Characterization of materials including elemental content analysis, surface area, zeta potential and particle size measurements were conducted and the results are summarized in Table 1. SWCNT (CNI, Houston, TX), MWCNT (MWNT-7, lot #05072001K28; Mitsui & Company, Tokyo, Japan), UFCB (Elftex 12; Cabot, Edison, NJ), and ASB (Crocidolite, CAS 12001-28-4; National Institute of Environmental Health Sciences, Research Triangle Park, NC) were analyzed for elemental contents by nitric acid dissolution and inductive coupled plasma-atomic emission spectroscopy. Surface area of SWCNT, MWCNT, UFCB, and ASB was analyzed by Brunauer Emmett Teller (BET) nitrogen adsorption technique. Particle dimensions were assessed by electron microscopy. Dynamic light scattering (DLS) measurements of average hydrodynamic diameter were performed with a NanoSight NS300 (Malvern Instrument, Worcestershire, UK). All particles were dispersed in cell culture medium and DLS measurements were conducted using scattering angle of 90° with an argon ion laser set at excitation wavelength of 488 nm. Zeta potential was measured using a Zetasizer Nano ZS90 (Malvern Instrument). Particle were dispersed in cell culture medium and equilibrated inside the instrument for 2 min, and five measurements (10 sec delay between measurements) each consisting of five runs (2 sec delay between runs) were recorded.

All test materials were kept as a dry powder in tightly closed containers at 4 °C and freshly dispersed in phosphate-buffered saline (PBS) to obtain stock solutions at 0.1 mg/mL. A natural lung surfactant, Survanta^® (Abbott Laboratories, Columbus, OH, USA), was then added to SWCNT, MWCNT, and UFCB stock solutions at the concentration of 150 μg/mL to aid particle dispersion. All particle preparations were sonicated using light sonication (Sonic Vibra Cell Sonicator, Sonic & Material Inc., Newtown, CT, USA) with the power, frequency and amplitude settings of 130 W, 20 kHz and 60% for 10 seconds, followed by a dilution with cell culture medium to the exposure dose. The concentration of Survanta^® used in this study was shown to cause no cellular toxicity or neoplastic transformation.⁴⁷

Primary human small airway epithelial cells immortalized with hTERT (SAECs) were kindly provided by Dr. Tom Hei (Columbia University, New York, NY, USA).⁴⁸ SAECs were grown in small airway epithelial cell growth basal media (SABM) supplemented with Clonetics SAGM SingleQuots (Lonza, Walkersville, MD, USA) containing 0.4% bovine pituitary extract, 0.1% insulin, 0.1% hydrocortisone, 0.1% retinoic acid, 1% bovine serum albumin, 0.1% transferrin, 0.1% triiodothyronine, 0.1% epinephrine, 0.1% human epidermal growth factor and 0.1% gentamicin. Cells were continuously exposed to 0.02 μg/cm² of SWCNT, MWCNT, UFCB or ASB for 6 months and maintained at 37 °C in a humidified 5% CO₂ as previously reported.²⁰ The cells were passaged weekly at pre-confluent densities using a solution containing 0.05% trypsin and 0.5 mM EDTA (Invitrogen, Carlsbad, CA). Passage-matched cells were exposed to the dispersant for 6 months utilized as control cells.

The bottom layer of agar medium (0.5%) was prepared by mixing 1:1 mixture of 1% agarose and 2× concentrate MEM medium (Lonza) containing 15% fetal bovine serum (FBS) and 1% gentamicin at 44 °C. Each growth factor from the SAGM SingleQuots kit (Lonza) was added to the warm agar to obtain the corresponding concentration of their normal growth medium. The bottom layer of agar medium (600 μL) was added and allowed to solidify for 30 minutes in a 24-well plate. The upper layer was prepared by suspending 3×10³ cells in 0.33% agar medium (200 μL). Next, normal growth medium of SAECs (300 μL) was added over the upper layer after it was solidified at 37 °C and replaced with fresh culture medium every three days. Colony formation was captured under a microscope (Keyence BZ-X700; Osaka, Japan). Relative colony number and surface area were calculated by dividing the values of particle-exposed cells by that of passage-matched control cells.

