3-Dimensional porous nanocomposite sca ﬀ olds based on cellulose nano ﬁ bers for cartilage tissue engineering: tailoring of porosity and mechanical performance †

Fully bio-based 3-dimensional porous sca ﬀ olds based on freeze-dried cellulose nano ﬁ bers (70 – 90 wt%) stabilized using a genipin crosslinked matrix of gelatin and chitosan were prepared. Morphology studies using scanning electron microscopy showed that the sca ﬀ olds have interconnected pores with average pore diameters of 75 – 200 m m and nanoscaled pore wall roughness, both favorable for cell interactions with cartilage repair. X-ray tomography con ﬁ rmed the 3-dimensional homogeneity and interconnectivity of the pores as well as the ﬁ brillar structure of the sca ﬀ olds. The compression modulus of the sca ﬀ olds (1 – 3 MPa) at room conditions was higher than natural cartilage ( z 1 MPa). The lowered compression modulus of 10 – 60 kPa in phosphate bu ﬀ ered saline (PBS) at 37 (cid:1) C was considered favorable for chondrogenesis. The current study therefore successfully addressed the challenge of tailoring the pore structure and mechanical properties simultaneously for cartilage regeneration. Furthermore, the sca ﬀ olds' high porosity ( z 95%), high PBS uptake and good cytocompatibility towards chondrocytes are considered bene ﬁ cial for cell attachment and extracellular matrix (ECM) production.


Introduction
Articular cartilage is an avascular, non-innervated tissue composed mostly of extracellular matrix (ECM) with a sparse population of chondrocytes distributed throughout the tissue and 70-85 wt% of water. 1 Due to its poor cell density and lack of blood vessels, cartilage has a very limited capacity to repair itself from defects caused by trauma or aging. In this respect, developing new tissue engineering approaches to repair cartilage defects and to restore cartilage function are of great interest. 2,3 Due to good biocompatibility, natural biomaterials such as chitosan, gelatin, alginate and collagen have been used as raw materials for scaffolds in so tissue engineering. [3][4][5][6][7][8][9] These natural polymers support the cell growth and regeneration. However, their use is limited because of their lower mechanical properties when compared to synthetic polymers, especially load behaviour requirements necessary to allow proper cell proliferation. [10][11][12] In cartilage tissue engineering, scaffolds are expected to imitate the functions of damaged cartilage and provide a 3dimensional (3D) environment for cell growth and ECM production. 3,4 Typically, the tailoring of scaffold porosity with pore sizes in the range of 200-300 mm is considered a benchmark for cartilage regeneration. 4,13 Such porous scaffolds can be prepared using many different processes, such as freeze-drying, CO 2 foaming, electrospinning or cryogelation, with a wide variety of polymers as found in literature. 5,[14][15][16] Chitosan and gelatin based scaffolds have been reported to be benecial for so load bearing tissues. Positive charges on the surface of chitosan can promote chondrocyte growth thus benecial for cartilage repair. 3,4 Gelatin is partially derived from collagen, which is the main protein component of connective tissues as cartilage, skin and bone. Furthermore, a blend of these polymers absorbs water and forms a hydrogel whereby allowing uid to be retained in the scaffold structure leading to a higher compression modulus similar to natural so tissue. 4,13,17,18 A recent study on chitosan/gelatin blends with random and aligned pore structures prepared with a freezedrying process showed a compression modulus in the range of 5 kPa (for randomly aligned scaffolds) to 30 kPa (in the vertical direction of the aligned scaffold), when tested at 37 C aer conditioning in PBS medium. 18 To enhance the mechanical performance of biopolymers and their blends, different types of reinforcements obtained from natural materials have been used; the most important examples being derivatives of cellulose and chitin. 8,9,14,15 Our earlier studies have also demonstrated that solution cast brous nanocomposite structures with high cellulose nanobers concentrations (75 wt%) and collagen provide mechanical performance suitable for ligaments. 8,9 Also, our previous studies have demonstrated the noncytotoxicity of nanocellulose from different sources and their potential in medical applications. 15,19,20 In the current study, 3D nanocomposite scaffolds for cartilage regeneration were processed via freeze-drying technique, where porous cellulose nanober structures were bound together and mechanically/dimensionally stabilized using low amounts of crosslinked chitosan/gelatin blend system. The pore structure developed during the processing is expected to impact the moisture uptake and mechanical performance of the resultant scaffolds. Moreover, the pore structure and sizes, which favor the movement of uids through the scaffold, creating drag forces, 21 as well as the development of ECM, is known to contribute to the load bearing under in vivo conditions. Though tailoring of mechanical properties using nanocellulose have received some attention in tissue engineering, 8,9,22 limited studies are available on tailoring the pore structure in bionanocomposites and understanding the effect of porosity on the mechanical performance. It is highly challenging to tailor the pore structure and mechanical properties required for tissue regeneration, simultaneously, as increase in porosity generally decreases the mechanical properties and vice versa. The pore structure and porosity of the scaffold was tailored in the current study by varying the suspension concentrations, bers/matrix ratio and crosslinking with the aim to obtain optimal mechanical properties and chondrocyte attachment and proliferation to obtain extracellular matrix of optimal quality. It was expected that these 3D porous structures could act as templates for the formation of new tissue and act as guidance for cell growth while facilitating nutrient and oxygen transport.
