Ion-Specific Gelation and Internal Dynamics of Nanocellulose Biocompatible Hybrid Hydrogels: Insights from Fluctuation Analysis

1. Introduction

Hydrogels are defined as three-dimensional networks capable of absorbing a large quantity of liquids. Given such broad definition, their composition and physico-chemical properties are highly tunable, as are their preparation and characterization methods[1]. So hydrogels are considered versatile materials with wide range of applications[2]. For instance, by using the appropriate components, smart functional hydrogels can respond to various external stimuli (such as pH, temperature, electric and mechanical stimuli) [3,4], or enable applications in advanced electronics [5], food industry [6] and biomedicine[7,8,9]. The expected increase in large scale applications requires to move toward sustainability and the use of natural polymers as hydrogel components offers an eco-friendly opportunity. In this regard, nanomaterials derived from cellulose sources are ideal candidates [10,11,12,13] with cellulose nanocrystals (CNC) showing a polymer-like behavior which quickly jellify upon addition of multivalent cations. The ionotropic gelation process is mediated by the interaction of the cations with the negatively charged or polar groups of CNC (mainly hydroxyls, carboxyls, carbonyls, sulfates, depending on the CNC’s synthesis procedure) and allow precise control over the macroscopic gel structure [14,15,16] and the synthesis of hydrogels with controlled properties[17,18,19,20,21]. However, to overcome the intrinsic CNC stiffness and to obtain the desired mechanical properties, CNCs are often used as fillers of softer polymers [22,23,24,25,26,27]. The development of nanocellulose hybrid gels as substrates for cell cultures requires both an optimized synthesis process and a thorough characterization of their physical properties. In this study, we present a simplified synthesis approach that enhances the reproducibility and scalability of the hybrid material while maintaining its structural integrity and biocompatibility. To further investigate the gel’s physicochemical properties, we introduce a novel analysis of the correlation function obtained from dynamic light scattering (DLS) measurements. This refined analysis provides deeper insight into the gel network dynamics, shedding light on particle interactions and aggregation behavior. By linking the synthesis process with advanced characterization, we establish a more comprehensive understanding of how material properties influence cell culture performance, paving the way for improved biomaterials in tissue engineering and regenerative medicine applications. A simplified synthesis to fabricate an hybrid gel made of CNC and aminated PEG is an important result since this formulation strategically integrates materials with complementary chemical and mechanical properties and it enables the creation of versatile hydrogel platforms. In fact, while cellulose contributes with carboxyl groups, amino-PEG introduces valuable amine functionalities to the matrix, establishing a dual-functionality system. The presence of both amine and carboxyl groups is particularly advantageous for bio-applications, as these functional groups can facilitate cell adhesion, protein binding, and potential crosslinking through various chemical reactions such as amide bond formation. Amines functionalities can be directly added to CNC[28,29] but the reaction is time-consuming and care should be taken to avoid unexpected gelation and the cross-linking of amine and carboxylic groups. Thus, this strategy is mainly limited to dilute nanocellulosic solutions. Several studies report on the use of amino-PEG and CNC to form hybrid gels but they mostly use diacrylate-PEG and they require a dedicated step to cross-polymerize PEG and CNC. For example in [30,31], the authors report the synthesis of a photopolymerizable 3D PEG diacrylate (PEGDA) ink in which CNC is used as a filler to increase the mechanical properties. Similarly, [32,33] report the use of PEGDA-CNC composites to form gels with variable Poisson ratios. In [34], a 3D printable bioink is formed by mixing a nanocellulosic solution with a 4-arms start PEG polymer and by photopolymerizing the resulting viscous fluid. Ref. [35] demonstrates the synthesis of an hybrid hydrogel by crosslinking PEG and CNC with acrylamide and N,N’-methylene bis-acrylamide. In [36], 8-arm branched amino PEG polymer is acrylated and photopolymerized under UV irradiation. [37] is the only report that demonstrated the possibility to obtain hydrogels by direct mixing micron-sized cellulosic fibers with branched, tetra-arms amino-PEG. Our approach takes a step forward compared to what has been shown so far and demonstrates the possibility to synthesize biocompatible hybrid gels by simply mixing CNCs and linear amino-PEG. Moreover, although the biocompatibility of hybrid gels made of CNCs and PEG have been demonstrated [38,39,40], similar results are unreported for amino-PEG and CNC hybrid gels.

