Controlled Polyelectrolyte Association of Chitosan and Carboxylated Nano-Fibrillated Cellulose by Desalting

We prepared chitosan (CHI) hydrogels reinforced with highly charged cellulose nanofibrils (CNF) by the desalting method. To this end, the screening of electrostatic interactions between CHI polycation and CNF polyanion was performed by adding NaCl at 0.4 mol/L to the chitosan acetate solution and to the cellulose nanofibrils suspension. The polyelectrolyte complexation between CHI polycation and CNF polyanion was then triggered by desalting the CHI/CNF aqueous mixture by multistep dialysis, in large excess of chitosan. Further gelation of non-complexed CHI was performed by alkaline neutralization of the polymer, yielding high reinforcement effects as probed by the viscoelastic properties of the final hydrogel. The results showed that polyelectrolyte association by desalting can be achieved with a polyanionic nanoparticle partner. Beyond obtaining hydrogel with improved mechanical performance, these composite hydrogels may serve as precursor for dried solid forms with high mechanical properties.


Introduction
Natural polymers such as polysaccharides are gaining interest and application niches over synthetic polymers for a variety of reasons. They are dispersible or soluble in aqueous solvents, they are non-toxic, issued from renewable sources, and can be degraded in different biological media.
However, in the particular field of biomaterials, polysaccharides are of major interest for their cytocompatibility and resorption as materials constituting implants. They exhibit a wide range of physico-chemical properties with specific biological behaviors and can be reproducibly processed in absence of organic solvent and toxic crosslinkers, the use of the latter complicating the marking procedures of medical devices.

Chitosan Sourcing and Characterization
Chitosan from squid pen chitin was supplied by Mahtani Chitosan (Veraval, India; batch type 144). 1 H NMR analysis was used to determine the degree of acetylation (DA) close to 4.0 ± 0.5% following the methodology of Hirai et al. [42]. A weight-average molar mass Mw of 580 ± 50 kg/mol and a dispersity Ð of 1.5 ± 0.3 were determined as previously described [43], using size exclusion chromatography (SEC) coupled online with a differential refractometer (Optilab T-rEX, Wyatt; λ = 658 nm) and with a multi-angle laser light scattering (SEC-MALLS) detector (Dawn-HELOES II, Wyatt; λ = 664 nm) at the Chromatography Center of Institute of Chemistry of Lyon (ICL).
Preparation of chitosan acetate solutions: Chitosan powder was dispersed in deionized water. Then, a stoichiometric amount of acetic acid vs. amine groups was added to protonate chitosan, accounting for a DA value of 4% and a water content close to 8% w/w, determined by thermogravimetry analysis (mass loss between 30 • C and 200 • C). This acetic acid amount allowed complete solubilization of chitosan under mechanical stirring in a closed reactor for 12 h. Chitosan solutions were obtained at different concentrations, ranging from 1.7% w/w to 5% w/w.

Cellulose Nanofibers
Carboxylated nanofibrillated cellulose suspensions were provided by the Centre Technique du Papier (CTP, Grenoble, France). These suspensions were obtained from a bisulfite birch pulp which was pretreated with TEMPO to introduce surface carboxylic groups, and nanofibrils were subsequently mechanically individualized through two passes in a homogenizer at 1500 bar. The resulting suspension had a gel like texture and a solid content of 1.2 % w/w. The suspension diluted in deionized water at a concentration of 0.44% w/w was further purified by dialysis against deionized water using a membrane with cut-off of 6-8 kD, until the conductivity of the dialysis bath was constant close to 0.9 mS/cm.
Carboxylate content of CNF: The carboxylate content of the nanofibrillated cellulose was determined by conductimetric titration with NaOH solution in the presence of an excess of hydrochloric acid (pH~2.5). The low conductivity plateau ∆V was measured and allowed to calculate the ionic exchange capacity [18]. The carboxylate content can also be expressed as the molar ratio of oxidized repeat units of cellulose, i.e., degree of oxidation (DO) and calculated from: where m AGU = 162 g/mol is the molar mass of the anhydroglucose repeat unit of cellulose, m Ox . AGU = 198 g/mol is the molar mass of oxidized repeat unit, C NaOH is the molar concentration of NaOH used for the titration, ∆V = V 2 − V 1 is the volume extent of the low conductivity plateau, and m (g) is the mass of the dry CNF sample. The value of the DO was close to 0.22 mol/mol and the corresponding equivalent ionized carboxylate content (ion exchange capacity) of 1.3 meq/g. For hardwood pulp, it was reported that the reaction reached a threshold at 1.6 meq/g corresponding to the oxidation of all the primary hydroxyls lying on the surface of the nanofibrils. Herein, about 75% of the primary surface hydroxyls were oxidized to carboxylate. Attenuated total reflection Fourier transform infrared spectroscopy (ATR-FTIR): ATR-FTIR analyses were performed on dry CNF films. FTIR spectra were recorded with a Thermoscientific iS-10 operating in ATR mode with a germanium crystal (64 scans, wavenumber range 670 to 4000 cm −1 , normal contact, resolution of 4 cm −1 ).
Transmission electron microscopy observation of CNFs: Drops of about 0.001 wt% of aqueous suspension of CNCs were deposited onto glow-discharged carbon-coated 3 mm grids for transmission electron microscopy (TEM). After 2 min, the liquid in excess was wicked off with a filter paper and a drop of Uranyless staining solution was deposited on the specimen. After 2 min, the stain in excess was wicked off and the remaining thin liquid film allowed to dry. The specimens were observed with a Philips CM120 CRYO TEM operated at 120 kV.

