Nanotoxicology and nanomedicine make extensive use of in vitro cellular assays that were developed prior to the nanotechnology era. The introduction of nanomaterials to these standard assays causes problems that are currently limiting progress in the field. Nanoparticles are often difficult to disperse; they can interfere with optical measurements through light absorption, and they can interact with dyes used as molecular probes of cellular integrity. In some cases the resulting artifacts can lead to gross misinterpretation of effects on cell viability and cytotoxicity. Because sp2-hybridized carbon materials are near-universal sorbents for organic compounds in aqueous phases, and in light of a recent report of favorable noncovalent interactions between small-aromatic-molecule therapeutic agents and single-walled carbon nanotubes (SWNTs), we hypothesize that SWNTs will adsorb a wide variety of small organic solutes from biological media, not limited to indicator dyes or their water-insoluble reduction products.
Herein, we use biochemical profiling techniques and UV/visible spectroscopy to show that SWNTs cause dose-dependent adsorption and depletion of over 14 amino acids and vitamins from RPMI cell culture medium. HepG2 cells cultured in these depleted media show significantly reduced viability, which can be largely restored by replenishment of folate. We have thus demonstrated a fundamentally new mechanism through which hydrophobic nanotube formulations influence cell behavior indirectly—that is, without physical contact between the nanotubes and cells. This mechanism, based on depletion of essential micronutrients, must be carefully considered when reporting and interpreting in vitro cytotoxicity assays.
Several recent studies have reported artifacts in nanotoxicology assays arising from interactions between nanomaterials and chemical or fluorescent probes. For example, Worle-Knirsch and co-workers reported that the common 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay for cell viability gives false indication of cytotoxicity when applied to carbon nanotubes (CNTs). In this assay, the MTT salt is reduced by mitochondrial dehydrogenases to the water-insoluble MTT formazan, which is then extracted and photometrically quantified at 550 nm as a measure of cell viability. Worle-Knirsch et al. reported that the MTT formazan crystals associate with SWNTs and are not effectively extracted through the standard protocol, thus leading to low optical absorbance (60% of control) and a false indication of low viability or high cytotoxicity. Casey et al. recently presented Raman spectroscopic evidence for nanotube interactions with cell culture medium and reported that nanotube–dye interactions make a number of standard assays unsuitable for quantitative assessment of cytotoxicity. They also suggested that these interactions could produce a secondary cytotoxicity through medium alteration.
The problems are not limited to CNTs, but have also been reported for carbon black used as a negative control in cytotoxicity assays. Monteiro et al. used four sources of carbon black as negative control samples in a study of human keratinocyte cytotoxicity. Conflicting results were obtained across all cytotoxicity endpoints due to problems associated with adsorption of viability indicators onto carbon black surfaces. The same group reported that cytokines IL-6, IL-8, IL-10, and TNF-alpha can also be adsorbed by water-soluble SWNTs at doses from 0.0005 to 0.5 mg mL−1, which may cause underreporting of cytokine release.
In each of these cases the underlying phenomena involve some interaction between nanomaterials and organic small-molecule solutes in multicomponent biological phases. While there has been great interest in the fundamental interaction of CNTs with biological macromolecules, less attention has been paid to the fundamental interactions of nanotubes with small biomolecules in complex physiological fluid phases (for example, see Roman et al. or Liu et al.). The objectives of the present study are therefore threefold: 1) to quantify the effect of CNT exposure on a wide range of small-molecule solutes in cell culture medium; 2) to identify the primary mechanism of the nanotube–biomolecule interaction; and 3) to characterize the effect of medium depletion on cell viability and its implications for nanotoxicity testing.
Our interest in nanotube–media interactions began during a study of metal catalyst residues in SWNTs with the observation that small quantities of nanotubes removed the color associated with the pH indicator dye, phenol red, in cell culture medium (also reported by Casey et al.). Such small-solute interactions can be conveniently studied by incubating purified SWNTs (see Figure 1) in medium and removing them completely by centrifugal ultrafiltration followed by spent medium biochemical analysis, UV/visible adsorption, and exposure to target cells in culture (see Experimental Section). Optical adsorption at 560 nm was used to quantify phenol red removal, which showed 90% depletion of the dye after addition of only 80 μg mL−1 of SWNTs to RPMI cell culture medium.
