All tested particles were previously characterised by other studies and their physiochemical properties are summarised in Table I. Briefly, SWCNTs, synthesised using high-pressure carbon monoxide disproportionate process (HiPCO), were obtained from carbon nanotechnology (CNI, Houston, TX). SWCNTs were acid treated to remove a significant portion of metal catalyst contaminants. UFCB, SWCNT and MWCNT elemental analysis was performed using nitric acid dissolution followed by inductively coupled plasma-atomic emission spectroscopy while particle surface area was determined using Brunauer Emmet Teller nitrogen absorption-desorption technique at −196°C using a SA3100 Surface Area and Pore Size Analyzer (Beckman Coulter, Fullerton, CA, USA). Purified SWCNT contained less than 1% w/w of contaminants (Table I; Wang et al. 2011b). MWCNTs were provided by Mitsui & Company (MWNT-7, lot #05072001K28) containing 0.41% w/w metal impurity (Porter et al. 2010). Both SWCNTs and MWCNTs possessed ≤0.32% Fe and ≤0.41% Na ion impurities. Crocidolite asbestos (CAS# 12001-28-4) was originally obtained from the Kalahari Desert in South Africa by the National Institute of Environmental Health Sciences (Research Triangle Park, NC, USA). Crocidolite fibres had a mean length of 10 μm, mean diameter of 0.21 μm and surface area of 9.8 m²/g (Msiskaetal. 2010). Ultrafine carbon black (Elftex 12) was obtained from Cabot (Edison, NJ, USA) and contained <1% w/w metal impurities. All particles possessed surface areas between 9.8 and 43 m²/g except for SWCNT which possessed 10–100-fold greater surface area (400–1040 m²/g).

Particles were suspended in phosphate-buffered saline (PBS) to acquire stock solutions of 0.1 mg/ml. Next, a natural lung surfactant, Survanta (Abbott Laboratories, Columbus, OH, USA), was added to UFCB, SWCNT and MWCNT stock solutions to aid in their dispersion. The Survanta concentration in particle stock solutions was 150 μg/ml. All particles were dispersed 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 sec followed by dilution with cell culture medium to exposure dose. Final Survanta concentration in exposure medium was 0.15 μg/ml. Our Survanta dispersion method was previously optimised for improved dispersion of CNT and UFCB in cell culture media to mimic structure size distribution obtained by aerosolisation of dry CNT (Wang et al. 2010). Survanta exposure alone and dispersed UFCB does not cause acute cell proliferation or cytotoxicity; however, dispersed CNTs exhibited significant increases in both proliferation and cytotoxicity compared to non-dispersed agglomerated CNTs (Wang et al. 2010; Mishra et al. 2012).

Dimensions and imaging of dispersed particles in medium were determined using field emission scanning electron microscopy methods as previously described (Wang et al. 2010; Mishra et al. 2012). Briefly, dispersed particles were suspended in medium (0.02 μg/cm²) and filtered through polycarbonate filter to collect dispersed particles. Dried samples were mounted, gold/palladium sputter coated and imaged at 400 and 30,000 magnification. Length and width measurements of >300 particles from three independent replicates were performed. Survanta-dispersed SWCNT structure exhibited mean count dimensions of 1.08 μm × 0.27 μm (Table I; Wang et al. 2010), while MWCNT structure exhibited 5.1 μm × 0.078 μm mean dimensions (Figure 1; Mishra et al. 2012). Since asbestos readily disperses in PBS with light sonication, no Survanta addition was performed. The same concentration of Survanta alone (0.15 μg/ml) and PBS alone were used as dispersant or saline (no treatment) controls.

Primary human SAECs immortalised with hTERT were kindly provided by Dr. Tom Hei (Piao et al. 2005). SAECs were chosen as an ideal model since aspiration of CNT results in substantial deposition in terminal bronchioles and alveolar region of exposed mice (Mercer et al. 2010). SAECs were cultured in SABM medium supplemented with Clonetics SAGM SingleQuots (Lonza, Walkersville, MD, USA) which contained 0.4% v/v 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 cultured at 37° C in a humidified 5% CO₂ incubator during and following exposure.

SAECs were continuously exposed for 6 months to a subtoxic dose (0.02 μg/cm², equivalent to 0.1 μg/ml) of dispersed UFCB, SWCNT, MWCNT or crocidolite asbestos in a six-well culture plate in triplicate. Each replicate was exposed and independently assayed throughout the study. This relatively low concentration was chosen due to its relevance to in vivo SWCNT and MWCNT exposure dose of 10 μg/mouse previously reported (Stone et al. 1992; Mercer et al. 2008; Porter et al. 2010). Every 3–4 days exposure media were removed, cells were triplicate washed with PBS and resupplied with media dosed with dispersed test particles. Cells were passaged weekly at pre-confluent densities using a solution containing 0.05% w/v trypsin and 0.5 mM EDTA (Invitrogen, Carlsbad, CA, USA). Parallel cultures grown in the same background concentration of dispersant or saline medium provided passage-matched controls and were designated as DISP and SAL cells, respectively. Following long-term exposure, particle-exposed cells were referred to as D-UFCB, D-SWCNT, D-MWCNT and ASB cells, respectively.

Suspended parental SAECs (2 × 10⁴ cells) were seeded onto sterile glass coverslips into a 24-well plate overnight in triplicate. Cells were exposed to each dispersed particle (0.02 μg/cm²) or unexposed medium for 24–48 h to evaluate cell uptake and co-localisation with major cell organelles. Following exposure, cells were fixed, stained and examined for cell/particle co-localisation under an Olympus BX-41 scope using dark-field enhanced microscopy (CytoViva, Auburn, AL, USA) using previously described methods (Mercer et al. 2011). Briefly, cells were PBS washed, fixed in 10% neutral buffered formalin, PBS washed again, stained with 0.1% w/v toluidine blue and rinsed thrice in distilled water. Stained cells were dehydrated in ethanol, cleared in xylene and mounted to glass slides with Permount. Since ENMs possess several attributes ideal for producing high quantities of scattered light (Mercer et al. 2011), enhanced dark-field microscopy images ENM scattered light compared to low amounts of scattered light from surrounding tissue.

All cell morphology, cancer hallmark behaviour and genome expression signature assessments were performed between 1 and 15 passages post-exposure (Table II). Immediately following 6-month exposure, suspended cells were seeded in duplicate at 1 × 10⁵ cells/well in a 6-well plate and cultured up to 72 h to visually assess proliferation. Cells were then digitally photographed using reverse phase contrast on an Olympus IX70 inverted microscope equipped with a Retiga 2000R digital camera (QImaging, Surrey, BC).

Following exposure, colony-forming efficiency (CFE) was performed to assess cell proliferation using methods for MWCNT exposure (Ponti et al. 2012). Briefly, 300 cells were plated in triplicate to 60-mm plates (28.2 cm² growth area) and cultured for 7 days with one medium change occurring on Day 4. Surviving cells were washed in PBS, fixed in 4% paraformaldehyde, stained with 1% w/v crystal violet, rinsed twice with MilliQ water and allowed to dry. Plates with colonies were digitally photographed. Five independent experimental assays were performed. Colony counts were visually scored under an inverted scope and mean CFE calculated, where CFE (%) = (average treated cells/average of DISP cells × 100).

