Development of high throughput, high precision synthesis platforms and characterization methodologies for toxicological studies of nanocellulose

Abstract

Cellulose is one of the most abundant natural polymers, is readily available, biodegradable, and inexpensive. Recently, interest is growing around nanoscale cellulose due to the sustainability of these materials, the novel properties, and the overall low environmental impact. The rapid expansion of nanocellulose uses in various applications makes the study of the toxicological properties of these materials of great importance to public health regulators. However, most of the current toxicological studies are highly conflicting, inconclusive, and contradictory. The major reasons for these discrepancies are the lack of standardized methods to produce industry-relevant reference nanocellulose and relevant characterization that will expand beyond the traditional cellulose characterization for applications. In order to address these issues, industry-relevant synthesis platforms were developed to produce nanocellulose of controlled properties that can be used as reference materials in toxicological studies. Herein, two types of nanocellulose were synthesized, cellulose nanofibrils (CNF) and cellulose nanocrystals (CNC) using the friction grinding platform and an acid hydrolysis approach respectively. The nanocellulose structures were characterized extensively regarding their physicochemical properties, including testing for endotoxins and bacteria contamination.

1. Introduction

Cellulose is the most abundant natural polymer on earth and the structural features of most plants are comprised of it. This renewable and sustainable material has been used for many years in applications such as construction, packaging, paper, and textiles. Recently, interest is growing around cellulose nanomaterials due to the sustainability of these materials, the novel properties that can be controlled, and the overall environmental impact (Shatkin and Kim 2015). The nomenclature of these materials is still developing and past publications have referred to these materials in various ways (Moon et al. 2011; Dufresne 2012). Currently, cellulose nanomaterials are classified into three major groups based on the synthesis approach: 1) cellulose nanocrystals (CNC), produced by acid hydrolysis, 2) cellulose nanofibrils (CNF), produced by mechanical action such as homogenizers, grinders or refiners, and 3) bacterial cellulose (BC) produced by fermentation of glucose by bacterial such as Gluconacetobacter xylinus. Moon et al. have reviewed the various forms of cellulose nanomaterials and their properties in detail (2011).

These materials have already found some practical applications (Posteck et al. 2013), including novel packaging systems that reduce oxygen permeability (Nair et al. 2014; Padberg et al. 2016), polymer reinforcement (Siró and Plackett 2010; Abdul Khalil et al. 2012), rheology modifiers (Liu et al. 2017), hydrogels and aerogels (De France et al. 2017), thin films (Wei et al. 2014), coatings for food preservation (Zhao et al. 2016), additive as a lipolysis retardant (Torcello-Gómez and Foster 2016), food packaging (Paunonen 2013; Lavoine et al. 2014), electro-conductive substrates (Shi et al.2013), and as a binder in structural systems (Tajvidi et al. 2016). More applications are on the horizon due to the low production cost, the high volume of production and the versatility of the nanoscale cellulose materials that can be formed into films and fibers.

The rapid expansion of nanocellulose uses in various applications makes the assessment of potential toxicological properties of these materials very important for public health regulators (Ong et al. 2017). Given the increasing literature on potential health effects of ingested engineered nanomaterials (ENMs) (McClements et al. 2016; Guo et al. 2017; Nallanthighal et al. 2017; Schoepf et al. 2017; Yao et al. 2017; Chen et al. 2017), there is a critical need to understand whether these cellulosic materials pose certain health risks or if they can be generally regarded as safe (GRAS), the same way the micro and macro scale cellulose products are regarded (International Food Information Council (IFIC)U.S. Food and Drug Administration (FDA) 2004). Such toxicological studies need a proper and detailed characterization of the physical and chemical properties of pristine cellulose ENMs, but should also include the potential property transformations and interactions in biological media (McClements et al. 2016; Deloid et al. 2017b).

Currently, only a limited number of toxicological studies for cellulose nanomaterials exist and they display conflicting results. While most of these studies found that nanocellulose materials are nontoxic, other studies indicate potential adverse health effects (Roman 2015; Camarero-Espinosa et al. 2016). Since inhalation is an important route for exposures, occurring mainly at the occupational level, a large number of these studies were focused on pulmonary toxicity (Endes et al. 2016). Some of these in-vivo studies found elevated toxicity markers like cellular toxicity, tissue damage, inflammatory effects, systemic immune response, reproductive alterations, or genotoxicity following inhalation of different CNCs or CNFs (O’Connor et al. 2014; Yanamala et al. 2014; Shvedova et al. 2016; Farcas et al. 2016). Some in vitro studies showed potential adverse effects (Clift et al. 2011; Hannukainen et al. 2012; Catalán et al. 2015). More specifically, Clift et al. observed proinflammatory and cytotoxic responses of CNCs using a 3D in vitro model of human lung epithelial tissue (Clift et al. 2011). Others also reported cytotoxicity and genotoxicity using BEAS-2B cells exposed to high doses of CNCs and CNFs (Hannukainen et al. 2012; Catalán et al. 2015). Other similar in-vivo and in-vitro pulmonary studies did not show any toxic effect (O’Connor et al. 2014; Shatkin and Kim 2015; Endes et al. 2016).

In addition to potential inhalation exposures to nanocellulose, oral and dermal routes are also inevitable due to the large-scale use of nanocellulose in the food and cosmetic industry. However, data for these important routes of exposure are fragmentary. Dermal studies that focused on skin-sensitizing, corrosion, and irritating potency were carried out following Organization for Economic Co-operation (OECD) guidelines where adverse health effects were reported (O’Connor et al. 2014). Regarding nanocellulose oral exposure, two in vivo acute animal studies have been performed using rats where CNCs were administered by oral gavage. No adverse health effects were observed after gross necropsy (O’Connor et al. 2014). In summary, the existing literature on the toxicological implications of various forms of nanocellulose is flooded with contradictory results, making it difficult for regulators to assess potential health risks (Hanif et al. 2014).

