The CoMoCAT method of SWNT synthesis involves a bimetallic Co-Mo catalyst supported on a silicon dioxide substrate [99-103]. The purification procedure includes removal of amorphous carbon by low-temperature oxidation, removal of the SiO₂ substrate with HF, and removal of metals by HCl. The final product, a SWNT-containing powder, is rinsed with deionized water until its pH is neutral [104]. Visible microscopic examination of the lot used in this work revealed that the fine, black, fluffy powder comprised irregularly shaped particles with dimensions ranging from 5–50 μm. SEM revealed that the majority of these particles comprised tightly entangled networks of SWNTs, similar to those observed by Resasco and co-workers [105], and that these networks comprised small bundles of SWNTs with 5–20 nm diameters (Figure 1).

TGA of the as-received SWNT-containing powder was performed to assess the powder's composition with respect to metals, SWNTs, and non-tubular carbon (NTC) species such as amorphous carbons, fullerenes, carbides, graphitic nanoparticles, etc. TGA measurements of the SWNT-containing powder were performed under the assumption that upon heating to 1000°C in O₂, all carbon and metals were converted to their corresponding oxides, and that the presence of other trace elements could introduce small errors to calculated metal contents [106]. Figure 2 shows the weight percent decrease as a function of temperature (red trace) and the first derivative of the weight percent curve (blue trace) for the as-received SWNT-containing powder. The identities of the components corresponding to the three main peaks observed in the derivative plot were determined in experiments whereby the residues in the TGA pan were recovered and analyzed by Raman spectroscopy and/or XPS before, during, and after peak onset. In brief, peak 'a' at ~410°C was determined to comprise SWNTs based on the appearance of a strong G-band – a Raman resonance uniquely associated with SWNTs. The oxidation temperature of the SWNTs ranged between 375–450°C and was consistent with the oxidation temperature of CoMoCAT SWNTs observed by Resasco and co-workers [105]. Peak 'b' at ~505°C was determined to comprise NTCs based on the disappearance of the G-band and an increase of the D-band – a Raman resonance uniquely associated with miscellaneous forms of disordered carbon. Peak 'c' at ~700°C, 9% weight loss, was determined to comprise MoO₃ by XPS and was supported by the ~700°C sublimation temperature of MoO₃. XPS experiments also ruled out the presence of residual SiO₂ in the as-received SWNT-containing powder. The remaining 5% mass at 1000°C (Figure 2, red trace) was considered to be oxidized metals of Co and Mo, most likely CoMoO₄ and MoO₂. In summary, the oxidized SWNT-containing powder was classified as comprising ~70% SWNTs, ~7% NTC, and ~14% oxidized metals.

SWNT dispersions were prepared using a sonication and centrifugation protocol and DMEM supplemented with 5% FBS (DMEM/FBS). The resulting DM-SWNTs were homogeneous in appearance and could be stored for 30 d at 4°C before any SWNTs were observed to precipitate. The final concentration of SWNTs in DMEM/FBS was estimated to be ~50 μg/mL (Additional File 1) and SWNT lengths were estimated to be 100–400 nm (Additional File 2). Figure 3 shows the absorption spectrum of a representative DM-SWNT dispersion. The observed spectral profiles of DM-SWNTs were similar to the spectra of CoMoCAT SWNTs dispersed in sodium dodecyl sulfate (SDS) as prepared by Resasco and co-workers [103] and Stupp and co-workers [41], where the two predominant semi-conducting SWNT types present were (6,5) and (7,5) tubes with an average diameter of 0.8 nm. Specifically, the DM-SWNT peak observed at ~569 nm corresponds to the S₂₂ optical transition of (6,5) tubes, the shoulder observed at ~587 nm corresponds to the S₂₂ optical transition of (8,4) tubes, the peak observed at ~652 nm corresponds to the S₂₂ optical transitions of (7,5) tubes at 644 nm and (7,6) tubes at 647 nm, the broad peak at ~1011 nm corresponds to S₁₁ optical transitions of (6,5) tubes at 975 nm and (7,5) tubes at 1025, and the peak at ~1120 nm corresponds to the S₁₁ optical transitions of (8,4) tubes at 1113 nm and (7,6) tubes at 1122 nm, which are all in accordance with spectroscopic assignments by Bachilo et al. [102]. In summary, the data indicate that CoMoCAT SWNTs dispersed in media supplemented with serum retain their optical transitions between van Hove singularities in the electronic density of states.