Cells were seeded on a cover slip at the density of 1.5×10⁴ cells per well in 24-well plates with a total volume of 500 μL. After a 48-h incubation period, cells were fixed with 4% paraformaldehyde and permeabilized with 0.25% Triton-X for 10 min at room temperature. After that, the cells were blocked with a medium containing 0.5% saponin, 1% FBS, and 1.5% goat serum, and incubated with primary antibodies against CD133 (1:200 dilution, Thermo Fisher Scientific, Waltham, MA), phosphorylated histone H2AX (γ-H2AX; 1:300 dilution, Millipore Sigma, Burlington, MA), and β-actin (1:3,000 dilution, Cell Signaling Technology, Danvers, MA) overnight at 4 °C. The cells were then incubated with AlexaFluor 488 and 647-conjugated secondary antibodies (1:500 dilution, Thermo Fisher Scientific, Waltham, MA) for 1 h at room temperature, and mounted in DAPI-containing medium (Vector Laboratories Inc, Burlingame, CA). Immunofluorescence images were acquired using Keyence BZ-X700 fluorescence microscope and quantified using an image J analysis software.

Cells were seeded at the density of 1.5×10⁵ cells per well in 6-well plates with a total volume of 2 mL. After a 48-hour incubation period, the cells were lysed with ice-cold lysis buffer containing 20 mM Tris–HCl (pH 7.5), 1% Triton X-100, 150 mM NaCl, 10% glycerol, 1 mM Na₃VO₄, 50 mM NaF, 100 mM phenylmethylsulfonyl fluoride, and a commercial protease inhibitor mixture (Roche Molecular Biochemicals, Indianapolis, IN) for 40 minutes. Cell lysates were determined for protein content using the Bradford assay kit (Bio-Rad Laboratories, Hercules, CA). Equal amounts of denatured proteins (40 μg) were loaded onto 7.5–10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred onto nitrocellulose membranes (Bio-Rad). The transferred membranes were blocked in 5% nonfat dry milk in TBST (25 mM Tris–HCl, pH 7.4, 125 mM NaCl, 0.05% Tween 20) for 1 hour and incubated with primary antibodies against CD133, p53, CD44, ABCG2, OCT4, NANOG, SOX2, ZEB1, SNAI1, SNAI2 and GAPDH (Cell Signaling Technology, Danvers, MA) at 4°C overnight. Membranes were washed twice with TBST for 10 minutes and incubated with HRP-coupled isotype-specific secondary antibodies (Cell Signaling Technology, Danvers, MA) for 1 hour at room temperature. The immune complexes were detected by an enhanced chemiluminescence detection system (Millipore Corporation, Billerica, MA) and quantified using image J densitometry software.

Cells were detached and made into single-cell suspension in growth factor-free SABM medium. The suspended cells were then plated onto a 6-well ultralow attachment plate at the density of 1×10⁵ cell/mL with a total volume of 2 mL and incubated for 60 h. The cells were harvested at various times (0–60 h) to assess cell viability using MTS assay (Promega, Madison, WI), according to the manufacturer’s protocol. Apoptosis of the suspended cells was determined by Hoechst 33342 assay (Molecular Probes, Eugene, OR). After 12 h, the above single-cell suspension in 6-well ultralow attachment plate were seeded on to 96-well ultralow attachment plate with a total volume of 200 μL. Cells were incubated with 10 μg/mL Hoechst 33342 for 30 min at 37 °C and visualized under a fluorescence microscope. Cells with intense nuclear fluorescence were considered apoptotic. . The percentage of apoptosis was calculated from 10 random fields from approximately 1,000 nuclei for each sample.

Spheroid formation was performed under stem cell-selective culture conditions as previously described.⁴⁹ Cells were detached and resuspended in 0.8% methylcellulose (MC)-based serum-free medium (Stem Cell Technologies, Vancouver, Canada) containing 20 ng/mL epidermal growth factor (EGF; BD Biosciences, San Jose, CA), basic fibroblast growth factor (bFGF), and 4 μg/mL insulin (Sigma). Cells at the density of 1×10⁴ cell/well were plated onto an ultralow attachment 24-well plate with a total volume of 600 μL and cultured for 4 weeks. EGF and bFGF (20 ng/mL) and insulin (4 μg/mL) were added every three days. The culture medium was replaced with fresh 0.8% MC-based serum-free medium supplemented with growth factors as described above once a week. Spheres were harvested, dissociated into single cell suspensions, and cultured under stem cell-selective conditions to form secondary spheroids. Relative spheroid numbers and areas were calculated by dividing the values of particle-exposed cells by passage control cells

Whole genome mRNA microarray datasets from our previous work²⁰ were uploaded into Ingenuity Pathway Analysis (Qiagen) to investigate the roles of SOX2-associated CSC- and EMT-associated signaling in CNM-exposed cells. Roles for SOX2 in ranked cellular functions, upstream regulator, and signaling networks were searched within the results of the Core Analysis. For those functions and networks with SOX2, all genes associated with the function were plotted to investigate signaling relationships and potential relationships with previously identified pro-cancer signaling in each cell type.²⁰ For those transformed cells that did not show over-expressed SOX2, we investigated other transcriptional regulators including SNAI1. Meaningful signaling networks were overlaid with mRNA differential expression and Z-score-based activation/inhibition prediction values (Z ≥ ±2).