There are many reasons for these inconsistencies and conflicting results. The two most important reasons are i) the lack of a standard reproducible reference nanocellulose materials; and ii) comprehensive and detailed characterization of nanocellulose materials due to the lack of standardized characterization methodologies. For example, in the cases where mechanical action is used to synthesize CNF, there are variations in the equipment used, such as refiners, homogenizers, or fine friction grinders (Siró and Plackett 2010; Siqueira et al. 2010; Abdul Khalil et al. 2012), which may result in significant variations in the properties of nanocellulose. Further, the mechanical action may be performed at different temperatures or the fibrils may undergo chemical pre-treatment (Missoum et al. 2013). The most common chemical modification of cellulose prior to nanocellulose fabrication is the TEMPO (2,2,6,6-tetramethylpiperidine-1-oxyl radical) (Saito and Isogai 2004; Saito et al. 2006a; Saito et al. 2006b; Saito et al. 2007) that has a double effect on the nanocellulose: i) reduces the energy required for milling by a factor of a ~100 (Eichhorn et al. 2009), and ii) prepares the surface for other modifications post grinding (Missoum et al. 2013). Other less widely used methods for chemical modification pre-fabrication include enzymatic modification (Henriksson et al. 2007; Endes et al. 2016), carboxymethylation (Aulin et al. 2010) and acetylation (Okahisa et al. 2009). For acid hydrolysis, the type of acid used, the duration, the temperature, and the purification steps will result in different properties (Wang et al. 2012; Chen et al. 2015). Even if all process parameters are standardized, there may be other variations that come from the cellulose source, such as species, storage, and preparation steps that could lead to variable properties of materials. For instance, wood (Nyström et al. 2010), cotton (Morais et al. 2013), rice husk and straws (Ludueña et al. 2011) were routinely used in commercial refiners and this might result in nanocellulose with source-to-source and batch-to-batch variability in terms of properties.

Furthermore, the complex nature of the nanocellulose ENMs makes the characterization a very challenging and daunting task. CNFs have a complicated morphology where fibrils are entangled, resembling web-like structures, or bundled with others, forming long rope-like fibers. The diameter of these fibers is likely related to the fibrils that are known to wrap the cell wall in wood fibers and are in the range of 10–100 nm. During the CNCs synthesis, the acid hydrolysis removes most of the amorphous cellulose leaving the crystalline form of cellulose (Siqueira et al. 2010). CNCs have the morphology of rice with a fairly uniform diameter and length depending on the fiber source (Habibi et al. 2010). Typical dimensions are 3–30 nm wide by 50–500 nm in length (Wang et al. 2012; Chen et al. 2015). The unique shape of the CNC can induce self-assembly behavior that can lead to novel structures and innovative applications (Lam et al. 2012a), including a number of biomedical type applications (Lam et al. 2012b).

Due to their nanoscale size and complex morphologies, traditional methods for characterizing fibrils are not suitable (Robertson et al. 1999). For example, light scattering methods are not reliable because of the high aspect ratio of the particles and the often entangled nature of the fibrils (Dufresne 2012). Further fiber analyzer tools, used extensively to characterize fiber length, relying on polarized light, are not suitable for fibrils with a diameter less than the utilized wavelength (Robertson et al. 1999). Although several attempts have been made to characterize nanocellulose intrinsic properties (Salas et al. 2014) using AFM (Usov et al. 2015), TEM (Wang et al. 2012), and SEM (Chen et al. 2015), to this day, there are no standardized methodologies. In particular, CNFs have a complicated structure with the formation of bundles and tangled fibrils that makes measuring the length a challenge. Similarly, although CNCs are simpler in terms of physical characterization, there are still significant challenges due to possible complex surface chemistry. Furthermore, it is not clear what properties are considered essential as part of their characterization, and, in particular, which properties are essential to their toxicological implications. Therefore, it is important to develop universal methodologies and protocols for the characterization of the cellulose structures in particular to correlate these properties to biological responses.

Considering both the source-to-source and batch-to-batch variability of properties of commercially available nanocellulose that can be procured and used in nanotoxicology studies and the need for proper characterization, it has become apparent that there is a need to develop industry-relevant synthesis methods to synthesize nanocellulose materials with controlled properties (i.e, size, diameter, surface chemistry, etc) in a highly precise and reproducible manner, which can then be used as reference materials in toxicological studies. Here, we report the development of synthesis platforms and detailed characterization of reference nanocellulose structures of chemically high purity with tunable intrinsic properties suitable for nanotoxicology research. The CNF material was produced with a lab-scale friction grinding setup that mimics the most commonly used method for CNF fabrication (Vartiainen et al. 2011). The CNCs were produced with a lab-scale acid hydrolysis reactor similar to what is mainly used in the large-scale production of CNC (George and S N 2015) Further, the produced ENMs were characterized to satisfy all the criteria required for ENMs to be used as reference ENMs (Beltran Huarac et al. 2018; Cohen et al. 2018). Such ENMs were developed as part of the Harvard-National Institute for Environmental Health Sciences (NIEHS) Reference ENM Repository established at the Harvard T. H. Chan School of Public Health. The reference ENMs have been utilized for nanotoxicology research as part of the Nanotechnology Health Implications Research (NHIR) consortium established by NIEHS in 2016. Here, the lab-scale reactors and methods used to generate CNCs and CNFs and developed methodologies to characterize the most important intrinsic properties such as diameter, length, structure complexity, density, and surface chemistry are reported. In addition, the sterility and endotoxin levels of generated ENMs were measured to ensure that there are no biologicals that can interfere with the biological assessment of nanocellulose.