From the perspective of characterization methods, we have developed an innovative approach to analyze DLS measurements that exploits the statistical properties of the 2^nd order correlation function measured from a commercial, DLS instruments at single scattering angle to correlate information about the sol-gel transition and the material composition. To demonstrate this possibility we exploit the peculiar characteristic of alginates to jellify in the presence of $C{a}^{2+}$ but not with $M{g}^{2+}$[41,42] and we show that the statistical properties of the measured 2^nd order correlation function contains information about the sol-gel transition dynamics and also about the material composition. We define a phenomenological Figure of Merit (FOM), based on the short-time values of the 2^nd order correlation function, and the comparison among the two families of materials (i.e. the $C{a}^{2+}$ crosslinked gel and the $M{g}^{2+}$ suspension) unveil the differences in their internal dynamics.

The cation complexes of alginates have an intriguing nature and their role in the hydrogel network formation is currently matter of research [43].

Different methods spanning from the atomic scale up to the microscale are necessary to characterize the structural morphology, composition, and mechanical properties of hydrogels. The advantages and disadvantages of these methods have been recently reviewed concluding that the hydrogel characterization remains challenging also because of the hierarchical length scales of the network structure [44]. In general, the most popular approach to investigate the macroscopic mechanical properties is the rheological analysis. This method allows for the investigation of sol-gel transition and provides quantitative information about the macroscopic properties of the material [45]. Yet, routine rheology experiments cannot sense the fine details of the gel structures and different materials can share similar macroscopic mechanical properties (like the elastic, and viscous modulus) while differing in their local structure. Advanced rheological techniques such as dynamic shear rheology and large amplitude oscillatory shear experiments are very promising investigative tools of the dynamics of gelation of thixotropic suspensions. However, the complexity of data analysis and the necessity of integrative structural information restrict their use to few material and require skilled researchers [46].

Dynamic Light Scattering (DLS) that is a family of techniques that exploit scattering of light to investigate structural and dynamical details of colloidal particles and gel dynamics. Despite their basic theory has been thoroughly developed starting from 1950’s [47,48,49,50], the complex interpretation of the data and the strong assumptions on which models are based make DLS a lively research field even today with several variants proposed in recent years to overcome some of the limits of the techniques [51,52,53]. One of the weak points of experimental data interpretation of any DLS technique is the assumption about the type of motion of the particles and/or their size distribution. In fact, these methods require a transformation from the experimentally measured quantity (the light intensity that produces the 2^nd order correlation function) to the 1^st order correlation function that relates how the electromagnetic field interacts and is scattered by the illuminated sample structure. Standard DLS measure cannot handle highly turbid samples (as often gels are), since they assume single scattering event and are thus compatible only with dilute solutions. Moreover systems undergoing irreversible dynamics (non ergodic) require specialized setups where scattering is averaged by moving the sample and investigating different positions. On the other hand, diffuse wave spectroscopy provides information about highly scattering samples, but in this case, a large number of scattering events is required to properly describe the light scattering as a diffusive process[54,55,56]. The intermediate range, that is the dynamic one formed during the sol-gel transition, is poorly investigated since there are no analytical models able to handle the complexity of the transient between the sol and the gel regimes. In order to fill this gap and investigate the dynamics at the nanoscale of a solgel transition we used the statistical properties of the light scattered during a DLS experiment and proved tha they contain relevant information about the dynamics of the sol-gel transition and from its analysis we can distinguish samples that differ in the relative amounts of their components.