Data are presented as the means S.D. from at least three independent experiments. Statistical differences were determined using two-way ANOVA and post hoc Tukey’s test for multiple comparison at a significance level (α) of 0.05.

Physiochemical properties of all test nanoparticles are summarized in Table 1. Electron microscopic characterization of all test particles was performed and reported previously by our group.³⁰ SWCNT (CNT, Houston, TX) were produced by high-pressure carbon monoxide (HiPco) technique. They were purified to remove metal contaminants using acid treatment. Elemental analysis by nitric acid dissolution and inductive coupled plasma-atomic emission spectroscopy showed that the SWCNT contained 99% elemental carbon and 0.23% iron. Surface area was assessed by Brunauer Emmet Teller (BET) method and was 400–1,040 m²/g. The length and width of individual SWCNT as determined by field emission scanning electron microscopy were 1–4 μm and 1–4 nm, respectively. Particle agglomerate or hydrodynamic diameter measured by DLS and zeta potential of SWCNT in culture medium were 122.5 nm and −17.1 mV, respectively. The nanomaterial was previously characterized by Wang et al.⁴⁷ and Sargent et al.¹³ MWCNT (MWNT-7, lot #05072001K28; Mitsui & Company, Tokyo, Japan) was synthesized by chemical vapor deposition. Elemental analysis of MWCNT indicated 99% elemental carbon and 0.32% iron. BET surface area was 26 m²/g. Length and width obtained from field emission scanning electron microscopy of individual MWCNT was 8.19 ± 1.7 μm and 81 ± 5 nm, respectively. Hydrodynamic diameter and zeta potential of MWCNT in culture medium were 163.3 nm and −17.6 mV, respectively. Hydrodynamic size of non-spherical particles such as CNTs and asbestos fibers should be taken with caution and only as an estimate of their hydrodynamic size. Detailed characterization of MWCNT was described by Pacurari et al.⁵⁰, Porter et al.⁵¹, and Mishra et al.⁵² UFCB (Elftex 12; Cabot, Edison, NJ) was synthesized by vapor-phase pyrolysis. Content of UFCB was 99% elemental carbon and 0.0011% iron. BET surface area and width of individual UFCB were 43 m²/g and 37 nm, respectively. Length was not applicable (i.e. spherical shaped based on electron microscopic observations). Hydrodynamic diameter and zeta potential of UFCB in culture medium were 126.7 nm and −15.4 mV, respectively. UFCB was previously characterized by Wang et al.⁴⁷ ASB (Crocidolite, CAS 12001-28-4) was from the National Institute of Environmental Health Sciences (Research Triangle Park, NC) and was characterized by Msiska et al.⁵³ Elemental analysis of ASB was 50.9% SiO₂, 31.8% iron and less than 1% carbon. BET surface area, length and width of individual ASB were 9.8 m²/g, 10 μm and 210 nm, respectively. Hydrodynamic diameter and zeta potential of ASB in culture medium were 144.9 nm and −18.4 mV, respectively.

Our group previously demonstrated that Survanta^®-dispersed SWCNT (SD-SWCNT) have a smaller sized structure compared to PBS-dispersed particle (ND-SWCNT). SD-SWCNT elicited a cell growth-stimulating effect but induced a suppressive effect at high doses. On the other hand, ND-SWCNT showed no effect, suggesting that dispersion status of particles is essential to their bioactivities. We also found that the concentration of Survanta^® used in this study does not mask the bioactivity of SWCNT or cause cytotoxicity.⁴⁷ Therefore, Survanta^® dispersion method was employed to improve the dispersion of all tested CNMs. The dispersion of CNMs in culture medium was observed under a microscope and an average agglomerate size and zeta potential of the particles were determined using DLS as shown in Table 1. Human small airway epithelial cells (SAECs) were continuously exposed to non-cytotoxic concentration of SWCNT, MWCNT, UFCB or ASB at 0.02 μg/cm² and passaged weekly for a period of 6 months as previously described.³⁰ The nanomaterial dose used in this study was calculated based on the lowest in vivo dose of CNTs that caused biological response in mice (10 μg/mouse)^{5, 7} normalized to the average mouse alveolar surface area (~500 cm²)⁵⁴, indicating an in vitro surface area dose of 0.02 μg/cm². ASB was used in this study as a carcinogenic particle control with similar morphology to CNTs. To assess the carcinogenic potential of CNMs, the exposed cells were examined for their ability to form colonies under non-adherent conditions, which is a phenotypic characteristic of cancer cells.⁵⁵ The cells were grown on soft agar and colony formation was determined at 28 days post-plating. Consistent with the previous reports by our group in human bronchial (BEAS-2B) and small airway epithelial cells.^{29, 30} SWCNT-exposed BEAS-2B cells, SWCNT- and MWCNT-exposed SAECs were able to form significantly more and larger colonies as compared to control cells which formed very few small colonies at 6 months post-exposure. ASB- and SWCNT-exposed cells formed the highest number of colonies, followed by MWCNT- and UFCB-exposed cells, respectively (Fig. 1A–C). Less colony formation of the control cells was first observed at 14 days post-plating. This is likely due to an inability of normal cells to grow and form spheroids under non-adherent conditions, whereas transformed cells possess this ability due to the activation of oncogenic signaling pathways.⁵⁶ Since the exposed cells were maintained in normal cell culture medium without the particles for at least ten passages before testing, these results indicate that long-term exposure to CNMs induced irreversible neoplastic-like transformation in human small airway epithelial cells similar to that observed with ASB.