2. Materials and Methods

The raw material for the synthesis of cellulose ENMs was a softwood bleached kraft fiber that was supplied in the form of dried sheets (St. Felicien Mill, Canada). These fibers were shipped and stored in dry form as standard pulp sheets. The material has been kept in isolation and all batches, current and future, will be produced from this particular batch of raw material.

2.1. Nanocellulose synthesis platforms

2.1.1. Cellulose nanofibrils

The synthesis platform for the production of cellulose nanofibrils uses an ultra-fine friction grinder manufactured by Masuko Sangyo Co (Kawaguchi, Japan). The grinding system includes two ceramic nonporous grinder stones, each designed with a series of grooves or channels (Masuko Sangyo) is depicted in Figure 1a. The clearance distance or vertical gap between the two grinders is adjustable with a rotating handle and moves in 10 μm increments. The batch mode is used in this production. This grinding method is similar to what was described by Taniguchi and Okamura (Taniguchi and Okamura 1998). While a loop system is available, the equipment is operated in a batch mode to make sure that all of the material is forced through the grinding region; a continuous loop may have some material that becomes stagnant in the collection tank and not circulated.

Figure 1:

Experimental setup for the synthesis of the cellulose nanomaterials. a) friction grinder for the synthesis of CNFs. b) Acid hydrolysis setup used for the synthesis of the CNCs.

The moisture content of the fiber sheets was determined by drying a sample on a hot plate at 177°C for two minutes, weighing the sample before and after drying. The amount of fiber sheet (dry basis) and the amount of reverse osmosis (RO) water needed to make the pulp suspension for 2.5% solids were combined and soaked for 10 minutes. The fiber sheets along with the RO water were then dispersed using a disintegrator (Noremac, Labtech Inc., Hopkinton, MA), which consists of an agitator with a variable rotation setting. The disintegrator was set for 10,000 revolutions.

The grinder was operated in batch mode at 1250 RPM. The zero clearance position was determined by rotating the top stone and adjusting the gap until a faint noise is heard; this is the position where the stones just touch with no sample in the grinder. The clearance distance between the grinding stones was adjusted to 100 μm for the first pass, to break up any large flocs of material. The fiber suspension was fed to the device in a batch mode, using plastic buckets to collect and feed the material. The gap was reduced to 80 μm, 50 μm, 30 μm, and finally 0 μm for the first series of passes of the material through the device. The true gap was not actually zero because the presence of material between the stones forces the stones apart. The gap setting was adjusted to −50 μm, −100 μm, and − 150 μm for the next three passes. Again, this gap setting does not represent the actual gap but is an indication of the force applied to the stones. The pulp suspension was forced into the gap between the two grinding stones by centrifugal force. The fibers were ground into fine particles by shearing and friction forces. The resulting material was collected and fed back to the hopper for additional passes through the grinder. As the CNF suspension increases in viscosity, the material does not flow through the gap easily. When the flow through the grinder slows, the gap was adjusted to −80 μm for the remaining passes. These gap setting may vary due to the grinding stone wear and even for the same device, so the suspension needs to be tested after each pass through the grinder.

After each pass, 40 g of the suspension was collected to measure in terms of fiber length with a fiber analysis device (Morfi Compact, Techpap Inc. Gières, France). From the collected sample, 20 g of 2.5% material is diluted with 1 L of water. From this suspension, 100 ml is removed and diluted with another 900 ml of water. This liter of sample is used as the feed for the fiber analysis. The fiber lengths are determined by image analysis as the material flows through a fine capillary. Fibers less than 200 μm in length are considered “fines.”

Two different diameters of cellulose were synthesized, 50 nm and 80 nm. For the 50-nm CNFs (CNF- 50), the process is complete when the sample has over 93% content by weight, less than 200 μm in length. This may require 30–40 passes through the grinder to obtain this result. For the 80-nm CNF (CNF-80), the process is complete when it reaches 82–85% fines content by weight, which typically requires 10–20 passes. It should be noted that the fiber length analysis done here is a quick way to measure the degree of grinding; most of the material is below the normal detection level. For the raw material used here, this correlation of fines content and the CNF average diameter seems to hold, but other fiber sources may have different relationships.

After the final pass, the material was transferred and placed in an autoclave (HY-85, Hirayama) at 121°C for one hour. The samples were then aliquot to 500 ml clean, sterile hHigh-Ddensity-pPoly-Eethylene (HDPE) bottles.

To verify the consistency of the dispersed pulp suspension, approximately 30 g of the pulp suspension was dried on a hot plate, and the consistency was calculated from the difference between the wet and dry weight of the sample.

2.1.2. Cellulose nanocrystals synthesis platform

Cellulose nanocrystals (CNC) were produced using acid hydrolysis. The general acid hydrolysis method follows the descriptions as given by Wang et al. (2012) and Chen et al. (2015). Sulfuric acid at 72% (w/w) from Ricca Chemical Company was used in our platform. Sodium hydroxide at 4 wt% was prepared from NaOH pellets from Fisher Chemical. All acid hydrolysis experiments were conducted at a ratio of 1:10 (g/ml) of dried fiber to the acid solution.

Pulp sheets were cut into approximately 1 cm × 1 cm and placed in a glass beaker. In a separate 500-ml beaker, the sulfuric acid was diluted with RO water to a final concentration of 60 wt% and was heated to 55°C in a water bath. The complete setup is shown in Figure 1b. The pulp sheets were then added to the temperature-controlled sulfuric acid solution and were dissolved for 102 minutes. The pulp suspension was stirred using a glass rod during hydrolysis. At the end of the hydrolysis, the CNC suspension was transferred in equal parts to two 2000-ml beakers. The reaction was quenched by diluting with deionized water at 8:1 ratio (water:acid) and stirred for about ten minutes. The pulp suspension was then neutralized by the slow addition of 4 wt% NaOH and checked with pH measurements.