2. Materials and Methods

2.1. Materials

Commercial wood pulp Celeste 85 has been kindly provided by SCA (Sweden). TEMPO 2,2,6,6-tetramehylpiperidine-1-oxyl (TEMPO) 98+% purchased by Alfa Aesar; NaBr 99% and NaOH 1 M produced by Carlo Erba and NaClO solution with 6-14% active chlorine produced by EMPLURA were used. Bis(3-aminopropyl) terminated PEG, with average molecular weight of ∼ 1500 Da, purchased by Sigma-Aldrich and alginic acid sodium salt solution with very low viscosity, purchased by Thermo Fisher Scientific Chemicals were used to realize CNC hybrid hydrogels. All chemicals have been used as received, without further purification. MilliQ grade water has been used for all preparations and processes. HeLa cells were cultivated in DMEM containing phenol red and supplemented with 10% Fetal Bovine Serum (FBS), 2 mM glutamine, 100 U/ml penicillin and 100 $\mu$g/ml Streptomyosin (Pen-Strep). PBS (pH 7.4), Trypsin-EDTA 0.25% and phenol red were used to wash cells before dissociation and for trypsinization, respectively. Trypan Blue Solution, 0.4% was used to stain alive cells to be counted in a Burker chamber hemocytometer. For MTT assays, phenol red-free DMEM supplemented with FBS, Pen-Strep and L-Glutamine was used since it is proven that phenol red compromises the outcome of MTT assays. CyQuanTTM MTT kit, was used to assess cell viability. All the listed chemicals were purchased by Gibco-Thermo Fisher Scientific.

2.2. Synthesis of CNC

TEMPO-oxidized CNC has been produced with a slightly modified receipt from[57]. Briefly: 10 g of cellulose pulp are swollen in 500 ml of distilled water for 5 hours at room temperature under stirring and then sonicated in bath for 30 min. 160 mg of TEMPO and 1 g of NaBr salt are added and the final volume is brought to 1 l. The oxidation is started by addition of 35 ml of NaClO and left to proceed for 3 hours keeping the pH at 10.5 by addition of 1 M NaOH. The slurry is rinsed 5 times by plenty of distilled water and divided in 35 ml aliquots. Each aliquot is sonicated with a Bandelin Sonopuls HD 2200 ultrasonic homogeneizer equipped with a 13 mm tip, settled at 160W and then vacuum filtered over a 5 $\mu$m cellulose acetate filter (Sartorius, Germany). The CNC suspension is heated at 70 °C in a water bath under magnetic stirring until a concentration of 7 mg/mL is obtained. Finally, the suspension is again vacuum filtered to remove residual impurities.

2.3. Synthesis of Hybrid Gels

CNC-PEG hybrids (NP) were prepared in the following way: 200 mg of bis(3-aminopropyl) terminated PEG were dissolved in 10 ml of deionized water and stirred until a homogeneous solution was achieved. Subsequently, HCl (1 M) was added until the pH reaches 7.2 ÷ 7.4. The resulting suspension was then filtered using a 0.22 $\mu$m pore-size filter with a PES membrane. Following filtration, it was mixed with a 7 mg/ml CNC solution in two different volume ratios: CNC:PEG = 2:1 and CNC:PEG = 3:1. CNC-alginate hybrids (NA) were similarly prepared: 2.5 g of alginic acid sodium salt were dissolved in 100 ml of DI water, to get a 2.5% w/v aqueous solution. It was then filtered with a 0.45 $\mu$m pore-size filter with PES membrane and later mixed with the CNC solution in two different volume ratios: CNC:Alginate = 2:1 and CNC:Alginate = 3:1. The several compositions of the hydrogels under study are summarized in Table 1. Hybrid hydrogels were crosslinked by dropping a 1 M $MgC{l}_2$ solution with a pipette along the inner border of polystyrene cuvettes/multiwells already containing the CNC-based solution, such that the final $M{g}^{2+}$ concentration is 47.6 mM (we will round it up to 50 mM for the rest of the paper).

When used for cell culture experiments, all CNC-based solutions (pure CNC and hybrids) were sterilized in an autoclave at 1 bar and 121 ° C for 1 h.

2.4. Hydrogels Characterization

CNC was structurally characterized via Atomic Force Microscopy. NT-MDT SMENA head equipped with NSG-10 tip (elastic const.: 11.8 N/m and typical resonant freq.: 240 kHz) was used to measure the average CNC dimensions.

Fourier Transform Infrared spectra of both pure CNC and hybrid materials were acquired with a Nicolet FTIR microscope at standard spectral resolution (4 cm⁻¹) and a suitable number of acquisitions (typically 64) for a good signal-to-noise value.

