TGF-β has been shown to involve ROS formation via NOX4, whereas ROS also stimulate the epithelial excretion of TGF-β (Latella, 2018). ROS can also convert latent TGF-β to its active form by oxidizing the latency-associated peptide (LAP) (Liu and Gaston Pravia, 2010; Montorfano et al., 2014; Jaffer et al., 2015). Excessive ROS derived from protein oxidation can induce TβRI/II expression and further initiate TGF-β/SMAD signaling, which induces EMT in hepatocytes in vivo and in vitro (Sun et al., 2017). These studies indicate that ROS can activate TGF-β/SMAD signaling at different points along the pathway, including ligand-receptor interactions and intracellular signaling.

An in vivo study demonstrated that long-term PM_2.5 exposure induced increased TGF-β1 expression, SMAD2/3 phosphorylation, and collagen accumulation in mice lungs, indicating potential links among PM_2.5 exposure, the TGF-β1/SMAD2/3 pathway, and pro-fibrotic EMT (Gu et al., 2017). Another study supported this pathogenesis in mice model: Maternal exposure to PM_2.5 could also result in significant oxidative stress, activation of TGF-β/SMAD3 signaling, and EMT in fetal mice lung, which contributed to postnatal pulmonary dysfunction (Tang et al., 2017). In addition, exposure of mice to PM_2.5 for 4 weeks not only elicited SMAD3 phosphorylation and cardiac fibrosis but also induced excessive ROS accumulation and NOX4 expression in heart tissues, indicating that the potential occurrence of PM_2.5-induced EMT or endothelial-mesenchymal transition (EndMT) might be ROS dependent (Qin et al., 2018). Our previous in vitro study validated this mechanism: The release of TGF-β1, the activation of the TGF-β1/SMAD3 pathway, and the altered expression of hallmark EMT proteins in human bronchial epithelial cells were observed after 30-passage low-dose PM_2.5 exposure (Xu et al., 2019a). In another study, the exposure of alveolar epithelial cells to PM_2.5 not only induced the activation of latent TGF-β but also increased cell contractility and elongated cellular morphology in a dose-dependent manner. These effects could be reversed by NAC treatment, demonstrating the role of ROS in this pathogenic event (Dysart et al., 2014). In addition, increased TGF-β1 mRNA expression was observed in bronchial epithelial cells after 7 days’ consecutive exposure to PM_2.5, and treatment with the antioxidant NAC could also effectively offset the toxicity of PM_2.5 exposure (Tripathi et al., 2018). However, the detailed EMT transcription patterns induced by PM_2.5-induced ROS through the SMAD2/3-dependent pathway remain unknown.

In summary, these data demonstrate that PM_2.5 might induce EMT via ROS in a TGF-β/SMAD2/3 signaling-dependent manner. However, additional research is needed to further establish a firm causal link.

The activation of NF-κB can be evoked by ROS. NF-κB is activated as the response of cellular oxidative stress, responsible for inflammation, apoptosis, and differentiation. Moreover, the activation of NF-κB by ROS is also correlated with Snail upregulation and EMT phenotype transformation in epithelial cells (Valavanidis et al., 2013; Cichon and Radisky, 2014). Different antioxidant treatments targeting ROS in human non-small cell lung carcinoma cells, pancreatic cancer cells, and hepatocellular carcinomatous cells all effectively inhibit NF-κB activation, which further represses matrix metalloproteinase (MMP) expression together with cellular migration and invasion (Lee and Lim, 2011; Chao et al., 2017; Huang and Xin, 2018). In addition, IL-6 and IL-8, which are transcriptionally upregulated by ROS-activated NF-κB, also exert pro-EMT abilities via downstream JAK/STAT3 signaling (Gupta et al., 2012). In gastric cancer cells, even the regulatory effects of IL-6/JAK/STAT3 on expression of pro-EMT TFs and EMT markers relied on NOX4, indicating the potential comprehensive role of ROS in the transduction of NF-κB/IL-6 signaling (Gao et al., 2017). NF-κB is also able to induce TNF-α expression in response to ROS stimulation, and activated TNF-α receptors in turn upregulate NF-κB transcription with further Snail expression, implying positive feedback loops underlying ROS-induced EMT via NF-κB (Dong et al., 2007).