Confocal microRaman spectrometer acquisition methods and the interpretation of the Raman spectra of various SWNT dispersions prepared using our sonication and centrifugation protocol have been detailed previously [19,45,47,49]. A representative Raman spectrum for a DM-SWNT dispersion is shown in Figure 4 (blue spectrum; DMEM + 5% FBS). The spectrum clearly shows a number of well characterized SWNT resonances [100,107,108], in particular, two predominant radial breathing modes at ~281 and ~301 cm^-1, the D-band at ~1303 cm^-1, and the G-band in the 1550–1610 cm^-1 region. Control spectra of DMEM/FBS without SWNTs did not display detectable resonances under these operating conditions (data not shown). Spectrometer stability was assessed by monitoring the reproducibility of the G-band peak intensity at ~1590 cm^-1 since it is the most prominent Raman peak indicative of intrinsic SWNT features [109]. In brief, the relative standard deviation (RSD) of G-band peak intensities acquired from the same region of a SWNT dispersion was <1%, the RSD of G-band peak intensities acquired from four different regions of a SWNT dispersion was <10%, and the correlation coefficient for the linear relationship between the G-band peak intensity and relative SWNT concentration was 0.982 (Figures S3-S5 in Additional File 3). In summary, the data indicates that the FBS components coating the SWNTs did not significantly affect the G-band profile of SWNTs dispersed in this fashion, which is in agreement with previous reports using non-covalently modified SWNTs dispersed in aqueous solutions of peptides [19,41,45] and proteins [53,62].

While a variety of cell types have been cultured in pristine or functionalized CNTs solubilized in various growth media formulations [19,43,65-72], only a few of these reports have emphasized the important role that the added serum plays. The importance of FBS in dispersing SWNTs is evident in the series of Raman spectra shown in Figure 4. In brief, DMEM comprises inorganic salts, amino acids, buffers, vitamins, and minerals with the three major components being glucose (4500 mg/L), sodium bicarbonate (3700 mg/L), and sodium chloride (6400 mg/L). FBS is also a multi-component mixture comprising many low and high molecular weight substances. The major dissolved substances are proteins, lipids, steroid hormones, minerals, and metabolites. The most notable FBS components known to solubilize CNTs are BSA and phospholipids [10,53,58]. Using our sonication and centrifugation procedure, DMEM without FBS did not support SWNT dispersion as observed by the lack of detectable SWNT Raman resonances (Figure 4; gold spectrum). Conversely, aqueous 5% FBS solutions (without DMEM) were quite effective in dispersing SWNTs (Figure 4; red spectrum).

Previous elemental analyses of peptide-coated SWNT dispersions prepared using our sonication and centrifugation protocol revealed only trace amounts of metal catalysts even though the as-received HiPco SWNT-containing powder contained ~32% metals by weight [19]. Herein, we further characterize this protocol's ability to effectively remove toxic materials by analyzing CoMoCAT DM-SWNT dispersions. CoMoCAT SWNTs are made from a process that uses Co and Mo as catalysts rather than Fe, and thus, exposing CoMoCAT SWNTs to cells avoids the known cellular toxicity that Fe can impact to a CNT preparation [66,70]. The ICP-MS analyses of DM-SWNT dispersions revealed 6.64 ppm Mo and 1.55 ppm Co, and that the only Fe in our DM-SWNT dispersions was from the 0.10 ppm ferric nitrate in DMEM. For comparative purposes in the absence of EC50 values for dispersed SWNTs, the metal levels observed in the DM-SWNT dispersions were well below the 90 ppm EC50 of mammalian stem cells exposed to 30-nm MoO₃ particles (as determined by MTS assays) [110], and the 19 ppm EC50 of murine fibroblasts exposed to Co (as determined by MTT assays) [111]. Since >99% of the Mo and Co present in the as-received SWNT-containing powder was not detected in the DM-SWNT dispersions (relative to oxidized metal levels from the TGA of the SWNT-containing powder), these data again demonstrate that our sonication and centrifugation protocol is an effective method for removing the heavier metal-containing SWNTs and bundles. Such results are important to note since it has not been made clear in all previous published reports of cells being exposed to CNTs if such metal-removing measures were implemented before the CNT cytotoxicity was assessed.