The CNC suspension was allowed to settle for approximately one hour, after which the top phase was decanted off. The CNC suspension was further diluted with deionized water and then centrifuged (accSpin 400, Fisher Scientific) at 4000 rpm for ten minutes. At the end of the first centrifuge cycle, the supernatant was decanted. More deionized water was added to disperse the CNC and the material was centrifuged again. This washing/centrifuge procedure was repeated for ten cycles.

The solids content of the CNC suspension was measured with a solids balance, and the material was transferred and placed in an autoclave (HY-85, Hirayama) at 121°C for one hour. The samples were then transferred to cleaned and sterile HDPE bottles.

2.2. Nanocellulose Characterization

2.2.1. Physical and morphological characterization of CNF

The dimensions of the produced cellulose nanofibrils were analyzed both with Transmission Electron Microscopy (TEM) and Scanning Electron Microscopy (SEM). TEM is used to more precisely measure the diameter (smaller dimension) and SEM is used to more precisely measure the length (longer dimension).

Transmission Eelectron Mmicroscopy:

The CNF suspension was diluted down to a final concentration of 0.875 mg/ml by adding the appropriate amount of DI water. The suspension was then placed on a magnetic stirrer at medium speed (150 rpm) for 15 hours to untangle the CNF fibrils. Following, one drop of the suspension was dropped on a lacey carbon TEM grid (#01829, TedPellaInc, Redding, CA) and was left there for 10 minutes. The excess was removed with the filter paper (Whatman, Maidstone, UK). The grid was left in the air to dry.

For the imaging, the JEOL 2100 (Jeol, Tokyo, Japan) was used. The accelerating voltage was set to 80 keV to avoid damaging the sample. Images were taken at different magnifications. The images were analyzed with ImageJ (NIH) software. The diameter was measured for each fiber at a minimum of five points across the length of the fiber. The length of the fiber cannot be measured with TEM since the field of view cannot fit an entire single image. A minimum of 100 fibrils was analyzed.

Scanning Electron Microscopy:

The CNF suspension was diluted down to a final concentration of 0.875 mg/ml by adding the appropriate amount of water. As before, the suspension was placed on a magnetic stirrer at medium speed for 15 hours to untangle the CNF fibrils. For SEM imaging, a mica substrate (Ted Pella Inc, Redding, CA) modified with poly-L-lysine, was used. The poly-L-lysine was employed to enhance the fibril adhesion. To prepare the substrate, a freshly cleaved mica was placed on a glass slide, and the surface was covered by an aqueous poly-L-lysine solution at 0.1% (w/v) (#18026, Ted Pella Inc, Redding, CA). After five minutes, the substrate was dried with compressed air and the surface was covered by the CNF suspension. After five minutes, the excess suspension was shaken off and it was rinsed with DI water. The mica substrate was then fixed on the SEM stub and was sputter coated with an 80:20 Pt/Pd to a thickness of 1 nm.

For the imaging, the Zeiss Supra55VP was used. The accelerating voltage was set to 5 keV. Images were taken at different magnifications. The images were analyzed with ImageJ (NIH) software. The diameter was measured for each fiber at a minimum of five points across the length of the fiber. The length was traced from end to end for each fibril. Only fibrils with distinct and defined outlines were measured. A minimum of 100 fibrils was analyzed.

For cryo SEM imaging, a Zeiss Nvision FIB (Zeiss, Oberkochen, Germany) was used. The SEM is attached to a Leica EM VCT contamination-free cryo-transfer system, which cross-links preparation units with various analysis systems via a transfer shuttle connected to a load-lock. The hydrated suspension was frozen in Liquid Nitrogen slush, freeze fractured to expose a fresh surface, followed by sublimation at −90C for 5 minutes, followed by metal coating (Pt/Pd), then transferred frozen SEM with vacuum transfer system. Samples were imaged at 2kV to show the surface morphology.

2.2.1. Physical and morphological characterization of CNC

Although typical CNCs have small dimensions that can be measured with the TEM, TEM tends to bias the length towards smaller sizes, due to the limited number of larger crystals that can be seen in each image. Therefore, the dimensions of the produced cellulose nanocrystals were analyzed both with TEM and SEM. TEM is used to more precisely measure the diameter (smaller dimension) while the SEM is used to more precisely measure the length (longer dimension).

Transmission Electron Microscopy:

Prior to imaging, the CNCs were stained with Uranyl Acetate to enhance contrast under the TEM. The CNCs suspension was diluted down to a final concentration of 0.1 mg/ml by adding the appropriate amount of water. Uranyl acetate was added to the suspension to a final concentration of 1% (w/v). The suspension was then placed on a magnetic stirrer at medium speed for 15 hours to separate the CNCs and allow time for the uranyl acetate to absorb. The suspension was centrifuged at 3000 rpm for 10 minutes. The supernatant was removed and replaced with an equal amount of DI water. The suspension was vortexed for 1 minute. The process was repeated three times. Following, one drop of the suspension was dropped on the TEM grid and was left there for 10 minutes. The excess amount was removed with the filter paper (Whatman, Maidstone, UK). The grids were left to air dry.

For the imaging, the JEOL 2100 (Jeol, Tokyo, Japan) was used. The accelerating voltage was set at 120 keV. Images taken at different magnifications were analyzed with ImageJ (NIH) software. The crystals’ perimeter was traced, and the Minimum Feret Diameter (MFD) was used as the diameter and the Feret Diameter (FD) was used as the length. For each crystal, the aspect ratio was defined as the ratio of the FD to MFD. A minimum of 100 fibrils was analyzed.

