Production potential of cellulose nanobrils from transgenic trees with reduced expression of cellulose synthase interacting 1

Cellulose nanobrils can be derived from the native load-bearing cellulose microbrils in wood. These microbrils are synthesized by a cellulose synthase enzyme complex residing in the plasma membrane of developing wood cells. In addition to the cellulose synthase complex several other proteins, including cellulose synthase interacting 1 (CSI1), facilitate cellulose biosynthesis. It was previously shown that transgenic hybrid aspen trees with reduced expression of CSI1 exhibit changes in wood mechanics and cellulose microbril properties. We hypothesized that these changes in native cellulose may impact the quality of corresponding nanobrils. To investigate this hypothesis wood from wild type and transgenic trees with reduced expression of CSI1 was subjected to oxidative nanobril isolation processes. The transgenic wood extracted nanobrils exhibited a signicantly lower suspension viscosity and estimated surface area. Furthermore, the manufactured nanobril networks exhibited a higher stiffness, reduced water-uptake, tensile strength, strain-to-break and degree of polymerization compared to wild type derived nanobrils. It was concluded that the difference in wood properties caused by decreased expression of CSI1 gave rise to nanobrils with distinctive qualities. The observed changes in the physicochemical properties suggest that the differences are caused by changes in apparent nanobril aspect ratio and surface accessibility. This study demonstrates the possibility to inuence wood derived nanobril quality through genetic engineering of trees and opens for breeding of wood properties for extraction of nanomaterials with dened characteristics. Author


Introduction
Wood is a complex biological structure providing trees with the mechanical support needed to ensure upright growth whilst simultaneously facilitating the transport of nutrients and water (Groover et al., 2010). An important role of mechanical reinforcement in wood is performed by cellulose micro brils (CMFs) which are hierarchical structures of β(1→4) linked D-glucose units which form glucan chains of nanoscale semi-crystalline brils in a bottom-up fashion (Fernandes et al., 2011). The CMFs are positioned throughout the wood cell walls where they are embedded in a hydrated matrix comprising primarily of hemicellulose and lignin (Knox, 2008). The resulting natural nanocomposite structure gives rise to the distinct properties of wood (Gibson, 1992).
The prospect of extracting CMFs as a value-added product from wood has gained attention over the past two decades. CMFs that have been subjected to isolation and puri cation processes are referred to as cellulose nano brils (CNFs) and inherit many of the distinct properties of CMFs, including high tensile strength (Wu et al., 2014), stiffness (Iwamoto et al., 2009), aspect ratio (Varanasi et al., 2013) and low coe cient of thermal expansion ).
Development of various processing techniques have further enabled successful isolation of CNFs that closely resembles the most basic crystalline structures of the CMFs (Saito and Isogai, 2004;Saito et al., 2009;Wågberg et al., 2008). Through careful processing it is also possible to retain high degree of polymerization (DP) of the cellulose ), which is manifested as longer CNFs (Henriksson et al., 2008;Shinoda et al., 2012). These process developments have further led to the prospect where the characteristics of CMFs are potentially in uencing the isolated CNFs. Thus, understanding of how to alter CMFs in the original feedstock opens for the possibility to control nal CNF characteristics.
The understanding of how CMFs are synthesized in plants has been described as one of the main challenges in plant biology (Saxena and Brown Jr, 2005). Research of the CMF synthesis machinery has revealed multiple proteins that are involved in this process (Polko and Kieber, 2019). One of the cellulose biosynthesis associated proteins that plays a crucial role in the alignment of the nascent CMFs into the cell wall is the cellulose synthase interactive 1 (CSI1) protein (Gu et al., 2010;Li et al., 2012). The plasma membrane localized cellulose synthase complex (CSC) moves in the plasma membrane during cellulose biosynthesis (Paredez et al., 2006). CSI1 guides the CSC along the cortical microtubules (cMTs) to align CMFs during primary cell wall biosynthesis (Bringmann et al., 2012;Li et al., 2012), and during the initial phase of secondary cell wall formation (Schneider et al., 2017). From a study of transgenic trees with reduced expression of CSI1 it was observed that both stiffness and strength of the wood was decreased, as well as the cellulose DP . There were no apparent structural or compositional changes in the wood cell wall of the transgenic lines which led to the hypothesis that a reduction of CSI1 may impact the mechanical properties of the wood by a reduction of the cellulose DP and thus altered CMF characteristics. It is thus of interest to isolate CNFs from these transgenic trees with a reduced level of CSI1 and assess the possible in uence of the genetic modi cation on CNF properties.
To compare CNFs isolated from wild type (WT) and transgenic wood, mild (pH = 6.8) direct oxidation using the catalytic system 2,2,6,6-tetramethylpiperidin-1-yl) oxyl (TEMPO) was employed together with high-pressure homogenization . This process has the bene t of allowing for direct ). Two transgenic lines with reduced expression of CSI1 were studied together with the control tree (WT) with normal CSI1 expression. The isolated CNFs were characterized in the dispersion state using viscosity, conductimetric titration, atomic force microscopy, yield-and surface area estimations. Networks were then manufactured from the dispersions and tested for DP, water-uptake and mechanical behavior. The characteristics of the CNFs and their networks are discussed in the context of initial wood properties.

