Increasingly, there is demand for products made of nonpetroleum based materials with low-environmental impact and low-health risk to humans. Cellulose nanomaterials could be a building block for developing such materials (Moon et al., 2011). Cellulose (C₆H₁₀O₅)_n is one of the main molecules in the cell walls of trees and plants and is also formed by some fungi, bacteria, algae and marine tunicates. Cellulose is a biopolymer consisting of a linear chain of several hundred to thousands of glucose (C₆H₁₂O₆) molecules. Each glucose molecule contains a carboxyl functional group and a hydroxyl functional group which undergo nucleophilic addition to form a hemiacetal ring (Figure 1). During biosynthesis, adjacent glucose molecules bond with each other via 1,4-β-glycoside linkages (i.e. the hydroxyl group at carbon in position 1 of a molecule bonds with the carbon at position 4 of the next molecule) so that neighboring units are corkscrewed 180° with hydroxyl groups in the ring plane (Habibi et al., 2010; Moon et al., 2011). Hydrogen bonding of adjacent glucose molecules results in stacking of cellulose chains to form elementary fibrils which in turn form larger microfibrils, the latter which have diameters up to 50 nm and lengths of several micrometer (Moon et al., 2011). Within elementary fibrils, highly ordered repeating patterns of glucose molecules form regions of crystalline cellulose.

Cellulose nanomaterials are derived by extracting the crystalline regions of microfibrils. Depending upon the method of biosynthesis of the cellulose microfibrils (CMF) and the extraction process used to isolate the crystals, the type of cellulose obtained can be cellulose nanocrystals (CNC), cellulose nanofibrils (CNF) or CMF. Extraction of wood fibers, non-wood plant fibers and CMF by acid hydrolysis of amorphous domains yields whisker-shaped CNCs (Bai et al., 2009). Generally, CNCs possess a high degree of crystallinity (~54–88%) and high aspect ratios (Moon et al., 2011). Extraction of purified wood and non-wood plant fibers via mechanical (e.g. homogenizers to shear microfibrils) processing, often with chemical pre-treatment, yields high aspect ratio CNF (Moon et al., 2011; Saito et al., 2009). Mechanical refining of wood pulp or non-wood plant pulp yields high aspect ratio CMF having nanoscale width and micron-scale length. In addition to cellulose nanomaterials, purified wood pulp can also be pulverized and subjected to acid hydrolysis to form cellulose microcrystals (CMC). These fibers typically have widths of 10–50 μm and crystallinity of ~80–85% (Moon et al., 2011; Seehra et al., 2012). Cellulose nanomaterials possess high strength and directional rigidity, high surface area, high aspect ratio, low density and good thermal stability which make them ideal for use in advanced composites (Habibi et al., 2010; Lee et al., 2012; Ruiz et al., 2000), electronic storage devices (Hubbe et al., 2008), and as biocompatible and biodegradable scaffolds to direct tissue or bone growth (Future Markets Inc., 2012; Mathew et al., 2013; Sehaqui et al., 2011; Siqueira et al., 2013; Zoppe et al., 2009). CMC is used as a low-calorie substitute for fat and fillers in some foods and an excipient agent in pharmaceutical tablet drugs administered via ingestion (French et al., 2003).

Inhalation of CMC is associated with adverse effects in rats. Under overload conditions in the lungs, rats instilled with cellulose were observed to have transient inflammation (Cullen et al., 2000) and engulfment by alveolar macrophage cells followed by formation of granulomas (Muhle et al., 1997; Tatrai et al., 1995) and long-term retention (clearance half-time of >500 days). Understanding of potential adverse effects from exposure to cellulose nanomaterials is limited. Male et al. (2012) observed that CNC derived from flax, hemp and other biomass materials did not cause significant cytotoxicity in vitro. Vartiainen et al. (2011) reported that birch wood CMF did not induce release of inflammatory cytokines or cause significant in vitro cytotoxicity to mouse macrophages and human monocyte-derived macrophage cells. Studies by Pereira et al. (2013) and Alexandrescu et al. (2013) suggest that cotton- and wood-derived CNF do not cause significant cytotoxicity to fibroblast cells in vitro, though viability and altered expression of genes associated with cell stress are observed at high CNF concentrations.