The structural morphology of the produced scaffolds was observed by scanning electron microscopy (SEM) and X-ray tomography. Mechanical properties in room condition and in PBS medium, moisture uptake, density and porosity, as well as in vitro biodegradation and cytocompatibility towards chondrocytes were also investigated.

Materials
High-purity cellulose from sowood bers (Norwegian spruce) with high cellulose content (95% cellulose, 4.5% hemicellulose and 0.1% lignin content as provided by Domsjö Fabriker AB, Sweden) was used as starting material for the production of cellulose nanobers. Medium M w chitosan (DD z 75-85%), acetic acid, gelatin, as well as phosphate buffered saline (PBS) and genipin were purchased from Sigma-Aldrich, Germany.

Methods
Processing of cellulose nanobers. Cellulose bers were dispersed in distilled water at a concentration of 2 wt% using a mechanical blender, Silverson L4RT (England), at 6000 rpm for 15 min. Then, the suspension was ground using an ultra-ne grinder, MKCA 6-3 from Masuko (Tokyo, Japan) to obtain nanobers (CNF), following the procedure reported by Mathew et al. 9 As the pore structure required for the cartilage application is in micron scale range, 4,13,23,24 the brillation process was aimed towards obtaining relatively coarser brils as they are expected to give larger pore sizes for the scaffold compared to their ner counterparts.
The atomic force microscopy of the prepared nanobers ( Fig. 1a) showed nanosized bers with diameters in the range of 19-38 nm (Fig. 1b), based on the measurements using the Nanoscope V soware (Santa Barbara, CA, USA). The diameter of nanobers was measured from the height to compensate the tip broadening effect. The photograph of the gels of CNF obtained aer grinding is also shown (Fig. 1c).
The suspension was centrifuged for 30 min and the rotor speed was 1500 rpm to remove water and obtain highly concentrated gel, typically between 7 and 10 wt%, which was used for scaffold processing.
Processing of the nanocomposite scaffolds. The matrix (gelatin/chitosan), the nanobers and the genipin were mixed in a one step process and crosslinked in situ. The gelatin (G)/chitosan (Ch) mixture in a ratio of (9 : 1) was dissolved in 0.01% acetic acid medium containing 0.004 M genipin solution. Gelatin/chitosan mixture in the ratio of 9 : 1 is referred to as the matrix (M) throughout the manuscript. The nanober suspensions were mixed with this solution in appropriate amounts to obtain nal nanocomposites with different compositions. All samples were placed in plastic Petri dishes, frozen at À30 C and freeze-dried at À70 C in a vacuum (0.0026 mbar). The complete process with a photograph of the produced scaffold is shown in Fig. 1d and e, respectively. All scaffolds prepared in this work and their compositions are listed in Table 1.

