Native cellulose (10 g, Whatman No. 1 filter paper) was cut into small pieces and immersed in deionized water (DI) water to make 1 L of 1% w/w cellulose pulp slurry. In addition, 0.189 g of TEMPO (Sigma Aldrich, Canada), 1.178 g NaBr (Sigma Aldrich, Canada), and 50 ml of 1.8 M NaClO (Sigma Aldrich, Canada) were added into the cellulose slurry under continuous stirring until dissolved. The pH of the reactants in the flask was maintained at pH = 10.5 by adding 0.5 M NaOH until the pH of the reactant were stable at 10 (Shinoda et al., 2012). The TEMPO-oxidized cellulose was then washed with DI water for 15 times until the conductivity for supernatant was constant (∼12.5 µs/cm). The TEMPO-oxidized cellulose nanofiber (TCNF) was prepared by blending the TEMPO-oxidized cellulose for 7 min. The carboxylate content for the TCNF was 0.84 mmol/g. The TCNF was centrifuged and adjusted to 3.5% w/v solid content. Freeze-dried cellulose pulp slurry (pure cellulose) and TCNF were characterized by Fourier Transform Infrared spectroscopy (FTIR).

Sodium alginate (ALG, Alfa Aesar, J61887, United States) was prepared into a 3.5% w/v solution. The concentrated TCNF and the ALG solution were sterilized by autoclave. The TCNF and ALG are mixed into various printing formulations (Table 1) with a final solid content of 3.5% w/v.

The rheological properties of all the TCNF/ALG precursors were characterized using a rotatory rheometer (AR-G2, TA Instrument, United States) with a 25 mm parallel-plate geometry. The steady-state flow sweep was characterized under the shear rate from 0.001 to 1,000 s⁻¹ at room temperature. The oscillatory frequency sweep was measured under 1% strain (within the linear viscoelastic region) and 1–100 rad/s frequency. The thixotropy tests were done by first fixing the shear rate at 1 s⁻¹ for 100 s, followed by a sudden increase in the shear rate to 1,000 s⁻¹, and then a sudden drop in the shear rate to the initial state (1 s⁻¹). The rheology data was analyzed using TRIOS software (TA Instruments, United States).

To assess the printability of the TCNF/ALG precursors, 20 mm × 20 mm × 3 mm blocks with a 30% infill were bioprinted using an extrusion-based bioprinter (INKREDIBLE+, CELLINK, Sweden). The geometry and printing parameters of the printed constructs were predefined in commercial design software (Slic3r, United States). The needle inner diameter was 0.413 cm (22G). The printing speed was 10 mm/s with each layer height of 0.5 mm (6 layers in total). After bioprinting, a 100 mM CaCl₂ solution was added over the bioprinted TCNF/ALG constructs for 3 min followed by a phosphate buffered saline (PBS) rinse. After crosslinking, the precursors with the best printing fidelity were used for further biological functionality evaluation. A commercially available bovine collagen I gel with similar solid content to the TCNF/ALG precursors was used as a control (3.5% w/v, Advanced Biomatrix, United States). The collagen I gel was printed at room temperature and spontaneously polymerized at 37°C. The fidelity parameter (side length of the square blocks (L) before and after the crosslinking, filament diameter) were quantified using ImageJ (Fiji, United States). The contractility from the crosslinking process was calculated as: contractility = ( L before crosslinking − L after crosslinking ) / L before crosslinking

Meniscus specimens from three (3) male donors were collected from partial meniscectomy surgeries with the approval of the University of Alberta’s Health Research Ethics Board - Biomedical Panel (Study ID: Pro00018778). Human meniscus fibrochondrocytes (hMFCs) were isolated enzymatically using collagenase digestion as previously described (Liang et al., 2017). The primary hMFCs were plated at a density of 10⁴ cells/cm² and expanded to passage 2 (P2) in monolayer culture in high glucose Dulbecco’s Modified Eagle Medium (DMEM, Sigma Aldrich, Canada) supplemented with 10 v/v (%) fetal bovine serum (FBS), 1 ng/ml of TGF-β1, and 5 ng/ml of FGF-2 at 37°C in normoxia (∼20% O₂) conditions (Liang et al., 2017). Donor information is shown in Table 2. The cumulative population doublings of the hMFCs at the end of P2 was 7.0 ± 0.2 (mean ± standard deviation).

