SWCNT were obtained from Carbon Nanotechnology (CNI, Houston, Texas) and were purified by acid treatment to remove metal contaminates, while MWCNT were obtained from Mitsui & Company (New York, NY). Elemental analysis by nitric acid dissolution and inductive coupled plasma-atomic emission spectrometry (ICP-AES) showed that purified SWCNT and MWCNT contained 99% elemental carbon and less than 1% of contaminants. The metal residues were mostly iron (Fe) at 0.23 and 0.32% by weight in SWCNT and MWCNT, respectively. The BET surface area, length (L), and width (W) of individual dry CNT were 400–1040 m²/g, 0.6 ± 0.5 μm (L), and 1 ± 0.2 nm (W) for SWCNT, and 26 m²/g, 8.19 ± 1.7 μm (L), and 81 ± 5 nm (W) for MWCNT.^23,24 CNTs were dispersed by using bovine serum albumin (BSA) and were lightly sonicated prior to culture exposure. The CNT doses used in the in vitro exposure studies were calculated based on in vivo CNT exposure data normalized to alveolar surface area in mice. For example, the doses that induced positive in vivo fibrogenic response were 40–80 μg/mouse lung (0.5 mg/kg body weight).^5,6,25 Dividing this dose by the average alveolar surface area in mice (∼500 cm²)²⁶ indicates the in vitro surface area dose of 0.08–0.16 μg/cm². Extrapolation of the experimental dose to human exposure scenarios in the workplace can be evaluated using the approach adopted by the National Institute for Occupational Safety and Health.^25,27 Assuming the alveolar surface area in humans of 100 m², the human burden is equal to 80–160 mg/Lung. Considering a lung deposition fraction of 30% and a working day inhalation of 10 m³ (a minute ventilation of 20 L/min), the experimental dose could be reached within 3–6 months of human inhalation exposure at 400 μg/m³ (high CNT level reported in a research facility)²⁸ or about 10–20 years at 10 μg/m³ (average CNT level in U.S. facilities)²⁹ (see Supporting Information Figure S1).

A key cellular mechanism of fibrogenesis is fibroblast activation and subsequent induction of ECM production and accumulation.^30,31 To investigate the effect of CNTs on the fibrogenic process, primary human lung fibroblasts were treated with various concentrations (0–0.15 μg/cm² or 0–0.71 μg/mL) of SWCNT and MWCNT (see Supporting Information Table S1 for dosimetry) and were evaluated for cell proliferation and collagen production using MTT and Western blot assays. Figure 1A shows that SWCNT and MWCNT had no significant effects on cell proliferation and cytotoxicity at the indicated concentrations, the results that were confirmed by CyQuant cell proliferation assay (Invitrogen, Carlsbad, CA) (data not shown). Figure 1B,C shows that SWCNT, and to a lesser extent MWCNT, strongly induced type I collagen production at the dose as low as 0.08 μg/cm². Transforming growth factor (TGF)-β, a known fibrosis inducer that was used in this study as a positive control, similarly induced collagen production at the concentration of 1 ng/mL. Figure 1B,D shows an increased expression of myofibroblast marker α-smooth muscle actin (α-SMA) in SWCNT-, MWCNT- and TGF-β-treated cells, indicating the transformation of fibroblasts to myofibroblasts, which is known to be the main source of collagen production. Analysis of soluble collagen in the cell culture medium by Sircol assay further showed that both SWCNT and MWCNT induced collagen secretion from the treated cells in a dose-dependent manner (Figure 1E).

An important pathological feature of lung fibrosis is the presence of fibroblastic foci, which are aggregates of lung fibroblasts and myofibroblasts and the newly deposited collagen.^11,15 In this study, we developed a 3D biomimetic model of fibroblastic foci by growing primary human lung fibroblasts in culture on an adherent substrate consisting of poly-l-lysine in the presence of TGF-β, a known inducer of fibrosis. Figure 2A shows the formation of 3D cell clusters that are referred to as fibroblastic nodules after treatment of the cells with TGF-β (1 ng/mL) for 16 h. Interestingly, treatment of the cells with SWCNT or MWCNT similarly induced the fibroblastic nodules, whereas vehicle-treated cells showed minimal nodule formation (Figure 2A). Quantitative analysis for the number of fibroblastic nodules demonstrated the dose-dependent effect of SWCNT and MWCNT treatment (Figure 2B). This number correlates well with the level of collagen produced by the cells as indicated by the correlation plot depicted in Figure 2C showing the linear relationship between the two parameters with a goodness of fit (R²) of 0.8629. To aid visualization of the fibroblastic nodules, the cells were stained with Hoechst 33342 dye and observed under a confocal fluorescence microscope. Figure 2D shows the nodule-forming cell aggregates induced by SWCNT, MWCNT, and TGF-β but not by control treatment. The orthogonal views and reconstructed Z-stack image series display 3D morphology of the fibroblastic nodules (Figure 3 and Supporting Information Video S1 and S2). These results support the potential utility of the fibroblastic nodules as an in vitro 3D model for fibrogenicity testing of nanomaterials.