With the rapid interest in commercial applications of cellulose nanomaterials and recent efforts to scale production from laboratory scale to pilot plant, there may be opportunity for inhalation exposure to workers. Vartiainen et al. (2011) evaluated friction grinding and spray drying of birch wood CMF using non-specific real-time particle counting instruments. During wet grinding of CMF suspensions, the concentration of airborne nanoparticles in workplace air was above background though total airborne particle number concentration was generally low. During spray drying of CMF suspensions, total particle number concentration in workplace air was a factor of three higher than background levels. Martinez et al. (2013) monitored exposure to airborne CNCs at a pilot-scale production facility. In that study, CNCs were tagged with cesium metal which allowed for their collection onto filters and subsequent analysis for cesium content using inductively coupled plasma-atomic emission spectroscopy. Cesium was detected on filter samples which indicated that CNCs were aerosolized during centrifugation of product slurry and subsequent removal of dry product from a freeze dryer. The morphology of airborne particles generated these tasks was not reported so it is unknown whether the cesium tag corresponded to dispersed nanoscale CNCs or aggregated micron-scale CNCs.

While cellulose nanomaterials hold great commercial promise, data are needed to understand their potential safety. The purpose of this study was to evaluate the lung biodurability and free radical generation capacity (as a precursor of inflammation responses) of CNC and CNF powders and compare data to a commercially available benchmark CMC material. The hypothesis of this research was that because of their higher surface areas, inhaled dispersed CNC and CNF materials will have different biodurability than CMC in artificial lung fluids.

Six materials were evaluated in this study: three CNC, two CNF and one CMC (used as a benchmark because it is made of cellulose and has high crystallinity but different physical dimensions than the test materials). CNC and CNF were chosen for this study as these are the most commercially produced types of cellulose nanomaterials (Future Markets Inc., 2012). Two private sector companies each donated a CNC material in powder form. One CNC and two CNF materials were donated by the US Department of Agriculture, Forest Products Laboratory. A CMC powder material was purchased on the open market (Product No. 8-01018; Carbomer Inc., San Diego, CA) for use as a benchmark. For purposes of our study, the CNC powders are designated A, B or C. The CNC-A powder was made by acid hydrolysis using sulfuric acid followed by spray drying and was in the Na⁺ salt form. The CNC-B powder was prepared by physical processing without sulfur and was in the Na⁺ salt form. The CNC-C (6.6% by weight suspension) was produced by acid hydrolysis using sulfuric acid followed by counter-ion exchange to yield a Na⁺ salt form. One CNF material (0.84% by weight suspension) was made by pre-treatment with 2,2,6,6-tetramethylpiperidine-1-oxyl radical (TEMPO) and herein referred to as CNF-T. The other CNF (0.5% by weight suspension) was made by mechanical homogenization (referred to as CNF-H).

Table 1 summarizes the physical characteristics of the as-received cellulose materials. Values of SSA for the as-received cellulose powders ranged from 0.6 to 1.4 m²/g. During manufacturing, CNC suspensions are dried for storage and transport which results in the formation of large aggregates via hydrogen bonding (Beck et al., 2012). Hence, because of their nanoscale physical dimensions, the individual CNCs are expected to have significantly higher SSA than measured for the aggregated material. Measured densities (1.6 g/cm³) match published values (Moon et al., 2011). Values of D_H ranged from 80 (CNC-C) to 417 nm (CNF-H). Generally, the zeta potential of cellulose materials ranged from −30 to −50 mV, except for CNF-H which had potential of −11 mV. A summary of the measured values of carbon, hydrogen, oxygen and sulfur content for the cellulose materials is provided in Appendix A.

Figure 2 is micrographs of the as-received materials. The benchmark CMC powder was elongated flat particles that had physical length <10 μm and diameter <2 μm (mean aspect ratio = 7; orientation-averaged aerodynamic diameter, D_ae = 3 ± 1 μm). CNC-A powder was aggregates of semi-spherical particles with average D_ae = 9 μm). CNC-B powder was a mixture of aggregates of semi-spherical particles and elongated flat particles; the mean AR of all particles was 2.5 (D_ae = 10 μm). The CNC-C was ellipse-shaped elongated particles with average diameters of ~10 nm (at widest cross-section) and average lengths of ~105 nm (aspect ratio = 12; orientation-averaged D_ae = 25 nm). CNF-T was relatively straight elongated particles with average diameter of ~11 nm and length of ~165 nm (aspect ratio = 14; orientation-averaged D_ae = 32 nm). An image that was clear enough to discern external dimensions could not be obtained for the CNF-H material.