Characterization
Scanning electron microscopy (SEM). The samples were cryogenically fractured to preserve the structure and SEM micrographs were acquired with a SEM JEOL JSM-6460LV at voltages of 5 and 15 kV. All samples were sputter-coated with gold for 50 s at 50 mA to avoid electron charging. The pore sizes were measured from the SEM images using SemAfore.
For high-resolution images, MAGELLAN 400 XHR-SEM (FEI Company, Eindhoven, The Netherlands) was used. The samples were placed on carbon tape, coated with tungsten, and observed under the SEM at an acceleration voltage of 3 kV.
X-ray tomography. 3D structure of pores was reconstructed using a Zeiss XRadia XRM 520 X-ray tomograph. For these images, an X-ray tube voltage of 40 keV was used, which results in a broad spectrum of X-ray energies, with a maximum of about 40 keV. No lter was used on the source. The beam produced by the source is a cone-beam, which provides a geometrical magnication of the image depending on the source-detector distance and the position of the sample between the two. In this case the sample was placed at 9 mm from the source and 9 mm from the detector. 1601 radiographs were acquired over 360 with an exposure time of 2 s and 12 s per projection, respectively for the 4Â and 20Â images. The tomographic reconstruction was performed using the Zeiss reconstructor soware with a correction for the center of rotation.
Density and porosity. The density was calculated by cutting approximately cubic samples, measuring all dimensions, weighing them and dividing the weight by the volume. Each measurement was taken three times and the results reported are based on the average values. Porosity of the scaffolds was evaluated based on the weight and density of the scaffolds. The porosity was dened as the volume fraction of the voids (V v ) and was calculated using the following equation. 22 1where r e is the experimental density of the scaffold and r t is the theoretical density of a non-porous scaffold. The densities of CNF, gelatin and chitosan were taken as 1.54, 0.98 and 0.235 g cm À3 , respectively. Nanoscaled pores were measured using a Micromeritics ASAP 2000 instrument and the average pore diameters were determined from nitrogen adsorption measurements at 77 K using the BET method. The measurements were performed aer degassing the samples at 100 C for 48 h in dry N 2 ow.
Moisture uptake. The moisture uptake measurements were performed in PBS medium. The samples were dried overnight in a vacuum oven at 80 C, immediately weighed (W d ) and thereaer immersed in PBS. Weights were taken at different time intervals (t), i.e. 30 s aer immersion and another one 3 days later (W t ). Every time the excess water was removed by gently tapping the samples on a dry so tissue paper. The moisture uptake was calculated according to the following equation.
In order to assess the swelling of the matrix phase, samples were dried in the vacuum oven and weighed as described above and placed in a 95% moisture desiccator. The samples were weighed every two days for two weeks in order to follow the weight gain until equilibrium. The moisture uptake was calculated using the same equation as above.
Compression tests. All compression tests were performed at 37 C using a TA Instruments DMA Q-800 (New Castle, DE, USA), according to an adapted version of the D11621-94 standard test method. The samples were cut into square pieces with sides between 5 and 10 mm. The thickness was about 10 mm for dry tests and the displacement rate was 100 mm min À1 with the contact force of 0.05 N. The compression moduli were calculated as the slope of the stress-strain curve in the linear region, below 15% strain.
For tests in PBS, the sample thickness was about 5 mm and the cut samples were immersed in PBS for 24 h before testing. The displacement rate for moduli calculation was 400 mm min À1 and the initial contact force was 0.02 N, slightly adapted according to the specic rigidity of each type of sample (up to 0.05 N). Compression tests were also performed in PBS at varying strain rates ranging from 100 mm min À1 to 400 mm min À1 to evaluate viscoelasticity. Each test was performed at least ve times and the average values were reported.
In vitro biodegradation. The biodegradation of scaffolds was investigated using conventional technique. In this method, the dried specimens were immersed in PBS under pH 7.4 and stored in a thermostatically shaking water bath at 37 C for up to 28 days. The samples were removed at different times. The pH of the PBS solution was measured by pH meter and then replaced with fresh PBS each time. Aer removal of samples, the surface of the specimens was gently blotted by so tissue paper in order to remove water. The specimens were then completely dried in a vacuum oven at 50 C until a constant weight was achieved, and then they were weighed. The weight loss was calculated using the following equation.