Starting at second passage post-6-month exposure, passage-matched treated and control cells were plated in sextuplicate in 100 μl normal growth medium to 96-well plate at 2 × 10³ cells/well and incubated for 24–48 h. Next, 10 μl of WST-1 reagent was added to each well (Mishra et al. 2012) and spectrophotometrically assayed at 450 nm with SpectraMaxPlus 384 (Molecular Devices, Sunnyvale, CA, USA). To validate that increased mitochondrial activity correlated with live cell number, trypan exclusion assay was performed by seeding 1 × 10⁵ cells in a 6-well plate in duplicate and incubating for 24 and 48 h. Next, cells were collected via trypsinisation, stained with trypan blue and counted using a Countess cell counter (Invitrogen). A minimum of three independent experiments were performed for both WST-1 and exclusion assays. Exclusion assay data were pooled across experiments prior to statistical analysis.

A morphological transformation assay for foci was conducted based on previously described methods for MWCNT exposure (Ponti et al. 2012). Briefly, all six exposed cell types were seeded at 3 × 10⁴ cells/dish to 60-mm plates in triplicate. Cells were cultured for 14 days with a PBS wash and fresh growth medium supplied every 3 days until a uniform monolayer was established. Following medium change at Day 14, cell medium was not changed for 1 week. On Day 21, cells were PBS washed, fixed in 10% formalin and stained with 1% crystal violet to visualise foci. The assay was repeated three times and Type III foci were scored from each treatment under an inverted microscope using cell transformation criteria described by Sasaki et al. 2012. Transformation frequency (TF) was calculated as TF = [foci count/(CFE (%) × seeding efficiency × cells seeded × replicate plates)].

All SAEC lines were assayed for anchorage-independent cell proliferation using a previously described soft agar assay method (Azad et al. 2010). Briefly, 0.5% w/v agar medium was obtained by mixing Difco agar, 15% v/v foetal bovine serum (FBS) and 1% gentamicin with 2× concentrate MEM medium (Lonza) at 44° C. Appropriate amounts of each growth factor from the SAGM SingleQuots kit (Lonza) that equalled their respective normal growth medium concentrations were added to the warm agar since all cell lines did not survive in preliminary soft agar assay trials. Next, all SAEC lines were suspended in 0.33% agar at 1 × 10⁴ cells/well and slowly layered onto precast agar in triplicate into six-well plates. After 14 days incubation, colonies were examined and digitally photographed under light microscopy. Only colonies with >50 μm diameter were scored as positive in five replicate photographs per well. A minimum of three independent soft agar assays were performed.

Following exposure all exposed cells were subjected to chemotaxis Transwell invasion and migration assays (Azad et al. 2010). Invasion Matrigel Transwell inserts (8 μm pores; BD Biosciences) contained a thin layer of Matrigel to simulate extracellular matrix while migration control inserts (8 μm pores; BD Biosciences) did not. Briefly, cells were suspended in basal SABM medium without SingleQuot growth factors and seeded at 1.5 × 10⁴ and 3 × 10⁴ cells/insert into the upper chamber of preconditioned Transwell inserts for invasion and control inserts for migration, respectively. Duplicate inserts each containing 250 μl of cell suspension were immediately placed into individual wells containing 750 μl of normal SAEC growth medium in a 24-well plate. Invasion and migration assays were cultured for 48 or 24 h, respectively, to assess cell movement from upper to lower chambers. Next, a sterile cotton swab was used to remove all non-mobile cells from the inside of each insert. Adherent cells on underside of membrane of each insert were fixed, stained with Diff-Qik solutions according to manufacturer's instructions, rinsed in water and allowed to dry. Lastly, inserts were visualised in five random fields of view and scored under an inverted microscope. Three independent experimental runs were carried out for both assays.

Primary human vascular endothelial cells (HUVECs) were used to assay angiogenic potential of conditioned media from transformed cells using a two-dimensional angiogenic assay previously described (Azad et al. 2012). Briefly, low-passage parental SAECs and all six exposed cell types were plated at 5 × 10⁵ cells/plate into 60-mm plates and held overnight. Next, cells were washed in sterile PBS and allowed to culture in 1 ml of fresh base SABM medium for 24 h. Conditioned medium was then removed from each cell line and frozen at −80° C until needed. HUVEC cells acquired from ATCC were cultured in MCDB 131 medium supplemented with 25% v/v heat-inactivated FBS, 0.05% bovine brain extract, 0.25% w/v endothelial growth factor, 0.1% heparin, 1% L-glutamine and 1% gentamicin. Subconfluent HUVECs between second and seventh passage were suspended in a 1:1 ratio mix of SABM conditioned media and HUVEC reduced medium (2.5% dialysed FBS). HUVEC suspensions (4.5 × 10⁴ cells/well) were seeded in duplicate onto reduced growth factor Matrigel-coated wells in 48-well plates. After 18 h, five replicate photos per well were digitally photographed and the number of endothelial cell tube nodes were scored. Three independent experimental runs were performed.

To determine mechanisms of action driving CNT-driven neoplastic-like transformation, collected mRNA from all six exposed cell types at second passage post-exposure was subjected to whole-genome expression microarray analysis following MIAME guidelines. All exposed and passage control SAECs were seeded in triplicate at 5 × 10⁵ cells to 60-mm plates in normal growth medium and incubated in normal growth conditions for 1–2 d. Then, sub-confluent plates were lysed with ice-cold TRIzol Reagent (Invitrogen) following manufacturer's instructions to collect mRNA. Lysates were then immediately frozen at −80° C for 24 h and shipped on dry ice to ArrayStar (Rockville, MD, USA) for mRNA processing, microarray hybridisation and probe expression normalisation. On arrival, mRNA was extracted following manufacturer's protocol and evaluated for RNA quality and quantity using a Bioanalyzer 2100 (Agilent) and Nanodrop ND-1000. mRNA was then reverse-transcribed with Superscript double-stranded-cDNA synthesis kit (Invitrogen). cDNA samples were then amplified, Cy3-labeled with NimbleGen one-colour DNA labelling kit and hybridised to NimbleGen whole human genome 12 × 135k microarrays in NimbleGen Hybridisation System. Following microarray plate image capture with Axon GenePix 4000B (Molecular Devices), images were uploaded into NimbleScan (v2.5), and Cy3 label intensities determined. Intensities were quantile normalised using Robust Multichip Average method included in NimbleScan. Only genes expressing ≥50.0 normalised intensity in all 18 samples were subjected to further analysis. All normalised gene expression data were deposited to NCBI Gene Expression Omnibus archive and are accessible via accession number (GenBank ID: GSE41178).

Differentially expressed genes (DEGs) for each treatment were determined in Agilent GeneSpring GX (v11.5.1) by comparing each gene's normalised intensity to DISP SAEC control intensity using a t-test (α = 0.05) and a ±2.5-fold change screen. DISP SAECs served as an appropriate control since D-SWCNT, D-MWCNT and D-UFCB SAEC cells were exposed to lung surfactant and allowed for identification of particle-specific effects. Fold change and p-values for each DEG were saved in a tab-delimited text file for upload into Ingenuity pathway analysis (IPA) ver. 9.0 (Ingenuity Systems, Redwood City, CA, USA).