Sustained DNA damage is frequently associated with carcinogenic transformation¹⁶ and failure to repair DNA lesions, particularly double-strand breaks (DSBs), may contribute to genomic instability and cancer development.¹⁶ In case of particle-induced carcinogenesis, it was shown that ASB was able to induce DNA strand breaks in pleural mesothelial cells,⁵⁷ which is believed to contribute to its pathogenicity. In this study, we investigated the ability of CNMs to induce DNA strand breaks in the exposed lung cells by measuring phosphorylated histone H2AX (γ-H2AX) foci, a widely accepted biomarker for DSB.⁵⁸ Fig. 1D,E shows that like ASB-exposed cells, SWCNT-, MWCNT- and UFCB-exposed cells exhibited a significant increase in γ-H2AX-positive cells (>5 foci/cell)⁵⁹ as compared to control cells. The magnitude of induction was quite comparable among the tested particles, with SWCNT-exposed cells exhibiting the most significant change. SWCNT and MWCNT were previously shown to induce mitotic spindle alterations and chromosomal aneuploidy in primary and immortalized human airway epithelial cells.^{13, 14} Likewise, pulmonary administration of UFCB (Printex 90) in C57BL/6 mice caused substantial DNA damage in bronchoalveolar lavage and lung tissue cells.¹⁵ These findings support the genotoxicity and carcinogenic potential of the tested CNMs. Indeed, MWCNT (Mitsui-7) and UFCB are currently classified as a group 2B carcinogen by the International Agency for Research on Cancer, whereas other CNTs were non-classifiable due to the lack of sufficient evidence for their carcinogenicity.

p53 is a tumor suppressor protein that is involved in DNA repair process and functions to protect the genome from mutations and genomic aberrations.⁶⁰ Loss of function and/or mutation of p53 has been regarded as a predisposing factor for tumor initiation.⁶¹ In this study, we investigated the effect of long-term CNM exposure on p53 expression in human SAECs by immunoblotting. Fig. 1F,G shows that p53 expression was substantially downregulated in SWCNT-, MWCNT-, UFCB- and ASB-exposed cells as compared to control cells, with the effect being most pronounced in SWCNT-exposed cells. These results are in good agreement with our previous finding showing a significant decrease in basal p53 expression in CNT- and ASB-exposed cells.³⁰ However, the expression of phospho-p53 in these cells remained intact, suggesting the preservation of p53 functionality in these cells. In bronchial epithelial BEAS-2B cells, however, the level of phospho-p53 was decreased upon long-term exposure to SWCNT.²⁹ Since BEAS-2B cells have dysregulated p53 expression as a result of SV40-mediated immortalization, these results suggest that cell immortalization could result in attenuated phospho-p53 in the exposed cells. Together, our results demonstrated the DNA-damaging effect of chronic CNM exposure in SAECs and provided a possible mechanism of particle-induced DNA damage through p53 downregulation.

