Main

Single-walled carbon nanotubes (SWNTs) are hydrophobic, tubular nanostructures with diameters of only 0.6–3.5 nm, and are therefore much smaller than the nuclear pore complex but within the size range of other pores such as ion channels and gap junction complexes that allow molecules to move freely into and out of cells. It has therefore become of great concern that SWNTs may also enter cells through the lipid bilayer and interact with organelles or even enter the nucleus11,12. SWNTs are routinely synthesized using a metal catalyst, which also has the potential to have toxic effects on cells1,8,13. Internalization of fluorochrome-labelled SWNTs into cells has been observed with confocal microscopy9. However, tracing the same process using unlabelled SWNTs is challenging, because it is difficult to distinguish carbon-based nanotubes from carbon-rich organelles due to similarities in composition and dimensions1. Visualizing SWNTs inside cells will help us understand how SWNTs enter cells, where they migrate to, and their fate after uptake. Previously we have demonstrated that low-loss energy-filtered transmission electron microscopy (EFTEM) enables clear differentiation between C60 and cellular compartments in unstained sections14; however, there has been no information about the pathological effect of internalized C60 on cell morphology and survival. Here we visualized individual SWNTs in the cell using EFTEM in combination with electron energy loss (EEL) spectrum imaging—a method that yields characteristic energy-loss information—and mapped the distribution of SWNTs in the cell. This technique gives a good image contrast of unstained sections that cannot be achieved using conventional imaging techniques15,16,17. We also present a simple method for imaging intracellular SWNTs using the confocal microscope by filling the nanotubes with silver iodide (AgI) as described in ref. 18. Encapsulating AgI inside SWNTs increases the reflectance from SWNTs without chemically modifying the surface and also removes toxicity issues19,20. Structural changes seen in the cell were correlated to the distribution of internalized SWNTs, and these changes were compared with the results obtained from conventional cell viability assays.

Macrophages, which are cells of the immune system that can cause inflammation, form the first line of defence against foreign materials in many tissues, including the lung. A significant proportion of nanoparticles entering the pulmonary airways are likely to be ingested by macrophages, where they are concentrated into membrane-bound organelles, phagosomes or lysosomes, before entering the blood and lymph circulation. Therefore, we chose human monocyte-derived macrophages (HMMs) as an in vitro model for the exposure of SWNTs to cells. The HMMs were treated with SWNTs for 2 and 4 days at concentrations of 0–10 μg ml−1. TEM studies were performed at a concentration of 5 μg ml−1. Following incubation with SWNTs, cell viability was measured in parallel using neutral red (NR) (3-amino-7-dimethylamino-2-methyl-phenazine hydrochloride) and MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide] assays. The HMMs exposed to concentrations up to 10 μg ml−1 after 2 days showed no significant toxicity with the NR assay, but significantly fewer cells survived after 4 days in both assays. The NR assay showed a significant decrease in cell viability at the highest concentrations of SWNTs. In the MTT assay a significant decrease in cell viability was found at all concentrations above 0.3 μg ml−1. The localized effect of the uptake of SWNTs on cell death was quantified by indexing 100 cells in each group as either apoptotic, necrotic or healthy. No significant difference in cell death was observed between the control and SWNT-treated cells at 2 days. This difference became significant after 4 days (Fig. 1b).

Figure 1: Toxicity of SWNTs.
figure 1

a, SWNT-treated cells show decreased viability in the neutral red (NR) assay at 4 days when compared to control cells (NA) (ANOVA, F = 4.406 and p < 0.001). Fishers PLSD indicated significantly decreased viability at 2.5 and 10 μg ml−1 when compared with equivalent controls (groups labelled with the same letters indicate homogeneous means). With the MTT assay, at all concentrations, SWNT-treated cells show significantly decreased viability when compared to control cells (NA) at 4 days (ANOVA, F = 16.67, p < 0.0001). b, TEM index of the analysis of cell death (apoptosis/necrosis; ±statistical error %) between control (NA) and SWNT-treated cells (5 μg ml−1). Numbers indicate % dead cells.

Full size image

Traditional TEM studies use heavy metal stains to generate contrast and reveal the structure of membranes and subcellular organelles. This obscures visualization of the intracellular SWNTs. Typically, we study the morphology of the cells in stained sections, which are compared with unstained sections. Achieving contrast from unstained sections is technically demanding, but using low-loss EEL spectrum imaging (Fig. 2a) and low-loss (0–78 eV) EFTEM with a small energy window (2–6 eV) (Fig. 2c–f), we achieved improved contrast between SWNTs and certain cell organelles (plasma membrane, vesicles and the nucleus) without staining (Fig. 2a–f). High-resolution bright-field imaging (Fig. 2b) and low-loss EELS (see Supplementary Information, Figs S1, S2) confirmed that these structures were SWNTs. In EFTEM we observed SWNTs forming bundles with dark contrast at 0 eV and reverse bright contrast at all energy losses up to 78 eV (see Supplementary Information, Fig. S3). We expect that contrast between the SWNTs and the cell arises primarily from the higher carbon atom density of the SWNTs. In EFTEM, this contrast was enhanced at ∼26 eV, corresponding to the higher σ + π bulk plasmon from the graphitic nanotubes with respect to the amorphous cell (see Supplementary Information, Fig. S2)15,16,17.

