Towards the scalable isolation of cellulose nanocrystals from tunicates

In order for sustainable nanomaterials such as cellulose nanocrystals (CNCs) to be utilized in industrial applications, a large-scale production capacity for CNCs must exist. Currently the only CNCs available commercially in kilogram scale are obtained from wood pulp (W-CNCs). Scaling the production capacity of W-CNCs isolation has led to their use in broader applications and captured the interest of researchers, industries and governments alike. Another source of CNCs with potential for commercial scale production are tunicates, a species of marine animal. Tunicate derived CNCs (T-CNCs) are a high aspect ratio CNC, which can complement commercially available W-CNCs in the growing global CNC market. Herein we report the isolation and characterization of T-CNCs from the tunicate Styela clava, an invasive species currently causing significant harm to local aquaculture communities. The reported procedure utilizes scalable CNC processing techniques and is based on our experiences from laboratory scale T-CNC isolation and pilot scale W-CNC isolation. To our best knowledge, this study represents the largest scale where T-CNCs have been isolated from any tunicate species, under any reaction conditions. Demonstrating a significant step towards commercial scale isolation of T-CNCs, and offering a potential solution to the numerous challenges which invasive tunicates pose to global aquaculture communities.


Scientific Reports
| (2020) 10:19090 | https://doi.org/10.1038/s41598-020-76144-9 www.nature.com/scientificreports/ characterizations were performed to better understand the behavior and challenges of preparation as well as the attributes of the final T-CNCs. Experiences from large-scale preparation of W-CNCs using established protocols 37 as well as the ultimate characteristics of W-CNCs and other nanocelluloses provided useful comparisons.

Results and discussion
Preparation of a tunicate-derived cellulose feedstock. Harvesting. The starting material for the pilot-scale production of W-CNCs is a high-purity commercial cellulose pulp prepared by well-established wood pulping protocols. Obviously, the preparation of a similar cellulose feedstock from tunicates is necessarily a very different process. To prepare a relatively large quantity of tunicate cellulose feedstock, we began by manually harvesting approximately 20 kg of invasive Styela clava tunicates from waterways surrounding PEI. Manual harvesting is a viable process to collect commercial scale quantities of tunicates. In fact, it is estimated that over a million pounds of Styela clava (wet weight) are cultivated and harvested annually from waters around South Korea, where they are consumed as a seafood delicacy known locally as "mideuduck" [50][51][52] . These have primarily been manual tunicate harvesting methods similar to those employed here. Although we posit efficient automated processes may lower harvesting costs. Recently, Ocean Bergen AS implemented an automated approach for harvesting tunicates from Norwegian waters to extract protein for animal feed 49 . For these reasons, we are currently developing an automated harvesting process for collecting invasive tunicates on PEI, based on their success and the findings of this study. The processing of the tunicates after harvesting is schematically shown in Fig. 1. Important aspects are discussed herein and a more detailed description of the T-CNC and W-CNC isolation process is provided in section (S1).
Tunic preparation. Once harvested, the cellulose-containing tunics were manually separated from the proteinrich internal organs. We are currently investigating more economically viable approaches including automated tunic separation and a biorefinery-type approach which utilizes the entire tunicate as a process input. The manually prepared tunics used here were washed, dried and ground as described in S1. While others have used the internal organs to prepare animal feed 48 or to ferment bioethanol 53 , we chose to focus on T-CNC isolation and simply disposed of the internal organs. The use of such byproducts is left for future work. Generally, one half of a tunicates weight is its tunic, although this varies with tunicate species, environmental factors and life cycle stage. We found that our ~ 20 kg of Styela clava tunicates harvested from PEI waters resulted in ~ 10 kg of tunic, which were ~ 90% water, yielding approximately 1 kg of tunic powder when dried.
Tunic pretreatment. To isolate T-CNCs from this tunic powder, the cellulose must be purified and the noncellulose components removed to prepare a high cellulose feedstock for acid hydrolysis. To accomplish this, the tunic powder was shipped to the Forest Products Laboratory where it was further processed by alkaline deproteination treatments and bleaching following the protocols described by van den Berg et al. 2 , with modifications as described in S1. The overall yield for the deproteination and bleaching steps was ~ 31%, comparable to the yields reported in Table 1 for similar processes at lab scale. The final bleached material was used as the feedstock for preparing T-CNCs by acid hydrolysis.