Cancer stem cells (CSCs) are defined by their fundamental properties of self-renewal, tumor-initiating capacity, and ability to give rise to more differentiated progeny.⁶² It was evidenced that CSCs are able to survive and proliferate to form spheroid structures under non-adherence in serum-free media known as stem-cell selective conditions.⁴⁹ In this study, we examined the effect of CNMs on CSC-like phenotypes in SAECs by anoikis (detachment-induced apoptosis) and spheroid formation assays. In anoikis assay, cells were detached, suspended in serum-free medium, and incubated for 60 hours in non-adherent poly-HEMA-coated plates. The cells were then collected and evaluated for cell survival at different time points by MTS assay. Fig. 2A shows that as compared to passage-matched control cells whose survival gradually decreased over time, SWCNT-, MWCNT- and UFCB-exposed cells exhibited death (anoikis) resistance, as indicated by their increased survival under non-adherent conditions. Similar to ASB-exposed cells, SWCNT- and MWCNT-exposed cells were able to retain anoikis resistance at 9–60 hours after cell detachment, whereas UFCB-exposed cells exhibited anoikis resistance at 9–12 hours (Fig. 2A). We also evaluated apoptosis of the exposed cells by Hoechst 33342 assay, which detects DNA condensation and fragmentation, a morphologic characteristic of apoptotic cells. Fig. 2B,C shows that all CNM- and ASB-exposed cells exhibited apoptosis resistance relative to control cells, as indicated by their fewer number of Hoechst 33342-positive cells at 12 hours post-detachment. ASB- and SWCNT-exposed cells were found to be most resistant, followed by MWCNT- and UFCB-exposed cells, respectively (Fig. 2A–C). These results are consistent with our previous studies showing apoptosis resistance of SWCNT-transformed cells to various agents including etoposide, antimycin A, Fas ligand, and tumor necrosis factor-α (TNF-α).^{29, 63}

To substantiate the effect of CNMs on stem properties of SAECs, we tested the ability of CNM- and ABS-exposed cells to induce 3D spheroid formation in culture. The nanomaterial-exposed cells were detached and seeded at a low density onto non-adherent plates. The primary spheroids were allowed to form in serum-free media for 21 and 28 days. Fig. 2D–F shows that SWCNT-, MWCNT-, UFCB- and ASB-exposed cells were able to form significantly more spheroids than control cells. We also determined the effect of long-term particle exposure on secondary spheroid formation. In this study, primary spheroids were detached, resuspended in serum free media, and grown under non-adherent conditions for another 21 and 28 days. Fig. 2G–I depicts the number of secondary spheroids formed at the two time points, showing a significantly higher number of spheroids formed by ASB- and SWCNT-exposed cells, followed by MWCNT- and UFCB-exposed cells, as compared to control cells. This is in accordance with our previous findings that formation of primary and secondary spheroids was strikingly increased in chronic SWCNT-exposed BEAS-B cells at 14 days after plating.⁶⁴ These results support our earlier finding and indicate the ability of ABS- and CNM-exposed cells to self-renew and exhibit CSC-like properties.

CD133 is a commonly used biomarker for lung CSCs,⁶⁵ and its high expression in lung cancer patients has been associated with poor survival outcome⁶⁵ and short disease free period.⁶⁶ CD133-positive lung cancer cells were reported to possess self-renewal and unlimited proliferative capabilities, whereas these characteristics were absent in CD133-negative cells.⁶⁷ In this study, we evaluated the expression of CD133 in CNM- and ASB-exposed cells to assess their stem properties by immunofluorescence staining. Fig. 3A,B shows that the expression of CD133 was significantly elevated in the secondary spheroid of CNM- and ASB-exposed cells, as compared to control cells, with the expression in SWCNT-exposed cells being the highest, followed closely by ASB- and MWCNT-exposed cells. UFCB-exposed cells expressed the lowest level, although their CD133 expression was still higher than that of the control cells. In an effort to determine the localization and distribution of stem cells in the spheroids, we performed z-stacking image analysis of fluorescently stained spheroids. Orthogonal views and 3D images were created and analyzed for CD133 expression along the z-axis. Fig. 4A,B shows that the majority of CD133-expressing cells resided in the central area of the spheroids. Such localization could be a result of less oxygen being able to diffuse through the central area of the spheroids, creating an advantageous condition for preserving CSC-like characteristic.⁶⁸

To validate the CSC-inducing effect of CNMs, the CNM-exposed cells were analyzed for CD133 and other known CSC markers, including CD44⁶⁹ and ABCG2,⁷⁰ by Western blotting. CD44 and ABCG2 have been linked to the occurrence of lung cancer,^{71, 72} but their involvement in nanoparticle-induced pathogenesis is not known. Consistent with the immunofluorescence results, our Western blot results showed an increased expression of CD133 in CNM- and ASB-exposed cells relative to control cells (Fig. 3C,D). The expression levels of CD44 and ABCG2 in these cells, however, showed no clear discernible pattern with the expression of CD44 being relatively unchanged in all CNM-exposed cells, whereas ABCG2 expression was upregulated in all but MWCNT-exposed cells. These results indicate particle type-specific regulation of stem cell markers, with CD133 as the most predictive marker of CNM-induced CSCs.