Rheological properties of hydrogels were evaluated with a Malvern Panalytic KINEXUS PRO+ rheometer equipped with a plate-plate geometry (diameter: 20 mm). The upper geometry plate was lowered until it was in conformal contact with the top surface of the hydrogel, corresponding to gap distance of 1mm. Samples were preformed, directly transfer on the bottom plate, and equilibrated for 5-minute resting time in the instrument. Dynamic moduli and elastic response under oscillatory stresses were assessed through strain sweep experiments to estimate the LVER (frequency = 1 Hz, strain 0.01% to 10%) and frequency sweep (frequency ranging from 0.1 Hz to 100 Hz, constant stress chosen within the LVER). The share rate dependence of viscosity was studied within a shear ramp $\gamma$ = 1-100 $s^{-1}$. All measurements were repeated at least three times and carried out at a constant temperature of 25 °C. Dynamic Light Scattering (DLS) characterization was carried out with an Anton-Paar Litesizer 500 DLS machine, used in the 90° angle configuration. In this case, 100 $\mu$l of 1 M $MgC{l}_2$ aqueous solution is added directly into the DLS cuvette that contains 2 ml of any of the CNC suspension. The DLS measure starts soon after the addition of the cations and last for hours.

2.5. Cell Culture and Cell Viability

Hela cells were used, since this cell line is considered a versatile model to test the biocompatibility of the substrates[58]. The cells were directly seeded on the material under study and the survival and growth rate were quantified by colorimetric, fluorimetric and cytometric assays.

To assess cell survival on the hydrogels, two 48-wells plates were used to culture HeLa cells, extracted from the same cell line, to perform MTT assays after 48 and 96 hours of incubation. Triplicates of the five CNC-based hydrogels under study were prepared in each multiwell plate by adding 20 $\mu$l of a sterile 1 M MgCl₂ aqueous solution to each well containing 400 $\mu$l of CNC-based solution, for each composition. Prior to use, the five CNC-based suspensions were autoclaved, while the MgCl₂ aqueous solution was filtered with a 0.22 $\mu$m pore size filter. The multiwell plates containing the hydrogels were left unperturbed overnight to complete gelation and then they were rinsed 5 times with sterile DI water before adding 2 ml of pure phenol-red DMEM to each well. In this way, DMEM could diffuse inside the hydrogels and exchange with water under the effect of the concentration gradient. After at least 3 hours of incubation at 37 °C and in a 5% v/v CO₂ atmosphere, the excess supernatant above each hydrogel was removed and substituted with 2 ml of freshly supplemented phenol red-free DMEM to increase the amount of nutrients stored inside the gel. After a second incubation of at least 3 hours, another exchange with supplemented phenol red-free DMEM was done, such that the concentrations of FBS, Pen-Strep and L-Glutamine inside the gels could increase up to the standard ones. HeLa cells were seeded on the hydrogels with an areal cell density of 10³ cells/ml, which was spotted to be the optimal one during a preliminary trial. The same areal cell density was seeded in triplicate on pure polystyrene as reference. MTT assay were performed at two different incubation times (48 h and 96 h) to verify cells interaction with the substrates and their growth rate. After incubation, the MTT protocol was followed to quantify the metabolic activity of HeLa cells: 1 ml of sterile PBS, thermalized at 37 °C, was added to one vial containing 5 mg of MTT, to prepare the 12 mM MTT solution to drop in each well with a cell culture. Indeed, after having substituted the old culture medium with 300 $\mu$l of fresh supplemented DMEM without phenol red in each well, 30 $\mu$l of the MTT solution were added to each well to test. After 4 hours of incubation, 300 $\mu$l of 0.1 mg/ml SDS solutionat pH 2 were added to the culture medium to dissolve the formazan crystals formed by alive cells. After 15 hours of incubation, absorbance at 570 nm of the culture medium in each well was read with an Infinite M200 PRO microplate reader by Tecan. For the reading, the area of each well was sampled 25 times.

2.6. DLS Analysis

According to the objective of our research, we investigated the possibility of using DLS data to unravel the nanoscale dynamics of the 3D hydrogel network formation during an ionotropic gelation process. Below a brief survey of the relevant equations is reported to clarify the idea behind the DLS data analysis performed, while more detailed descriptions can be found in dedicated literature cited above.