A novel publication confirmed the positive role of NF-κB in PM_2.5-induced EMT. Organic extracts of PM_2.5 enhanced the binding of NF-κB with the promoter of long non-coding RNA (lncRNA) MALAT1, and MALAT1 relieved the expression silencing of ZEB1 by sponging miR-204 targeting ZEB1 mRNA, resulting in the mesenchymal phenotype transition of pulmonary epithelial cells (Luo et al., 2019). Another novel study showed the correlation between PM2.5, ROS, NF-κB, and pro-fibrotic EMT in vivo. The oxidative stress within the rat lungs exacerbated during PM2.5 exposure but disappeared after the termination of exposure, and the upregulation of RelA/p65, deteriorated pulmonary fibrosis, and EMT in rat lungs were observed in the post-exposure phase, indicating that PM2.5 might lead to pulmonary fibrosis by oxidative stress-initiated NF-κB/inflammation/EMT pathway (Sun et al., 2019). Moreover, a previous study demonstrated that NF-κB activation following PM_2.5 exposure can be ROS dependent: The inhibition of ROS effectively abrogated the sustained NF-κB activation in BEAS-2B bronchial epithelial cells induced by PM_2.5 exposure (Song et al., 2017). Mice exposed to ultrafine carbon black particles showed oxidant-dependent NF-κB activation and increased expression of NF-κB responsive genes including TNF-α and IL-6 in lung tissues (Shukla et al., 2000). Silica particles, a kind of well-described pro-fibrotic agent, were able to induce a similar pathological response in silicosis murine models, whereas non-fibrogenic particles failed to induce this effect. These results implied that ROS-derived NF-κB and its downstream inflammatory mediators actively participate in fibrogenesis, in which EMT might act as a major participant (Hubbard et al., 2002). Additionally, an in vitro study showed increased IL-6 level and downstream JAK2/STAT3 signaling activation concomitant with the emergence of oxidative stress in PM_2.5-treated alveolar cancer cells, whereas a similar pattern was also observed in another in vitro study, in which ROS accumulation, IL-6/IL-8 release, and JAK2/STAT3 signaling pathway activation were observed to be mediated by cytochrome P450, family 1, member A1/B1 (CYP1A1/B1) after PM_2.5 exposure to human bronchial epithelial cells (Yuan et al., 2019). This evidence partly substantiated the hypothesis that PM_2.5 might induce EMT via ROS/NF-κB/IL-6/8/JAK/STAT3 axis (Xu et al., 2015). Although the NF-κB activation and following release of IL-6, IL-8, TNF-α, or mitochondrial dysfunction in epithelial tissues induced by PM_2.5-derived ROS have been validated to be significant in lung inflammation and cell death, the in-depth regulatory role of this shared pathway in EMT is largely unknown (Bourgeois and Owens, 2014; Leclercq et al., 2018; Liu et al., 2018; Wang et al., 2018; Xin et al., 2018). Currently, the link between cigarette smoke extract (CSE) and pulmonary inflammation together with EMT in airway epithelia has been preliminarily verified. CSE-induced inflammatory responses in human bronchial cells include NF-κB activation and further increased IL-6 transcription, which promote phenotype transformation via downregulation of miR-200-c, a proven anti-EMT micro-RNA (miRNA) family member. Hence, ROS in conjunction with NF-κB might lie at the intersection of PM_2.5-derived inflammation and EMT, although the exact mechanisms remain to be elucidated (Zhao et al., 2013). Controversy on this issue also existed. Fe₃O₄, one of the major anthropogenic components in PM, could result in a delayed NF-κB response in A549 neoplastic pulmonary epithelial cells as a result of the decrease in IκB degradation induced by excessive ROS (Konczol et al., 2011).

In summary, whether PM_2.5-intracellular ROS induced NF-κB phosphorylation could initiate EMT is still obscure, as the small amount of peripheral proof is too weak to conclusively establish this stand.

ROS are strong initiators of pro-EMT PI3K/Akt and Ras/ERK signaling. ROS have been shown to directly induce the phosphorylation of EGF and PDGF receptors in the absence of ligand and to enhance VEGF-induced receptor phosphorylation via oxidative repression of the intracellular inhibitors protein tyrosine phosphatase-1B (PTP1B) and density-enhanced phosphatase-1 (DEP1) (Lei and Kazlauskas, 2009; Oshikawa et al., 2010; Leon-Buitimea et al., 2012). The ROS-induced conversion of PIP₂ to PIP₃ also initiates Akt signaling events (Zhang et al., 2016b). Additionally, glutathione peroxidase inactivation-induced ROS accumulation resulted in the acquisition of mesenchymal characteristics of pancreatic cancer cells by the Akt/GSK-3β/Snail regulation (Meng et al., 2018). Likewise, the inhibition of GSK-3β, which increases Snail transcriptional activity and an EMT-like phenotype following ROS-stimulated ERK activation instead of Akt, was observed in Helicobacter pylori-treated gastric cancer cells (Ngo et al., 2016). Besides, the definitive role of ROS in activating ERK to promote the invasion of vascular smooth muscle cells and keratinocytes has also been unfolded. ROS-mediated ERK phosphorylation is also a significant alternative pathway for TGF-β-induced EMT in renal tubular epithelial cells (Rhyu et al., 2005; Wu, 2006).