Type I collagen is the most abundant ECM proteins robustly expressed in lung fibrosis.^32,33 To further validate the clinical resemblance of the 3D fibroblastic model, type I collagen expression was evaluated in SWCNT-, MWCNT- and TGF-β-induced fibroblastic nodules using immunofluorescence in comparison with the monolayer of vehicle-treated control fibroblasts. Figure 4A shows that the fibroblastic nodules induced by CNTs or TGF-β were enriched with type I collagen. Together with the cell aggregation data in Figure 2D, we demonstrate here that the fibroblastic nodules have important compositional and morphological features of the fibroblastic foci observed in fibrosis patients. The greater number of fibroblastic nodules in SWCNT as compared to MWCNT (Figure 2) are consistent with previous in vivo findings that indicated a greater fibrogenicity of SWCNT than MWCNT,^5,6 thus validating the potential utility of the 3D fibroblastic nodules as a functional assay for CNT fibrogenicity. Overall, the advantages of this 3D model over conventional cell monolayer models include (i) clinical resemblance to human lung fibrotic lesions; (ii) rapid and quantitative or semiquantitative analysis, that is, by colony (nodule) counting as opposed to the more laborious immunological and biochemical assays; (iii) fewer cell numbers needed per assay; and (iv) potential for high-throughput screening, that is, by using automated colony counter.

Evolving research indicates the presence of putative stem cells residing in the lungs to be significant in an early development of lung fibrosis.²¹ To investigate the existence of stemlike fibroblasts, we performed immunofluorescence assays determining the expression of universal stem cell markers ABCG2 and ALDH1A1^34,35 in fibroblastic nodules induced by SWCNT, MWCNT, and TGF-β. Figure 4B shows that these fibroblastic nodules expressed a high level of ABCG2 and ALDH1A1 stem cell markers as compared to vehicle-treated control cells. These results suggest the presence of stemlike cells in the fibroblastic nodules and support their role in fibrogenesis. We also observed CNT deposition in the fibroblastic nodules induced by SWCNT and MWCNT (Figure 4A and B, arrows) with minimal presence of CNTs outside the nodules.

Adult stem cells are known to efflux Hoechst dye slowly due to the high expression of ABCG2. Flow cytometric analysis was used to identify these stem cells with a distinct low Hoechst staining pattern referred to as side population (SP).^36,37 An increase in SP population was reported in the lung of mice with fibrosis, suggesting the involvement of SP-positive stem cells in lung fibrogenesis.²¹ To determine the potential role of stemlike fibroblasts in CNT-induced fibrogenesis, we first determined the change of SP subpopulation upon CNT treatment. Primary human lung fibroblasts were treated with SWCNT or MWCNT at the concentration of 0.15 μg/cm² for 48 h, after which they were incubated with 5 μg/mL of Hoechst 33342 in the presence or absence of 25 μM fumitremorgin C, an inhibitor of ABC transporter. Figure 5A shows that both SWCNT and MWCNT were able to induce the SP subpopulation. The percentage of SP was approximately 5% in SWCNT-treated fibroblasts and 3% in MWCNT-treated cells versus less than 0.5% in vehicle-treated control cells (Figure 5B).

To determine the role of stemlike fibroblasts in CNT-induced fibrogenesis, we isolated the stemlike cells from SWCNT-treated (0.15 μg/cm²) fibroblasts based on their SP phenotype using fluorescence-activated (flow cytometry-based) cell sorting (FACS). Sorted SP and non-SP fibroblasts were further evaluated for the stem cell marker ABCG2 and type I collagen expression by immunofluorescence. Figure 6A,B reveals a substantially higher expression of ABCG2 in the SP versus non-SP cells, thus confirming the stemlike phenotype of SP fibroblasts and the reliability of stem cell isolation by FACS. Importantly, type I collagen expression was significantly higher in the SP versus non-SP population. It is widely known that fibroblasts play a key role in fibrogenesis through its ability to synthesize and secrete ECM proteins including type I collagen, which characterizes fibrosis.^38,39 Our results thus indicate that the stemlike fibroblasts are a potential key source of collagen production and may play a crucial role in fibrogenesis.

To substantiate the functional role of stem-like fibroblasts in CNT fibrogenesis, the sorted SP and non-SP cells from CNT-treated fibroblasts were assessed for their ability to form fibroblastic nodules. Figure 6C shows that that SP cells had a substantially higher capability to form fibroblastic nodules than the non-SP fibroblasts. Altogether, these findings support the role of stemlike fibroblasts in CNT-induced fibrogenesis.

To provide a supporting evidence for the clinical relevance of stem cells in lung fibrosis, we performed an expression analysis of universal stem cell markers ALDH1A1 and ABCG2 in human clinical specimens from fibrotic and matched normal lung tissues (Origene, Rockville, MD) using immunohistochemistry and immunofluorescence, respectively. Figure 7A demonstrates for the first time an upregulation of the stem cell markers in human lung fibrosis tissues as compared to matched normal lung tissues. Quantitative analysis of the stem cell marker expression by Western blotting further showed an increased expression of ALDH1A1 and ABCG2 in the cell lysates obtained from lung fibrosis tissues versus matched normal lung tissues (Figure 7B). These data provide preliminary supporting evidence for the role of stem cells in human lung fibrosis. As the high expression of ALDH1A1 and ABCG2 was similarly observed in the CNT-fibrotic nodules, these findings strengthen the role of stemlike fibroblasts in CNT-induced fibrogenesis.

In summary, we have developed a 3D model of CNT lung fibrogenesis that is fast, robust, and resembles the clinical fibrotic foci of lung fibrosis. The model employs primary human lung fibroblasts that form a collagen-rich 3D structure upon stimulation with CNTs or TGF-β. Using this model, we unveiled the presence of fibroblast stemlike cells in the fibroblastic nodules and demonstrated its role in CNT-induced fibrogenesis. The developed model could potentially be used as an alternative assay to predict the fibrogenicity of CNTs and other nanomaterials for their safer design and risk assessment. In addition, the model could be used to aid mechanistic investigations of the cellular and molecular events leading to fibrogenesis.