Receptor-γ Ligands Have Potent Antifibrotic Effects in Human Lung Fibroblasts. Am. J. Respir. Cell Mol. Biol. 41 (6), 722−730. (20) Cox, T. R., and Erler, J. T. (2011) Remodeling and Homeostasis of the Extracellular Matrix: Implications for Fibrotic Diseases and Cancer. Dis. Models & Mech. 4 (2), 165−178. (21) Blanquart, C., Barbier, O., Fruchart, J. C., Staels, B., and Glineur, C. (2003) Peroxisome Proliferator-Activated Receptors: Regulation of Transcriptional Activities and Roles in Inflammation. J. Steroid Biochem. Mol. Biol. 85 (2−5), 267−273. (22) Rizzo, G., and Fiorucci, S. (2006) PPARs and Other Nuclear Receptors in Inflammation. Curr. Opin. Pharmacol. 6 (4), 421−427. (23) Kawaguchi, K., Sakaida, I., Tsuchiya, M., Omori, K., Takami, T., and Okita, K. (2004) Pioglitazone Prevents Hepatic Steatosis, Fibrosis, and Enzyme-Altered Lesions in Rat Liver Cirrhosis Induced by a Choline-Deficient L-Amino Acid-Defined Diet. Biochem. Biophys. Res. Commun. 315 (1), 187−195. (24) Uto, H., Nakanishi, C., Ido, A., Hasuike, S., Kusumoto, K., Abe, H., Numata, M., Nagata, K., Hayashi, K., and Tsubouchi, H. (2005) The Peroxisome Proliferator-Activated Receptor-γ Agonist, Pioglitazone, Inhibits Fat Accumulation and Fibrosis in the Livers of Rats Fed a Choline-Deficient, l-Amino Acid-Defined Diet. Hepatol. Res. 32 (4), 235−242. (25) Dantas, A. T., Pereira, M. C., de Melo Rego, M. J., da Rocha, L. F. J., Pitta Ida, R., Marques, C. D., Duarte, A. L., and Pitta, M. G. (2015) The Role of PPAR Gamma in Systemic Sclerosis. PPAR Res. 2015, 124624. (26) Ruzehaji, N., Frantz, C., Ponsoye, M., Avouac, J., Pezet, S., Guilbert, T., Luccarini, J. M., Broqua, P., Junien, J. L., and Allanore, Y. (2016) Pan PPAR Agonist IVA337 Is Effective in Prevention and Treatment of Experimental Skin Fibrosis. Ann. Rheum. Dis. 75 (12), 2175−2183. (27) Yoon, Y. S., Kim, S. Y., Kim, M. J., Lim, J. H., Cho, M. S., and Kang, J. L. (2015) PPARγ Activation Following Apoptotic Cell Instillation Promotes Resolution of Lung Inflammation and Fibrosis via Regulation of Efferocytosis and Proresolving Cytokines. Mucosal Immunol. 8 (5), 1031−1046. (28) Milam, J. E., Keshamouni, V. G., Phan, S. H., Hu, B., Gangireddy, S. R., Hogaboam, C. M., Standiford, T. J., Thannickal, V. J., and Reddy, R. C. (2008) PPAR-γ Agonists Inhibit Profibrotic Phenotypes in Human Lung Fibroblasts and Bleomycin-Induced Pulmonary Fibrosis. Am. J. Physiol. Cell. Mol. Physiol. 294 (5), L891− L901. (29) Wu, M., Melichian, D. S., Chang, E., Warner-Blankenship, M., Ghosh, A. K., and Varga, J. (2009) Rosiglitazone Abrogates Bleomycin-Induced Scleroderma and Blocks Profibrotic Responses through Peroxisome Proliferator-Activated Receptor-β. Am. J. Pathol. 174 (2), 519−533.