To further delineate whole-genome expression signature differences between treatments using cluster analysis techniques, normalised unscreened fold change data in a tab-delimited file were uploaded into MultiExperimental Viewer (Saeed et al. 2006). Due to the size of the data set, the lowest 20% of the least variable genes across all treatments were removed from the data set. Next, a one-way ANOVA (α = 0.05) was used to identify genes whose expression differed among treatments. Both filters resulted in 20,497 probes for cluster analyses. Unsupervised hierarchical cluster analysis was performed on screened genes using Pearson's correlation and average linkage to develop a dendrogram and fold-change heat map. To validate the cluster analysis, the same screened data set was subjected to principal component analysis (PCA) to evaluate inter- and intra-sample clustering. The top three principal components were used to map all 18 samples in a three-dimensional space.

DEG data from each treatment were uploaded into IPA for Core Analysis (Ingenuity Systems, Redwood City, CA, USA) to determine major alterations in cell functions, gene signalling networks (GSNs) and key genome alterations promoting neoplastic transformation in CNT-exposed cells. As previously described (Stueckle et al. 2012), negative logarithm p-values from IPA were used to rank cellular/molecular functions, disease functions, canonical pathways and signalling networks to identify potential genes and/or GSNs that promoted neoplastic transformation in D-SWCNT, D-MWCNT and ASB SAECs. Gene IDs, t-test p-values and fold expression data in tab-delimited text files were uploaded in IPA. Using IPA-generated negative logarithm p-values, score rankings were used to rank cellular/molecular functions, disease functions, canonical pathways and signalling networks to identify genes and novel signalling pathways potentially promoting observed neoplastic transformation. P values and GSN scores reflected likelihood tests of a gene occurring in a given pathway versus other pathways based on pure chance. Z scores (≥±2) were used to make predictions of activation/inhibition of cellular functions, respectively.

To determine potential novel signalling mechanisms and identify differences between CNT- and asbestos-induced tumour-promoting signalling, IPA-generated p values and GSN scores were determined, which reflected likelihood tests of a gene occurring in a given pathway versus other pathways based on pure chance. Initial evaluation of IPA functions and networks identified cancer-related disease and gene signalling functions in CNT-exposed cells. Therefore, ranked GSNs were mapped and cancer-related genes and signalling pathways were identified. To further depict cancer-related signalling, CNT cancer GSNs were created by filtering the DEG data set for those genes with a cancer-related role. Lastly, cancer cells typically exhibit increased proliferation, tumour formation, migration, invasion and anti-apoptosis abilities. Cell behaviour GSNs (i.e. pro-cancer GSN) were developed for each of these phenotypes to assist in identifying signalling patterns. Genes passed the filter if they promoted the behaviour and were up-regulated or antagonised the behaviour and were down-regulated.

To confirm microarray gene expression data, a subset of key genes of interest was selected to validate their mRNA expression. 5 × 10⁵ cells from each treatment were seeded in 60-mm² dishes and allowed to proliferate for 1–2 days. mRNA from sub-confluent cells was collected via TRIzol procedures (described above) and stored for 48 h at −80°C prior to rtPCR analysis. Intron-spanning primers and probes were designed using ProbeFinder 2.45 (Roche) and obtained from Operon (Table S1). RNA sample quality was assessed using NanoDrop 1000 (Thermo Scientific), reverse transcribed, and amplified using 2720 Thermo Cycler (Applied Biosystems) in triplicate. Quantitative rtPCR was conducted on an ABI 7500. Each thermocycling reaction used 25 μl total volume containing 7.25 μl cDNA, 2.5 μl primers and 12.5 μl Roche Taqman Master Mix. Relative expression levels to GAPDH were determined using 2^−ΔΔct and compared to microarray values using a t-test.

Following 6-month exposure, cells were incubated in 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) at 4°C for 20 min. The lysate was collected and determined for protein content using the Bradford method (Bio-Rad Laboratories, Hercules, CA, USA). Proteins (40 μg) were resolved under denaturing conditions by 10% sodium dodecyl sulphate-polyacrylamide gel electrophoresis and transferred onto nitrocellulose membranes (Bio-Rad). The transferred membranes were blocked for 1 h in 5% nonfat dry milk in TBST (25 mM Tris–HCl, pH 7.4,125 mM NaCl, 0.05% Tween 20) and incubated with the appropriate primary antibodies at 4°C overnight. p53, IL-1B and IL8 anti-bodies were purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA) while all other antibodies were acquired from Cell Signaling Technology (Danvers, MA, USA). Membranes were washed twice with TBST for 10 min and incubated with horseradish peroxidase-coupled isotype-specific secondary antibodies (Santa Cruz) for 1 h at room temperature. The immune complexes were detected by enhanced chemiluminescence detection system (Amersham Biosciences, Piscataway, NJ, USA) and quantified using analyst/PC densitometry software (Bio-Rad Laboratories, Hercules, CA, USA).

DISP cells served as negative passage control for both cell behaviour and microarray statistical comparisons among treatment groups. All cell behaviour assay data were pooled from three independent exposure replicates and analysed using one- or two-way ANOVA (α = 0.05) to determine differences between chronic exposure treatment groups. Post hoc Tukey–Kramer honestly significant difference tests were used to identify those treatments that were different from each other. Microarray and rtPCR mRNA expression values for each gene were compared using a two-way Student's t-test. All statistical analyses were performed in JMP 9.0 (SAS Institute).

Scanning electron microscopy displayed dispersed particles before exposure (Figure 1A) while hyperspectral dark-field visualisation of 24-h-exposed SAECs showed that dispersed UFCB co-localised in both the cytoplasm and nucleus (Figure 1B). Dispersed SWCNT, MWCNT and asbestos exhibited fibre-like morphologies either co-localised in the cytoplasm of cells or puncturing the cellular or nuclear membranes. UFCB, SWCNT and MWCNT also formed large micron-sized aggregates that appeared to make contact with the cell membrane. Adjustment of the focal plane to acquire Z-stack images (data not shown) suggested that these large micron-sized particles were potentially lying on top of cells and not within the cytoplasm. At equal cell seeding density, more dispersed SWCNT and MWCNT particles were observed co-localised with SAECs than asbestos fibres.

After 24 weeks of exposure, D-SWCNT and D-MWCNT cells visually exhibited greater proliferation over 48 h (Figure 2A). Both D-SWCNT and D-MWCNT exhibited significantly greater CFE (Figure 2B) and morphological transformation evidenced by Type III foci than DISP control cells (Figure 2C). Parental SAECs possessed significantly lower CFE than all other cells. Type III foci exhibited deep basophilic staining, multilayered mounding with different cell orientations and invasive growth at colony edges. Both D-CNT SAECs demonstrated cells with numerous intracellular vesicles compared to DISP and SAL control cells while ASB periodically displayed this morphological characteristic. Attempts to identify these vesicles by staining lipid droplet, endoplasmic reticulum, Golgi apparatus, lysosome and peroxisome were all negative (Figure S1). In addition, all six passaged cell lines possessed pleomorphic giant cells possibly due to long-term passage. In summary, increased CFE, morphological transformation and visual observations of altered morphology suggested a moderate CNT-induced lung epithelial morphological transformation.