To understand the molecular mechanisms of CNM-induced CSCs, we investigated self-renewal and EMT transcription factors, which are known to be essential to maintaining CSC functions,^{33, 73} in CNM-exposed cells. Immunoblot analysis showed a substantial increase in stem cell transcription factor SOX2 in SWCNT-, MWCNT- and ASB-exposed cells, with a negative effect in UFCB-exposed cells (Fig. 5A,B). The magnitude of upregulation in SWCNT-, MWCNT- and ASB-exposed cells were 15, 2.5 and 16-fold over a control level, respectively. The expression of other self-renewal transcription factors, including OCT4 and NANOG, were relatively unchanged in all exposed cells (Fig. 5A,B), suggesting their limited role in controlling the stem properties of these cells.

In the lung, SOX2 expressing cells were shown to be responsible for tissue homeostasis and cell proliferation during injury repair.^{74, 75} SOX2 expression was also reported to be elevated in lung cancer cells,⁴² which promoted survival, oncogenic phenotype, and tumorigenesis of lung CSCs possibly through the upregulation of multiple oncogenes including c-Myc, Wnt1, Wnt2, and Notch1.⁴⁴ Similarly, SOX2 overexpression was found to enhance oncogenic phenotypes of lung cancer cells via a positive feedback loop of epidermal growth factor receptor and BCL2L1.⁷⁶ A recent study suggested that SOX2 overexpression led to bronchial epithelial cell dysplasia through an upregulation of oncogenic proteins such as cyclin D1 and BCL2, which control cell cycle progression, apoptotic cell death, and tumor suppressor protein p21.⁷⁷ Therefore, strong evidence supports self-renewal and oncogenic properties of SOX2 in human lung cells, which may drive the pathogenic development of CNM- and ASB-exposed cells.

Increasing evidence also supports the role of EMT in CSC regulation and tumor progression in various cancers.^{34, 78–80} EMT is a cellular transformation process that allows epithelial cells to adopt mesenchymal phenotype to increase their motility and invasiveness, which is crucial to cancer development and progression.⁸¹ Key EMT transcription factors such as ZEB1, SNAI2, and SNAI1 have been associated with poor prognosis of lung cancer.^82–84 In this study, we evaluated the expression of ZEB1, SNAI1, and SNAI2 in CNM- and ASB-exposed cells by immunoblotting. Fig. 5C,D shows that ZEB1 was weakly upregulated in SWCNT-, MWCNT- and ASB-exposed cells, and minimally expressed in UFCB-exposed cells, as compared to control cells. SNAI2 expression was also weakly upregulated in SWCNT- and MWCNT-exposed cells, but downregulated in ASB-exposed cells (Fig. 5C,D). On the other hand, SNAI1 was highly upregulated in SWCNT-, MWCNT- and UFCB-exposed cells, but downregulated in ASB-exposed cells. These results indicate the variability and complexity of EMT-mediated regulation of CSCs by nanoparticle-induced transcription factors. Most EMT transcription factors appeared to be upregulated in SWCNT- and MWCNT-exposed cells, which may attribute to their neoplastic and CSC-like properties. ZEB1 has been shown to positively regulate anchorage-independent cell growth in lung cancer cells.⁸⁵ It also suppresses tumor suppressor genes such as Semaphorin 3F.⁸⁶ SNAI2 was reported to promote tumor angiogenesis, metastasis, and immune suppression in lung cancer cells via downregulation of 15-hydroxyprostaglandin dehydrogenase, an enzyme responsible for prostaglandin E2 degradation.⁸⁷ SNAI1 was shown to be involved in cell malignancies and its genetic knockdown repressed the proliferation rate and tumorigenicity of lung cancer cells.⁴⁶ Together, these studies support for the role of EMT-related transcription factors in neoplastic and CSC-like transformation of SAECs after a chronic exposure to CNMs and ASB.