According to Zhao and Li, generally tunic possesses a ~ 50:50 weight ratio of carbohydrates to proteins, where between 75 and 95% of the carbohydrate fraction is glucose, and of the glucose fraction, between 50 and 75% is cellulose 10 . Although their work focuses on four different tunicate species, we feel that their general conclusions  www.nature.com/scientificreports/ are applicable to our processing. Therefore, this suggests that the 1 kg of dried tunic powder prepared for this work likely possesses only ~ 19-36% cellulose. Given this estimate, coupled with the findings reported in Table 1, our overall yield of ~ 31% for the deproteination and bleaching steps seems reasonable. While the additional non-cellulose tunicate components present a challenge when isolating T-CNCs, these additional components have intrinsic value and may be recoverable. Although not the focus of this study, we suggest that additional value-added product streams, including protein 63 and heavy metal recovery (See S2) [64][65][66][67] , may be feasible if tunicates are processed to T-CNC in a biorefinery-type approach. This requires thoroughly understanding the components of waste streams generated in T-CNC isolation and determining their recoverability, an active area of investigation in our group. CNC preparation. Wood derived W-CNCs are prepared from high purity cellulose wood pulp (≥ 97% cellulose) in the Nanocellulose Pilot Plant at the Forest Product Laboratory using standard protocols 37 . The main steps in the process are: (1) sulfuric acid hydrolysis, (2) diafiltration to remove by-products, and (3) concentration of the resulting aqueous CNC suspension. Tunicate derived T-CNCs were prepared similarly, albeit on a smaller scale, and with necessary changes to accommodate differences in the source materials. Our experiences during the various steps of the T-CNC preparation are discussed below along with relevant comparisons to W-CNC processing and proposed changes to protocols that may improve the process.
Sulfuric acid hydrolysis. Hydrolysis of the tunicate cellulose was accomplished using 64% H 2 SO 4 for 2 h with additional details described in S1. The hydrolysis yield was ~ 42% for T-CNCs, compared to ~ 50% for the optimized W-CNC isolation, resulting in aspect ratios of 65 and 12 respectively (See Fig. 2, S4 and S5). For additional context, we have summarized the resulting aspect ratios and yields reported in numerous studies where similar cellulose sources and processing conditions were utilized to isolate CNCs at differing scales (See Table 2).
In many reports, information such as yield and precise processing conditions unfortunately are omitted. However, we note that the T-CNCs prepared here display properties consistent with previous T-CNCs isolated at laboratory scale. Indicating that the impressive properties attributed to T-CNCs can, as pioneered in the development of large-scale W-CNC isolation, be preserved when T-CNC isolation is scaled up. At this time, replicate experiments and concurrent process optimization of T-CNC isolation at this scale remain future areas of study. Also, as discussed later, some material was lost during diafiltration, which adversely affected the T-CNC yield. Therefore, with further improvement of protocols, the T-CNC yield could very well approach that of the W-CNCs.