PM_2.5 has been extensively regarded to stimulate the ERK or Akt pathway. PM_2.5 could induce MMP-13 expression, invasion, and migration of hepatocellular carcinoma cells via Akt (Zhang et al., 2017a). The increased secretion of EGF by human alveolar epithelial cells after PM_2.5 exposure and the further initiation of intracellular EGFR/ERK/NF-κB have also been observed (Jeong et al., 2017). In addition, a microfluidic system-based study on PM_2.5-treated bronchial epithelial cells showed significant activation of PI3K/Akt pathway, FGF/FGFR/MAPK/VEGF signaling, and the JAK/STAT pathway, which led to cell proliferation and apoptosis evasion. And increased intracellular and mitochondrial ROS levels implicated that PM_2.5-derived ROS could be correlated with the activation of these signaling series (Zheng et al., 2017).

In addition to ERK, other members of the MAPK family have been shown to participate in PM_2.5 responses. The coordination of p38 MAPK with proliferating cell nuclear antigen (PCNA) was responsible for PM_2.5-induced vascular smooth muscle cell proliferation, and the positive role of ROS was also implied by upregulation of superoxide dismutase (SOD), a cellular oxidative stress biomarker (Wan et al., 2018). Nevertheless, current research on the toxicology of PM_2.5-ROS also primarily focuses on the linkage between Akt or MAPK activation and inflammation or apoptosis, suggesting a double-edged role of PM_2.5-ROS that could either induce cell death or enhance cell viability (An et al., 2018; Li et al., 2018a). Abundant in PM_2.5, Ni²⁺ was shown to be responsible for cellular oxidative stress and the phenotype transformation of alveolar epithelial cells, and ROS/Akt-dependent MMP-9 and COL1 induction. Similarly, Ni²⁺/ROS/Akt also led to the inflammation of pulmonary epithelia and mice lung tissues (Yang et al., 2018). Hence, there is also crosstalk between PM_2.5-ROS-induced inflammation and EMT via RTK-related cascades. However, further research is needed to clarify the interplay of PM_2.5, ROS, Akt, and ERK with EMT.

Increase of cell invasion/migration is the characteristic of EMT regulated by subtle cytoskeleton remodeling and shifting in adhesion. TGF-β-induced intracellular ROS via NOX4 accelerates the polymerization of globular actin (G-actin) to fibrous actin (F-actin) in human umbilical vein endothelial cells, which is essential for lamellipodia and filopodia formation (Hu et al., 2005; Lamouille et al., 2014). ROS generated by NOX1 was also able to induce the spindle-like cellular morphology, EMT, and differentiation of human gingival epithelial cells via the post-transcriptional accumulation of CK18, a kind of pro-EMT intermediate filament (Sattayakhom et al., 2014). And ROS/ErbB2 signaling was also reported to be responsible for the aggressiveness of breast epithelial cells by phosphorylating FAK and activating RhoC (Wang et al., 2017). In another study, although mere H₂O₂ treatment was not sufficient to induce complete EMT in retinal pigment epithelial cells due to the transiency of the stimulation, ROS could also initiate Rho-GTP binding, resulting in stress fiber formation (Inumaru et al., 2009).

PM_2.5 has been verified as the modulator of cytoskeleton inducing incomplete contact between adjacent human bronchial epithelial cells, in which strong oxidative stress was also observed (Longhin et al., 2016). Besides, long-term chronic PM_2.5 exposure to human bronchial epithelial cells could enhance RhoA-GTP binding and stress fiber formation, although no increase of cellular proliferation was observed (Dornhof et al., 2017). In another research, however, PM_2.5 that stimulated ROS generation in corneal epithelial cells failed to promote EMT via cytoskeletal regulation. On the contrary, PM_2.5 inhibited FAK and paxillin phosphorylation, which further repressed RhoA activity and epithelial cell migration. Now that these inhibitory processes could be reversed with NAC treatment, the discrepant roles of PM_2.5-derived ROS in cytoskeleton regulation have been identified (Cui et al., 2018). In summary, the lack of comprehensive data leaves the role of PM_2.5 in cytoskeleton modulation shadowy.