Materials
Wood from hybrid aspen trees (Populus tremula x tremuloides) was used as feedstock for CNF isolation. Two transgenic lines (T1+T2) with a signi cantly reduced expression of CSI1 and wild type trees (WT) were used in this study. The details of the genetic modi cation and characterization of these trees are described elsewhere . The chemical and physical wood characteristics of the transgenic trees as measured by Bünder et al. (2020) are summarized in Table 1. .
According to the previous study the transgenic trees exhibit no signi cant differences in wood chemical composition, cellulose micro bril angle, wood anatomy and cell wall ultrastructure compared to WT.

Wood oxidation and brillation
The wood samples were ground and sieved through a 300 -500 µm mesh in to obtain powders with uniform particle size. The wood powder (2 g per sample) was dried in oven over night and then soaked in distilled water for 24 hours prior to starting the oxidation. Sodium chlorite (5.0 g/g wood) was added together with TEMPO to the vessel (250 ml) containing the soaked wood powder and a phosphate buffer (pH = 6.8,). The vessels were submerged in a shaking water bath (Cole-Parmer Stuart, Staffordshire, UK) at 60°C for 1h to dissolve the reagents. Reaction was started by adding sodium hypochlorite (2 ml/g wood) and then resubmerged into the shaking bath for 48 hours. This procedure is an adaptation of the methodology described elsewhere  with modi cations to the amount of reagent used and cellulose being in its native wood context instead of in wood pulp.
The nal deligni ed and relatively swollen wood bers were washed to remove reagent residues, dissolved lignin and hemicellulose. This was monitored by reaching a constant conductivity of the washing water. The oxidized samples were then adjusted to identical concentrations of 0.2 wt %.
The suspensions were then homogenized one pass without recirculation using an APV-2000 highpressure homogenizer (SPX Flow Inc, Silkesborg, Denmark) with an average ow rate of 4 mL/s at a pressure of 1000 bar. The suspensions were accurately standardized with respect to solid content after homogenization prior to further characterization.

Viscosity measurement
The viscosity of the oxidized and homogenized suspensions was measured at various concentrations using a Vibro Viscometer (SV-10, A&D Company Limited, Tokyo, Japan) with a tuning fork vibration method at a vibrational frequency of 30 Hz. Concentrated suspensions (0.30 wt %) were analyzed after dilution was performed for respective sample in 0.03 wt % increments down to 0.12 wt %.

Carboxylate content
Carboxylate content was analyzed through the electric conductivity titration method adapted from (Saito and Isogai, 2004). 150 mL of respective CNF suspensions (≈0.2 wt %) were protonated through addition of 0.1 M hydrochloric acid and 0.01 M sodium chloride. The suspensions were titrated with fresh 0.01 M sodium hydroxide through 0.5 mL increments until pH 10 was reached. The amount of carboxylate groups (mmol·g −1 ) induced from TEMPO-oxidation was calculated through Equation (1) below where c is the concentration of the sodium hydroxide, V 2 and V 1 is the volume of added sodium hydroxide at the end and start, respectively, m is the mass of the cellulosic material in the sample, calculated through subtraction of added acid, base and salt mass from the oven dried suspension. The measurements were repeated three times and presented as mean.
Atomic force microscopy Tapping-mode atomic force microscopy (AFM) Veeco MultiMode scanning probe microscope, (Santa Barbara, CA, USA) was used to con rm the presence of nano brils and analyze the morphology.
Antimony-doped silicon cantilevers (TESPA-V2, Bruker, Camarillo, California) with a spring constant of 42 Nm −1 and a nominal tip radius of 8 nm were used for the analysis. Samples were prepared by depositing a small droplet of the CNF suspension (0.001 wt %) on a freshly cleaved mica plate and letting it air dry for ≥ 5 h. Around 100 fully individualized nano brils, total from four micrographs were analyzed for each sample and presented as a mean with corresponding standard deviation.