W 0 is the sample's original weight, and W t is the weight of the specimen at time (t).
Cytocompatibility studies Cytocompatibility of the CNF. CNF lms were xed to cell culture dishes and the cells (adipose derived stem cells (ASCs) and L929 cell line) were seeded evenly throughout the cell culture dish. The impact of the biomaterial on cell growth and morphology was monitored and documented with photographs.
Cytocompatibility of the scaffolds. Cytocompatibility of the scaffolds was monitored in a direct contact testing system according to ISO 10993. The biomaterials were xed in cell culture vessels and cells, namely chondrocytes, were seeded on the biomaterial, (0.5 Â 106 cells per ml) in cell culture media supplemented with 10% FBS (Fetal Bovine Serum). The scaffolds with cells were incubated at 37 C in 5% CO 2 for 7 days. Aer 7 days the biomaterials were stained with MTT (3-(4,5dimethylthiazolyl-2)-2,5-diphenyltetrazolium bromide) and inspected for the presence of live cells on the upper surface of the scaffolds. Biomaterial with no seeded cells was regarded as a negative control.

Morphology and pore structure of scaffolds
The SEM images were evaluated to understand the effect of crosslinking and suspension composition on scaffold morphology. CNF 4 -M 0.4 and X-CNF 4 -M 0.4 ( Fig. 2a and b) showed the pore sizes in the micrometer range in both cases. Uncrosslinked scaffolds had a wide distribution in pore size (Fig. 2a) while crosslinking slightly enhanced the overall homogeneity of the structure and decreased the pore size (Fig. 2b).
The morphologies of X-CNF 4 -M 0.4 , X-CNF 4 -M 1.1 , and X-CNF 4 -M 2.2 with varying the matrix content are compared in Fig. 2c-e. Fig. 2c shows CNF bound by the matrix forming an interconnected pore structure with single and bundled bers emerging from and embedded in the matrix. X-CNF 4 -M 1.1 and X-CNF 4 -M 2.2 ( Fig. 2d and e) show at and layered structures and resemble self-assembly behaviour similar to that of pure gelatin, as previously found in literature. 13,25 As the matrix content increased, smoother and thicker wall structures and fewer pores were observed, which was not considered favorable for the pore sizes required for cartilage applications.
When comparing the scaffolds with different initial suspension concentrations, X-CNF 4 -M 0.4 showed a pore structure which is relatively homogeneous and interconnected, whereas X-CNF 5 -M 0.5 and X-CNF 6 -M 0.6 showed more of a layered structure than a brillar structure and formed denser structures with fewer pores and cellulose nanobers coated with matrix compared to X-CNF 4 -M 0.4 (images given in ESI, S1 †). Earlier studies also have demonstrated the controlling the cell size and foam density by changing the suspension concentration. 26,27 The pore sizes measured from SEM for all developed scaffolds are summarized in Table 1. In all cases the standard deviations are high and pore sizes in the range of 20 mm to 200 mm were observed. The highest average pore size (100-200 mm) was with the uncrosslinked system and the pore size decreased with crosslinking (20-120 mm), increased matrix content (20-75 mm) and increased suspension concentration (40-115 mm).
The X-CNF 4 -M 0.4 system was considered optimal for the cartilage tissue engineering based on pore structure, homogeneity of pores and average pore sizes. When examined using high-resolution microscopy (Fig. 2f), X-CNF 4 -M 0.4 showed a highly entangled network of CNF on the pore walls. These brous nanostructures and the pore wall roughness is expected to aid cell xation and extracellular matrix (ECM) development aer implantation because the rougher surface improves vascularization, diffusion rates to and from the scaffold for oxygen/nutrients supply and removal of waste. 23,28 (For this system, the nanoscaled pores measured from micromeritics porosity analyser were in the range of 12-14 nm).