To further investigate the potential roles of CSC- and EMT-associated genes in CNM-exposed cells, pathway analysis was conducted using whole genome mRNA expression from each exposed cell type. Based on the whole genome expression profile, Upstream Analysis in Ingenuity Pathway Analysis (IPA) predicted an activation of SOX2-OCT4-NANOG complex (Z = 2.0) in SWCNT-exposed cells (Fig. 6A). SOX2 was overexpressed by 14.3-fold while POU5F1 (OCT4) was under-expressed 6.4-fold compared to control cells. To investigate the gene signaling responsible for this prediction, all associated genes both up and downstream of this complex were plotted. Downstream over-expressed SMARCAD1 (10.1-fold) and RIF1 (5.8-fold) and downregulated PAX6 (−6.9-fold) and HOXB1 (−6.3-fold) were identified (Fig. 6A). The proteins associated with CSC signaling were added to this network (PROM1 [CD133], ABCG2, SNAI1, and SNAI2) to understand the signaling relationships with SOX2 signaling network. Activated SOX2 caused a direct overexpression of downstream SMARCAD1, PROM1, SNAI1, and NANOG with indirect activation of RIF1. Both SMARCAD1 and RIF1 are involved in Ataxia telangiectasia mutated (ATM)-mediated DNA double strand break (DSB) repair and DNA replication repair at the 5’ end.^{88, 89} SMARCAD1 can contribute to naïve pluripotency in early embryo development,⁹⁰ hence suggesting that SOX2-SMARCAD1 signaling may play an important role in CSC-like phenotype in CNT-exposed cells. RIF1 is involved in stem cell pluripotency and its overexpression is associated with poor prognosis in non-small cell lung cancer (NSCLC) due to suppressed homologous recombination following DSB repair, genetic instability, and mutagenesis.⁹¹ These findings suggest that SOX2 overexpression and activation are in response to DSB repair following chronic CNT exposure and may contribute to their neoplastic CSC-like phenotype.

Based on the overexpression of SOX2 in ASB-exposed cells, the signaling relationships associated with SOX2 involving oncogenesis and stem cell development were investigated in IPA. Cancer was identified as the top ranked cellular function in ASB-exposed cells with 236 differentially expressed genes and a Log p-value = 9.5E-6. In this signaling network, an overexpressed SOX2 was found to activate CD44 and ETV5. To further understand SOX2-associated signaling, the IPA knowledge base was queried for all known genes associated with SOX2. This network displayed unique crosstalk between cell differentiation and pro-cancer signaling (Fig. 6B). Overexpressed CD44 with overexpressed SOX2 was found in lung cancer cells following PM_2.5 exposure,⁹² and promoted lung cancer in animal models by enhancing angiogenesis, migration, and apoptotic signaling.^{93, 94} In addition, overexpressed ATF3, a transcriptional regulator that maintains genetic integrity under stressful conditions, indicates ASB potential damage to SAEC DNA. ATF3 overexpression was observed in NSCLC patient tumor samples.⁹⁵ Upregulated ETV5 downstream of SOX2 was found to have critical function for retaining alveolar Type II cell phenotype and associated with lung tumor initiation.⁹⁶

Downregulated FGFR2, FN1, and LOX appeared to assist in downregulating SNAI1, suggesting that extracellular matrix influenced SOX2 activity. Decreased FGFR2 expression is associated with maintaining undifferentiated cells during lung branching development by repressing SOX2, and plays a role in cancer development.⁹⁷ LOX crosslinks collagen/elastin and may stabilize microtubules in the mitotic spindle during mitosis, thus acting as a tumor suppressor⁹⁸ while FN1 is associated with cell migration and adhesion. Overexpressed GATA3, a T-cell developmental and endothelial cell regulator, was associated with FN1 and decreased LOX and NF2. NF2 possesses tumor suppression activity, regulates cytoskeleton dynamic, and is downregulated in malignant mesothelioma, suggesting a strong link with asbestos-induced cancer.⁹⁹ Interestingly, the loss of NF2 expression may have promoted overexpressed SRC, a proto-oncogene and cancer cell invasion signaling in many cancers.¹⁰⁰

Gene expression analysis showed no differential expression in SOX2 in MWCNT-exposed cells, while protein expression was mildly overexpressed (Fig. 5A,B). Based on the weak SOX2 expression and strong overexpression of SNAI1 (Fig. 5A–D), we investigated the role of SNAI1 in stem cell-associated signaling pathways and their potential role in cancer signaling. Both ‘cancer of cells’ and ‘morphology of embryonic tissue’ functions possessed overexpressed SNAI1 as a gene signaling hub in both functions’ signaling networks (Fig. 6C). Only those genes associated with SNAI1 were kept in the cancer signaling network. Overexpressed MYC, FOXO1, and SNAI1 served as key hub genes in the morphology of embryonic tissue network, while SNAI1 and downregulated CXCL8 and CCL2 served as key hub genes in the cancer of cells network. To investigate the potential crosstalk between these two networks, both networks were plotted with all known associations in IPA database. Along with the overexpressed SNAI1, downregulated FOXA2, FGFR2, COL2A1, and overexpressed RNF2 were shared between the two networks.