Rheological Study
The viscoelastic properties of mixtures of CNF suspensions and chitosan solutions as well as hydrogels produced after neutralization in NaOH solution 1 M were studied by dynamic-mechanical rheological measurements. These measurements were carried out using an ARES rheometer (TA Instruments) operating with a plate-plate geometry (diameter of 25 mm) at room temperature. The strain amplitude was monitored to ensure the measurements, which were carried out within the linear viscoelastic region, resulting in a storage modulus (G ) and loss modulus (G") independent of the strain amplitude. Thus, we carried out angular frequency sweep measurements at an applied strain of 1%. We repeated these analyses at least three times for each type of hydrogel. Figure 1 shows the evolution of suspensions of chitosan with increasing NaCl content, for a solution at a chitosan concentration of 3% (w/w). Screening effects at high salt concentration induces a precipitation of chitosan [40]. Thus, the accessible NaCl concentration was restricted to be below 0.4 M. Above this limit, the solution became highly turbid and scattered light at 650 nm and 400 nm. film allowed to dry. The specimens were observed with a Philips CM120 CRYO TEM operated at 120 kV.

Rheological Study
The viscoelastic properties of mixtures of CNF suspensions and chitosan solutions as well as hydrogels produced after neutralization in NaOH solution 1 M were studied by dynamic-mechanical rheological measurements. These measurements were carried out using an ARES rheometer (TA Instruments) operating with a plate-plate geometry (diameter of 25 mm) at room temperature. The strain amplitude was monitored to ensure the measurements, which were carried out within the linear viscoelastic region, resulting in a storage modulus (G') and loss modulus (G'') independent of the strain amplitude. Thus, we carried out angular frequency sweep measurements at an applied strain of 1%. We repeated these analyses at least three times for each type of hydrogel. Figure 1 shows the evolution of suspensions of chitosan with increasing NaCl content, for a solution at a chitosan concentration of 3% (w/w). Screening effects at high salt concentration induces a precipitation of chitosan [40]. Thus, the accessible NaCl concentration was restricted to be below 0.4 M. Above this limit, the solution became highly turbid and scattered light at 650 nm and 400 nm. Solubility study of chitosan acetate solutions with added NaCl. The concentration of chitosan was 3% w/w. Solubility is quantified by the evolution of absorbance at 400 nm (adapted from [40]), and precipitation is displayed by images of the samples submitted to a laser beam (λ = 650 nm). Solubility of chitosan acetate in saline aqueous solution was limited to the NaCl concentration range from 0 to 0.4 M.

Cellulose Nanofibers (CNF) Microstructure and Stability of Suspensions in the Presence of NaCl Salt
The chemical structure of the TEMPO oxidized CNF and the starting non-oxidized cellulose nanofibers were investigated by ATR-FTIR (Figure 2a). The spectra displayed a Figure 1. Solubility study of chitosan acetate solutions with added NaCl. The concentration of chitosan was 3% w/w. Solubility is quantified by the evolution of absorbance at 400 nm (adapted from [40]), and precipitation is displayed by images of the samples submitted to a laser beam (λ = 650 nm). Solubility of chitosan acetate in saline aqueous solution was limited to the NaCl concentration range from 0 to 0.4 M.