Nano bril and process yield
Process yield was calculated as the washed gravimetric yield after the chemical treatment relative dry wood mass. The fraction of the process yield that comprised individual, colloidally stable nano brils were further quanti ed through centrifugation of the suspensions at 12000 x g (Beckman Coulter J25i, Beckman Coulter AB, Bromma, Sweden) for 20 min at an approximate consistency of 0.2 wt %. The suspensions were decanted, and the solids retained in the sediment were dried for 24 h at 95°C. This was repeated three times and the nano bril fraction was then calculated according to Equation (2) where m p and m s is the dry precipitate and supernatant mass in sample, respectively. The nano bril fraction ( OE\P ) was presented as fraction of the process yield.
Nano bril surface area estimation Surface area of the nano brils were estimated in the never dried state by analyzing Congo red adsorption of the cellulosic bers as described by (Spence et al., 2010). Brie y, the absorbed dye was quanti ed using a UV-vis spectrophotometer (GENESYS, 10 UV, Thermo Scienti c, Schwerte, Germany) at the absorption maxima 500 nm and then translated to surface area (SSA) using Equation (3) and (4)  Network manufacturing and characterization CNF suspensions were degassed for 30 min in a vacuum oven prior to vacuum ltration on hardened lter paper (Whatman, Grade 52, GE Healthcare, Machelen, Belgium, Pore size: 7 µm). The wet networks were carefully peeled from the lter paper and dried to around 14% solid content and characterized for water-uptake or further pressed (2 kNm −2 ) for 10 h. The dry networks were lastly compression molded using Fontijne Grotnes LPC-300 (Vlaardingen, Netherlands) between two mylar lms (Lohmann Technologies, Knowl Hill, UK) at a pressure of 0.32 MPa and at a temperature of 120°C.
The wet networks were fully swelled in distilled water for 24 h prior to air drying in room temperature where after water-uptake was studied as a function of time when going from an air-dried state to fully hydrated. 0.1 g sections of the hydrogel were used and gravimetrically monitored over time.

Mechanical testing
The dry networks were cut into rectangular samples (40 mm x 5 mm) using a mechanical punch. The samples were then stored at 50% RH for at least two days prior to mechanical characterization. Mechanical testing was performed using a Shimadzu AG-X universal testing machine (Kyoto, Japan) with a 500 N load cell. Testing was performed at a cross-head speed of 10%min −1 , with the strain being measured using a video extensometer (high-speed camera, HPV-X2). The gauge length was set to 20 mm for each measurement. Seven specimens were analyzed for each network batch. The tensile strength was reported as the maximum strength at break. Young's modulus was calculated through the slope of the linear (R 2 = 0.95-0.99) portion in the elastic region (around 0.1-0.5% strain).
Cellulose, porosity, and moisture analysis Cellulose content of the nal networks was estimated through soaking in 17.5 NaOH according to (TAPPI 1999a) where the dissolved fraction was estimated as hemicellulose (and partly degraded low molecular weight cellulose). Porosity (P) of the networks was estimated using Equation (5) where ρ s and ρ c is the density of the sample and theoretical density of cellulose (1.5 g·cm −3 ), respectively. Moisture content of the networks prior to mechanical testing was estimated as the difference in weight before and after 24 h oven-drying at 105°C.