In DLS an electromagnetic field is used to track the dynamics of scatterers. If there is a relative movement between the scatterers and the illuminated volume, then the scattered light fluctuates in time. The detector of the DLS tracks these fluctuations and constructs the so called time-average normalized intensity correlation function: $g^{\left(2\right)}(k,\tau )={\displaystyle \frac{{〈I(k,0)I(k,\tau )〉}_T}{{〈I(k,0)〉}^2}}$, where $k={\displaystyle \frac{4\pi n}{\lambda }}sin(\theta /2)$ is the k-vector of the scattered light, n is the solvent refractive index, $\lambda$ is the laser wavelength (elastic scattering is assumed) and $\theta$ is the angle at which light is collected with respect to the incident laser direction. $<\dots {>}_T$ indicates a time average, that is: each measurement (run) is divided in N temporal bins and the numerator of $g^{\left(2\right)}\left(\tau \right)$ is ${\displaystyle \frac{1}{N}}{\sum }_{i=1}^N\chi (i·T)·\chi (i·T+\tau )$, where $\chi$ is the number of photons collected within the $iT$-th bin (the normalization value at the denominator is of no relevance for our purposes). This analysis is performed by a dedicated electronic (the correlator) [47].

In the case of dilute colloidal solution, where scatterers freely diffuse with Brownian dynamics, long enough measures allow to average over all possible configurations of the colloids, that is $g^{\left(2\right)}\left(\tau \right)$ contains all the information to fully describe the average system dynamics. On the contrary, in systems where scatterers are not free to move (eg. within a gel), the temporal averaged $g^{\left(2\right)}\left(\tau \right)$ does not sample all the possible structural configurations of the scatterers and it provides a biased estimation (ie. incomplete) of the system dynamics. These systems are said non-ergodic and a proper sampling of their properties requires an ensemble averaging that is a time average replicated in several different spatial positions, to be sure to collect the dynamics from most of the possible structural configurations and, thus, to get a reasonable estimate of the system dynamics.

The statistical behavior of $g^{\left(2\right)}(\tau \to 0)$ changes depending if the system is (non)-ergodic. In particular[48]:

• $g^{\left(2\right)})(\tau \to 0)=2$ in solutions (ergodic systems)
• $1<g^{\left(2\right)}(\tau \to 0)<2$ in gels
• $g^{\left(2\right)}\left(\tau \right)\sim 1$ in glasses (completely freezed systems)

Generally DLS retries information about the particle size by assuming dilute and ergodic samples. In this case the field scattering time-correlation function is given by the Siegert relation: $g^{\left(2\right)}\left(\tau \right)={\left|g^{(1}\left(\tau \right)\right|}^2+1$, where $g^{(1}\left(\tau \right)$ is the Laplace transform of the characteristic decay rates: $g^{(1}\left(\tau \right)={\int }_0^{\infty }G(\Gamma )·exp(-D\Gamma \tau )d\tau$, where D is the characteristic decay time (ie. the time the colloid takes to flow across the illuminated volume and it depends on both the solvent viscosity and the colloid size). In the case of monodispersed samples there is just a single D and $g^{\left(1\right)}(k,\tau )=exp(-D{k}^2\tau )$. By exploiting the Stokes-Einstein equation, the hydrodynamic radius of the colloid is extracted: $R_H={\displaystyle \frac{k_{B}T}{6\pi \eta D}}$.

The gel building blocks are not free to diffuse. A basic model describes the gel scatterers as fluctuating around an average stable position. The overall light scattering is the superposition of two contributions:

• one coming from the "freezed", averaged configurations of the scatterers,
• one due to the fluctuations around their average position.

Despite this apparently simple decomposition of the motion, the formulation of a model able to quantitatively describe the gel case is a formidable task, yet to be solved and it requires a detailed knowledge of the actual system under study (both physical and chemical details).