Figure 3 shows the baseline powder XRD patterns of each of the study materials before exposure to artificial lung fluids. A reliable spectrum could not be obtained from the CNF-H material. Following the discussion in the article by Mansikkamaki et al. (2007), the structure of CNC-C corresponds to cellulose II, whereas the spectra for the other four samples correspond to cellulose I. For CNF-T, the lines are broader because of the smaller crystallite size. Both cellulose I and cellulose II have monoclinic structures but different unit cell parameters. An important signature of cellulose II is the strong Bragg line at 2θ ~ 12° which is absent in the cellulose I structure. Since the length of the c-axis is nearly the same in cellulose I and cellulose II, the (004) line is a common feature of the two structures. However, the Miller indices of the remaining lines are quite different as evident in the spectra.

The as-received CMC and cellulose nanomaterials had distinctly different physicochemical properties, including D_H and zeta potential (Table 1), morphology and size (Figure 2), x_CR and D (Appendix B and Figure 5), and T_p (Figure 6). Of particular interest are the crystalline properties (structure, x_CR and D) and values of T_p. Values of T_p were ~50 °C lower for the CNC and CNF relative to the benchmark CMC material. The lower values of T_p of the cellulose nanomaterials vis-a-vis CMC in TGA and their correlation with ESR intensity and D_H (Figure 5) are interpreted in terms of the surface area effect which increases inversely with the decrease in particle size. The atoms on the surface are weakly bound as compared to those in the bulk of a particle and they often have unsaturated bonds. These effects lead to lower T_p and higher ESR activity with decrease in particle size. The surface areas of the dispersed CNCs in artificial lung fluids are expected to be significantly higher than measured for the as-received aggregated powder material because of their smaller external dimensions.

Generally, it can be stated that both CMC and the cellulose nanomaterials investigated here were quite biodurable in SUF and PSF. All study materials except CNC-C were observed to have the cellulose I structure of cellulose; CNC-C was determined to have cellulose II structure. There was no observed change in crystal structure with exposure to either SUF for up to 7 days or PSF for up to 9 months; in essence, the cellulose I or cellulose II structure remained unchanged with exposure time. For the benchmark CMC material, there were no measurable changes in x_CR and D with exposure time to either SUF (Appendix B) or PSF (Figure 5). In SUF, there were no dramatic changes of x_CR or D (Appendix B) for cellulose nanomaterials, though a slight decrease in these parameters with increase in exposure time was indicated for CNC-A. In the acidic PSF, there appeared to be a slight increase in the magnitude of x_CR of CNCs with exposure time. Acid hydrolysis of CMF preferentially degrades amorphous domains to yield whisker-shaped CNCs (Bai et al., 2009). Hence, this observed increase in x_CR with exposure to PSF could be understandable in terms of the acidic nature of the fluid which could have dissolved some of the amorphous components of cellulose. In PSF, there were no systematic trends in changes of D with exposure time for any of the CNC samples.