Cancer hallmark transformation phenotypes of the particle-exposed cells were determined by several well-established methods and compared to passaged DISP control cells. D-SWCNT, D-MWCNT and ASB cells exhibited significantly greater cell proliferation at 48 h compared to SAL, DISP and D-UFCB cells (Figure 3A-1). Similarly, D-SWCNT, D-MWCNT and ASB cells showed a significant increase in live cell number over 48 h compared to SAL and DISP controls. D-UFCB cells displayed a significantly lower number of cells compared to controls indicating a reduced ability to attach and proliferate following chronic UFCB exposure (Figure 3A-2). A 2-week proliferation analysis during subchronic exposure showed that D-UFCB cells recovered by 72 h post-seeding, but exhibited lower viable cell counts at 7 days post-seeding than all other treatments (Figure S6).

Next, anchorage-independent cell growth was determined by assessing the number of isolated colonies on soft agar to confirm SWCNT- and MWCNT-induced neoplastic-like transformation compared to asbestos. D-SWCNT, D-MWCNT and ASB SAECs exhibited significant increases in number of colonies formed compared to control SAECs (Figure 3B). D-SWCNT and D-MWNCT SAECs displayed a 2.2-fold increase compared to DISP control cells, while ASB-SAECs had a significant 1.3-fold increase in formed colonies over SAL control cells.

To test for enhanced cell motility, all subchronic SAEC lines were assessed for invasion and migration ability. Both D-SWCNT and D-MWCNT cells demonstrated a significant 4.7-fold increase in 48-h cell invasion and a 2–2.5-fold increase in 24-h cell migration as compared to the control cells (Figure 3C and Figure S2). ASB SAECs showed a significant increase in invasion ability but no significant change in migration ability compared to DISP cells.

Lastly, we assessed angiogenic potential of D-UFCB, D-SWCNT, D-MWCNT and ASB SAECs compared to parental, SAL and DISP control SAECs. Conditioned media from D-SWCNT, D-MWCNT, ASB and SAL SAECs caused a significant increase in tube formation while UFCB and parental SAECs exhibited a significant decrease compared to DISP SAECs (Figure 3D). Increased angiogenic potential in SAL cells compared to DISP control cells was possibly due to increased mRNA expression of HIF1α, VEGF and p65NF-κB (microarray data described below) which was not observed for CNT-exposed cells. In summary, the increased expression of several cancer hallmarks (i.e. cell proliferation, soft agar colony formation, migration, invasion and angiogenesis) in both D-SWCNT- and D-MWCNT-exposed cells suggests that subchronic CNT exposure elicits a more aggressive epithelial lung cell neoplastic-like transformation effect than asbestos, suggesting increased risk for CNT-induced carcinogenesis.

To assess potential neoplastic-like transformation mechanisms and to compare genome expression signatures of SAECs subchronically exposed to CNTs vs. asbestos, collected mRNA from all exposed and control SAECs were subjected to whole-genome expression microarray profiling. Following mRNA hybridisation and DEG determination, hierarchical clustering (HC) heat map expression signature and PCA showed that D-SWCNT- and D-MWCNT-transformed SAECs possessed the most similar toxicogenomic signatures, which were clearly distinct from ASB cells (Figures 4A and B). In addition, DISP and D-UFCB SAECs showed similar expression signatures, which were loosely clustered with ASB SAECs. Conversely, SAL SAECs possessed a unique expression signature, which differed from all other cell lines. PCA explained 75.8% of the variance in the first three components and confirmed the HC analysis of treated SAECs.

To identify signalling pathways associated with CNT-induced potential neoplastic-like transformation compared to asbestos, all DEG data were subjected to IPA Core Analysis. Top-ranked cellular/molecular functions, disease functions and signalling networks were used to identify GSNs contributing to a potential neoplastic-like transformation phenotype. Both D-SWCNT and D-MWCNT SAECs displayed large alterations in DEGs associated with cell proliferation, movement, cell death, development and cell–cell signalling (Figure 5A). Cell development signalling associated with connective tissue development was predicted as activated in D-SWCNT SAECs (*, Z ≥ 2). Further analysis of the cell death function revealed that a large majority of the altered DEGs were associated with apoptosis. In addition, a majority of cell movement and development signalling in both D-SWCNT and D-MWCNT was associated with decreased gene expression of inflammatory cytokines and chemokines involved with leukocyte recruitment (†; Z ≤ −2). ASB SAECs showed similar pattern in cell-death- and proliferation-related DEGs. Upon further analysis of those functions predicted as activated in ASB (*), a majority of these genes were associated with cell function (phagocytosis), inflammatory signalling and leukocyte cell functions. SAL cells exhibited changes in cell proliferation, death and development with predicted activation of endothelial cell development (*). Lastly, D-UFCB cells possessed large changes in cell proliferation, cell assembly, cell function and cell death signalling with a majority of these genes down-regulated. Cell invasion/migration, cell proliferation, neoplasia and cell attachment signalling were predicted as inhibited (†) while cell death/senescence and homing of phagocytes/neutrophils were predicted as activated (*, Figure 5A).

In evaluating top-ranked molecular functions, D-SWCNT cells exhibited significant alterations in gene expression, post-translational modification and lipid metabolism (Figure 5B). D-MWCNT showed the most significant alterations to gene expression and predicted activation of lipid metabolism/transport (*). Conversely, ASB cells displayed a 2–5-fold decrease in DEGs associated with gene expression and a comparable number of post-translational modification, small molecule metabolism and transport DEGs to D-MWCNT. SAL and D-UFCB cells exhibited the largest number of DEGs in gene expression. Only D-UFCB and ASB cells possessed a substantial number of DEGs associated with DNA recombination and repair. ASB SAEC post-translational modification, amino acid metabolism, antigen presentation and molecular transportation were all predicted as activated (*). In summary, HAR fibres showed similar alterations in cellular functions, but exhibited strikingly different molecular functions, which were primarily driven by differences in lipid metabolism and inflammatory signalling.