Autophagy is a cellular degradation process that is activated in response to environmental stressors such as GF depletion and hypoxia (Yang et al., 2015; Marcucci et al., 2017). Recent evidence has linked autophagy and EMT through shared signaling pathways and functional causality in cancer progression and metastasis (Lv et al., 2012; Li et al., 2013; Catalano et al., 2015; Gugnoni et al., 2016; Zhang et al., 2017b). A recent study verified that PM_2.5 participated in liver fibrosis via ROS-mediated mitophagy. Hepatic stellate cells, a principal liver cell type with EMT plasticity, displayed enhanced α-SMA and COL1 expression after PM_2.5 treatment. These processes were mediated by ROS-induced PTEN-induced kinase 1 (PINK1) and Parkin, which signified mitophagy (Choi et al., 2009; Qiu et al., 2019). In lung cancer cells, PM_2.5-induced cell migration and invasion depended on autophagy, which was initially induced by increased intracellular ROS and its downstream effector lncRNA, loc146880 (Deng et al., 2017). Another study revealed an alternative mechanism underlying PM_2.5-autophagy-EMT in lung cancer cells, in which upregulated lncRNA LCPAT1 served as a messenger that mediated autophagy and cell migration by interacting with RCC2 after PM_2.5 exposure (Lin et al., 2018). This evidence not only validates the correlation between autophagy and PM_2.5-induced epithelia phenotype transformation but also provides new insight into the epigenetic toxicology of PM_2.5.

PM_2.5 can also activate EMT via diverse signaling pathways, including HMGB1-receptor for advanced glycation end-products (RAGE), Shh, Wnt/β-catenin, and Notch pathway together with SMAD1 signaling, in which the role of ROS has not been elucidated. HMGB1 and its cognate receptor RAGE were both induced in human bronchial epithelial cells and in the murine airway in response to PM_2.5 exposure. HMGB1 binding further activated TGF-β1 and PDGF release, providing strong evidence that HMGB1 is a mediator of PM_2.5-induced EMT (Zou et al., 2018). Increased Shh expression activated downstream Shh/Gli1 signal transduction and further enhanced Snail expression as well as cell migration after low dose PM_2.5 exposure to human bronchial smooth muscle cells (Ye et al., 2018). Moreover, the enhanced expression of Notch1 together with downstream TFs Hes1, Snail, Slug and biomarkers of EMT was detected in lungs of mice after intratracheal instillation of PM_2.5 with varied concentrations. The induction of cancer stem cells markers including OCT4 and SOX2 in A549 neoplastic pulmonary epithelial cells following the activation of PM_2.5/Notch1 pathway further validated the pro-EMT property of PM_2.5 (Wang et al., 2019). Furthermore, unlike the SMADs-dependent EMT mechanisms stated above, SMAD1 activity could also be indirectly induced by repressing expression of SMAD6/7 after PM_2.5 treatment on neoplastic pulmonary epithelial cells, resulting in the inhibition of SMAD1 degradation and induction of the spindle-like cellular morphology with cadherin switch (Yang et al., 2017). Interestingly, rather than directly responding to PM_2.5 exposure, Wnt/β-catenin signaling promoted the proliferation of human alveolar cancer cells and xenograft tumor growth via PM_2.5-induced Wnt3a-enriched exosomes (Xu et al., 2018). Our recent research was focused on exosomal miRNAs of EMT-like pulmonary epithelia after PM_2.5 exposure. It predicted the abnormal regulation of multiple cancer-related pathways by exosomal miRNAs in a systematic perspective, further emphasizing the significance of diverse interactions between exosomes and signaling pathways in the process of PM_2.5-induced EMT (Xu et al., 2019b). ATP citrate lyase (ACLY) linking glycolytic and lipidic metabolism has been considered as a novel anticancer target, and recently, it has also been found responsible for PM_2.5-induced EMT (Granchi, 2018). Increase of intracellular ACLY expression and citrate acid levels synchronized with EMT phenotype transformation of human bronchial epithelial cells after long-term PM_2.5 exposure, and the knock down of ACLY significantly abrogated the in vivo metastasis ability of the cells (Fu et al., 2019).