Diafiltration and concentration. Following hydrolysis, the reaction is quenched and neutralized with aqueous NaOH. The resulting highly saline suspension leads to the association and settling of CNCs. Most hydrolysis byproducts could then be removed by decanting the supernatant, adding deionized water, again allowing CNCs to settle and repeating the process. Eventually as salinity decreased the CNCs began to suspend rather than settle, and a tubular ultrafiltration unit was used to complete by-product removal by tangential (cross) flow filtration (TFF) 37 . TFF reduces filter cake formation by creating turbulent flow, which improves flux rate compared to conventional dead-end filtration (See S3). As filtrate is removed, additional water is added and the hydrolysis byproducts are flushed from the CNCs in a process referred to as diafiltration. Unfortunately some residual, aggregated tunicate derivatives obstructed the circulation pump during diafiltration of T-CNCs, suggesting that improvements in our processing protocols are warranted. The suspension was filtered and centrifuged using a large, industrial centrifuge to remove the aggregated material (See S1 and S4). The diafiltration process was then www.nature.com/scientificreports/ completed. This additional product loss almost certainly contributed to the lower yield of the T-CNCs when compared to the W-CNCs. The make-up water was then shut off to concentrate the CNC suspension until its viscosity increase inhibited flow through the membranes tubes, after which point the system was back flushed to yield the concentrated CNCs. The CNC suspension viscosity is primarily governed by the aspect ratio of the CNCs and the CNC concentration, where the salinity of the suspension is assumed to be consistent since both W-CNCs and T-CNCs are neutralized prior to filtration. Figure 3 presents rheological properties of the 1 wt% W-CNC and T-CNC suspensions in water. The viscosity of the T-CNC suspension in the same concentration (1 wt%) is considerably higher than that of the W-CNC suspension (Fig. 3a). This is attributed largely to the higher aspect ratio of T-CNCs of 65 compared to about 15 for W-CNCs. For CNC suspensions, a shear-thinning behavior with increasing shear rate is expected due to the orientation of fibers. For the 1 wt% W-CNC suspension, a low, constant viscosity of less than 2 mPa.s is observed. This is in line with the results obtained by Lenfant et al. for a similar W-CNC suspension 74 . At this concentration of low aspect nanoparticles, Brownian motion prevents the orientation of the particles under flow 74 . In the case of the 1 wt% T-CNC suspension, the shear-thinning behavior is attributed to a gel-like structure formed by this suspension of large aspect ratio nanoparticles. With increasing shear rate, this structure is broken down explaining the decreasing viscosity although particle orientation could be partly responsible of the shear thinning. The presence of the gel structure is confirmed by the linear storage and loss moduli data of the T-CNC suspension presented in Fig. 3b. We observe a gel-like or viscoelastic solid-like behavior where the storage modulus (G'), is much greater than the loss modulus (G") and both moduli are relatively independent of frequency 75 . Gelation of the T-CNC suspension at low concentration is due to its high aspect ratio and this behavior was observed previously at much higher concentration for W-CNC aqueous suspensions (~ 10 wt%) 74 . www.nature.com/scientificreports/ As observed previously by others [76][77][78] , high aspect ratio CNC suspensions display considerably higher viscosity values than lower aspect ratio CNC suspensions. In our case, the maximum concentration of the high aspect ratio T-CNC was ~ 1.3 wt%. This is far below the ~ 10 wt% achievable with lower aspect ratio W-CNCs, but similar to that found for TEMPO pretreated wood-derived cellulose nanofibrils 79 .
Suggested improvements to processing protocols. Several future procedural improvements are proposed based on our findings from the isolation of T-CNCs performed. Among the strongest recommendations is that, from extraction of the tunicates to isolation of the T-CNC, the cellulose containing material should remain wet as drying should be avoided to prevent hornification 80 . Hornification results from the formation of hydrogen bonded networks during drying that are only partially reversible. Here, we initially dried the tunic and between each process step the yield was determined by drying the intermediate product. This may have led to a compounding of the hornification of cellulosic material, resulting in reduced efficacy of chemical treatments and a lower process yield. If drying is necessary, lyophilization is preferable to air or oven drying to lessen the effect of hornification. Traces of color were still observed after the bleaching step, which we attempted to remedy with additional bleaching after acid hydrolysis. To obtain whiter and cleaner materials, alternating between acid chlorite bleaching and alkaline extraction may be a beneficial procedural improvement. The level of calcium in the final T-CNC product is quite high at 0.054 wt%, as typical levels observed in W-CNC processing are less than 0.002 wt% (See S2). We expect that the source is likely the tunicates' natural calcium-rich environment. The T-CNC hydrolysis is highly acidic and when the reaction was neutralized, it likely caused association of the negatively charged CNC sulfate groups with calcium cations. The calcium level may be reduced by the addition of an acid wash after bleaching, by decanting the acidic T-CNC solution after hydrolysis but before neutralization, or with suitable chelation treatments. After the hydrolysis and neutralization, some aggregates were observed which  www.nature.com/scientificreports/ interfered with subsequent ultrafiltration and purification steps. Simple screening for these aggregates prior to filtration may improve ultrafiltration efficacy. With scaling, it may also be advantageous to replace the large-scale centrifuging, if it is still found necessary after other improvements, with mechanical homogenization to improve yield, processing efficiency and overall consistency of the T-CNC product. Given the very different size distributions of the two types of CNCs, it may be possible to increase the T-CNC concentration efficiency by optimizing the pore size of the ultrafiltration membranes.