To further uncover altered signalling pathways, top-ranked disease functions and canonical pathways were investigated for promotion potential of neoplastic-like phenotype. Cancer was the top-ranked disease for all five cell lines compared to DISP control cells (Table S2); however, genetic disorder was pronounced for both D-SWCNT and D-MWCNT genome expression signatures. All top-ranked organ system diseases were specific cancer types for all exposure treatments except ASB cells. Inflammatory response, arthritis disease and immunological disease were highly ranked in ASB cells, but were absent from top-ranked CNT disease functions. IPA predicted cancer inhibition (†) in D-UFCB cells while it predicted activation (*) of inflammatory response/disease in ASB cells. Investigation of top-ranked canonical pathways revealed that D-SWCNT SAECs exhibited changes in altered apoptotic, decreased inflammatory (NFKB2, Ikk) and cancer-related signalling (Table III). For both D-SWCNT and D-MWCNT cells, CASP2, CASP3 and CASP8 were over-expressed, while pro-apoptotic BID, BAD and HTRA2 were under-expressed, thus suggesting altered apoptosis signalling. Down-regulated p100/p52 NF-κB (NFKB2) was a common player in many of the top-ranked canonical pathways for both D-SWCNT and D-MWCNT. D-MWCNT showed similar rankings; however, down-regulation of many genes in the complement system (BF, C3b, C4b; over-expressed CD55) and chemokines in IL-17 signaling (CXCL1–3, IL8, CCL2) took precedence. Conversely, ASBSAECs exhibited altered fibrosis, TREM and clathrin-mediated endocytosis signalling. Increased NF-κB inflammatory fibrosis signalling was evident resulting in increased cytokine expression (IL8, CCL2). Over-expressed ITGB2, SRC and PI3K with under-expressed clathrins indicated an integrin-mediated uptake of asbestos fibres. Over-expressed interleukins, tolllike receptors and phospholipase C (e.g. IL1B, TLR2, PLCγ2) indicated activation of NF-κB inflammatory signalling. Surprisingly, several complement proteins down-regulated in CNT-exposed cells were also down-regulated in ASB cells. SAL cells possessed over-expressed fibrosis signalling (COL1A1, PDGFRα/β, FGFR, VEGF, MMP2) and inflammation (IL6, IL8, RELA, MCP-1) that dominated a majority of the top-ranked pathways. Conversely, D-UFCB demonstrated down-regulation of fibrosis (TGFBR1, SMAD2, IL6, VEGF, CSF1, CCL2 and MMP2), WNT pathway and polo-like kinase signaling (APC, CCNB2, PLK, SLK) indicating a reduced ability to proliferate and contribute to a particle-induced fibrotic response. In summary, both D-SWCNT and D-MWCNT SAECs exhibited altered apoptosis signalling, cancer-related signalling and decreased inflammatory signalling, which differed from ASB and SAL SAEC pro-inflammatory signalling.

Evaluation of top-ranked GSNs revealed several gene hubs known to exhibit pro-cancer or tumour-associated signalling in both D-SWCNT and D-MWCNT (Table IV). Ap-1 (FOS)-, MYC-, NFKB2-, PPARG- and INHBA-centred GSNs were common in both D-SWCNT and D-MWCNT SAECs promoting altered gene expression, lipid metabolism, cell proliferation and cell movement. Although more associated with D-MWCNT, both D-CNT SAECs displayed GSNs with altered actin, tubulin and collagen expression, suggesting a remodelling of intra- and extracellular structure and function. In addition, several GSNs centred on altered histone expression (e.g. histone deacetylases, histones 1 and 3), suggesting either epigenetic modifications and/or alterations to chromosome structure and integrity. ASB SAECs exhibited a drastically different ranked GSN profile compared to both D-CNT SAECs. ASB cells exhibited over-expressed SRC, SPI1, TREM1 and NF-κB signalling with decreased hCG and PRKACB GSN hubs associated with cell assembly/maintenance, molecular transport, inflammatory response and cell–cell signalling. SAL SAECs exhibited a few similar expression profiles of several genes (Table IV; e.g. over-expressed FOS); however, over-expression of RELA and TP53 in SAL SAECs appeared to dominate several top-ranked SAL GSNs. Top-ranked GSNs in D-UFCB highlighted decreased cell development (GAS7), cell maintenance, lipid metabolism and cell cycle (TP53). With a lack of neoplastic-like transformation behaviour, the SAL SAECs appeared to undergo a moderate in vitro non-neoplastic transformation associated with increased p53 and inflammatory NF-κB signalling following 6 months of continuous passage.

Based on aggressive neoplastic-like transformation phenotype and identification of several signalling pathways associated with carcinogenesis, cancer-associated GSNs were constructed in IPA to further identify potential complex signalling pathways in both D-SWCNT and D-MWCNT compared to ASB SAECs. Large complex pro-cancer networks for both D-SWNCT and D-MWCNT were characterised by proliferation, tissue development, cell movement and suppressed immune signalling, while ASB cell signalling was primarily driven by pro-inflammatory signalling (Figures S3–S5). By applying a strict filter to show only gene expression solely promoting cancer, D-SWCNT pro-cancer GSN revealed over-expressed NRAS, MYC, FOS, CASP8, COL18A1, CDKN2A and under-expressed CSF2, CCL2 hub genes associated with cell proliferation and tissue morphology (Figure 6A). D-MWCNT pro-cancer GSN revealed a smaller network centred on over-expressed MYC associated with cell proliferation and cell movement (Figure 6B). Conversely, the ASB SAEC pro-cancer GSN possessed over-expressed IL1B, SPI1, CASP8, CD44, several inflammatory chemokines, including CCL2, IFNA21, and BCL2L11, with under-expressed CXCL2, FN1, MMP2 and CSF1 (Figure 6C). Cell proliferation, cancer and tissue morphology behaviour were associated with the ASB pro-cancer GSN. In summary, subchronic D-SWCNT and D-MWCNT exposure resulted in cancer-promoting signalling profile with known oncogenes in SAECs which differed from a pro-inflammatory signaling prolife in asbestos-exposed SAECs.

Following IPA analysis and identification of hub genes associated with potential tumourigenesis, a subset of key signalling molecules of interest for D-SWCNT, D-MWCNT and ASB cells were chosen to further investigate whether their protein expression matched the microarray and their potential as key players contributing to lung epithelial cell neoplastic-like transformation and potential lung carcinogenesis. rtPCR validation (Figure 7A) confirmed a majority of microarray expression values of key genes. Other significantly different rtPCR values were noted but were either over- or under-expressed in large magnitude, which matched the expression trends in the microarray.

Western blot analysis showed that D-SWCNT, D-MWCNT and ASB possessed over-expressed cFOS, c-Myc and PPARγ levels (Figure 7B), while the tumour suppressors Inhibin α and p53 were under-expressed (Figure 7C). We probed for p-p53 to evaluate if the p53 decrease was a function of loss of phosphorylation at serine 15 previously shown for SWCNT-transformed lung epithelial cells (Wang et al. 2011b). P-p53^Ser15 levels were increased above SAL, DISP and D-UFCB levels, suggesting partial protection against p53 proteasomal degradation in D-SWNCT, D-MWCNT and ASB SAECs. Given caspase 8′s prominence in the GSN analyses and known roles in cancer signalling, blots for pro-caspase 8 (57 kD) showed over-expression in D-SWCNT, D-MWCNT and ASB compared to both SAL and DISP control cells. Over-expression of N-terminal cleavage product (43 kDa) above pro-caspase 8 levels was apparent across all cell types except D-UFCB cells. Active caspase 8 (18 kDa) was not detected in any cell types. Next, we evaluated the status of NF-κB levels in the CNT- and ASB-transformed SAECs to evaluate whether differences between ASB and CNT inflammatory signalling were partially due to NF-κB expression. Both p100 and p52 NFKB2 subunits were diminished in D-MWCNT and ASB cells compared to SAL and DISP cells (Figure 7D). D-SWCNT cells exhibited a substantial reduction in the p100 subunit compared to all other cell types. In agreement with microarray data, we found that both SAL and DISP had greater p65 NF-κB (RELA) levels than D-SWNCT, D-MWCNT and ASB cells. In addition, increased phosphorylation of IKKα/β was found in both D-SWCNT and D-MWCNT, suggesting degradation of NF-κB. To evaluate potential pro-inflammatory signalling in ASB cells, we evaluated protein expression for several key gene signalling hubs. D-UFCB, D-SWCNT, D-MWCNT and ASB cells possessed over-expressed levels of IL-1β compared to both SAL and DISP controls with ASB displaying the highest IL-1β expression. ASB cells over-expressed both IL-8 and SPI1 compared to DISP cells, while both D-SWCNT and D-MWCNT possessed basal low expression level (Figure 7E). In summary, protein expression analysis confirmed a majority of the microarray expression data and displayed several differences between CNT SAECs and ASB SAECs.