The observed lung biodurability of our CMC is consistent with previous in vivo studies. Tatrai et al. (1995) reported that rats instilled with a single dose of CMC dust developed granulomas (macrophage cells that group tightly together to sequester and wall off foreign bodies to protect lung tissue) in their alveoli. Muhle et al. (1997) investigated the biodurability of two types of CMC fibers in rat lungs following intratracheal installation. The instilled fibers were phagocytized by alveolar macrophages and by 3 and 6 months, fiber-associated granulomas had formed in exposed rat lungs; fiber clearance half-times from the alveoli were on the order of hundreds of days to years. In our study, cellulose nanomaterials (CNCs and CNF-T) were observed to have biodurability similar to that of a benchmark CMC in both SUF and PSF. Upon inhalation, large particles such as the CMC and aggregated CNCs would deposit in the conducting airways and be immersed in airway lining fluid which we modeled using SUF. Given the prolonged biodurability observed in this study, the CMC and CNCs would not undergo rapid biodegradation in airway lining fluid; rather mechanical clearance via the mucocilliary escalator to the gastrointestinal tract would be dominant transport mechanism. Clearance via the mucocilliary escalator occurs on the order of several hours. In contrast, small particles (e.g. dispersed CNCs or CNFs) that are inhaled will deposit in the non-ciliated alveolar region of the lung where gas exchange occurs. Upon deposition in the alveoli, foreign bodies are rapidly phagocytized by macrophage cells (Lehnert & Morrow, 1985) and sequestered in vesicles called phagolysosomes wherein the cells attempt to degrade the foreign material in an acidic milieu. Results from this study indicate that cellulose nanomaterials that reach the lung alveoli and are engulfed by macrophages would persist for hundreds of days to years. The observed biodurability indicates that the acidic pH of the macrophage phaoglysosome is insufficient to degrade cellulose. Hence, in the alveoli, cellulose nanomaterials are likely to be cleared by mechanical movement of macrophage cells out of the alveoli and eventually to the mucocilliary escalator, a process that can take month to years. This comparison of existing literature on the in vivo toxicity of CMC fibers to our in vitro data suggests that the hazard potential of high aspect ratio cellulose nanomaterials could reasonably be expected to be at least equivalent to micron-scale cellulose fibers. No data is available on the long-term persistence of CNCs in vivo, though our in vitro data indicate that in a simulant of macrophage phagolysosomal fluid (PSF), CNCs are just as biodurable as CMC fibers. Given this similarity, it is likely that CNCs will have higher biodurability than some ceramic fibers (Muhle et al., 1997) but have less inflammatory potential than crocidolite asbestos (Cullen et al., 2000).

Prolonged retention of biopersistent, long, high aspect ratio particles in the lung is of interest for potential pulmonary inflammation toxicity. As such, we evaluated the potential for the benchmark CMC and the cellulose nanomaterials to generate ^●OH radicals, which are important for pulmonary inflammatory responses. Cullen et al. (2000) reported that rats exposed to CMC via inhalation exhibited an inflammatory response in the lungs after the first exposure; however, this response declined during the remainder of the 14-day exposure regimen. Vartiainen et al. (2011) reported that in vitro birch wood CMF did not induce secretion of inflammatory cytokines by mouse macrophages and human monocyte-derived macrophage cells. In our study, using a cell-free H₂O₂ reaction system the CNCs and CNF-T materials produced significantly more free radicals than that of the essentially inert CMC (Figure 6). Further studies are required to determine if these cellulose nanomaterials could be quenching some of their own radical production as seen in previous studies of other nanomaterials (Fenoglio et al., 2006; Krusic et al., 1991). Exposing particles to H₂O₂ directly models what would happen once the particles were engulfed by lung macrophages and is a first step in measuring particle potential for generation of reactive oxygen species (ROS). If the particle surface has transition metal present, such as Fe, then it may generate radicals through a Fenton-like reaction. Silica particles, chromium and welding fumes all generated free radical production under similar exposure conditions (Leonard et al., 2000, 2010; Vallyathan et al., 1999). In order for particles to generate free radicals in a cellular system, the cell must first recognize and react with the particles. Roser et al. (1998) reported that in vitro the phagocytosis of nanoparticles by U-937 monocyte macrophage-like cells and primary peritoneal mouse macrophages was increased as zeta potential became more negative or more positive and was minimal for nanoparticles with zeta potential near 0 mV. Our measured values of zeta potential for the cellulose materials (Table 1) spanned from −10 to −48 mV, which is favorable for phagocytosis. However, our short exposure time in the cellular experiment did not lead to measurable radical generation. The absence of a cellular reaction may indicate a slower reaction with the RAW cells or lack of cellular interaction. Hence, longer-term exposure times may be needed to elucidate whether cellulose nanomaterials can lead to engulfment and intracellular ^●OH radical production. Additionally, though outside the scope of our study, it may be useful to evaluate DNA fragmentation by COMET assay and cellular membrane damage by lipid peroxidation as measures of cellular ROS damage. Results from workplace monitoring indicate that cellulose nanomaterials can become airborne during production and handling (Martinez et al., 2013; Vartiainen et al., 2011). This evidence of inhalation exposure potential coupled with our data indicating that cellulose nanomaterials have prolonged lung biodurability and potential to generate ^●OH radicals suggests that in vivo studies could be helpful to more fully evaluate whether their inhalation presents an important health hazard.