hMFCs at P2 were resuspended in a standard serum-free chondrogenic medium containing 10 ng/ml TGF-β3 (Liang et al., 2017). The cell suspension was mixed with 7030, 8020 precursors (7030, 8020 bioinks), and collagen I gel (COL bioink) to a cell density of 10⁷ cells/mL.

To investigate the fibrocartilage formation of the bioprinted constructs after in vitro chondrogenic culture, hMFC-laden TCNF/ALG bioinks and COL bioink were bioprinted into cylindrical shapes (7 mm diameter, 3.5 mm height) with an infill rate of 70%. Then bioprinted constructs were crosslinked as before. At the same time, the hMFC-laden COL bioinks were kept in a 37°C incubator for 30 min and then placed in a serum-free chondrogenic medium containing 10 ng/ml TGF-β3 (2 ml/construct, twice per week) for 6 weeks under hypoxic conditions (3% O₂).

hMFCs viability was assessed using Syto 13/Propidium iodide (PI) staining. Syto 13 (S7575, ThermoFisher, United States) is a green live-cell fluorescent nucleic acid stain, and PI (P3566, ThermoFisher) is a red dead-cell fluorescent nuclear and chromosome counterstain. The Syto 13 and PI concentrations in PBS were 6.25 and 15.0 μM, respectively.

Bioprinted constructs of TCNF/ALG (8020, 7030 bioinks) and COL bioinks after 1-day culture were incubated in Syto 13 and PI solution at room temperature for 30 min in the dark. The cell viabilities were viewed under a Nikon confocal laser scanning microscope (Leica TCS SP5). Fluorescence was quantified using Python.

Scanning electron microscopey (SEM., Zeiss Sigma 300 VP-FESEM) was used to observe the TCNF fibre diameter and the ECM formation of the bioprinted constructs after 6 weeks of in vitro culture. Each construct was fixed in sodium cacodylate trihydrate buffer containing 2% v/v glutaraldehyde and 2.5% v/v paraformaldehyde overnight. The samples were then cut in half and washed twice with deionized water to wash away the fixation solution. The samples were further treated with osmium tetroxide and tannic acid before SEM observation. All the reagents were purchased from Electron Microscopy Sciences (Pennsylvania, United States).

Bioprinted constructs for three experiment groups (8020, 7030, COL) from the same donor were fixed in 10% (v/v) neutral buffered formalin, dehydrated through a series of alcohol washes, and then embedded in paraffin. The embedded samples were cut into 5 μm sections cross-sectionally.

For histological assessments, the sections were deparaffinized by the xylene substitute, then rehydrated and stained with Safranin-O/Fast Green. Collagens I, II, and X as well as aggrecan were examined by immunofluorescence. In brief, sections were prepared as above. Collagens I and II were counterstained in one slide. Collagen X and aggrecan were stained separately in two slides. The primary antibodies for collagens I and II were rabbit anti-human collagen I (CL50111AP-1, Cedarlane, Canada) and mouse anti-human collagen II (II-II6B3, Develop-mental Studies Hybridoma Bank [DSHB], United States). The primary antibody for type X collagen (collagen X) was rabbit anti-human collagen X (rabbit polyclonal to collagen X, ab58632, Abcam, United States), and the primary antibody for aggrecan was a rabbit anti-human aggrecan (recombinant monoclonal to aggrecan, MA5-32695, ThermoFisher, United States).