Actual DLS correlators, routinely record delays down to 100’s ns. These timescales are much faster than the typical transit time of a colloidal particle over the volume illuminated by the laser. In fact, consider a spherical colloid of about 100 nm size, freely diffusing in water at 298 K. From the Stokes-Einstein equation, D $\sim {10}^{-12}$ m²/s. For a Brownian motion in 1D systems, the average displacement is: $<x^2>=2Dt$. Considering a 20 $\mu$m diameter illuminated volume, the time the colloid takes to cross it is: $t={\displaystyle \frac{<x^2>}{2D}}\sim 10s$. This demonstrates the idea that the first points of the $g^{\left(2\right)}(\tau \to 0)$ contains information about the fast gel dynamics and can be used to monitor its structural evolution as well as its internal dynamics. To ease the notation, in the rest of the paper we define ${\sigma }^2=g^{\left(2\right)}(k,\tau \to 0)$, since this is the relevant value to define our FOM. It should be remarked that the information provided by our approach is complementary and not-overlapping to those provided by rheology. In fact while DLS investigates the microscopic system dynamic evolution, rheology monitors the macroscopic gel slow timescale evolution, even at the sol-gel transition point.

DLS measurements were performed as follows: each gel composition has been measured in triplicate and for each measurement we recorded several hundreds of autocorrelation runs, each of $5s$ of duration. The exact number of runs depends on the sample since the induction time before ${\sigma }^2$ decreases cannot be controlled (it depends on how $MgC{l}_2$ diffuses in the cuvette).

The FOM is defined as the frequency of the fluctuations of ${\sigma }^2$ and it is calculated as follows: we average the first 20 points of each runs to calculate the value of the ${\sigma }^2$ (these points correspond to the time interval between 200 ns and 1 $\mu$s). Then, we define a threshold value (see below) for the ${\sigma }^2$ function (called T) and we locate the time when ${\sigma }^2$ decreases below this threshold (at time $t=t_0$). To normalize the frequency vs the duration of the experiment, we calculate the frequency at which ${\sigma }^2$ crosses T as the number of crossing (c) divided by the duration of the measurement $t_T<t<t_f$: $FOM=c/(t_f-t_T)$, with $t_T$ is the time corresponding to the threshold and $t_f$ the total duration of the experiment. The optimal threshold has been selected using a dynamic programming algorithm [59]: it aims at minimizing a cost function with respect to the optimal partitioning of the dataset:

$$V(\mathcal{T},y):=\sum _{k=0}^{K}c\left(y_{t_k}^{t_{k+1}}\right))$$

(1)

where V is the quantitative criterion function to be optimized (minimized in our case), $\mathcal{T}=\{t_1,t_2,...t_k\}$ is the set of indexes that partition the signal y. Thus, V is given by the sum of the cost functions (denoted by c) for each signal partition $y_{t_k}^{t_{k+1}}$. Since we know the number of thresholds in our data (a single threshold), the optimal solution is computed by:

$$\underset{\left|\mathcal{T}\right|=K}{min}V(\mathcal{T},y)=\underset{0=t_0<t_1<...t_K<t_{K+1}=T}{min}\sum _{k=0}^{K}c\left(y_{t_k}^{t_{k+1}}\right))$$

(2)

3. Results and Discussion

CNC was structurally characterized with AFM and a representative image of CNCs is reported in Figure 1(a). The results of the statistical analysis of CNC sizes is shown in Figure 1(b). Nanocrystals have an average length of about 170 nm, broadly distributed with a St.Dev. of about 120 nm, while their height lies between 2 and 5 nm (averaged over more than 300 CNCs). DLS analysis confirms this data as reported in Figure 1(c). The structure is compatible with the usual one obtained from the TEMPO-mediated oxidation synthetic pathway.

The presence of carboxylic groups on the surface of CNC is confirmed by FTIR spectroscopy and reported in Figure 2: the spectrum of pure and hybrid CNC gels (NA3 and NP3, see Table 1 for the sample composition) have been vertically shifted and rescaled to make relevant spectral signatures overlap. No baseline corrections have been applied. The significant spectral region is below 1800 cm⁻¹. Looking at CNC spectrum, the strong peak at 1600 cm⁻¹ marks the $CO{O}^{-}$ asymmetric stretching, while the one at 1420 cm⁻¹ refers to the symmetric one. The $C-O-C$ stretching is underlined by the peak at 1160 cm⁻¹ [60,61]. The triplet between 1030 and 1110 cm⁻¹ confirms the high crystallinity degree of the CNC [61]. A more detailed assignation of FTIR peaks can be found in the ESI.