(30) Kang, H. R., Cho, S. J., Lee, C. G., Homer, R. J., and Elias, J. A. (2007) Transforming Growth Factor (TGF)-Β1 Stimulates Pulmonary Fibrosis and Inflammation via a Bax-Dependent, Bid-Activated Pathway That Involves Matrix Metalloproteinase-12. J. Biol. Chem. 282 (10), 7723−7732. (31) Wei, J., Bhattacharyya, S., and Varga, J. (2010) Peroxisome Proliferator-Activated Receptor γ: Innate Protection from Excessive Fibrogenesis and Potential Therapeutic Target in Systemic Sclerosis. Curr. Opin. Rheumatol. 22 (6), 671−676. (32) Massagué, J., Seoane, J., and Wotton, D. (2005) Smad Transcription Factors. Genes Dev. 19 (23), 2783−2810. (33) Chen, S. J., Yuan, W., Mori, Y., Levenson, A., Trojanowska, M., and Varga, J. (1999) Stimulation of Type I Collagen Transcription in Human Skin Fibroblasts by TGF-β: Involvement of Smad 3. J. Invest. Dermatol. 112 (1), 49−57. (34) Mori, Y., Chen, S. J., and Varga, J. (2000) Modulation of Endogenous Smad Expression in Normal Skin Fibroblasts by Transforming Growth Factor-β. Exp. Cell Res. 258 (2), 374−383. (35) Ghosh, A. K., Yuan, W., Mori, Y., and Varga, J. (2000) SmadDependent Stimulation of Type I Collagen Gene Expression in Human Skin Fibroblasts by TGF-β Involves Functional Cooperation with P300/CBP Transcriptional Coactivators. Oncogene 19 (31), 3546−3555. (36) Ghosh, A. K., Yuan, W., Mori, Y., Chen, S. J., and Varga, J. (2001) Antagonistic Regulation of Type I Collagen Gene Expression by Interferon-γ and Transforming Growth Factor-β: Integration at the Level of P300/CBP Transcriptional Coactivators. J. Biol. Chem. 276 (14), 11041−11048. (37) Ghosh, A. K., Bhattacharyya, S., and Varga, J. (2004) The Tumor Suppressor P53 Abrogates Smad-Dependent Collagen Gene Induction in Mesenchymal Cells. J. Biol. Chem. 279 (46), 47455− 47463. (38) Gu, L., Zhu, Y. J., Yang, X., Guo, Z. J., Xu, W. B., and Tian, X. L. (2007) Effect of TGF-β/Smad Signaling Pathway on Lung Myofibroblast Differentiation. Acta Pharmacol. Sin. 28 (3), 382−391. (39) Kang, H. (2017) Role of Micrornas in TGF-β Signaling Pathway-Mediated Pulmonary Fibrosis. Int. J. Mol. Sci. 18 (12), 2527. (40) Giri, S. N., Hyde, D. M., and Hollinger, M. A. (1993) Effect of Antibody to Transforming Growth Factor /B on Bleomycin Induced Accumulation of Lung Collagen in Mice. Thorax 48 (10), 959−966. (41) Nakao, A., Fujii, M., Matsumura, R., Kumano, K., Saito, Y., Miyazono, K., and Iwamoto, I. (1999) Transient Gene Transfer and Expression of Smad7 Prevents Bleomycin-Induced Lung Fibrosis in Mice. J. Clin. Invest. 104 (1), 5−11. (42) Small, E. M., Thatcher, J. E., Sutherland, L. B., Kinoshita, H., Gerard, R. D., Richardson, J. A., Dimaio, J. M., Sadek, H., Kuwahara, K., and Olson, E. N. (2010) Myocardin-Related Transcription Factora Controls Myofibroblast Activation and Fibrosis in Response to Myocardial Infarction. Circ. Res. 107 (2), 294−304. (43) Yuan, Z., Chen, J., Chen, D., Xu, G., Xia, M., Xu, Y., and Gao, Y. (2014) Megakaryocytic Leukemia 1 (MKL1) Regulates Hypoxia Induced Pulmonary Hypertension in Rats. PLoS One 9 (3), No. e83895. (44) Zhou, Y., Huang, X., Hecker, L., Kurundkar, D., Kurundkar, A., Liu, H., Jin, T. H., Desai, L., Bernard, K., and Thannickal, V. J. (2013) Inhibition of Mechanosensitive Signaling in Myofibroblasts Ameliorates Experimental Pulmonary Fibrosis. J. Clin. Invest. 123 (3), 1096− 1108. (45) Shiwen, X., Stratton, R., Nikitorowicz-Buniak, J., Ahmed-Abdi, B., Ponticos, M., Denton, C., Abraham, D., Takahashi, A., Suki, B., Layne, M. D., et al. (2015) A Role of Myocardin Related Transcription Factor-A (MRTF-A) in Scleroderma Related Fibrosis. PLoS One 10 (5), e0126015. (46) Khalil, N., O'Connor, R. N., Unruh, H. W., Warren, P. W., Flanders, K. C., Kemp, A., Bereznay, O. H., and Greenberg, A. H. (1991) Increased Production and Immunohistochemical Localization of Transforming Growth Factor-Beta in Idiopathic Pulmonary Fibrosis. Am. J. Respir. Cell Mol. Biol. 5 (2), 155−162. 1221