After incubation with primary antibodies, the slides were incubated with secondary antibodies for 45 min. The secondary antibodies used for collagens I and II are goat anti-rabbit IgG Alexa Fluor 594 (ab150080, Abcam, United States) and goat anti-mouse IgG Alexa Fluor 488 (ab150117, Abcam, United States). The secondary antibodies used for collagen X and aggrecan were goat anti-rabbit IgG Alexa Fluor 594 (ab150080, Abcam, United States). In addition, all the slides were also stained with 4′,6-diamidino-2′- phenylindole (DAPI, Thermo Fisher Scientific) for 20 min at room temperature to examine the nuclei of hMFCs in each sample. A Nikon Eclipse Ti-S microscope coupled to a DS-U3/Fi2 Color CCD camera with 100x and 200x total magnification was used to capture the histological and immunofluorescent images.

Quantitative real-time polymerase chain reaction (qRT-PCR) was used to analyze expression of genes that are chondrogenic and fibrochondrogenic specific (ACAN, COL1A2, COL2A1, SOX9) and chondrocyte hypertrophy-related (COL10A1, MMP13, ALPL, RUNX2). The analysis was conducted after 6 weeks (42 days) of in vitro chondrogenic culture. Total RNA was extracted using TRIzol (Life Technologies, United States). The complementary DNA (cDNA) was synthesized from 100 ng of total RNA using GoScript, Reverse Transcriptase kit (Promega, United States) and 1 µg of oligo (dT) primer. Primer sequences for qPCR were designed using Primer Express 3.0.1 (Thermo Fisher Scientific). Transcript levels for the interested genes were normalized to the housekeeping genes: β-actin, B2M, and YWHAZ using the delta CT method (2^−ΔCT ). Primer sequences are shown in Table 3.

The mechanical properties of bioprinted constructs were assessed using a stepwise stress relaxation test using a Biodynamic 5,210 system (TA Instruments, United States). For each experimental group, constructs from n = 3 donors were tested. The constructs were preconditioned by 15 cycles of sine wave dynamic compressive loading with an amplitude of 5% tissue height at a frequency of 1 Hz. The following stress relaxation test consisted of 4 incremental strain steps. In each step, the constructs were subjected to a 10% strain ramp at the rate of 50% strain/s and followed by 20 min relaxation under constant strain. All tested constructs reached equilibrium in the given relaxation period. Forces were recorded as a function of time, and stress was calculated by normalizing force to construct cross-section area. The peak modulus was calculated by dividing the maximum stress increment immediately after the compression increment by 10%, the strain increment.

GraphPad Prism 8 was used to perform statistical analysis. The paired two-sample t-test was used to analyze the significance level between each sample group (8020 vs. COL, 7030 vs. COL, 7030 vs. 8020) in cell viability and in gene expression after taking the replicate means within donors. Two-way ANOVA was used to determine the statistical differences in stress relaxation tests for various bioinks at different strains with Tukey’s test for multiple comparisons. The results were presented as mean ± standard deviation (SD).

Figure 2 presents the FTIR spectra of unmodified pure cellulose and TCNF. After TEMPO-oxidation, the characteristic absorption band of carboxyl group (C==O) stretching appeared around 1,650 cm⁻¹, which is assigned to the formation of COO groups after the oxidation and release of nanofibers (El Bakkari et al., 2019). The enhancement of the -OH stretching group around 340 cm⁻¹, C-H at around 2,900 cm⁻¹, and C-O-C at around 1,050 cm⁻¹ were also observed in TCNF, which indicated more -OH, C-O-C group are exposed.