CNF properties
The viscosity of the different CNF suspensions obtained after homogenization was evaluated between 0.12 -0.30 wt % and shown in Fig. 1. The viscosity did not differ between suspensions up to approximately 0.18 wt % (see Fig. 1d). At the higher concentrations, the viscosity of WT (Fig. 1a) suspensions was higher compared to T1 (Fig. 1b) and T2 (Fig. 1c). At 0.30 wt % the viscosity plateaued at 52, 37 and 33 mPas for WT, T1 and T2, respectively. The presented viscosity trends indicate differences between CNFs from WT compared to T1 and T2. Factors that can in uence viscosity include aspect ratio of CNFs ( To identify the origin of the higher viscosity in WT the nano bril yield, carboxylate content, suspension conductivity and surface area were assessed. A higher suspension carboxylate content was detected for WT ( Table 2). The process yield from the initial wood mass showed no signi cant differences between the samples. The percentage of the yield that corresponded to ne nano brils (as determined using centrifugation) was similar for all samples at approximately 50 wt %. However, statistical analysis indicates a small difference for WT compared to T1 and T2 (Table 2). Suspension conductivity was indistinguishable for all samples at around 40 µS/cm. Surface area as estimated using Congo red, which binds to beta-glucan surface, was higher for WT compared to T1 and T2 with a difference in surface area of approximately 20 m 2 ·g −1 . This data suggest that the increased viscosity of WT observed in Fig. 1 originates from a difference in aspect ratio of the CNFs which manifests as a more rapid increase in WT sample viscosity with increasing concentration due to a lower percolation threshold. The transition into a more viscous state at lower concentrations has been demonstrated when comparing relatively short cellulose nanocrystals compared to CNFs similar to the ones isolated in this study (Moberg et al., 2017). The nano bril yield indicates only a slight increase for WT compared to T1 and T2. It is thus likely that the fraction being discarded during ne nano bril yield calculation contributes to the increased brillation metrics since the small difference in estimated CNF yield is unlikely on its own to give rise to the more distinct viscosity and surface area differences. Factors such as suspension concentration and centrifugation parameters affect the estimated fraction of nano brils. It has also been shown that there is a level of experimental variation regarding the ne CNF yield for TEMPO-oxidized CNFs as a function of carboxylate content (Isogai et al., 2011). Based on the carboxylate content observed in this study it is apparent that the chemical treatment has resulted in more oxidation of the retained solids for WT. This is interesting since there were no differences in the chemical composition and thus supports the hypothesis that the CMFs of T1 and T2 are less susceptible to brillation following mild TEMPO-oxidation. The conductivity of corresponding CNF suspensions was similar across all samples indicating no differences in ionic strength of the suspensions and, therefore, that appropriate ltration/puri cation was performed after TEMPO-oxidation. The surface area was approximately 20% higher for WT compared to T1 and T2 which aligns with the observed higher carboxylate content in WT. In accordance with previous reports this is indicative of a greater accessibility of the cellulosic surfaces that gradually becomes higher with increasing degree of brillation (Nge et  To further elucidate the CNF characteristics, AFM analysis was performed. The height of the nano brils comprising the suspensions is shown in Fig. 2 and revealed an average CNF diameter in the range of 0.5 -3.5 nm with no clear consistent differences between WT and T1/T2 regarding both presence of larger aggregates or differences in height of the smallest observed CNFs. From the size distribution (Fig. 2c) height ranges between <1 nm to around 3 nm which is indicative that surface peeling or splitting of the native elementary CMFs l (≈ 3 nm) is taking place (Li and Renneckar, 2009;Usov et al., 2015). This can be explained by a disconnect between degree of brillation and bril height (on single nanometer levels) (Lindström, 2017), which makes the observed CNF diameter more complicated than just being a generic brillation metrics. An example of this effect is the observation of a lower CNF diameter for samples which had an overall reduced degree of brillation (Jonasson et al., 2021). Another interesting aspect that can be observed from the micrographs is the indication of shorter bril fragments from the transgenic lines. This is exempli ed in Fig. 2 (a and b) for CNF from T1 where individual submicron brils are apparent to a larger extent compared to WT where the CNFs appear more seamless and interconnected. This could be a manifestation from variation in DP of the cellulose that comprises these brils, where DP is inherently correlated to nano bril length (Shinoda et al., 2012). This would explain the difference in viscosity despite the relatively similar nano bril yield, ionic strength, pH and volume fraction. This is further elaborated upon in the next section when analyzing corresponding CNF networks where network behavior is signi cantly in uenced by DP of the CNFs that comprise them and subsequently a good tool for assessing variation (Henriksson et al., 2008).