The spectra of hybrid materials maintain the typical feature of pure polymers, suggesting that the interactions that form the gel do not significantly modify the vibrational degrees of freedom of the polymer chains. Consequently, the main role in forming the gel seems to be due to the CNC, while CNC-polymers interaction is of less importance. This hypothesis is supported also by the evidence that the stiffer gels are those formed by pure CNC (see Figure 3(a)).

Finally, the zeta potential of TEMPO-oxidized CNC is shown in Figure 1(d) and clearly shows a large negative value of the surface potential, compatible with those typically reported in literature and mainly ascribed to the presence of deprotonated carboxylic groups. Rheological analysis characterized the elastic and viscous moduli ($G^{\prime }\mathrm{and}{G}^{\prime \prime }$) of the gels and the results within the LVER are reported in Figure 3(a). Generally, both moduli increase in the order CNC > NP > NA and for both CNC and NA the elastic modulus increases proportionally with the concentration of CNC. This fact is compatible with an overall stiffening of the gel framework, despite the poor affinity of Mg²⁺ with alginates in the case of NA hybrids. On the other hand, for a given ration CNC/PEG, NP samples show a nearly constant $G^{\prime }$ up to an $MgC{l}_2$ concentration of $30mM$ and only above this value their $G^{\prime }$ increases with a slope similar to that of the other compositions. Moreover, NP samples have slightly larger G moduli for low ionic strength compared with alginate hybrids. These considerations are confirmed by the viscosity characterization reported in Figure 3(b). All hybrids show the shear thinning behaviour typical of CNC suspensions [62], with a marked reduction of their viscosity over a range of about two orders of magnitudes at large shear frequencies. Pure CNC is the most viscous sample since the electrostatic interaction among CNCs is the strongest and the mesh size in the gel network is the smallest. All other samples show similar viscosities.

It should be noted that, despite the literature reports a variation of the elastic modulus over more than an order of magnitude (from ${10}^3\mathrm{to}{10}^4$ Pa [14,63]), the effectively exploitable range is limited by the equilibrium condition established within the gel during its actual use. Indeed, as already pointed out[15], the concentration of the crosslinking cation reaches a steady state value, independently of the amount of cations used during the gelation step. Moreover, increasing the mechanical properties of the gel by using a large excess of cations might produce unwanted results (e.g. toxic gels which release excess ions when used).

For applications such as cell cultures and, in particular, in the field of tissue engineering, where cells are embedded into an artificial ECM, it is of paramount importance to understand the actual state of the material internal architecture. DLS can provides information about the dynamics of the gels during their ageing and show how materials with similar macroscopic mechanical behaviour might have very different microstructures and dynamics, which can impact on their biocompatibility.

Figure 4 reports the dynamic of (${\sigma }^2$) for representative samples for each investigated composition (data for all samples can be found in the ESI). Initially, the systems are in a colloidal suspension state of barely interacting species and the ${\sigma }^2$ tends to the ideal value of 2, typical of ergodic systems. Upon the addition of $MgC{l}_2$, cations crosslink the species into clusters of increasing size and the suspension immediately jellifies, creating a solidified blob within the cuvette. Since the $MgC{l}_2$ solution is added at once within the DLS cuvette the dynamics of cation diffusion is stochastic in nature and the induction time before ${\sigma }^2$ decreases cannot be estimated in advance. Yet, all samples show a sharp drop of ${\sigma }^2$ by roughly an order of magnitude after a variable induction time. The optimal threshold for each sample is indicated by the different background color of Figure 4 panels. This phenomenon is the signature of a strongly reduced mobility of the particles (regardless if they are CNCs or polymer molecules), which is induced by the large concentration gradient of cations that quickly rearrange the structure of the system, from a solution into a gel network. The fast dynamics within the scattering volume destroys the coherence required to form significant autocorrelation, thus ${\sigma }^2$ collapses. Moreover the ${\sigma }^2$ fluctuations testify that the initially formed gel network is a metastable structure slowing evolving towards its equilibrium configuration. In the end, by tracking ${\sigma }^2$ it is possible to have a hint on the dynamic of the sol-gel transition of the system.