DM-SWNTs were additionally analyzed for the presence of Ti since it is possible that this metal could be introduced through the use of Ti-coated probe sonicator tips. ICP-MS analyses of DM-SWNT dispersions prepared using a probe tip that had been used for >20 non-continuous hours revealed 0.15 ppm Ti. For comparison, this level is well below the 250 ppm EC50 of rat liver cells exposed to 40-nm TiO₂ particles (as determined by MTT assays) [112]. To our knowledge, this is the first report of such an analysis amongst the previous reports of cells exposed to SWNT dispersions prepared using probe tip sonication.

Herein, the presence of CoMoCAT G-band intensities emanating from inside living cells incubated in DM-SWNT dispersions was evaluated using confocal microRaman spectroscopy. In the first series of experiments, cells were incubated in DM-SWNT dispersions for 60 h at 37°C. A representative transmitted white-light image of a single HeLa cell acquired through the Raman microscope is shown in Figure 5. Typical HeLa cells were observed to possess 10–30-μm widths and 40–70-μm lengths. The relatively large dimensions of HeLa cells, coupled with the 4-μm lateral resolution of the confocal microscope system, permitted Raman spectra to be acquired from distinct cellular regions [19]. For example, Figure 5 shows Raman spectra acquired from a cell that was incubated in a DM-SWNT dispersion. Intense G-band signals were observed from both cytoplasmic (Figure 5A) and nuclear (Figure 5B) regions. In the latter case, it should not be implied that SWNTs are in the nucleus because the detected G-band resonances could emanate from SWNTs located in the perinuclear region and/or in the cytoplasm immediately above or below the nucleus. Finally, control cells incubated in DMEM/FBS (without DM-SWNTs) had no detectable SWNT Raman signatures under these conditions (data not shown), and no SWNT resonances were detected from cell-free regions of the dish adjacent (≤5 μm) to cells (Figures 5A and 5B, dark blue spectra).

If the intense G-band signals emanated from DM-SWNTs inside cells, most likely the result of an active uptake process such as endocytosis, then the signals should be absent in cells exposed to DM-SWNTs at 4°C where energy-dependent uptake practically ceases. Figure 6 shows representative Raman spectra acquired from HeLa cells incubated in a DM-SWNT dispersion at 4°C. The peaks detected at ~1608 and 1651 cm^-1 in the spectrum acquired from the cytoplasm are presumed to emanate from proteinaceous material, as denoted by the amide-I band at 1650–1659 cm^-1 [115-117]. More importantly, the G-band intensities at ~1590 cm^-1 recorded from cytoplasmic and nuclear regions were 99.9% less than those recorded from cells incubated at 37°C (Figure 5). In summary, the lack of detectable G-band signals from HeLa cells incubated in DM-SWNT dispersion at 4°C indicates that HeLa cells do not uptake detectable levels of DM-SWNTs when their metabolic activity is low. In addition, the lack of G-band signals from cells incubated at 4°C indicates that there was negligible nonspecific adherence of SWNTs to HeLa cells (i.e., the rinsing procedures were sufficient to remove DM-SWNTs that were on the exterior surface of the plasma membrane).