Figure 2: Intracellular distribution of SWNTs in unstained sections.
figure 2

a, HAADF-STEM image of SWNTs within a lysosome (2 days exposure). Box shows low-loss EELS image: Fe map (white); plasmon map (Ep) ranges from 17.5 eV (blue) to 21.5 eV (white). b, High-resolution bright-field (BF) image of SWNTs. Arrows indicate the diameter of SWNTs. c–f, Low-loss EFTEM images of cells (4 days exposure) at 26 eV, showing SWNTs fused with the plasma membrane (c); at 75 eV, showing SWNTs parallel with the plasma membrane and merging with lysosomes (d); at 26 eV, showing SWNTs crossing the nuclear membrane (e); at 0 eV, showing SWNTs inside the nucleus (f). Inset to f shows individual SWNTs. Labels: m, plasma membrane; l, lysosomes; nm, nuclear membrane; c, cytoplasm; n, nucleus.

Full size image

At 2 days, SWNTs were found within lysosomes (Fig. 2a,b). To map the SWNTs, the position of the bulk plasmon peak was extracted (see Supplementary Information, Fig. S1) and clearly shows features not visible in the associated image (Fig. 2a). Maps of the Fe M2,3 edge showed a significant number of iron clusters associated with the SWNTs (An EELS M2,3 edge corresponds to an energy loss of the primary beam because electrons from the 3p1/2 and 3p3/2 shells have been excited to empty states above the Fermi level.).

At 4 days exposure, SWNTs had fused and aligned with their long axes parallel to the plasma membrane (Fig. 2c). Bundles of SWNTs were also found within long tubular vesicles, which are probably early endosomes (Fig. 2d). In some regions these oriented bundles merged with lysosomes and fused with or crossed the lysosomal membrane (Fig. 2d; see also Supplementary Information, Fig. S1). Most significantly, SWNTs translocated across the nuclear membrane (Fig. 2e) and localized within the nucleus (Fig. 2f).

Selected sections were stained with uranyl acetate and lead citrate to enhance contrast from cell organelles and membranes. In stained sections we noted a time-dependent change in morphology. At 2 days (Fig. 3a) exposure to SWNTs, the majority of cells appeared healthy, and high-angle annular dark-field scanning TEM (HAADF-STEM) showed little alteration in morphology. In comparison, after 4 days exposure we observed features characteristic of cell death (Fig. 3b–d). Cell death occurred by two processes: apoptosis and necrosis. Apoptotic cells appeared shrunken, with heavily capped chromatin. Cytoplasmic and nuclear contents had leached out in necrotic cells (Fig. 3b–d). A marked increase in cell mortality could be correlated with regions of higher SWNT density.

Figure 3: Cell morphology after exposure to SWNTs.
figure 3

a, A healthy cell exposed to 5 μg ml−1 SWNTs for 2 days. b, Zero-loss EFTEM image of a necrotic cell at 4 days. Upper right: higher magnification image of SWNT (∼3–6 nm diameter) bundles around the plasma membrane (pm) shown in the boxed region of the main panel. c, HAADF-STEM images of aggregates of SWNTs within a necrotic cell, with opposite contrast to the zero-loss image in b. d, Higher magnification of boxed area in image c.

Full size image

The mechanism by which SWNTs enter the cell is not fully understood and reports have indicated that they could either traverse the cellular membrane by means of endocytosis12 or they insert into and diffuse through the lipid bilayer11. In our experiments, after 2 and 4 days exposure, the majority of SWNTs were located within phagosomes and lysosomes of healthy cells (Fig. 4a,b), suggesting uptake by phagocytosis—a process of engulfment by which macrophages ingest cellular fragments or micro-particles. SWNTs also translocated across the membrane into the neighbouring cytoplasm at 4 days (Fig. 4c), indicating passive uptake through the lipid bilayer. Furthermore, AgI-filled SWNTs (AgI@SWNTs) were seen in the nucleus after 3 days when imaged in reflectance mode on the confocal microscope (Fig. 4d). In most regions the lysosomal membrane remained continuous, but in some areas where SWNTs had fused with the membrane, disruption was evident (Fig. 4e). This is the first demonstration that individual SWNTs can cross lipid bilayers and enter the cytoplasm and nucleus of the cell.