CNC properties.
Once the T-CNCs and W-CNCs were prepared and their morphologies understood, we compared their crystallinity and thermal stabilities while contrasting our findings with past reports. What follows is our assessment of the results and how the properties of the obtained T-CNCs compare to that of W-CNCs prepared by an optimized process.  www.nature.com/scientificreports/ Crystallinity. We assessed the overall structural order of as produced CNCs utilizing two complimentary techniques: XRD and Raman. A summary of our findings and those reported by others for CNCs prepared by similar procedures is displayed in Table 3 and Figs. 4 and 5. As described in S6, FTIR spectroscopy was also performed to determine the Lateral Order Index (LOI) and Total Crystallinity Index (TCI) of the isolated T-CNCs and W-CNCs.
XRD. The overall structural order of the prepared CNCs was further assessed by calculating their percent crystallinity from the background-corrected experimental diffractograms in Fig. 4, consistent with our prior work 25 . In this way, the T-CNC was determined to be 75% crystalline whereas the W-CNCs were 66% crystalline. Evidence of uniplanarity is observed in the T-CNC diffractograms by comparing the relative intensities of the 1 1 0 reflection intensity. However, this has also been reported to result from CNC orientation induced either from drying kinetics or incomplete hydrolysis 69 . Our XRD samples were prepared by freezing aqueous suspensions of dilute CNC (~ 0.5 wt%) in liquid N 2 followed by freeze-drying. Therefore, we expect that if orientation of the CNCs is contributing to the enhanced 1 − 1 0 reflection intensity, it is more likely a result of incomplete hydrolysis than drying induced orientation. This is supported by TEM results which indicate unusually wide T-CNC crystallites (~ 20 nm), which has been attributed to small bundles of CNCs arising from hornification or other processes 17 . We have contrasted our findings with past reports in Table 3 and provide further assessment of experimental diffractograms in S7. We note that a wide range of values are reported for W-CNCs and T-CNCs, which result from the diverse methods used to calculate crystallinity from XRD diffractograms reported in literature 83,84 . To provide a more comprehensive understanding of the relative crystallinity of our CNCs, we utilized Raman spectroscopy.
Raman. Raman crystallinity of the wood and tunicate CNCs were determined using two methods-380-Raman (Agarwal et al. 2010;2013) and 93-Raman (Agarwal et al. 2018), and the values are reported in Table 3. The methods are based, respectively, on the band intensity ratios 380/1096 cm −1 and 93/1096 cm −1 in the Raman spectra of the CNCs visible in Fig. 5. The Raman crystallinity data in Table 3 indicated that in both the Raman methods, compared to the crystallinity of wood CNCs the crystallinity of tunicate CNCs was significantly higher. For 380-Raman and 93-Raman, the crystallinity was higher by 21% and 109%, respectively. Although it's not clear why the two methods differed so significantly with respect to the increase, the increases supported the observation based on XRD that T-CNCs were significantly more crystalline compared to W-CNCs. The highly crystalline nature of the T-CNCs mean that they are stronger and less sensitive to moisture than W-CNCs in various applications.
Thermal stability. To understand the thermal stability of the T-CNCs isolated in this work and how it compares to W-CNCs, we performed TGA in both an oxidizing (air) and an inert (nitrogen) environment. The resulting thermograms and their derivatives were obtained and compared with those of W-CNC analyzed in the same manner. As visualized in Fig. 6 and further described in S8, the isolated T-CNCs are more thermally stable than W-CNCs in an oxidizing environment.