Cellulose Nanofibers (CNF) Microstructure and Stability of Suspensions in the Presence of NaCl Salt
The chemical structure of the TEMPO oxidized CNF and the starting non-oxidized cellulose nanofibers were investigated by ATR-FTIR ( Figure 2a). The spectra displayed a broad absorption band around 3350-3150 cm −1 , corresponding to the O-H stretching, which indicates the formation of intermolecular H-bonds in both nanofiber samples. The peaks at around 2850-2930 cm −1 are attributed to the asymmetric and symmetric stretching vibrations of the C-H group, respectively. In the non-oxidized CNF, absorption around 1635 cm −1 was observed, attributed to -OH bending [44]. The ATR-FTIR spectrum of the TEMPO-mediated oxidized CNF showed the specific band of the C=O stretching signal, around 1600 cm −1 , evidencing the presence of carboxylate ions in the TEMPO-oxidized CNFs [45]. Then, the deformation vibrations of the CH, CH 2 groups were visible around 1300-1400 cm −1 in both spectra, but with reduced intensity of the asymmetric deformation of methylene normally appearing higher than 1400 cm −1 in the spectrum of TEMPOoxidized CNF as expected after carboxylation of the CH 2 OH groups of cellulose. Then, the asymmetric stretching of the C-O from primary, secondary alcohol, and (hemi)acetal groups presented their characteristic bands at 1000-1300 cm −1 , and the C-O-C symmetric stretching at around 800-900 cm −1 . Finally, the fingerprint pattern of polysaccharides was clearly observed at the range between 1300 and 800 cm −1 . Figure 3a displays the gel aspect of CNF suspensions. The tube inversion test showed a gel-like behavior for concentrations higher than~1% w/w, as a result of the formation of a network of interacting cellulose fibrils.
The TEM micrograph of the TEMPO-oxidized CNF (Figure 3b) shows well individualized fibrils of 200-300 nm length and 6-8 nm width, in good agreement with previous reports from the literature [35].
CNF suspension stability (pH = 5.1) was qualitatively studied as shown in Figure 4. Laser scattering appeared similar in the NaCl concentration range from 0 to 0.8 M. As a result, in order to obtain the highest screening effect without precipitation, a NaCl concentration of 0.4M could be determined. In these conditions, the stability of salted CNF suspensions and CHI solutions were obtained over 1 week.
Lastly, we investigated the impact of the pH of CNF suspensions and CHI solutions on their turbidity. Gelation of CNF suspensions occurred at pH~3 and below, in good agreement with the pKa range of glucuronic acids (typically between 2.7 to 3.3) [46,47]. The pH of CHI solutions ranged from 5.1 to 4 with concentrations ranging from 1.5 to 5% w/w. The stability of CHI was preserved up to a pH close to 5.8, a value slightly below the apparent pKa~6.2 of CHI at low DA [48]. Thus, the pH window for the association of chitosan and CNF without precipitation was located between their pKa values, typically between 4 and 5.5. As a result, in order to mix the CHI solution with the CNF suspension, the pH and NaCl concentration of both media should stay in a well-defined processing window.

Formulation and Desalting of Polyelectrolyte Mixtures of CHI Solutions and CNF Suspensions
We easily prepared homogeneous CHI/CNF mixtures by imposing a pH of 4 both on the CHI solution and on the CNF suspension by the addition of a diluted acetic acid solution. For a proof of concept, the values of the CHI and CNF concentration tested were chosen below the gelation threshold in both cases. This ensured excellent mixing conditions with magnetic stirring, but higher concentrations are assessable for both components with mechanical blending. The resulting systems are described in Table 1. As expected, when comparable systems were prepared without NaCl addition, precipitation resulted in a macroscopically heterogeneous system with no satisfactory dispersions of CNF within the chitosan solution.        Then, complete gelation was induced by submerging the dialysates in NaOH 1 M for 12 h, and further water washing for 24 h. Syneresis was negligible (as in the pure chitosan hydrogels of same low concentration, low DA, and high mean molar masses) and thus, the washed gels were obtained using C CHI = 1.01% and C CNF ranging from 0 to 0.15%. Figure 5 displays the viscoelastic behavior of the mixtures 1 and 2 before dialysis (circles) and after neutralization in the alkaline bath (squares). Similarly, all formulated systems exhibited a solution-like behavior after mixing in presence of NaCl, but exhibited a gel behavior after desalting and neutralization, with G >> G" in the entire investigated angular frequency range.