The effect on a broad spectrum of amino acids and vitamins is shown by the biochemical profiles in Figures 2 and and3.3. Some amino acids, including glycine (GLY), threonine (THR), and glutamic acid (GLU), are not significantly affected up to CNT doses of 10 mg mL−1, while others, including tyrosine (TYR), phenylalanine (PHE), and methionine (MET), are affected at 1 mg mL−1 and are largely removed at a SWNT dose of 10 mg mL−1. These doses are beyond the range used in most toxicity assays, but may be relevant to some nanomedicine applications in which cells are plated onto CNT scaffolds or nanostructured surfaces. The vitamin behavior is even more variable, and some vitamins are affected at the lower doses typical of toxicity assays, such as folate and riboflavin, which fall below the detection limit after only 0.1 and 0.01 mg mL−1 of CNT addition, respectively. The three compounds adsorbed most readily (folate, riboflavin, and thiamine) all contain planar conjugated units, which have a known high affinity for carbon surfaces driven by π–π interactions. The lower affinity for free amino acids is consistent with their structure, which lacks polyaromaticity and includes zwitterionic backbones at physiological pH values, thus causing end-group hydrophilicity that is an important factor in the solubility and adsorption of these low-molecular-weight biomolecules.
The trends within the amino acid set provide evidence for a physical driving force in the adsorption mechanism (Figure 4). The fractional amount adsorbed (ΔC/Co), where ΔC is the concentration change in solution upon adsorption and Co is the initial concentration, at a given CNT dose (Figure 4) shows a positive correlation with the hydrophobicity index of the side chain as calculated by Black and Mould, based on the hydrophobic fragmental constant approach of Rekker. Hydrophobic interactions are typically a significant driving force for the physical adsorption of organic molecules from solution onto carbon surfaces. There are important outliers in Figure 4, however, including leucine (LEU), isoleucine (ILE), and valine (VAL), which show an anomalous low adsorption for their hydrophobicity, and arginine (ARG) and histidine (HIS), which show anomalous high adsorption. Note that these outliers are consistent with two known factors in carbon surface science: 1) the higher affinity for planar aromatic or conjugated structures relative to aliphatics, and 2) electrostatic attraction between positively charged solutes and the negatively charged carbon surface (zeta potential −20.4 mV). The weak adsorbers (LEU, ILE, VAL) all have branched aliphatic side chains, which are a poor geometric match to carbon surfaces. Carbon surfaces have a known affinity for planar aromatic structures capable of π–π interactions with the graphene faces on carbon atoms. The strong adsorbers (ARG, HIS) are basic and carry a positive net charge at physiological pH, as well as having planar sp2-hybridized carbon structures in the side chain. The only other basic amino acid is lysine, but it also contains an aliphatic segment that makes it a hybrid of the strong and weak adsorbing subgroups.
The adsorptive depletion of small biomolecular solutes can be reproduced in simple, single-component experiments. Figure 5 shows the interaction of phenol red, tyrosine (an example amino acid), and folic acid (an example vitamin) in simple phosphate-buffered saline (PBS) using UV/visible spectroscopy. Phenol red adsorption is described by the adsorption isotherm in Figure 5a, which also shows the desorption branch determined by successive dilution, and the behavior of functionalized nanotubes with aryl sulfonate groups introduced by diazonium salt chemistry. The similarity of the adsorption and desorption results indicates reversibility, which is conclusive evidence of physical adsorption. Aryl sulfonation imparts a negative surface charge (reduction in zeta potential from −20.4 to −53.1 mV) and increased hydrophilicity, and its suppressive effect on adsorption is consistent with a physical adsorption process driven by hydrophobic interactions and opposed by electrostatic repulsion.