CNF network characterization
The manufactured networks were tested for their capacity to bind water prior to hot pressing into dry networks. Water uptake is shown in Fig. 3 together with the visual appearance before and after swelling to max capacity. Networks made from the CNF suspensions could absorb 900-1100% of their initial weight with a slightly higher capacity for WT compared to T2. The capacity of T1 was indistinguishable from both WT and T2 where values were within the error margins of T2 and WT. A higher capacity for WT is supported both by the slightly higher degree of carboxylation of the solids in the WT-CNF suspensions and their corresponding increased surface area ( Table 2).
CNFs with gradually increased degree of carboxylation have a higher a nity for swelling in water (Patiño-Masó et al., 2019). The increased water uptake combined with an overall higher surface area is indicative that the increased cellulose accessibility in WT-CNF suspensions persists throughout network formation after ltration and drying of the samples. This is attributed to a minimized degree of horni cation during network manufacturing. This is mediated by the absence of hot-pressing when evaluating these wet networks, which does not irreversibly reduce water-uptake upon drying as much as more aggressive drying methods, e.g., oven-drying or hot-pressing.
The tensile strength, elongation-at-break and Young's modulus of the nal dense and homogenous hotpressed networks is shown in Fig. 4a and b shows a higher tensile strength and elongation-at-break for WT compared to T2 and T1. Interestingly, Young's modulus (Fig. 4c) showed the opposite behavior with a lower Young's modulus for WT compared to T1/T2. The increased tensile strength and elongation-at-break for WT relative T1 and T2 is supported by the data where WT-CNF showed higher nano bril yield, viscosity, degree of oxidation and surface area ( Table 2). All being factors which are relatable to greater network performance (Meng and Wang, 2019). The increased modulus for T1/T2 compared to WT indicates an opposite trend and supports the interpretation of shorter nano brils in T1 and T2. Shortening of nano brils have been reported to increase the stiffness of corresponding networks (Henriksson et al., 2008). An extreme case of this effect has been observed when comparing stiffness of networks made from CNF suspensions obtained from harsh (TEMPO/NaBr/NaClO) and mild (TEMPO/NaClO/NaClO 2 ) oxidation of wood powder, where the harsh treatment resulted in a network modulus that was approximately 30% higher compared to the one from the mild treatment (Jonasson et al. 2020a). Furthermore, the DP was reduced already in the initial transgenic wood cellulose as determined by using size-exclusion chromatography . It is thus possible that the decreased native cellulose DP for the transgenic lines is preserved throughout processing and manifested in the nal networks as decreased strain-to-break and increased stiffness for corresponding networks.
To further elucidate the potential in uence of DP on the mechanical behavior of the networks the DP V was estimated after dissolution in Cuen -a system that is compatible with TEMPO-oxidized CNFs (Isogai et al., 2011). The results are presented in Table 4 together with moisture content, porosity and α-cellulose content. The results indicate differences between WT and T1/T2 where WT CNFs have a higher DP compared to T1 and T2 both prior to and after nano brillation. The moisture content, porosity and cellulose content of the network were all relatively similar and are therefore likely not contributing to the difference in mechanical behavior.

Conclusion
Wood from transgenic trees with reduced expression of CSI1 was subjected to nano brillation using direct TEMPO-oxidation (pH = 6.8) followed by high-pressure homogenization. The CNF suspensions and their corresponding networks were studied and compared with the CNFs from the wild type (control tree).
CNF suspensions derived from transgenic trees showed lower degree of carboxylation, similar process yield, slightly decreased nano bril yield and surface area. Viscosity increased more rapidly as a function of concentration for the wild type indicating a difference in aspect ratio and capacity for network formation. This was further detected in manufactured networks where the transgenic trees resulted in networks with reduced tensile strength, elongation-at-break and degree of polymerization, but with increased stiffness. Consequently, it was shown that the isolated CNFs were in uenced by the reduced expression of CSI1 in developing wood and the augmented wood properties of the transgenics. This study thus shows that it is possible to in uence the CNF quality by altering the wood properties of the feedstock and that native cellulose structure in uences the nal isolated CNF properties. Figure 1 Viscosity as a function of time for different CNF suspensions obtained from a) WT, b) T1, c) T2 and d)

Declarations
Corresponding individual value of viscosity as a function of suspension concentration.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.