The onset of thermal degradation for W-CNCs is clearly lower (Fig. 6a) than that of the T-CNCs in air. However, in an inert environment (Fig. 6b), this trend is less apparent. In air, both CNC materials displayed ~ 3% ash content. However, in inert nitrogen there is an increase in the ash content of the W-CNCs (19%) and, to a lesser extent, the T-CNC (8%). This indicates that W-CNCs have a higher content of nitrogen-stable components or www.nature.com/scientificreports/ thermal degradation products 85 . We posit that this may be linked to the plethora of ocean-derived elements (S2) present in the T-CNCs but not found in W-CNCs. We have summarized some of the sparsely reported thermal properties of T-CNCs prepared by similar acid hydrolysis procedures in Table 4, and contrasted these with W-CNCs. We assess that the observed differences in thermal stability result primarily from previously discussed variations in crystallinity, as well as the relative sulfur content and the surface area of wood and tunicate derived CNCs which are discussed in S8.

Perspective and outlook.
By processing roughly 20 kg of invasive tunicates to H 2 SO 4 hydrolyzed T-CNCs, this work accomplishes the largest scale isolation of T-CNCs reported to date. Learning from the pilot scale development of W-CNCs, we isolated T-CNCs using scalable techniques, with reasonable yield, and of similar properties to those reported for T-CNCs isolated at laboratory scale by others. This represents a significant step toward kilogram scale and eventual commercial scale isolation of T-CNCs on PEI, and in areas where similar tunicate densities are available in local waters. The overall yield of our pretreatment (31%) and acid hydrolysis (42%) of the tunic powder was within the range of values reported for laboratory scale tunicate to T-CNC processes. Overall, the yield of T-CNCs from our process was 12.2% based on the dry weight of the tunic powder and T-CNCs isolated therefrom. Experimentally determined aspect ratios, crystallinity and some thermal properties of the T-CNCs exceeded those of W-CNCs, as expected; and were similar to those found for T-CNCs prepared at laboratory scale by others. Replicate trials that implement the numerous potential process improvements described here would likely lead to a considerable increase in yield and quality of T-CNCs at this scale, and we feel that the proposed improvements themselves are scalable in nature. Other procedures for T-CNC isolation may be scalable, and modified from literature in a likewise manner. These may yield T-CNCs of similar or improved properties based on the process conditions, tunicate source and degree of process optimization. We posit that the future commercial scale isolation of tunicate derived CNC is feasible and that the unique properties of these T-CNCs, which complement the growing global utilization of nanocellulose materials, justify this pursuit. We chose an invasive species negatively affecting local aquaculture communities in PEI and across Atlantic Canada as the T-CNC source. This allows us to demonstrate the unique conditions that currently exist on PEI, which mitigate the historic challenges of tunicate harvesting and T-CNC isolation at commercial scale. These conditions are not limited to Atlantic Canada and entities around the globe are currently harvesting tunicates at commercial scale for their proteinaceous components. Regardless of the driving force, tunicates will ultimately be considered and perhaps utilized as a large-scale source of numerous value-added products, including their unique animal-derived high aspect ratio cellulose, for commercial T-CNC isolation. This study lays tangible groundwork towards that goal, directly demonstrating the feasibility and results of kilogram scale tunicates to T-CNC processing, and promoting the wide spread utilization of both invasive and native tunicates to produce useful and sustainable materials for the benefit of our growing global community.

Methods
A more detailed description of the T-CNC and W-CNC isolation process is provided in section (S1). What follows are general descriptions of the equipment and techniques utilized to obtain the reported experimental data and is complimented further in the Supplementary Information. Elemental analysis. The sulfur, sodium and calcium content of the prepared CNCs was determined using Inductively Coupled Plasma Optical Emission Spectroscopy (ICP-OES) (Ultima II, Horiba Jobin-Yvon, Edison, NJ, USA) using previously developed protocols 88 .
The qualitative elemental composition of the dried tunic powder used to prepare the T-CNCs was investigated with a JEOL JSM6400 Digital SEM, using the equipped EDX (Genesis) Energy Dispersive X-ray system. Digital X-ray maps were obtained from powdered samples which were mounted to carbon tape and carbon coated for conductivity prior to imaging.

Transmission electron microscopy (TEM).