A potentially significant contribution of this work to understanding potential health effects from inhalation of cellulose nanomaterials is the linkage of a specific particle property, T_p, to an observed toxicologic endpoint, ^●OH radical production. As shown in Figure 6(c), a direct correlation appears to be valid between the free radical density in the ESR response and the shift in the TGA-based values of T_p of the cellulose nanomaterials relative to that of the inert CMC. The underlying physical relationship between T_p and ^●OH radical production can be understood in terms of particle surface area as noted earlier because the surface area per unit volume increases inversely with particle diameter. Compared to the atoms in the interior of a nanoparticle, the atoms on the surface have fewer neighboring atoms to bind with resulting in lower cohesive energy of the surface atoms. Thus, the average cohesive energy per atom of a nanoparticle decreases with decreasing particle size (Qi & Wang, 2002) leading to size-dependent lower melting points of nanoscale solids (Sun et al., 2002). It is suggested that the lower magnitudes of T_p for the CNCs in Figure 6(a) are likely related to a similar surface effect. Values of D_H of the CNCs having cellulose I structure scale well with measured values of T_p (Figure 6d). For the CNC-C with cellulose II structure, T_p is higher than those for the other CNCs with cellulose I structure because the cellulose II structure is known to be somewhat more stable (Mansikkamaki et al., 2007).

The features of micron-scale fibers that dictate their pathogenicity as developed in the traditional fiber pathogenicity paradigm (narrow width, long length and prolonged durability) may need to be modified for high aspect ratio nanomaterials (Donaldson et al., 2011, 2013). For nanofibers such as CNCs or CNF, high aspect ratio may represent the biologically effective dose because their small aerodynamic diameter permits greater lung deposition in the alveoli. Even if a particle has length on the order of a few microns but still has a nanoscale diameter, it is a high aspect ratio nanomaterial that can be engulfed by macrophages. If the material is durable and not cleared effectively, the high aspect ratio nanomaterial may still be able to penetrate the lung lining and cause pathogenicity. Hence, quantitative data on nanoscale fiber biopersistence such as that provided herein is needed to augment the traditional fiber pathogenicity paradigm for nanomaterials, which in turn may help to develop high aspect ratio nanomaterials that are safe by design.

Powder XRD was observed to be a powerful tool for characterizing cellulosic materials, distinguishing between the different structures and for monitoring changes in x_CR and D with exposure time in SUF and PSF. A direct correlation was observed between free radical density and the shift in T_p of cellulose nanomaterials relative to that of the inert CMC, which indicates that TGA might be a useful technique for characterizing cellulose nanomaterials in toxicology studies. Measures of D_H and zeta potential were useful for monitoring agglomeration state and particle charge in suspension over time. These latter metrics are important considerations for future in vitro cellular studies of uptake and intracellular reactivity (Roser et al., 1998).

All cellulose materials were biodurable in SUF and PSF. The benchmark CMC material was essentially inert and there were no measurable changes in x_CR and D with exposure time to either SUF or PSF. In SUF, a slight decrease in values of x_CR and D were observed for CNC-A. In the acidic PSF, there was a slight increase in the magnitude of x_CR of CNCs with exposure time which is attributed to the acidic pH of this model lung fluid. Using a cell-free H₂O₂ reaction system, the CNCs and CNF-T materials produced significantly more ^●OH radicals than the benchmark CMC material. Hence, the hazard potential of high aspect ratio cellulose nanomaterials is expected to be at least equivalent to that reported for CMC fibers in vivo (Cullen et al., 2000; Muhle et al., 1997). An important outcome of this work was the observation that increased radical production was correlated with particle T_p and can be explained in terms of the particle surface area to volume ratio which increases as nanoparticle diameter decreases. With more atoms on the surface of a nanoparticle relative to the interior, the atoms on the surface have lower cohesive energy leading to size-dependent lower melting points. Our data suggest that cellulose nanomaterials may be biodurable in the human lung. Further studies are warranted to assess the potential of these nanomaterials to contribute to pulmonary inflammation.