All of the materials were highly porous (>93%) and had low densities, shown in Table 1, and which agree with literature values. 27 The crosslinking as well as the increase in matrix content had limited inuence on porosity. These samples had a similar density irrespective of the matrix content (samples 1-5), most likely due to the higher bulk density of cellulose (1.54 g cm À3 ) as cellulose is the main component in all of the samples. When the concentration of the freeze-drying suspension increased (samples 3, 6 and 7) while keeping the bers to matrix ratio constant (10 : 1), the density increased from 0.066 to 0.093 g cm À3 due to denser packing of ber network and as expected led to a decrease in porosity. 26,27 Fig. 3a-c shows cross-sectional images of the porous scaffold for the X-CNF 4 -M 0.4 system obtained using X-ray tomography. It can be observed that the scaffold has high porosity (conrming the porosity data in Table 1) as well as pores are uniformly distributed in the horizontal as well as vertical sections of the scans. Furthermore, X-ray tomography conrmed the interconnectivity of the pores as well as the brillar structure of the scaffolds.
The 3D interconnectivity of pores (Fig. 3d) throughout the scaffold, the high degree of porosity, the hierarchical pore structure with micron sized pores in the bulk and the nanoscaled pores on the walls of scaffolds were considered optimal for the cartilage tissue engineering. 26

Moisture uptake
As 70-85% of the weight of natural cartilaginous tissues is water, 1 it is important to understand the water uptake by the scaffolds. The PBS uptake values of the submerged samples were in the range of 1000-1677%, depending on the composition. The results of the moisture uptake measurements are shown in Fig. 4. The initial moisture uptake was instantaneous (30 s) and remained constant aer 3 days in PBS and the materials showed hydrogel behaviour.
It can be inferred that the open pore structure of the materials (shown in SEM) plays a role in the fast and high moisture uptake and moisture susceptibility in the scaffolds. It was found that the amount of matrix does not signicantly affect the uptake, as X-CNF 4 -M 0.4 , X-CNF 4 -M 1.1 and X-CNF 4 -M 2.2 showed similar water uptake. This indicates that a signicant proportion of the water uptake is due to the porous structure of the scaffolds. A higher CNF concentration in the suspension resulted in a decrease in the uptake of the resultant scaffolds due to a tighter, denser network with smaller pores, as expected. However, it was noted that porous CNF scaffolds without matrix (CNF 4 ) absorbed PBS quickly and disaggregated easily when manipulated since there is no matrix to bind the bers together. The uncrosslinked samples also proved to be unstable aer 3 days in PBS, while the crosslinked samples remained stable in moist conditions during the whole duration of the experiment.
The water uptake by the same scaffolds was monitored in 95% RH conditions (without immersion) and the maximum uptake was 30% of its original weight, even aer 15 days (shown by Fig. 4 in the inset). This shows that the adsorption due to the scaffold swelling is negligible. No difference has been observed  regardless of whether the samples are crosslinked or not, and the total uptake showed a tendency to decrease as the initial concentration increased, which is in correlation with the density of the material.

Mechanical properties
Compression is the preferred mode of mechanical testing for cartilage materials because the role of natural cartilage is to bear loads in compression 1,13 (see ESI for the representative load displacement curves, S2 †). In dry conditions and at 37 C (Fig. 5a), the compression modulus was in the range 1-3 MPa which agree with earlier reports of anisotropic CNF based foams. 26,27 No clear trend can be observed when varying the total concentration or the CNF/matrix ratio and the values do not follow the density as reported in some earlier literature, 26 most likely due to the crosslinking effect which is not considered in density calculation. Nonetheless, the presence of the matrix enhanced the mechanical properties in dry conditions when compared to CNF alone. However, high standard deviations were observed and may be due to the wide distribution of the pore sizes.
The aggregate compression modulus of articular cartilage is reported to be around 0.9 MPa by Martin et al. 29 and 0.5-0.1 MPa by Guilak et al. 30 Also, a wide range of values varying between 0.1 MPa and 2 MPa are reported as compression moduli for healthy cartilage 31-34 depending on the source and testing conditions. The values of compression moduli of the current scaffolds are slightly higher than that of natural cartilage, but it may be noted that the moisture content in natural cartilage is greater.