Viscoelastic Properties of Resulting Hydrogels
A comparison of the elastic shear moduli of hydrogel samples prepared from the mixtures detailed in Table 1 is given in Figure 6. The reinforcement effect is clear as the elastic modulus increased with CNF content. Such behavior was previously analyzed [32] through the Einstein's generalized relation (Equation (2)): where G 0 is the value of the modulus of the non-reinforced reference system, ϕ v,CNF is the volume fraction of CNF filler, and ϕ m,CNF = C CNF is the mass fraction of CNF filler, ρ CNF~1 .43 g/cm 3 is close to the density of nanofibrilated cellulose [49], and k E is the Einstein coefficient (~2.5 for spherical objects) quantitatively accounting for the reinforcement effect. The value found for k E was close to 120, corresponding to an extremely strong reinforcement effect. Since the CNF were not oriented by the hydrogel processing technique, the reinforcement effect should be related to crosslinking of the chitosan matrix by nanofibrils, as a result of CHI and CNF physical interactions. Chitosan in hydrogel samples is neutralized, hence its polyelectrolyte character is not directly responsible for these interactions. Nevertheless, electrostatic interactions may be understood as the first driving force for chitosan adsorption onto CNF nanofibers during the polyelectrolyte complexation. After neutralization of the complex, H-bonding and hydrophobic interactions must take over to ensure physical crosslinking, as is the case for the complexation of chitosans with other macromolecular polysaccharide polyanions [50]. In agreement with previous studies [31,32,36], the morphology of reinforced CNF/CHI hydrogels obtained in this work can be schematized as (i) a first fibrillar nanocellulose network resulting from H-bonds between nanofibers and chitosan chains bridging nearby nanofibers, and (ii) a second chitosan hydrogel network where isolated fibrils portions play the role of physical crosslinks of high functionality for chitosan chains. The formation of CHI/CNF interaction in both networks is driven by electrostatic interactions, that are in turn controlled by the charge density of the initial polyelectrolytes, i.e., by the degree of acetylation (DA) of chitosan and the degree of oxidation (DO) of nanocellulose. After the neutralization of chitosan, the remaining interactions of chitosan adsorbed onto the CNF nano-rods are due to H-bonds and hydrophobic interactions. A comparison of the elastic shear moduli of hydrogel samples prepared from th mixtures detailed in Table 1 is given in Figure 6. The reinforcement effect is clear as th elastic modulus increased with CNF content. Such behavior was previously analyzed [3 through the Einstein's generalized relation (Equation (2): where G0 is the value of the modulus of the non-reinforced reference system, φv,CNF is th volume fraction of CNF filler, and φm,CNF = CCNF is the mass fraction of CNF filler, ρCNF 1.43 g/cm 3 is close to the density of nanofibrilated cellulose [49], and kE is the Einste coefficient (~2.5 for spherical objects) quantitatively accounting for the reinforcement e fect. The value found for kE was close to 120, corresponding to an extremely strong rei forcement effect. Since the CNF were not oriented by the hydrogel processing techniqu the reinforcement effect should be related to crosslinking of the chitosan matrix by nan fibrils, as a result of CHI and CNF physical interactions. Chitosan in hydrogel samples neutralized, hence its polyelectrolyte character is not directly responsible for these inte actions. Nevertheless, electrostatic interactions may be understood as the first drivin force for chitosan adsorption onto CNF nanofibers during the polyelectrolyte complex tion. After neutralization of the complex, H-bonding and hydrophobic interactions mu The reinforcement effects of CNF/CHI hydrogels may also be compared to pure CHI hydrogels at same total polymer concentration. The elastic shear modulus of CHI hydrogels may vary with mass concentration according to: where G 00 should be close to the measured value at C CHI = 1% w/w (i.e., G 0~0 .31 kPa) and β is an exponent previously found close to 4.4 by AFM force spectroscopy [51] for a chitosan of same mean molar mass and DA. Our rheological analyses combined with previous viscoelastic measurements [52] (see Sample H3b: C CHI = 1.6% w/w, E = 6 kPa; sample H4: C CHI = 2.6% w/w, E = 31 kPa, and assuming E = G/3, i.e., the Poisson's coefficient ν = 0.5) yield similar values (G 00 = 