Scanning electron microscopy:

The process for imaging with the SEM was the same as the one used for the fibrils. For the image analysis, the same procedure as the TEM (described above) was used.

Density:

The CNFs and the CNCs were freeze-dried with a Labconco Lyophilizer (Labconco, Kansas City, MO) for 24 hours. The dried matter was ground with mortar and pestle for homogenization it and used for the density measurement. Based on the expected density (assumed close to the one of the bulk material), a mass of 100–400 mg was used to fill the holding cell between ½ and ¾ high. The mass was measured with a Mettler Toledo TLE104E scale (Columbus, OH) with 0.1 mg accuracy. The measurement was repeated three times. The Ultrapyc 1200e (Quantachrome, Boynton Beach, FL) was used to measure the material density. The samples were analyzed a total of 20 times and the average was used as the value of the skeletal density. The average of the volume was used as the material volume and the standard deviation as the error.

Specific Surface Area:

Brunauer-Emmett-Teller (BET) N2-adsorption at 77K was conducted to determine the powder-specific surface area (SSA) of the ENMs. The CNFs and the CNCs were freeze-dried with a Labconco Lyophilizer (Labconco, Kansas City, MO) for 24 hours. The dried fibrils were ground with mortar and pestle for homogenization and used for the SSA measurement. Approximately 500 mg of nanocellulose was vacuum degassed at 40 °C for >3 h. The surface area was measured with a high-throughput surface area and pore size analyzer (Quantachrome Instruments, NOVAtouch LX⁴) following a 5-point BET isotherm method.

The surface area can also be calculated based on the size distribution (diameter and length) of the ENMs measured by the TEM, as follows:

$$\text{SSA}\left[{\text{m}}^2/\text{g}\right]=\frac{{\sum }_{\text{i}=1}^{\text{n}}\left(\text{π}\cdot {\text{d}}_{\mathrm{i}}[\text{nm}]\cdot {\text{l}}_{\mathrm{i}}[\text{nm}]+\frac{\text{π}}{4}\cdot {\text{d}}_{\mathrm{i}}^2[\text{nm}]\right)}{\text{ρ}\left[\frac{\text{g}}{{\text{cm}}^3}\right]{\sum }_{\text{i}=1}^{\text{n}}\frac{\text{π}}{4}\cdot {\text{d}}_{\mathrm{i}}^2[\text{nm}]\cdot {\text{l}}_{\mathrm{i}}[\text{nm}]}\times {10}^3$$

where di and li is the diameter and length respectively, and ρ is the density.

2.2.2. Chemical and biological characterization

Inductively coupled plasma mass spectrometry (ICP-MS):

The suspension of fibrils was homogenized before use with a TisueLysser ball mill (Quangen, Venlo, Netherlands) with Teflon coated stainless steel balls to ensure sample uniformity during the trace metals analysis. Evaluation of elemental composition of the ENMs was performed following a protocol previously described by Herner et al. (2006). One mg of the ENM powder was placed in a Teflon pressure vessel with a mixture made from 1.5 mL of 16 N HNO3, 0.5mL of 28 N HF, and 0.2 mL of 12 N HCl. The samples were then heated in a microwave oven to 180 °C in 9 min, followed by a 10-minute hold at that temperature, and 1 h of ventilation/cooling. After cooling, digests were diluted to 30 mL with high purity water prior to ICPMS analysis. This digestion technique required adequate safety measures when working with HF at high temperature and pressure.

The digestion vessels also required rigorous cleaning between samples to avoid “carryover” contamination. All vessels were pre-cleaned using an acid solution consisting of 2.4N hydrochloric acid for two days, then 3.2 N HNO3 for two more days, and lastly, rinsed with high purity water. In addition to the collected samples, sample spikes, sample duplicates, blanks, standards and certified reference materials (NIST 2709, NIST 1648a, NIST 2556, NIST 2702) were also used in the chemical testing.

The digested sample and extracts underwent ICP-MS analysis (Thermo-Finnigan Element 2), which has very low detection limits and can resolve very small fractional mass-to-charge ratio (m/z) differences that can be used to separate analyte m/z from interfering plasma molecular ions. The analysis included all 51 metals and Si.

X-ray photoelectron spectroscopy (XPS):

A portion of the undiluted CNC and CNF suspension was dried in a particle-free environment. It was then cut into sections and used for the XPS analysis. Survey and elemental scans were done for all detected elements.

Infrared spectroscopy (IR):

As before, a portion of the undiluted CNC and CNF suspension was dried in a particle-free environment. It was then cut into sections and was used for the Fourier transform infrared (FTIR). The FTIR was executed in ATR mode (Attenuated Total Reflectance) with the Perkin Elmer Spectrum One (Perkin Elmer, Waltham, MA). The scan range was 600–4000 cm⁻¹.

Sterility/biological assessment:

For the sterility assessment, the WHO protocol was used. (World Health Organization 2012) In brief, the suspension of the nanocellulose material was adjusted to 1 mg/ml mixed with DI water. The fluid thioglycollate medium (FTM) (BD, #299502) was used for the bacteria culture experiments. The media was prepared according to the manufacturer’s instructions. After sterilization, the pH of the media was verified to be between 6.9 and 7.3. A total of 10 ml of the FTM media was prepared to which 100 μl of the suspension was added. The solution was transferred to test tubes, which were kept in an incubator at 30 C for 14 days. Every day, the test tubes were inspected for froth, turbidity, and color change, all indications of possible bacterial growth. Every third day, a 10-μl sample was removed and plated on Tryptic Soy Agar (TSA) to evaluate the presence of bacteria. The process was carried out for 14 consecutive days.