The performance of the scaffolds was evaluated in simulated body conditions (PBS medium and 37 C) to understand the effect of compression rate, crosslinking, bers/matrix ratio and initial suspension concentration; (see Fig. 5b-e). The compression modulus showed a clear tendency to increase as the compression rate increased from 100 to 400 mm min À1 (Fig. 5b), a sign of the viscoelasticity of the scaffolds. In dry conditions, the scaffolds exhibited an elastoplastic behaviour, but when submerged in PBS, viscoelastic behaviour was evident. The viscoelastic behaviour in PBS medium is partly due to the swelling of the matrix phase as well as uid ow through the pores of the scaffolds during compression. This tendency was reported for natural cartilage tissues when tested in compression mode and therefore considered favorable for load bearing by the scaffolds. 21 The inuence of crosslinking on compression modulus was evaluated for CNF 4 -M 0.4 and X-CNF 4 -M 0.4 , tested in PBS at 400 mm min À1 (Fig. 5c) in the strain region of 0-5% and 10-15%. No signicant improvement on the mechanical properties was achieved by crosslinking and the compression modulus was around 30 kPa. The low amount of matrix material available for crosslinking as well as the similar density and porosity observed for these scaffolds explains this trend.
The compression moduli of CNF 4 , X-CNF 4 -M 0.4 , X-CNF 4 -M 1.1 and X-CNF 4 -M 2.2 in the strain region of 0-5% and 10-15% are given in Fig. 5d. The values decreased from 62 kPa for pure CNF to 10 kPa for nanocomposites with the highest matrix concentration (X-CNF 4 -M 2.2 ). The results show that the scaffolds become weaker in wet conditions when the matrix content increases. One possible reason is that gelatin and chitosan are mechanically weaker in wet conditions than the CNF network. Nevertheless, a minimum concentration of matrix phase was necessary to bind the bers together and ensure the stability of the scaffold in wet medium. Furthermore, CNF 4 collapsed completely during compression tests in submersion mode aer 15% strain, while scaffolds with matrix phase showed better  mechanical and dimensional stability in spite of the lower compression moduli. Fig. 5e shows scaffolds prepared by increasing suspension concentrations (X-CNF 4 -M 0.4 , X-CNF 5 -M 0.5 and X-CNF 6 -M 0.6 ) but the same ratio of bers/matrix (10 : 1). The compression modulus increased from 18 to 60 kPa as the concentration increased from 2.2 to 5.5 but stabilized thereaer. The increase in density of the scaffolds did not affect the mechanical properties signicantly, especially in PBS medium.
In general, it can be seen that the compression modulus in wet conditions is lower (10-60 kPa) than in dry conditions (1)(2)(3). The decreased mechanical properties in wet conditions were expected due to the swelling and plasticisation of the scaffold pore walls with water. The hydrophilicity of the nanocellulose, gelatin and chitosan as well as the high water-binding capability of gelatin signicantly impacts the performance in aqueous medium. The compression moduli of the scaffolds with CNF as reinforcement in the chitosan/gelatin matrix is however signicantly higher than those reported for the chitosan/gelatin blend, 18 indicating that CNF acts as reinforcements in the porous scaffold.
The values presented here are lower than reported for natural cartilage tested in wet conditions. It may be noted that the current compression studies are performed submerged in PBS and at 37 C which can weaken the scaffold in comparison to conditioned natural cartilage tested at 37 C. More importantly, it was demonstrated earlier that in in vivo conditions, chondrocytes sense the mechanical properties of the substrate and so scaffolds (4 kPa) facilitate chondrogenesis where as stiffer scaffolds ($40 kPa) are shown to favor bone regeneration. 35,36 In the current study, X-CNF 4 -M 0.4 scaffolds have shown favorable mechanical properties (18-32 kPa) for chondrogenesis, stability in moist conditions and favorable pore sizes and pore wall morphology for cell adhesion. Therefore, these scaffolds are expected to develop extracellular matrix (ECM) and regenerate cartilage with the right mechanical properties aer implantation. The evaluation of the scaffold aer ECM development will be required to understand the performance of the scaffold in in vivo conditions and may be addressed in future.