Article

Chemical Research in Toxicology (47) Miyazaki, Y., Araki, K., Vesin, C., Garcia, I., Kapanci, Y., Whitsett, J. A., Piguet, P. F., and Vassalli, P. (1995) Expression of a Tumor Necrosis Factor-Alpha Transgene in Murine Lung Causes Lymphocytic and Fibrosing Alveolitis. A Mouse Model of Progressive Pulmonary Fibrosis. J. Clin. Invest. 96 (1), 250−259. (48) Mack, M. (2018) Inflammation and Fibrosis. Matrix Biol. 68− 69, 106−121. (49) Underwood, D. C., Osborn, R. R., Bochnowicz, S., Webb, E. F., Rieman, D. J., Lee, J. C., Romanic, A. M., Adams, J. L., Hay, D. W. P., and Griswold, D. E. (2000) SB 239063, a P38 MAPK Inhibitor, Reduces Neutrophilia, Inflammatory Cytokines, MMP-9, and Fibrosis in Lung. Am. J. Physiol. Cell. Mol. Physiol. 279 (5), L895−L902. (50) Yan, C., Grimm, W. A., Garner, W. L., Qin, L., Travis, T., Tan, N., and Han, Y. P. (2010) Epithelial to Mesenchymal Transition in Human Skin Wound Healing Is Induced by Tumor Necrosis Factor-α through Bone Morphogenic Protein-2. Am. J. Pathol. 176 (5), 2247− 2258. (51) Vittal, R., Fan, L., Greenspan, D. S., Mickler, E. A., Gopalakrishnan, B., Gu, H., Benson, H. L., Zhang, C., Burlingham, W., Cummings, O. W., et al. (2013) IL-17 Induces Type V Collagen Overexpression and EMT via TGF-β-Dependent Pathways in Obliterative Bronchiolitis. Am. J. Physiol. Cell. Mol. Physiol. 304 (6), L401−L414. (52) Kasai, H., Allen, J. T., Mason, R. M., Kamimura, T., and Zhang, Z. (2005) TGF-Β1 Induces Human Alveolar Epithelial to Mesenchymal Cell Transition (EMT). Respir. Res. 6, 1−15. (53) Stone, R. C., Pastar, I., Ojeh, N., Chen, V., Liu, S., Garzon, K. I., and Tomic-Canic, M. (2016) Epithelial-Mesenchymal Transition in Tissue Repair and Fibrosis. Cell Tissue Res. 365 (3), 495−506. (54) Roeb, E. (2018) Matrix Metalloproteinases and Liver Fibrosis (Translational Aspects). Matrix Biol. 68−69, 463−473. (55) Rout-Pitt, N., Farrow, N., Parsons, D., and Donnelley, M. (2018) Epithelial Mesenchymal Transition (EMT): A Universal Process in Lung Diseases with Implications for Cystic Fibrosis Pathophysiology. Respir. Res. 19 (1), 1−10. (56) Cabrera, S., Gaxiola, M., Arreola, J. L., Ramírez, R., Jara, P., D'Armiento, J., Richards, T., Selman, M., and Pardo, A. (2007) Overexpression of MMP9 in Macrophages Attenuates Pulmonary Fibrosis Induced by Bleomycin. Int. J. Biochem. Cell Biol. 39 (12), 2324−2338. (57) Chen, C. Z. C., and Raghunath, M. (2009) Focus on Collagen: In Vitro Systems to Study Fibrogenesis and Antifibrosis_state of the Art. Fibrog. Tissue Repair 2 (1), 1−10. (58) Biasin, V., Wygrecka, M., Marsh, L. M., Becker-Pauly, C., Brcic, L., Ghanim, B., Klepetko, W., Olschewski, A., and Kwapiszewska, G. (2017) Meprin β Contributes to Collagen Deposition in Lung Fibrosis. Sci. Rep. 7 (1), 1−12. (59) Burgoon, L., Perkins, E. J., and Garcia-Reyero, N. Manuscript in Preparation.

1222


Development of Adverse Outcome Pathway for PPARÎ³

Recommend Documents