This study provides novel evidence supporting CNT-associated lung neoplastic-like transformation upon long-term in vitro exposure which differed from crocidolite asbestos. Subchronic (6-month) exposure of 0.02 μg/cm² dispersed SWCNT and MWCNT to SAECs resulted in increases in several cancer hallmarks including proliferation, morphological transformation, anchorage-independent growth, invasion/migration and angiogenesis compared to DISP control cells. Aggressive neoplastic-like behaviour correlated with several cancer-associated GSNs, pathways and key proto-oncogenes associated with cell proliferation, cell movement, cell death and lipid metabolism which largely differed from pro-inflammatory genome expression signature and protein expression in ASB SAECs. The moderate inflammation signature in SAL control cells (NF-κB and VEGF) compared to DISP cells likely stemmed from long-term passage and the absence of lung surfactant, a known inflammation suppressor (Wright et al. 2004). Previously, both SWCNT and MWCNT particles used in this study were shown to cause centrosome fragmentation, mitotic spindle disruption and aneuploidy in human small airway and bronchial epithelial cells during acute exposures (Sargent et al. 2009; Sargent et al. 2012) which are typically observed in cancer cells. Furthermore, these same CNTs were found to induce K-ras mutations following inhalation exposure in mice (Shvedova et al. 2008). Collectively, these results satisfy three out of the four major criteria for neoplastic transformation (OECD 2007; Creton et al. 2012) minus in vivo tumourigenesis after injection. These xenograft studies are currently underway at NIOSH.

Increased TF and altered morphology of both D-CNT cells vs. ASB cells provided the first evidence of a distinct CNT transformation effect. Morphological transformation evidenced by small Type III foci in both D-SWCNT and D-MWCNT cells correlated with transformation frequencies of Balb/3T3 cells exposed to functionalised MWCNT for 72 h (Ponti et al. 2012). Gross morphological assessment revealed CNT exposure-associated increase in the number of SAECs possessing multiple intracellular vesicles surrounding the nucleus. Their close association to the nucleus, Golgi apparatus and lysosomes suggests that these bodies are potentially associated with endocytotic uptake response to nanomaterial exposure, general xenobiotic exposure (Al-Jamal et al. 2011) or associated with enhanced synthesis/trafficking of intracellular material (i.e. lipid metabolism). Further evaluation of the potential role of this unique morphological characteristic is currently underway in our laboratory.

Several recent reviews on ENM toxicology have called for use of whole-genome/proteome expression assessments to help identify potential health risks associated with ENM exposure. Phenotypic anchoring of organismal or tissue behaviour with whole-genome expression signature profiling can determine differences between xenobiotic exposures and identify novel mechanisms of action (Ganter & Giroux 2008). Here, employment of whole-genome expression signature profiling following subchronic in vitro CNT exposure revealed novel signalling networks with known neoplastic effects in D-CNT cells with some stark differences to ASB cell neoplastic-like transformation signalling. Prior studies with acute SWCNT- and MWCNT-exposed cells displayed DNA damage, activated MAPKs, AP-1, NF-κB and Akt, all of which recapitulate key molecular events involved in asbestos-induced mesothelioma (Sharma et al. 2007; Chou et al. 2008; Pacurari et al. 2008a; Pacurari et al. 2008b; Lindberg et al. 2009; Hirano et al. 2010; Ponti et al. 2012; Sargent et al. 2012). Acute studies directly comparing different CNT morphologies to asbestos report greater SWCNT cytotoxicity than MWCNT while MWCNTs exhibit similar fibrotic effects in fibroblasts (Tian et al. 2006; Mishra et al. 2012). In addition, increased inflammatory signalling, altered actin cytoskeleton and cell morphogenesis in bronchial epithelial cells are known (Kim et al. 2012). These studies suggest that initial, acute lung epithelial response is largely dominated by inflammatory and fibrosis signalling.

Here, subchronic CNT exposure resulted in alteration of proto-oncogenes both observed in asbestos-associated lung disease and unique to CNT-exposed cells. Over-expression of MYC, PPARG, CASP8 and COL18A1 as major hub genes in both D-CNT SAEC cancer GSNs suggests a proto-oncogene-dominated genotype compared to ASB SAEC's pro-inflammatory genotype centred on over-expressed IL-1B, CCL2, SRC and SPI1. Up-regulation of c-Myc, fra1 and bcat in mesothelioma in rats following asbestos intraperitoneal injection (Sandhu et al. 2000) parallels over-expression of MYC, FOS and BCAT1 in D-SWCNT cells and MYC in D-MWCNT cells in the present study. MYC over-expression potentially represents a common response to HAR particles or a generalised response to xenobiotics since its acute over-expression drives apoptosis and other cell stress responses. Its persistent over-expression, however, is a well-established cancer promotion signal (Varella-Garcia 2010).

Key genes identified in the D-CNT cell GSN analysis exhibited mRNA and protein expression correlating with known cancer function. Several top up- and down-regulated genes were associated with extracellular matrix, cell motility and morphology. ANTXR1 and LAMA4 over-expression potentially contributed to enhanced invasion ability in CNT and ASB SAECs (Huang et al. 2008; Cullen et al. 2009). Conversely, large down-regulation of COL1A1, a known lung fibrosis marker, and PLAU with over-expression of COL18A1 and several basal collagens (collagens IV, V and VI; data not shown) suggests a shift towards basal ECM adhesion and potential motility enhancement. Over-expression of COL18A1 and collagens IV, V and VI is associated with poor clinical outcome in NSCLC patients and promotes cancer cell survival (Iizasa et al. 2004). INHBA and ATM, two anti-proliferation tumour suppressors commonly mutated in lung cancer (Ding et al. 2008), were down-regulated in D-CNT SAECs. Over-expression of BCL2L2, CASP8, FOS and MYC suggests substantial changes to survival, cell stress and cell death signalling pathways in CNT- exposed SAECs. Although CASP8, FOS and MYC are involved in cellular stress response with known cell death signalling roles, they also display several tumour promotion functions including mitochondria stimulation, cell proliferation, migration and transformation and exhibit increased expression in lung cancer (Heintz et al. 2010; Varella-Garcia 2010; Kamp et al. 2011; Kober et al. 2011).