Endotoxin assessment:

For the endotoxin assessment, the Recombinant Factor C assay was used (Alwis and Milton 2006), accepted by the Food and Drug Administration (FDA), European Pharmacopeia, and other leading regulatory authorities. This method is considered comparable to LAL-based assays (Ding and Ho 2010).

A suspension of the ENMs was prepared using endotoxin-free water to a final suspension of 10 μg/ml. As negative controls, Fetal Bovine Serum and endotoxin-free water were used, while E. coli endotoxin was used as positive control and calibration standard. The standard was reconstituted in the indicated volume of endotoxin-free water and was vortexed at maximum speed for 10 minutes. The resulting solution contained 50 EU/ml endotoxin. The calibration curve standards were prepared with serial dilutions of the endotoxin stock at 5.0, 0.5, 0.05, 0.005, and 0 EU/ml. The pH of the ENM suspension was adjusted to pH 7.0 using endotoxin-free 0.01N or 0.001N NaOH and/or 0.01N or 001N HCl. The assay reagent was prepared immediately before use by mixing eight parts Assay Buffer, one part enzyme, and one part substrate. For each ENM sample, a companion sample was prepared spiked with endotoxin standard to a final concentration of 0.5 EU/ml (for detection of reaction interference).

For each sample (standard, negative and positive controls, and spiked samples), the zero time point reading was subtracted from 90-minute time point reading. Using the standards points, the plot of the standard curve (log(EU/ml) vs log(rfu)) was constructed and was fitted with a linear equation (Y = A + BX) and the regression coefficient was >0.98. From the standard curve, the endotoxins concentration was calculated as EU/ml. If the value of the RFU was outside the standards range, extrapolation was used.

The limit set for endotoxins is based on the lowest value of endotoxins reported in the literature (Morris et al. 1992). According to Morris et al., the lowest reported concentration is 0.5 ng/ml, which is approximately 5 EU/ml, and caused the production of interleukin 6 in macrophages (1992). Given that the most commonly used upper limit for toxicological studies is 100 μg/ml (Deloid et al. 2017a), the endotoxin limit is set to 50 EU/mg of ENMs.

3. Results

3.1. Physical and morphological characterization

Figure 2 shows a typical TEM image of both CNF and CNC reference ENMs synthesized using the developed platforms. Figure S1 shows the SEM imaging of the raw material used for the production of the reference ENMs. The raw material is ribbon-like with several tenths of microns wide (20–60 μm) and several millimeters long (3–8 mm). As others have shown, CNFs have a wide range of fibril sizes. More importantly, the fibrils create a web-like structure with the various fibrils to be connected to others at various node points. To characterize the web-like structure of the CNFs in addition to the diameter, two new metrics were used: 1) Node-to-Node length (NtN length), which is defined as the distance between the centers of two nodes; 2), Number of branches per node, which is defined as the number of individual fibrils per node (NFpN).

Figure 2:

The synthesized cellulose nanomaterials from TEM: (a) CNF 50 nm, (b) CNF 80 nm, and (c) CNC 25 nm × 250 nm.

In contrast, the CNC materials were found to be uniform with a narrow range of diameters and length, similar to what others have reported (George and Sabapathi 2015). The crystals do appear in bunches, but this is likely caused by the drying of the grid, where capillary forces bring the ENMs together. Furthermore, the CNCs appear as long, straight, and rigid structures. Occasionally, a faint image of a precipitate appears due to the imaging salts (uranyl acetate) used in the TEM preparation; increased washing of the CNC did not seem to reduce the amount of precipitate.

Figure S2 shows the TEM size analysis of the CNFs focusing on the diameter, NtN, and NFpN. The results are summarized in Table S1. The diameters were 49.92 ± 43.69 nm and 83.23 ± 57.02 nm for CNF 50 and CNF 80 respectively. Similarly, the NtN Length was 335.60 ± 232.66 nm and 571.86 ± 514.71 nm for CNF 50 and CNF 80 respectively. Both the NtN length and the diameter display a wide distribution as is indicated by the large standard deviation of the measurements. However, it should be pointed out that the two CNF reference materials synthesized here are statistically different (P<0.05). Further, the NtN length and the NFpN are not characteristics of CNFs but are reported here purely as descriptors of the web-like structure CNF ENMs. Figure S5 shows the TEM results for the CNCs. The CNCs are small enough to fit within the field of view of the TEM so both length and the diameter can be measured. The diameter was 23.24 ± 6.99 nm while the length was 287.98 ± 114.42 nm resulting to Aspect Ratio (AR) 12.93 ± 5.21.

Figure 3 shows the SEM imaging and data analysis of the two different CNF materials (CNF-50 and CNF-80). Figure 3a and 3b show a representative SEM image of the fibrils adhered on a mica substrate. Figure 3c and 3d show the image analysis results for the CNF-50 and CNF-80 respectively. The average length was 6.710 ± 5.611 μm and 6.805 ± 4.724 μm for the CNF-50 and CNF-80 respectively. It should be mentioned that in both cases there were fibrils that exceeded the 15 μm length as indicated by the large Geometric Standard Deviation of the distribution that is 1.7584 and 1.9140 for the CNF-50 and CNF-80 respectively. Similarly, the diameters of the two CNFs were 64 ± 29 nm and 78 ± 25 nm for the CNF-50 and CNF-80 respectively, following a much tighter distribution as compared to the length distribution. The aspect ratio was calculated for each individual fiber and the average was based on these values was 107.6 ± 54.5 and 85.9 ± 40.5 for the CNF-50 and CNF-80 respectively.