The FOM for the five different compositions, for the optimal T-value, is reported in Figure 5.

CNC and NP samples share similar dynamics, with ${\sigma }^2$ decreasing sharply and mostly assuming small values after the transition occurs. On the other hand, NA samples show a marked decrease of ${\sigma }^2$ from its initial value but then it continues to oscillate for the entire time window considered. Pure CNC samples are reported for illustrative purposes and their FOM is comparable to those of the NP samples, suggesting a similar gelation dynamics for CNCs and NPs. NAs have a marked different behaviour and, by increasing the amount of CNC in the hybrids, the FOM decreases, tending to the one of pure CNC. This fact supports the idea that our phenomenological parameter effectively represents a physical feature of the internal dynamics of the gel.

The dynamics of ${\sigma }^2$ can be understood by the following reasoning: upon addition of the crosslinker, $MgC{l}_2$, a non-equilibrium jellifying blob forms nearly instantaneously. Then, strong diffusive fluxes establish from the blob to equilibrate its composition. These diffusive processes destroy the coherence required to measure the typical (flipped) sigmoidal profile of $g^2\left(t\right)$. Moreover, since these fluxes affect the entire volume of the cuvette, multiscattering events take place along the optical path of the laser, thus, further decreasing the characteristic decay time and ${\sigma }^2$. The difference between CNCs/NPs and NAs is due to their different aptitude to jellify: while CNCs and NPs form "stable" structures upon coordination with $M{g}^{2+}$ cations, the NAs maintain their internal dynamics. Alginate chains act as a viscous liquid reservoir that traps jellified particles of CNC and these latter are responsible for the fluctuations in ${\sigma }^2$ function. Again, given the large concentration and the broad dispersion in size of these CNC particles, no quantitative analysis can be performed on the $g^2\left(\tau \right)$ runs.

To verify the reliability of our approach we repeat the analysis for the same samples composition but using $C{a}^{2+}$: in these cases the cation binds both PEG and alginates and the ${\sigma }^2$ always produced a small FOM, comparable to that of the pure CNC. This result confirms the strong affinity of $C{a}^{2+}$ with both cellulose and polymers and its role as effective crosslinker.

Finally, the hybrids were used as substrates for cell cultures. HeLa cells were used as model cell lines and were seeded with an initial areal cell density of $3.5·{10}^3{\displaystyle \frac{cells}{c{m}^2}}$.

Cell cultures were incubated up to 96 h. Cell viability was verified using standard MTT assay. The results are reported in Figure 6.

HeLa cells adhered within 24 hours and exhibited growth over time on any of the five CNC-based hydrogels, with varying growth rates. Compared to the reference polystyrene dishes, all gels show a slowed cell growth, yet both CNCs and NPs show a constant increase of the cell number, while NAs hybrids initially hinder cell viability, albeit they also show an increase in cell number with time.

It is well known that, apart from their chemical nature, mechanical properties of substrates strongly affect processes such as cell-matrix interaction, cell signaling and mechanotransduction [64,65]. Therefore, the viscoelastic properties of synthetic ECMs for cell culture must be carefully adjusted to promote cell growth.

Pure CNC hydrogels seem to be the most promising substrate for cell culture. Notably, the viscoelastic properties of the hybrids are rather similar since their $G^{\prime }$ values differ for no more than a factor of three (for the 50 mM MgCl₂ conc.) and yet they show rather different cell growth rates. Literature data for such soft substrates demonstrate that cell viability is not affected by the actual value of their elastic moduli [66,67,68] thus, given their similar viscoelastic properties, differences in cell spreading seem to be associated mainly with the microarchitecture of the gel network. Indeed, all three individual components (CNC, PEG and alginate) are known to be viable substrates for cell growth. Thus, while CNCs and NPs have a similar stiffness, they differ for a fundamental functional group (the amino group), and they show similar growth rates; on the other hand, NAs and CNCs have very similar chemical compositions but different microarchitectures and this is reflected on the largely different growth rates.