Decreased p53 expression with enhanced phospho-p53 (Ser15) suggests that subchronic CNT and ASB exposure suppressed basal p53 expression. Down-regulation (e.g. ATM) and up-regulation (e.g. POLK, MGMT) of DNA repair genes with decreased basal p53 expression in D-SWCNT SAECs suggest an altered ability to respond to potential DNA damage. However, enhanced Ser15 phosphorylation suggests an active tumour suppressor function by allowing binding to p300 or binding directly to DNA damage sites (Shanaz et al. 2005). This contradicts our previous finding in SWCNT-exposed BEAS-2B cells showing decreased phosphorylation (Wang et al. 2011b) potentially due to a weakened p53 following BEAS-2B immortalisation.

Subchronic SWCNT and MWCNT exposure caused large alterations in lipid metabolism suggesting that CNTs may drastically alter many energy, proliferation and oxidative stress pathways that rely on fatty acids and sterols. Few studies have acknowledged the role that CNT exposure may have in affecting lipid peroxidation, lipidomic response and lipid metabolic pathways in cells (Shvedova et al. 2012). Lipid peroxidation is known to activate peroxidation detoxification pathways including the peroxisome proliferator-activated receptors (PPARs). PPARs regulate lipid metabolism, retinoic acid signalling and are implicated in either promoting or inhibiting carcinogenic signalling in lung cancer. PPARG expression is typically observed in adipocytes, possesses potent anti-inflammatory properties and is a known prognostic marker in lung cancer with it over-expression indicating a favourable prognosis (Szanto & Nagy 2008; Peters et al. 2012). Over-expression of PPARG in both D-SWNCT and D-MWCNT SAECs suggests a potential role that lipid signalling and metabolism may play in these cells. Down-regulated NFKB2, a key hub gene in several top-ranked D-SWCNT and D-MWCNT GSNs associated with lipid metabolism (Table IV), suggests a potential relationship between PPARG, NFKB2 and an anti-inflammatory genome expression signature. In addition, PPAR family activation promotes expression of several genes (i.e. HMGCR) that contributes to cancer promotion via lipid and cholesterol synthesis (Peters et al. 2012). Future studies should evaluate both lipid metabolism and PPAR expression to evaluate the role that CNT-induced alterations to cellular lipid peroxidation levels, lipid metabolism and suppressed immune function have on pulmonary tissue.

Knowledge-based mapping of cancer GSNs identified a pro-inflammatory dominated genome expression signature in ASB SAECs which was largely absent in both D-CNT SAECs. Pulmonary exposure to HAR fibres results in an acute inflammatory response and infiltration of polymorphonuclear leukocytes and macrophages. Due to their long length, frustrated phagocytosis and activation of the NLRP3 inflammasome ensue (Palomaki et al. 2011). Protein over-expression of IL1B and CCL2 is typically observed in lung lavage fluid following asbestos and CNT in vivo exposures (Jacobsen et al. 2009; Shvedova et al. 2005; Shvedova et al. 2008). Prolonged release of pro-inflammatory mediators results in activation of NF-κB and its downstream targets. Present hypotheses for ASB-mediated carcinogenesis focus on activation of tumour promotion signalling pathways via ROS generation and chronic inflammation (Nymark et al. 2007; Kamp 2009; Heintz et al. 2010; Kamp et al. 2011). Toxicogenome expression analysis of acute in vitro asbestos-exposed lung epithelial and mesothelial cells identified altered or enhanced NF-κB regulation, thioredoxin, Bcl-2, integrin, collagen, p53, caspase 8, c-Myc and Nrf2 signalling, most of which are known to occur in asbestos-associated lung cancers (Nymark et al. 2007; Hevel et al. 2008). Here, our subchronic exposure study's findings largely support a pro-inflammation hypothesis for asbestos-induced neoplasm and supplies additional novel insight for several key GSNs associated with such response. Top-ranked ASB GSNs and canonical pathways with over-expression of several integrin signalling genes (ITGB2, SRC, TREM1, PI3K) with down-regulated clathrins suggested adaptation to active uptake of asbestos fibres via integrin signalling and activation of NF-κB inflammatory pathways (Kamp 2009). Cell proliferation, cancer and tissue morphology behaviour were associated with the ASB pro-cancer GSN, suggesting an up-regulated inflammation response (Dostert et al. 2008). SPI1 over-expression, a known regulator of haematopoiesis and leukocyte differentiation, with its association with several inflammatory genes suggests that it contributed to ASB cell neoplastic-like behaviour and was recently identified as a potential squamous lung cancer marker (Bai & Hu 2012). To our knowledge, this is the first time SPI1 has been implicated in an asbestos-induced neoplastic-like mechanism. Many of the major expression changes observed in D-CNT SAECs were also observed in ASB SAECs (e.g. over-expressed CASP8, MAP3K1, CD44 and c-Myc protein), albeit to a lesser extent. In conclusion, GSN analysis of the ASB cell signature overwhelming supported a pro-inflammatory signalling associated with several cancer hallmarks while the D-CNT cell signature favoured a large decrease in the innate immune system and NF-κB.

Conversely, subchronic UFCB exposure resulted in a non-neoplastic phenotype characterised by slow proliferation, low soft agar colony formation and a genome expression signature consistent with enhanced cell death, reduced neoplasia and reduced attachment ability. Prolonged UFCB exposure results in alveolar deposition, interstitial penetration, pulmonary inflammation and epithelial cytotoxicity leading to lesions and fibrosis (IARC 2010). UFCB is listed as a Group 2B possible human carcinogen with no clear risk of human lung cancer development in the occupational setting; however, female murine inhalation exposure studies found increased risk for lung cancer (Borm et al. 2004). Our in vitro study differs from previous human and murine bronchial epithelial cell in vitro studies. Following relatively high exposure doses, cells exhibited increased ROS production, proliferation, kinase activation, epithelial growth factor receptor activation and weak mutagenesis (Tamaoki et al. 2004; Jacobsen et al. 2007). Although these studies suggest toxicological effects typical of known carcinogens, their reported acute or subchronic exposure doses were 17.6–2000 times greater than those tested in the present study. UFCB-induced ROS production leading to DNA damage is more likely due to an indirect mechanism via the ‘particle overload’ phenomenon and low clearance rates in both in vivo murine and in vitro exposure assessments (Oberdorster 2002; IARC 2010). In this study, the subchronic, low-dose exposure regime potentially did not result in a UFCB ‘particle overload’ mechanism leading to a neoplastic-like phenotype.