Figure 3:

The SEM analysis of the CNFs. A representative SEM image for the (a) CNF-50 and CNF- 80 (b). (c) and (d) show the data analysis for the CNF-50 and CNF-80 respectively.

Figure 4a shows a representative SEM image of the CNCs and Figure 6b summarizes the results. The average length of the CNC is 267 ± 91 nm while the diameter is 25.2 nm with and 272 nm, respectively, obtained from TEM images. These dimensions result in an aspect ratio of 11.5 ± 3.2 (Table 1).

Figure 4:

The SEM analysis of the CNCs. (a)A representative SEM image (b) the data analysis.

Figure 6:

The FTIR spectrum of the cellulose nanomaterials and the raw cellulose material.

Table 1:

Summary of the morphological characterization data

ENMs	Physical Properties (SEM measurements)			Density (g/cm3)	Endotoxins (EU/mg)	Sterility (CFU/g)
	Length (nm)	Diameter (nm)	Aspect Ratio
CNF 50 nm	6710 ± 5611	64 ± 29	107.6 ± 54.5	1.3120 ± 0.0185	< LOD	0
CNF 80 nm	6805 ± 4724	78 ± 25	85.9 ± 40.5	1.3410 ± 0.0194	5.46	0
CNC25 nm	267 ± 91	25 ± 9	11.5 ± 3.2	1.5724 ± 0.0212	1.47	0

Endotoxin Assay LOD 0.5 EU/mg (−0.05 ng/ml)

Although several attempts were made to SSA for the freeze-dried CNC and CNF, it was not possible due to the very low apparent density of the sample that significantly limited the mass that can be used for the SSA measurement. As a result, no reliable measurements could be obtained. The SSA, however, can be calculated based on the dimensions of the fibrils and comes to 34 m²/g and 29 m²/g for the CNF-50 and CNF-80 respectively. For CNC, the theoretical value is 93 m²/g. Finally, the skeletal density of all the nanocellulose structures synthesized in this study was also measured and the values are listed in Table 1.

3.2. Chemical and biological/sterility characterization

The trace metal purity of the starting cellulose fibers, the CNF, and the CNCs was found to be less than 1% by weight as determined by ICP-MS. Silicon was the highest trace metal impurity that was also the predominant element in the fiber used for the synthesis. Other dominant metals found were Ca and Na, which are also in abundance in the soil. In the case of the CNCs, the dominant element was sulfur (0.08% w/w), which is expected since sulfuric acid was used for the synthesis. Further IC analysis showed that the sulfonates (– SO3⁻² –) account for 0.26% w/w of all soluble ions. Based on the molecular weight, it seems that the sulfur is in its majority part of the sulfonate groups. It should be mentioned here that although ICP-MS measures all metals, IC can account only for the soluble ions and therefore the direct comparison between the two methods are only for semi-quantifiable reasons. These results were also confirmed with the XPS that showed similar binding energy between the starting fibers and the CNF, with only traces amounts of other metals for the case of CNF-80 (Figure 5).

Figure 5:

The XPS elemental analysis of the cellulose nanomaterials.

The FTIR results for all nanoscale materials showed the same peaks as the raw material (Figure 6). The difference in intensity comes from sample preparation. The major peaks that observed were: 3346 cm⁻¹ corresponding to -OH stretching, 2904 cm⁻¹ corresponding to C-H stretching, 1645 cm⁻¹ corresponding to O-H bending of adsorbed H2O, 1429 cm⁻¹ corresponding to CH2 scissoring, 1371 cm⁻¹ corresponding to C-H bending, 1317 cm⁻¹ corresponding to CH2 wagging, 1161 cm⁻¹ C-C corresponding to stretching, 1058 cm⁻¹ C-O-C corresponding to pyranose ring stretching, 1035 cm⁻¹ corresponding to C-O-C pyranose ring stretching, and 898 cm⁻¹ corresponding to Cellulosic β- glycosidic linkage. The key point is that there are no new peaks introduced or peaks missing when comparing the starting material; there are no changes to the cellulose during the fibrillation or acid hydrolysis.

Finally, the bacteria and endotoxin assessment showed no bacteria and very low levels of endotoxins for both CNF and CNC materials synthesized (Table 1).

4. Discussion

While a number of publications have described the production of cellulose nanomaterials (Vartiainen et al. 2011; Khan et al. 2014), a clear standard method to produce materials in a property-controlled and reproducible manner is not set. Furthermore, there are no standardized methods available that can be used universally to compare the various nanocellulose structures. In this manuscript, the development of standardized industrial platforms, which can be used to produce reference nanocellulose materials suitable for nanotoxicology research, is presented. Further, these platforms are evaluated via the synthesis and the characterization of three different cellulose materials that can be used as reference materials for various applications and implications studies.

The morphology of the cellulose nanomaterials was assessed using both TEM and SEM. Regarding the overall CNF structure, TEM imaging revealed that the fibrils form a web-like structure. In some regions, the fibrils look more like ribbons of material. This ribbon morphology likely comes from a region of the cell wall from the wood fiber that has not been fully broken apart. It is apparent from Figure 1 that it is challenging to characterize CNF, which is why methods to report the length of CNF have not been given in the literature. SEM is suitable for characterizing CNF as long as the fibrils are properly untangled and dispersed as described. In this approach, due to the high dilution and long stirring time, the fibrils were untangled and it became possible to measure the length of the fibrils. Still, some of the fibrils are hard to measure as they are connected to each other in bundles or brunches.