Figure 4: Localization of intracellular SWNTs within stained cell sections.
figure 4

a, HAADF-STEM image showing SWNT bundles being actively ingested by a phagosome at 4 days. b, Higher magnification image of boxed area in image a illustrating that SWNTs were compartmentalized inside the phagosomal membrane (m). c, Bright-field STEM image of SWNTs translocating across the lipid bilayer into the neighbouring cytoplasm. d, Confocal microscope image of HMM exposed to AgI@SWNT at 3 days, confirming inclusion of SWNT bundles inside the nucleus (blue). e, HAADF-STEM image showing SWNTs within a lysosome, with membrane disruption at the region where SWNTs fused with the membrane.

Full size image

HAADF-STEM electron tomography of intracellular SWNTs successfully confirmed, in a spatially resolved three-dimensional image, that the SWNTs were located within cell organelles. Figure 5 shows a representative tomogram illustrating SWNT bundles within lysosomes.

Figure 5: HAADF-STEM tomogram of bundles of SWNTs within lysosomes at 4 days in a stained cell section.
figure 5

a–d, A series of horizontal slices through a HAADF-STEM reconstruction. The slices are 10 nm apart in the z-direction and illustrate SWNT bundles (red arrows) within lysosomes (l) in cells containing mitochondria that appear healthy. Variations in the mitochondrial matrix (mi) and SWNT bundles were observed at different heights within the section. Subsequent segmentation of SWNT bundles confirmed that they were continuous throughout the section (see Supplementary Information, Movie).

Full size image

In summary, we show that it is possible to map the location of intracellular SWNTs using TEM and confocal microscopy. We successfully imaged individual SWNTs within lysosomes and also crossing cell membranes. We demonstrate two possible pathways of entry of SWNTs into cells: energy-dependent phagocytosis or endocytosis and passive diffusion across lipid bilayers. SWNTs were found primarily within lysosomes, but more significantly within the cytoplasm and the nucleus. Uptake to these sites implies they may interact with intracellular proteins, organelles and DNA, which would greatly enhance their toxic potential. SWNTs also fused with the plasma membrane, where they have been shown to cause cell damage through lipid peroxidation and oxidative stress2,3. Localization of SWNTs at these sites was correlated to an increase in cell death in both cell viability assays and TEM analysis. However, TEM provided direct estimates of cell viability and the simultaneous presence of SWNTs in the same cell, while cell viability assays measure the change in survival from a larger cell population (n = 100 cells in TEM). NR uptake or MTT formazan formation are both compared to a control value set to 100%. These are both global assays of a large number of cells (n ˙ 0.05 × 106). The NR assay detects uptake of the dye by lysosomes and does not discriminate between the routes of cell death, merely giving an assay of viability. The MTT assay gives false positives, as the SWNTs bind to the insoluble formazan5, preventing its dissolution for subsequent spectrophotometric assay. In contrast, TEM determines the mode of cell death by clearly defined ultrastructural characteristics. To conclude, we show that direct imaging of SWNTs within cells is achievable and is essential to complement cytotoxicity assays to understand localized effects of SWNTs and establishment of their potential toxicity.

Methods

Materials

Owing to their widespread use and easy commercial availability, purified SWNTs (purified HiPco, <15 wt% ash content) were purchased from Carbon Nanotechnologies. These SWNTs were produced by chemical vapour deposition (CVD) by means of the HiPco process21. This process produces SWNTs with a gaussian distribution of diameters, with a maximum peak around 0.9–1.2 nm. SWNTs were fully characterized by high-resolution (HR)-TEM and EELS prior to exposure, which showed bundles of nanotubes decorated with iron catalyst particles and also onion-like structures around iron particles, and confirmed they were graphitic-like carbon.

Cell culture study

The detailed methods used for the in vitro cell culture study and electron microscopy have been published previously22. In brief, human monocyte-derived macrophages were treated with SWNTs for 2 and 4 days at concentrations of 0–10 μg ml−1. SWNTs were dispersed in tetrahydrofuran (THF), and bundles were broken up by ultrasonication for 10 min. The THF was used as a vehicle to hinder bacteria formation after 4 days in cell culture. SWNTs were freshly suspended in the THF by sonicating for 30 min. The suspension was immediately mixed into a serum free cell culture medium using a pipette. No dispersing reagent (such as a surfactant) was used to break up the as-prepared nanotube bundles, so as to maintain the intrinsic properties of the nanotubes. Following the incubation with SWNTs, cell viability was measured using two cell viability assay: (1) The NR assay (3-amino-7-dimethylamino-2-methyl-phenazine hydrochloride), which measures the lysosomal accumulation of NR dye in viable cells, and (2) the MTT assay, which measures the intracellular enzymatic conversion of MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide] by mitochondrial dehydrogenase into formazan in viable cells. In both assays, cell viability is given as a relative measure of toxicity, where the control sample is set to represent 100% cell viability. Each experiment was performed in triplicate. Results were calculated as mean ± standard deviation. Statistical analysis was performed using ANOVA to identify differences between groups followed by a post hoc Fishers protected least significance difference test (PLSD) with significance set at p < 0.01. Groups labelled with the same letters in Fig. 1a have homogeneous means.