Differences in exposed cell behaviour, toxicogenome signatures and signalling pathways are potentially due to several physicochemical differences among particle types. HAR, presence/absence of metal impurities, major components (carbon vs. silica), surface area, uptake ability and internal dosimetry possibly contributed to the observed differences among exposed cells. High doses of long, thin MWCNTs compared to short and tangled MWCNTs were shown to cause increased damage and increased risk for mesothelioma (Nagai et al. 2011). CNTs containing large or minute amounts of metal catalysts caused enhanced ROS, inflammation and cytotoxicity both in vitro and in vivo (Lam et al. 2004; Shvedova et al. 2009; Hirano et al. 2010; Azad et al. 2012; Shvedova et al. 2012). Recent studies evaluating ENM dosimetry have repeatedly shown that surface area is a contributing factor to observed lung inflammation, fibrosis and potential carcinogenesis (Borm et al. 2004; Duffin et al. 2007; Murray et al. 2012). Surface area-based ‘particle overload’ also may occur with CNTs since they possess low mass but high surface areas (Donaldson & Poland 2012). Here, dispersed SWCNTs possessed a 10–100-fold increase in surface area compared to all other dispersed particles, while MWCNTs possessed a 2.6-fold increase in surface area over asbestos fibres. A greater surface area for SWCNTs and MWCNTs potentially contributed to the enhanced cancer hallmark behaviour compared to other treatment groups. UFCB's low aspect ratio with comparable surface area to MWCNTs suggests that HAR is a compelling factor in CNT toxicity (Donaldson et al. 2010). Conversely, qualitative analysis of particle uptake showed that a high number of dispersed MWCNT were co-localised with SAEC compared to equal doses (μg/cm²) of other particles, which suggests the need to evaluate how ENM internal dosimetry drives short- and long-term health effects. Regardless of physicochemical and uptake differences, both D-CNT cells exhibited the most similar toxico-genome signatures and pro-cancer signalling networks. This suggests that a long-term, occupational exposure to a carbon-based nanofibre with relatively high surface area results in a neoplastic-like transformation effect that is unique from UFCB or asbestos fibres.

Both D-CNT SAECs exhibited decreased gene expression in complement proteins, NFKB2 and several known inflammation cytokines. Recent in vivo and in vitro studies have reported alterations and/or reductions in immune response following CNT exposure, possibly due to changes in lung cell signalling to the spleen, impaired macrophage function or excessive activation of complement proteins (Shvedova et al. 2008; Herzog et al. 2009; Mitchell et al. 2009; Tkach et al. 2011; Anderson et al. 2012). MCP-1 (CCL2), a pulmonary inflammatory marker, consistently exhibits over-expression in mouse lung lavage following acute CNT exposure; however, a previous in vitro study found MCP-1 under-expression following SWCNT exposure (Herzog et al. 2009). Its decreased gene expression in D-CNT SAECs, along with other known inflammatory cytokines, suggests that inflammatory-related markers may only be relevant within short time frames following exposure. This is not a surprise since several studies show that the inflammatory response to CNT exposure is transient in vivo (Shvedova et al. 2005; Shvedova et al. 2012; Porter et al. 2010). Increased p-IKKa/B expression resulting in targeting p65 NF-κB and NFKB2 down-regulation suggests that continuous, low-dose CNT exposure results in reduced lung epithelial cell innate and adaptive immune responses, respectively, potentially contributing to enhanced susceptibility to infection (Tkach et al. 2011). At present, the mechanism(s) driving a suppressed innate and inflammatory response following CNT exposure are unknown. Based on in vitro and in vivo results, future studies should focus on CNT and nanomaterial ability to weaken pulmonary immune response resulting in increased susceptibility to other insults.

Few in vitro studies have assessed prolonged, continuous exposure to CNTs compared to other well-categorised particles for potential health risks. This study contradicts the study by Thurnherr et al. (2011) that reported no long-term effects or adaptive mechanisms in A549 cells following continuous MWCNT exposure for 6 months at the 0.16 μg/cm². A549 cells were found to experience no changes in proliferation, morphology, ROS levels or cell death at 3 and 6 months post-exposure. Lung cancer, such as human A549 adenocarcinoma cells, exhibits several enhanced cancer hallmarks including enhanced cell proliferation and ROS levels (Azad et al. 2008). As such, they do not represent a normal, non-tumourigenic human airway epithelial cell to assess neoplastic transformation, but do serve as an appropriate acute exposure model for inflammation and cytotoxicity for particulate and CNTs (Herzog et al. 2009). Any MWCNT-induced changes were potentially masked or undetectable in an A549 cell model due to an already present malignant phenotype. Our previous study (Wang et al. 2011b) displaying SWCNT-induced neoplastic transformation in a non-tumourigenic bronchial epithelial cell (BEAS-2B) model coincides with the present study showing that the same SWCNT particles promoted increases in several cancer hallmarks.

Several recent published studies have reported in vivo proteomic and genetic responses following CNT exposure. Comparison of C57BL/6 mouse whole-lung proteomic responses to UFCB vs. SWCNT vs. asbestos at 24 h following a repeated aspiration exposure over 3 weeks found that SWCNT exhibited a more potent but similar inflammatory response to equal mass of crocidolite asbestos (Teeguarden et al. 2011). Major protein expression changes in haematopoiesis, endocytosis, chemotaxis, immune response and leukocyte activation for both SWCNT and asbestos coincide with this study's ASB SAEC cell toxico-genome signature. Agreement between both in vitro and in vivo profiling studies at 6-month and 3-week exposure time points, respectively, matches the chronic lung inflammation paradigm for asbestos and other HAR fibre lung disease (Kamp 2009; Broaddus et al. 2011). Two other biomarker studies screened known human lung cancer markers in whole mouse lung at 7–56-d post-exposure to aspirated MWCNT (Pacurari et al. 2011; Guo et al. 2012). These authors reported differential expression in 11 and 38 cancer-related gene subsets, respectively, in MWCNT-exposed lung and were associated with signal transduction, cell proliferation, cell cycle and cancer. Preliminary comparison of this study's CNT pro-cancer signalling results to gene markers from these two studies correlate poorly. This is due to potential differences in in vitro monolayer vs. whole lung in vivo response, species tested, time/dose differences and data set approach (i.e. whole-genome expression vs. human lung cancer marker set). Regardless, network analysis of an 11-gene biomarker set at 56-day post-exposure (Pacurari et al. 2011) did identify PPARG as a key signalling gene hub associated with other cancer and immune response signalling centred on tumour necrosis factor, interferon β and Bcl-2. Two genes, neuregulin (NRG1) and sonic hedgehog receptor (PTCH1), did exhibit differential expression in both this and the two in vivo biomarker studies following CNT exposure suggesting the potential importance of lung development signalling following CNT exposure. A more detailed in vitro to in vivo comparison is currently underway in a follow-up study using the reported D-SWNCT and D-MWCNT cell expression signatures to identify potential gene markers for use in CNT exposure risk assessment.

Although our chronic in vitro model successfully identified known carcinogenic gene signalling in ASB SAECs and distinct CNT neoplastic-like mechanisms, care must be taken in interpretation and applicability of these results. SAEC-hTERT cells were chosen based on their wild-type p53 status, known CNT interaction in vivo and non-tumourigenic status; however, they do exhibit drops in p16 expression after continuous long-term passage (Piao et al. 2005). These immortalised cell models, however, may represent the in vivo human lung epithelial cell condition since many lung cancers exhibit mutations in p53 and p16 (Ding et al. 2008). In addition, in vivo studies have shown that cell-to-cell signalling between epithelial, fibroblast and leukocyte cells following in vivo asbestos and CNT-exposure is largely inflammatory in nature (Shvedova et al. 2005; Porter et al. 2010; Matsuzaki et al. 2012). The absence of cells capable of releasing pro-inflammatory or other cell damage cytokines/chemokines (i.e. macrophages) in our system must be taken into account when considering the large down-regulation of inflammation-associated genes in D-SWCNT and D-MWCNT. Nevertheless, subchronic in vitro exposure models provide a rapid assessment tool to compare a multitude of different xenobiotics in a controlled manner and can assist in identifying mechanisms promoting disease.