Overexpressed genes in the embryonic tissue network were primarily involved with tumor promotion and suppression, cholesterol receptors, and differentiation, while downregulated genes were primarily associated with fibroblast and platelet-derived growth factor signaling, cytoskeleton dynamics, and differentiation. Of note, overexpressed transcriptional regulators NR4A3, FOXO1, and TCF7L1 indicate several different differentiation pathways with known roles in cancers^{101, 102} that may contribute to the SNAI1-associated neoplastic phenotype in MWCNT-exposed cells. Previously, we identified MYC, FOXO1, and SNAI1 as possessing a proto-oncogene signaling function in MWCNT-exposed human SAECs,³⁰ which is supported by their known roles in lung adenocarcinoma and CSC regulation.^103–105 Current results suggest that embryonic development signaling of the overexpressed MYC and FOXO1 and downregulated FGFR2 primarily suppress FOXA2, resulting in a loss of its inhibitory function on SNAI1. FOXA2, a transcriptional regulator involving lung and embryonic development, responds to IL-6 induced regulation of fibrinogen synthesis and lung tumor suppression.¹⁰⁶ Downregulated CDK5 and overexpressed NR4A3 and SCARB1 appeared to regulate the overexpressed FOXO1. CDK5 plays roles in cytoskeleton control, cell migration and apoptosis, whereas NR4A3 plays several roles in cancer signaling including transcriptional dysregulation and cell survival.¹⁰²

The overexpression of SNAI1 appeared to associate with signaling that would modulate cancer signaling in MWCNT-exposed cells and was associated with upstream EMT and DNA damage response signaling. Overexpressed ROR1 and under-expressed FOXA2 and DIXDC1, a Wnt pathway modulator, contributed to SNAI1 overexpression. ROR1 participates in Wnt and NFκB signaling pathways, contributes to EMT, and promotes several cancers.¹⁰⁷ Downregulation of ATM, COL1A1 and PLAU, associated with stress response and extracellular matrix processes, appeared to promote downregulation of inflammatory cytokines and mediators including CCL2, CXCL8, TNFAIP3, and POU5F1. Along with the under-expressed ATM, evidence of DNA damage response impacting SNAI1 was observed with overexpressed PRKDC and NDRG1, kinases that associate with DSB repair, mitotic spindle checkpoint, and metastasis disruption.^{108, 109} Collectively, these signaling pathways imply that long-term CNM exposure elicited significant DNA damage response¹⁴ and alterations in embryonic signaling, potentially via Wnt and β-catenin pathways that promoted MYC and SNAI1 proto-oncogene signaling that is observed in exposed in vitro and in vivo models.^110–112

In summary, we provided evidence that long-term low-dose exposure to CNMs and ASB caused substantial DNA damage and p53 dysregulation in human SAECs. The exposed lung cells exhibited neoplastic and CSC-like properties, as indicated by anchorage-independent colony formation, spheroid formation, anoikis resistance, and CSC markers expression. High aspect ratio materials including SWCNT, MWCNT and ASB exhibited strong neoplastic and CSC-like properties as compared to low aspect ratio UFCB particles. This result is in good agreement with previous toxicity studies showing enhanced cytotoxicity of high aspect ratio nanomaterials such as nanowires and nanotubes vs. low aspect ratio nanomaterials such as round and ringed particles.¹¹³ Likewise, animal studies showed a strong pulmonary inflammatory response to high aspect ratio MWCNT as compared to low aspect ratio MWCNT, which induced no inflammatory response.¹¹⁴ The impact of aspect ratio on toxicity was also observed in other types of nanoparticles. For example, rod-type aluminum oxide nanoparticles were found to induce a stronger immune response than their spherical counterpart.¹¹⁵ Despite a vast difference in chemical composition, both ABS (silica-based) and CNTs (carbon-based) induced a similar neoplastic and CSC-like transformation, whereas the low aspect ratio UFCB induced weak effect. Since both UFCB and CNTs used in this study contain >99% carbon and both have minimal metal impurities (i.e. <1%), these results suggest that aspect ratio of nanomaterials is a key determinant of their pathogenicity. Metal impurity and rigidity of nanoparticles may also play a role since ASB, which has a high metal impurity content (38%) and is highly rigid, induced a stronger effect on CSC-like transformation than MWCNT despite having a lower aspect ratio as shown in Table 1. Indeed, previous studies have suggested rigidity as a key factor underlying frustrated phagocytosis caused by rigid fiber-like particles, leading to an elevated and sustained release of pro-inflammatory cytokines and reactive oxygen species that contribute to disease pathogenesis.^{116, 117} Our study also unveiled potential molecular mechanisms of CNM- induced neoplastic and CSC-like transformation. It identified self-renewal and EMT-related transcription factors, notably SOX2 and SNAI1 that may serve as key drivers of oncogenic transformation induced by the CNMs.