Figure 5b shows the equilibrium adsorption isotherm for the amino acid tyrosine in PBS buffer. Here, there is hysteresis between the adsorption and desorption branches that suggests partial irreversibility on the experimental timescale, which is not uncommon in carbon adsorption from solution due to chemisorption, surface polymerization, or strong multisite physical adsorption or π–π interaction. Figure 5c shows the phenol red adsorption isotherm from Figure 5b (single-component data) superimposed on the same isotherm determined in cell culture medium. The difference shows the effect of the other (>40) components present in the medium, which according to Figure 5c reduces the uptake of phenol red through competition for surface sites. Figure 5d shows equilibrium adsorption isotherms for folic acid in PBS buffer on as-produced and aryl sulfonated SWNTs. Folic acid is taken up rapidly at low doses on the as-produced SWNTs, but to a much lesser extent after aryl sulfonation.
Nanotube adsorption causes profound changes in the composition of the medium, and thus has the potential to influence in vitro cell behavior through an indirect mechanism that does not involve physical interaction between nanotubes and target cells. Figure 6a shows that a nanotube dose of only 10 ppm (10 μg mL−1) followed by nanotube removal leads to a highly significant inhibitory effect on the viability of HepG2 liver cells. After exposure for 72 h to the depleted medium, the 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium inner salt (MTS) assay reports only 50% viability relative to the unaltered medium.
We hypothesized that folate would be one of the most biologically significant of the depleted micronutrients (see Figure 7). Folate deficiency has deleterious effects on DNA metabolism and this deficiency is exacerbated by deficiency of other vitamins including riboflavin. In proliferating cells in culture, short-term folate deficiency is known to reduce cell proliferation, induce cell-cycle arrest, trigger apoptosis, and alter expression of micro-RNAs associated with stress responses in primary human lymphocytes as well as lymphoblastoid and hepatocyte cell lines. To test the specific role of folate depletion in our experiments, we supplemented a variety of the depleted media with external folate and observed reversal of the effect on cell viability (Figure 6b–d). Folate supplementation caused a statistically significant increase in MTS signal after 48 or 72 h in all cases, and accounted for the majority (but not all) of the loss observed by the original multicomponent depletion.
In summary, SWNTs have sufficient surface area to alter the micronutrient content of cell culture medium through adsorption of small-molecule solutes at CNT doses as low as 0.01 mg mL−1. Because physical adsorption is only weakly specific, this is a general phenomenon that influences a broad spectrum of small-molecule solutes including amino acids, vitamins, and indicator/probe dyes. The extent of adsorption is greatest for solutes at low initial concentrations and with a molecular structure favoring adsorption on graphenic carbon—hydrophobicity, planarity/sp2 hydridization for π–π interactions, and positive charge (which is opposite to that of most carbon surfaces at neutral pH). Functionalization of the SWNTs with aryl sulfonate groups significantly suppressed adsorption in both cases studied here (phenol red, folic acid), which suggests the dominance of hydrophobic driving forces. The depletion of folate, as well as other essential micro-nutrients at doses of 0.01–0.1 mg mL−1, significantly reduced HepG2 cell viability by an indirect mechanism that does not involve physical nanotube–cell interaction. An important implication of this finding is that CNT toxicity, as assessed by commonly used endpoints, such as cell viability, DNA damage, and apoptosis, may be mistakenly attributed to direct toxicity when in fact it is a secondary effect of adsorption of essential micronutrients from cell culture medium. Based on our results, this effect is expected to be much more pronounced for hydrophobic nanotube formulations (as-produced, annealed, or with hydrophobic coatings) than with tubes functionalized for water solubility. In short-term toxicity assays, these endpoints are potentially reversible by repletion of folate, and cells may adapt to folate depletion by upregulating expression of genes related to folate metabolism. It is unknown at this time whether folate depletion or altered folate uptake and metabolism would be produced by in vivo exposure to CNTs. However, chronic folate depletion in experimental animals and in humans has been associated with developmental anomalies, particularly neural tube defects, and an increased risk of cardiovascular disease and cancer.