In another series of experiments, the time-dependence of DM-SWNT uptake was evaluated. First, the heterogeneous distribution of DM-SWNTs was taken into consideration. As shown in Figure 5, the G-band intensities detected from the cytoplasm ranged from 20 to 500 a.u. and those for the nuclear region ranged from 10 to 350 a.u. It was therefore decided to perform all time-dependent studies with the Raman laser focused on the center of a cell's nuclear region. This selection was influenced by our previous observations of SWNT accumulation around the nuclear region as revealed through confocal fluorescence imaging of HeLa cells exposed to SWNTs dispersed with a fluorescent-labeled peptide [118], and by Strano and co-workers through Raman spectral mapping of 3T3 cells exposed to SWNTs dispersed with DNA [9]. Figure 7 shows Raman spectra from HeLa cells that were incubated at 37°C in DM-SWNT dispersions for 12, 24, 36, 48, and 60 h. In all cases, the number of cells displaying detectable G-band signals increased as the DM-SWNT incubation time increased. Typically, the G-band intensities acquired from HeLa cells incubated in DM-SWNTs for 60 h was 90% greater than those detected at 12 h. Specifically, <10% of the cells analyzed after 12 h incubation displayed detectable G-band signals, while >90% of cells analyzed after 60 h displayed significant G-band signals (n = 40 cells analyzed). In summary, the combined Raman evidence indicated that the observed G-band intensities emanate from DM-SWNTs inside HeLa cells, and that the uptake of DM-SWNTs by HeLa cells is a time- and temperature-dependent process. While complete elucidation of the mechanism(s) of SWNT uptake by cells still requires further investigation, our results are consistent with the work of Dai and co-workers [12] and Cherukuri et al. [81] who have demonstrated that CNTs are transported inside cells via a temperature-dependent mechanism, and contrast the work of Bianco and co-workers who provide evidence that CNT uptake follows a temperature- and endocytosis-independent mechanism [14,37,43].

TEM was used to examine the intracellular distribution of DM-SWNTs. Figures 8 and 9 show electron micrographs of HeLa cells incubated at 37°C for 60 h in DMEM/FBS (no SWNTs) or DM-SWNT dispersions, respectively. Colored arrows are used to denote the nucleolus and nucleus, vacuoles/vesicles, Golgi bodies, and mitochondria. In addition, it is important to note that all micrographs shown in Figures 8 and 9 were acquired from cells sliced in the plane of the nucleolus, as denoted by the low-magnification micrograph shown in Figure 8A. The first observation from the comparison of control and DM-SWNT treated cells was the lack of any SWNT-like structures visible in or associated with Golgi bodies (compare Figure 8C with 9E) and mitochondria (compare Figures 8D,E with 9B). The most striking observations between control (n = 8) and DM-SWNT treated (n = 10) cells was the appearance of dense black aggregated material in the cytoplasmic vacuoles of the DM-SWNT treated cells (Figures 9A–D) that was not observed in control cell vacuoles (Figures 8D–F). In the highest magnification view of these material-filled vacuoles (Figure 9D), the observed material displays black features with 5–20 nm diameters and apparent lengths of 50–300 nm, which is similar to the dimensions of CoMoCAT SWNTs in our dispersions. Such observations are consistent with those of Dai and co-workers who used confocal fluorescence microscopy to image the co-localization of SWNTs coated with a dye conjugate of avidin and the fluorescent endocytosis marker FM 4–64 [12,13].

Conclusive evidence of SWNT-like structures in the nucleus was not observed (compare Figures 8A–E to Figures 9A,B,C,F). This is important to note since there is presently no consensus regarding the ability of SWNTs to enter the cellular nucleus or the mechanism for their entry. For example, data that SWNTs have crossed the nuclear membrane has been presented by Bianco and co-workers using TEM and 300–1000-nm long peptide functionalized multi-walled CNTs [15], and Lu et al. using radioactive labels and ~400-nm long RNA-modified SWNTs [18]. In contrast, Strano and co-workers used confocal Raman imaging to observe DNA-coated SWNTs in the perinuclear zone of 3T3 cells, but not in the nuclear envelope [9]. In summary, amongst reports presenting high-resolution TEM images of cultured cells and tissue exposed to CNTs [11,15,17,37,68-70,89,114], it is apparent that large multi-walled CNTs can be unmistakably identified in cells by visual observation. The situation is more difficult when cells have been exposed to SWNTs. In most cases, the purported SWNT material appears as a single, dense black mass of material and there are few structural features observable on-scale with the expected diameters of individual/bundled SWNTs (± coatings). In fact, when SWNTs have been observed to be densely internalized in cell vacuoles [69], there are no observable differences between those TEM images and TEM images of cells exposed to fullerenes, which also display vacuoles densely filled with black material [119]. Clearly, the development of complementary analyses capable of identifying SWNT and NTC species in such images is warranted.