To assess the morphology of the T-CNC and W-CNC, transmission electron microscopy (TEM) micrographs were obtained on a JEOL 2011 STEM instrument. Dilute www.nature.com/scientificreports/ (0.001 wt%) colloidal suspensions were cast onto etched copper coated grids and air-dried prior to imaging. The average length, width and aspect ratio were calculated from at least 50 measurements from 5-10 representative micrographs of each sample using Image J software.

Fourier-transform infrared spectrometry (FTIR).
Attenuated total reflectance Fourier transform infrared spectrometry (ATR-FTIR) was performed to understand the functional groups present, screen for impurities and to calculate the Lateral Order Index (LOI) and Total Crystallinity Index (TCI) of the isolated T-CNCs and W-CNCs. A Bruker Alpha FTIR spectrometer (Alpha-P) was utilized with OPUS software, 32 scans were averaged against background scans to yield the reported spectra in the range of 4000 to 500 cm −1 . The measured transmittance values were converted to absorbance and the magnitude of the absorbance at 2900, 1430, 1375 and 897 cm −1 was used to determine LOI and TCI.
Thermogravimetric analysis (TGA). Thermal properties were assessed with the aid of Thermogravimetric analysis (TGA) which yielded thermal decomposition profiles for T-NCC and W-CNCs, as well as their first derivative with respect to weight (DTGA) thermograms. Experiments were performed on a TA Instruments TGA Q500 under an oxidizing atmosphere (60 mL/min compressed air, 40 mL/min nitrogen) from room temperature to 700 °C, using a heating rate of 10 °C/min. Inert atmosphere thermograms were obtained by first purging the sealed sample chamber for 30 min with a 100 mL/min nitrogen flow rate, after which the sample was heated at 10 °C/min to 700 °C under nitrogen.
X-ray diffraction (XRD). X-ray diffraction (XRD) was performed to assess the crystallinity of the isolated T-CNC and W-CNC used in this work. Aqueous CNC samples (0.5 wt%) were flash frozen in liquid nitrogen prior to lyophilization to obtain the dry CNC sample for analysis. The utilized Bruker AXS D8 Advance instrument was equipped with a graphite monochromator, variable divergence slit, variable anti-scatter slit and a scintillation detector. Cu (Ka) was the radiation source used (k = 1.542 A°) and the measurements were performed on glass slides with a double-sided scotch tape adhesive, in air, at room temperature, from 2° to 60° (2θ).
Raman spectroscopy. For estimations of crystallinity by Raman spectroscopy methods (380-Raman and 93-Raman 89-91 ), sample pellets were prepared with a pellet-forming die. Approximately 100 mg of T-CNCs and W-CNCs were used for making pellets. The CNCs were analyzed with a Bruker (Billerica, MA) Multi-Ram equipped with a 1064-nm 1,000 mW continuous wave (CW) diode pumped Nd:YAG laser. Spectra were recorded from 2,048 co-added scans using 600 mW laser excitation, as reported previously 92 . In all cases, Bruker OPUS 7.2 software was used to process the spectral data which involved normalization, selection of a spectral region, background correction, and band integration. Background correction was performed using a 64 points OPUS "rubberband option". For plotting purposes, the spectra were converted to ASCII format and exported to Excel.
CNCs crystallinity was estimated using two Raman methods-380-Raman 89,90 and 93-Raman 91 . The following two equations were used to estimate these crystallinities.

Rheometry.
A stress-controlled Anton Paar rheometer (MCR 502) was used to carry out the rheological measurements at 25 °C. Couette and double-Couette flow geometries were used for different samples. The region of linear viscoelasticity was first determined by performing strain-sweep tests. The viscoelastic behavior of the suspensions was determined from frequency sweep tests in the linear regime. The steady shear test was performed from low to high shear rate. The reproducibility of all data was investigated by repeating the tests three times. To eliminate the history effect, all samples were pre-sheared at shear of 100 s −1 for 5 min followed by 30 min rest prior to all subsequent tests. To assure homogeneity of the suspensions and eliminate aging effect all the samples were ultrasonicated using a Sonics & Materials VCX500 probe, operating at 20 kHz, at a power of 60 W and energy of 10,000 J/g CNC , operated in pulses with suspensions placed in an ice bath to avoid overheating. www.nature.com/scientificreports/