Standardized procedures for image analysis method used for other ENMs relying on the Feret diameter (Pyrz and Buttrey 2008) cannot be used here to assess the diameter and the length. Typically, these methods require tracing the perimeter of the particles, either manually or automatically, and based on that, calculation of the minimum and maximum Feret Diameter as diameter and length respectively (Figure S5). However, if this is applied here it will overestimate the diameter and underestimate the length, as the Feret, by definition, does not account for curving and bending. Therefore, for these CNFs, it is recommended to measure the length by tracing a line across the fibril following the curvature of the fibrils and follow by measuring the diameter at 3–5 points across the length. To our knowledge, this is the first time the length of CNF has been obtained in this manner.

Comparing the CNF data regarding the diameter obtained by TEM and SEM, it seems that there is a difference for the CNF-50 material. Although this difference is statistically significant (P<0.005), it is attributed to the difference in resolution between TEM and SEM that have 0.02 nm/px and 5 nm/px respectively. Due to the resolution limitations for small particles (<50 nm), TEM biases the results towards smaller dimensions while SEM towards larger. For the CNF-80 that are bigger in size, this is not the case as the difference that is observed is not statistically significant (P-value 0.5078), confirming the above hypothesis. The SEM data also revealed that the fibril length does not depend on the grinder settings as the length is essentially the same for both cases of CNFs.

One important issue with imaging was the SEM sample preparation. Typically, SEM requires samples that have been dried out and are completely dehydrated. The drying process, however, can result in significant loss of hydration that can effectively change the diameter of the fibrils. For that reason, cryo-SEM was also done (Figure S4) to evaluate the effect of drying. Cryo-SEM can image the samples in their hydrated state. The results clearly showed that the there is no measurable effect from the drying process. The cryo-SEM showed diameter 64 ± 23 nm as compared to the regular SEM that showed 64 ± 29 nm indicating that if there is any effect from the drying process is negligible. The effect of drying on the chemical or morphological properties remains, however, to be further investigated.

On the contrary to CNFs, CNCs are less challenging to characterize using both TEM and SEM. For TEM, however, it is important to stain the CNCs with uranyl acetate as they have low contrast and are not possible to image. SEM and TEM results are in agreement, both regarding the diameter (P- Value 0.07, 0.90 considering the sputtering thickness) and length (P-Value 0.12) of the CNCs. Also, since CNCs are rigid needle-like structures, they can be analyzed by using standardized particle sizing descriptors such as Feret diameter.

All produced reference CNFs synthesized in this study, are also pure, retaining the purity of the raw material, indicating that the grinding stone is not wearing and adding any impurities. The trace metals found to account for more than 90% of total metals present are metals such as Si, Ca, and Na, all incorporated into the wood fiber cell wall during plant growth. These results are expected as plants take up minerals from the soil during growth. An exception is the CNCs where the dominant element was the sulfur due to the acid hydrolysis. However, FTIR that was not able to detect the characteristic strong IR-active bands produced by stretching, non-degenerate bending (A1-mode), and degenerate bending (E-mode) vibrations of the SO3^2- anion are typically in the ranges 910–980 cm⁻¹, 600–660 cm⁻¹, respectively (Miller and Wilkins 1952; Nakamoto 2008; Frost and Keeffe 2009). This could be either due to the low concentration of the SO3^2- anion or due to the strong background of the -CH, =CH2 and -CH3 vibration. Generally, FTIR showed that both the mechanical (grinding) and chemical (acid hydrolysis) did not change the structure of the cellulose.

Furthermore, an important requirement of nanotoxicology studies is to ensure the sterility of the ENMs (Li and Boraschi 2016). In this study, the synthesized CNFs and CNCs were assessed in terms of endotoxin and bacterial contamination. No bacteria were detected, and, more importantly, the endotoxins were at a low level (or below the limit of detection), significantly below what can trigger cellular responses (Morris et al. 1992). This is due to the careful choice of the raw material and also due to the sterile procedures followed throughout the process. The autoclave step at the end of the synthesis process, although it can help to mitigate or eliminate bacteria, cannot remove endotoxins (Gorbet and Sefton 2005). Our studies showed that handling the CNFs and CNCs outside sterile conditions increase the possibility of contamination due to airborne fungi, mold, or bacteria. The cellulose in combination with the water provides excellent conditions for the growth of these biological contaminants. It is therefore recommended to handle the cellulose ENMs in small batches under sterile conditions, such as Argon glovebox or biological hood.

The colloidal characterization of nanocellulose in various biological media is of interest in nano bio interaction studies. Potential transformations of ENMs in biological media, such as protein corona formations, may affect both bioactivity and also particle-kinetics and their fate and transport (Cohen et al. 2012; Cohen et al. 2014). For anisotropic ENMs, such as nanocellulose, there is no method that can provide the colloidal characterization in aqueous, physiological, or biological media. Although widely used in nanotoxicology research, Dynamic Light Scattering is not suitable for size characterization due to the fundamental assumption that the particles are spherical (Pal et al. 2015; Deloid et al. 2017a). Although for the case of the CNFs, the large size will most likely dominate their fate and result in rapid sedimentation (Deloid et al. 2015); developing accurate colloidal characterization methods for such anisotropic material is of great importance for nanotoxicology research.

5. Conclusions

Industry-relevant standardized methods to produce reference CNF and CNC materials with reproducible and consistent properties that can be used to study the toxicological properties of cellulose nanostructures were developed. These platforms were utilized to synthesize reference nanocellulose materials of controlled properties. Furthermore, novel methods to characterize CNF and CNC morphology were developed; a method to measure the length of CNF is described. All the nanocellulose reference ENMs synthesized were produced at high-volume rates and at nanometer precision. The synthesized CNFs and CNCs were free of endotoxins and bacteria, which is important in particular for cellulose materials that are produced from a natural product, which can contain biologicals. In future studies, other cellulose ENMs will be added in the repository, including TEMPO-modified cellulose and fluorescently tagged cellulose.