Electron microscopy

TEM studies were performed at a concentration of 5 μg ml−1 in accordance with previous work14,22. For electron microscopy studies, SWNTs were fixed in 4% gluteraldehyde in piperazine-1,4-bis(2-ethanesulphonic acid) (PIPES) buffer, dehydrated, embedded, and subsequently sectioned with an ultramicrotome at 20–40 nm for high-resolution STEM imaging, 70 nm for EFTEM studies and 300 nm for three-dimensional electron tomography. The TEM sections were cut onto lacy carbon film grids and bare 600-mesh copper TEM grids. Selected sections were bulk stained with osmium tetroxide and post-stained with uranyl acetate and lead citrate for 5 min in each to enhance contrast from cell membranes and organelles. Changes in cell morphology upon exposure to SWNTs were assessed by comparing the structure and morphology of exposed and unexposed cells. Quantitative analysis of cell death was performed on osmicated cells by indexing 100 cells in each group (2, 4 days exposed and controls) as either apoptotic, necrotic or healthy. The statistical error was estimated using a binomial approximation in which the error (e) for a confidence level p = 0.05 was estimated as e = 1.96 × SQRT[(P × 1–P)]/n, where n = 100 and P is the percent of cases observed. Apototic and necrotic cells were identified by comparing cell morphology to reference images from previous studies23. To categorize the cells, apoptotic cells were characterized as shrunken, containing vacuoles in a condensed cytoplasm, and showing heavily capped chromatin. Necrotic cells were more electron-lucent than either normal or apoptotic cells, and their cytoplasmic and nuclear contents appeared to be leached out.

All TEM observations were made after viewing several hundred cell profiles from three different exposures. EFTEM studies were performed on a Philips CM300 operated at 300 kV with a Gatan imaging filter (GIF) model 2002 using a 10 μm objective aperture to optimize spatial resolution24. To optimize contrast, zero-loss images of non-stained sections were taken using a 3 eV slit. Selecting an energy window on the higher side of the plasmon feature enhanced the contrast from the SWNTs. Zero-loss images of stained sections were taken using a 20 eV slit. Low-loss energy filtered series were recorded from 0 to 30 eV using a 2 eV slit and 2 eV step size14. The position of the bulk σ + π plasmon peak was extracted from the EFTEM series using interactive data language (IDL) image processing software.

High-resolution imaging and EELS were performed at the UK SuperSTEM laboratory on a 100 kV VG HB501 dedicated STEM fitted with a Nion second-generation spherical aberration corrector and a Gatan Enfina EELS. The convergence semi-angle of the electron probe was 24 mrad for both imaging and spectroscopy. The collection semi-angle for the EELS was 19 mrad and 70–210 mrad for HAADF imaging. Low-loss EELS spectrum images were acquired at every pixel over a user-defined area with an energy dispersion of 0.1 eV per channel, covering an energy range of ∼0–100 eV. By fitting a nonlinear least-squares (NLLS) fit to the plasmon peak, its peak position was extracted and mapped. The iron M2,3 edge occurs at 54 eV, and background fitting can be problematic due to effects from plural scattering and the plasmon tail. By removing plural scattering and extracting the imaginary part of the dielectric function, ɛ2, from a Kramers–Kronig analysis, an improved background fit was achieved. This in turn allows for a more accurate signal integral of the iron M2,3 edge. Medium-resolution HAADF-STEM of stained sections was performed on an FEI Tecnai F20 operated at 200 kV using a 30 μm condenser aperture and a camera length of 150 mm.

In electron tomography, projections of the object are acquired from several different orientations and reconstructed into a three-dimensional volume. HAADF-STEM tomographic data sets were acquired over a tilt range of −70° to +70° using a step size of 2°. Three-dimensional reconstruction was carried out using the simultaneous iterative reconstruction technique (SIRT)25,26 using Inspect 3D image processing software. Reconstructions were visualized by a voltex projection using Amira 3D visualization software (Mercury Computer Systems, France).

Confocal microscopy

For confocal microscopy, HMMs were exposed to AgI@SWNT at a concentration of 5 μg ml−1 for 3 days and imaged using a Leica TCS SP2 confocal microscope, in reflectance mode. Three-dimensional image stacks were taken though the z-direction of the cell to confirm inclusion of AgI@SWNTs within the cell volume. Cells were mounted in Vectashield containing 4′,6-diamidino-2-phenylindole (DAPI, Vectalabs), a fluorescent stain that binds strongly to DNA.