Biodegradability
In vitro degradation tests of the scaffolds were investigated up to 28 days and the results are shown in Fig. 6. The scaffold, which contained only CNF 4 (used as control), displayed signicant morphological changes during the degradation time. In agreement with the observations during water uptake studies, these scaffolds did not have dimensional stability in order to bind the bers together due to lack of matrix content. The scaffold comprising X-CNF 4 -M 2.2 showed around 27% decrease in weight, while the scaffold comprising X-CNF 4 -M 0.4 and X-CNF 6 -M 0.6 demonstrated 1.5% weight loss aer up to 4 weeks. Therefore, the greatest weight loss was obtained for the scaffold containing the higher amount of matrix.
An optimal rate of degradation crucial for cartilage regeneration is one which balances stable 3D structures that provide support with gradual development of ECM. 37 The short-term biodegradation studies showed that the rate of degradation of nanobrous scaffolds could be controlled by the bers/matrix ratio. The slow biodegradation tendency of biologically stable scaffolds in the current study, which provides an enduring support for the patient while also favoring ECM formation, can be considered benecial. However, further longterm investigation needs to be done to evaluate ECM development in correlation with biodegradation of the scaffold to ensure the potential of the produced scaffold.

Cytocompatibility studies
Cellulose nanobers. The cytocompatibility of CNF was evaluated using a direct contact testing system and the results are shown in Fig. 7a and b. The L929 cell line was chosen due to  ISO 10993 recommendations and (ASCs) were chosen due to their broblast morphology, sensibility and their role as a chondrocyte precursor. In addition to CNF lms, noncytotoxic (ÀK) and cytotoxic (++K) controls were also used to measure cytotoxicity effects. The cells in the presence of negative control (ÀK) exhibited efficient proliferation between day 1 and 8, whereas positive control (++K) showed no cell attachment or growth. This short-term cytocompatibility test results indicate that CNF support cell growth and can have potential in biomedical scaffold fabrication. We have also recently reported these CNF are cytocompatible according to current ISO criteria, with non-inammatory and non-immunogenic properties. 20 Higher concentrations were found to be tolerogenic to the immune system, a characteristic very desirable for implantable biomaterials, which justies the use of wood-based CNF in the current application.
3D porous scaffold. When cells seeded on biomaterial surface, they attached to the scaffold. However, the cells retained the circular morphology up to day 6, but showed no sign of zone of inhibition or reduced growth, as shown in Fig. 7c. At day 7, the samples were stained with MTT in order to highlight the live cells, as the enzymes in live cells catalyze the reaction, resulting in a purple colored product. All seeded samples were compared with samples without cells, as a negative control. The morphology of the cells was round, which indicates that the cells are being encapsulated in the scaffold, as also seen in some other hydrogels which are routinely and effectively used in clinical practice. 38 As shown in Fig. 7c, the chondrocytes in the scaffolds remained viable aer 7 days. Therefore, the material is regarded as non-cytotoxic and is suitable for further evaluation.

Conclusions
Nanocomposites of cellulose nanobers bound in a gelatin and chitosan matrix were prepared via freeze-drying and crosslinked using genipin to obtain highly porous (z95% porosity) 3D scaffolds with optimal pore size, porosity, pore interconnectivity as well as mechanical performance, moisture stability and cell interactions. The freeze-drying route resulted in isotropic scaffolds and the pore structure was most homogenous for nanocomposites with a low amount of crosslinked matrix that acted as binding phase and dimensional stabilizer in moist conditions. The compression moduli of the optimal scaffolds (X-CNF 4 -M 0.4 ) in dry conditions, at 37 C, were around 1 MPa and considered comparable to natural cartilaginous tissue. These scaffolds showed viscoelasticity when immersed in PBS, similar to that of natural cartilage, but lower compression modulus (18-32 kPa) which was considered favorable for so tissue regeneration. We have successfully tailored fully bio-based scaffolds with a high degree of bulk porosity, hierarchical pore structure, nanoscaled roughness and brillar structure of the pore walls combined with good mechanical properties and cytocompatibility with high potential for cartilage regeneration. Furthermore, the possibility to develop ECM and trap moisture in the interconnected pore structure is expected to bring the mechanical properties closer to those of natural cartilage aer implantation.