Showing high affinity for a variety of PAHs, ligand-bound aryl hydrocarbon receptor (AHR) is able to interact with Ah receptor nuclear translocator protein (ARNT) and further mediate the toxicities of PAHs by stimulating the expression of target genes such as CYP1A1, which oxidizes PAHs and generally enhances their toxicities (Safe et al., 2013; Noakes, 2015). The AHR-ARNT complex can also promote cell proliferation by binding to the Slug promoter, activating ERK in the absence of external GFs, and enhancing FGF, PDGFA, and PCNA production (Shimba et al., 2002; Diry et al., 2006; Randi et al., 2008; Wang et al., 2009; Safe et al., 2013; John et al., 2014). Because PAHs account for a large portion of the organic constituents in PM_2.5, their contribution to PM_2.5-induced EMT might be nonnegligible. The significant cyp1a1 upregulation following the exposure of HBE cells to diesel exhaust particles (DEPs) or the exposure of BEAS-2B and A549 cells to PM_2.5 proved the ability of PM to activate AHR signaling (Gualtieri et al., 2011, 2012; Rynning et al., 2018). Likewise, inhibiting AHR or CYP1A1 in BEAS-2B and A549 epithelial cells before PM_2.5 exposure could effectively rescue PM_2.5-induced E-cadherin downregulation and N-cadherin upregulation. In this study, lncRNA MALAT1 showed synchronized expression tendency with N-cadherin, indicating that it functions as an effector regulating EMT at downstream of PM_2.5/AHR signaling, in addition to its role in NF-κB pathway reviewed above (Huang et al., 2017). These studies indicated that PM_2.5/PAHs/AHR/CYP1A1 signaling significantly participates in EMT initiation, although more exploration is urgently needed.

Epigenetic regulations through lncRNAs and miRNAs underlying PM_2.5-induced EMT have been partly reviewed above. Besides that, miR-16-1-3p directly targeting Twist mRNA could be inhibited by PM_2.5 exposure, participating in PM_2.5 exposure-induced EMT of hepatocellular carcinoma cells (Zhang and Li, 2019). Moreover, PM_2.5 exposure also participates in initiation of EMT of bronchial epithelial cells by altering DNA methylation (Li et al., 2017). After 5 weeks of PM_2.5 exposure, hypomethylated CpG islands (CpGs) in the genome of bronchial epithelial cells were significantly enriched in genes associated with GTPase activity, extracellular matrix organization, and GF stimuli, whereas hypermethylated CpGs were clustered in genes responsible for cell adhesion and ion transportation (Hesselbach et al., 2017). Another set of DNA methylation analyses on bronchial epithelial cells after PM_2.5 exposure for 24 h showed comparable results, as Gene Ontology (GO) terms related with regulation on actin filament and adherin junction were significantly enriched in hypomethylated gene list (Shi et al., 2019). Because these annotated genes play dominant roles in EMT, these in silico analyses strongly indicated an association between PM_2.5 exposure and EMT from a systematic perspective.

Calcium signaling is highly correlated with multiple malignant cancers. Alterations in the expression of Ca²⁺ channels or transporters have been identified in tumor cells, indicating that ectopic Ca²⁺ signal transduction might play a role in EMT (Cui et al., 2017). For example, the activation of transient receptor potential cation channel 7 (TRPM7) and Ca²⁺ influx through TRPM7 indicates the initiation of EMT. Furthermore, EGF-induced TRPM7-Ca²⁺-dependent STAT3 activation significantly upregulated Twist, Snail, and EMT marker expression in breast cancer cells (Davis et al., 2014). Ca²⁺ is also able to elicit the expression of the ATP-binding cassette subfamily C member 3 (ABCC3), a promoter of Twist and EMT in breast cancer cells (Stewart et al., 2015). As Ca²⁺ has been well characterized as one of the most plentiful cations in PM_2.5, and PM_2.5 is able to penetrate into the bloodstream or be taken by cells, PM_2.5 exposure might result in increased Ca²⁺ concentration in microenvironment around the epithelia or cytoplasm, which could further regulate the activity of pro-EMT Ca²⁺ signaling (Huang et al., 2018; Qin et al., 2018; Xie and Zhao, 2018). In our recent study, pulmonary epithelial cells showed EMT-like phenotype after long-term low dose PM2.5 exposure, and a subset of the intracellular differentially expressed genes was significantly related to the cellular response to calcium ion (Xu et al., 2019b). However, although limited in vitro studies have reported that PM_2.5 exposure could raise the intracellular Ca²⁺ concentration of BEAS-2B cells and another non-epithelial cell line, the biological correlation of this event with EMT was yet to be clearly figured out (Zheng et al., 2017; Zhao et al., 2019).