0.29 kPa, β = 4.15, R 2 = 0.9998) after a linear regression in logarithmic scales. As a result, for a pure CHI hydrogel 1.2% w/w, G'/G 00 > 2 and the resulting shear modulus ratio G'/G 0 should largely exceed the values displayed in Figure 6. This can be explained, at a macromolecular level, by the involvement of all chitosan chains contributing to the CHI-CHI network via H-bonds, hydrophobic interactions, and the formation of nanocrystallites [53]. On the other hand, CNF physical crosslinking with CHI only occurs at the surface of the nanofibrils. Indeed, cellulose chains entrapped within the cellulose nanofiber cores should not contribute to crosslinking with the matrix: considering a given mass of polymer, the reinforcement effect of CNF in CHI hydrogels stays below the effect of CHI concentration. However, for materials obtained in the dry state, chitosan concentration of the parent hydrogel does not significantly impact the final hydration, density, interactions, and microstructure. As a result, composite reinforcement effects of CNF are more apparent in the solid (dry) state. Thus, the interest of CHI/CNF associations is particularly attractive in the case of films or fibers prepared from CHI/CNF associations [38].
Polymers 2021, 13, x FOR PEER REVIEW 10 of 14 tosan and the degree of oxidation (DO) of nanocellulose. After the neutralization of chitosan, the remaining interactions of chitosan adsorbed onto the CNF nano-rods are due to H-bonds and hydrophobic interactions. Figure 6. (*): Evolution of the elastic part of the shear modulus of hydrogels obtained from the controlled polyelectrolyte assembly of CHI and highly charged CNF by desalting, in normalized scales. The value of G0 for non-reinforced systems obtained in this work was close to 0.31 kPa. These results are compared with literature data (□) for CHI/CNF hydrogels obtained with weakly charged CNF [32].
The reinforcement effects of CNF/CHI hydrogels may also be compared to pure CHI hydrogels at same total polymer concentration. The elastic shear modulus of CHI hydrogels may vary with mass concentration according to: where G00 should be close to the measured value at CCHI = 1% w/w (i.e., G0 ~ 0.31 kPa) and β is an exponent previously found close to 4.4 by AFM force spectroscopy [51] for a chitosan of same mean molar mass and DA. Our rheological analyses combined with previous viscoelastic measurements [52] (see Sample H3b: CCHI = 1.6% w/w, E = 6 kPa; sample H4: CCHI = 2.6% w/w, E = 31 kPa, and assuming E = G/3, i.e., the Poisson's coefficient ν = 0.5) yield similar values (G00 = 0.29 kPa, β = 4.15, R 2 = 0.9998) after a linear regression in logarithmic scales. As a result, for a pure CHI hydrogel 1.2% w/w, G'/G00 > 2 and the resulting shear modulus ratio G'/G0 should largely exceed the values displayed in Figure 6. This can be explained, at a macromolecular level, by the involvement of all chitosan chains contributing to the CHI-CHI network via H-bonds, hydrophobic interactions, and the formation of nanocrystallites [53]. On the other hand, CNF physical crosslinking with CHI only occurs at the surface of the nanofibrils. Indeed, cellulose chains entrapped within the cellulose nanofiber cores should not contribute to crosslinking with the matrix: consider-

Conclusions
In this work, we designed a controlled polyelectrolyte association between highly charged nanofibrillated cellulose (CNF) and chitosan (CHI) of low DA in excess. It required evaluating the stability of CNF suspensions and CHI solutions in the presence of a screening salt (NaCl) and the use of a multiple step dialysis against aqueous solutions with decreasing salt concentrations.
In comparison with previous literature results, we obtained hydrogels with higher mechanical properties using a highly charged CNF vs. weakly charged cellulose nanofibrils [32]. This effect should reflect the high physical crosslink capacity of highly charged CNF interacting with chitosan chains adsorbed onto their surface, thanks to the initial polyelectrolyte complexation of CHI and CNF.
Future studies should focus on the preparation strategies of systems at higher concentrations (for CHI and CNF), including the processing of dry fibers or films, where the composite reinforcement effects are expected to yield higher mechanical properties. In particular, yarns prepared by stretching and drying reinforced hydrogels should yield bioinspired solid materials with enhanced mechanical properties, with practical interest in the field of textile knitting.