The rheology data of the TCNF/ALG precursors were measured to pre-estimate the printability as potential bioink materials. The viscosity vs. shear rate (steady-state flow sweep) of all the TCNF/ALG precursors is shown in Figure 3A. All TCNF/ALG precursors exhibited shear-thinning behaviour, in which the viscosity decreased as the shear rate increased. The Power Law model was also used to fit the viscosity vs shear rate curve (0.01–1,000 s⁻¹) to describe the shear-shinning behaviour. The printing pressure of each TCNF/ALG precursors are shown in Table 4. The viscosity and shear rate relationship can be described in the equation of the η = K γ ˙ ⁿ⁻¹, where η is viscosity, γ ˙ is the shear rate, K and n are the empirical curves fitting parameters, known as the flow consistency index and flow behaviour index (Middleman, 1977). The fitted power-law parameters (K and n) are shown in Table 4. For shear thinning fluid, n always less than 1. The lower the n, the better the shear-thinning behaviour responding to the increase of shear rate. TCNF, 8020, 9010 exhibited and the best shear thinning behaviour, and 5050 exhibited the worst shear-thinning behaviour.

The recovery and thixotropic properties were determined by applying a steady shear rate of 1 s⁻¹ for 100 s suddenly increasing the shear rate to 1,000 s⁻¹ for 100 s, and then reducing it to 1 s⁻¹ for 100 s. Figure 3B depicts the thixotropic behaviour of the prepared hydrogels. The viscosity of the all the TCNF/ALG precursors rapidly dropped with an applied shear (printing stage) and recovered quickly after the shear force was removed (post-printing stage). However, all the samples were thixotropic and decreased in viscosity after recovery. The average viscosities of the TCNF/ALG precursors under the three shear rate stages are recorded in Supplementary Table S1. The recovery rates (initial/final viscosity) are shown in Table 4.

The frequency sweep tests were performed within the linear viscoelastic deformation region. Loss tangent (tan δ), the ratio of loss moduli (G″), and storage moduli (G″) for TCNF/ALG precursors as a function of frequency are shown in Figure 3C. The loss tangent (tan δ) determines if a material is solid-like or liquid-like. All the TCNF/ALG precursors showed a solid-like behaviour (G’>G″, tan δ< 1). The pure TCNF, 8020, and 9010 showed the most solid-like behaviour followed by 6040, 7030, 2080, and pure ALG and 5050 showed the least solid-like behaviour.

The TCNF/ALG precursors were then bioprinted into mesh shapes and crosslinked using 100 mM of CaCl₂ solution. The bioprinted constructs before and after crosslinking are shown in Figure 4A. The printing pressure, filament diameter, and contractility after crosslinking are shown in Table 4. Apparent size and shape contractions are observed in the TCNF/ALG precursors with low or no TCNF content (0100, 2080, 5050, 6040). On the other hand, the 7030, 8020, and 9010 showed the lowest contractility (≤10%) and best printing fidelity (finest filament diameter). Due to the relatively low alginate content, the 9010 was too soft to retain its shape during transfer or movement of the construct. Therefore, the 7030 and 8020 formulations were selected as bioinks to mix with hMFCs for further biochemical and biomechanical evaluations.

The LIVE/DEAD assay images after bioprinting using a 22 G needle are shown in Figure 4B. No statistically significant difference was observed between 7030 and 8020 bioinks (Figure 4C). The collagen I bioink showed a significantly higher cell viability compared to the 8020 bioink (Figure 4C).

It is commonly understood that Safranin-O would only stain when sulfated GAG is present within the sample. Interestingly, we discovered that cellulose nanofibers (TCNF) could also be stained by Safranin-O despite lacking sulfated GAG, as shown in Supplementary Figure S1 in supplemental materials. Previous studies which used CNFs as biomaterials for cartilage tissue engineering did not report and discuss this phenomenon (Nguyen et al., 2017). Through comparing against cell-free CNF samples, it was still possible to distinguish the Safranin-O positive matrix throughout the histological sections of two TCNF/ALG bioink (Figures 3A,B). In contrast, the bioprinted COL bioink did not show any positive safranin-O positive matrix deposition throughout the section (Figures 5E,F).

Positive aggrecan immunofluorescent staining was shown in both TCNF/ALG bioinks (Figures 6A,B) but not in the COL bioink (Figure 6C), which was consistent with the Safranin-O staining results. The immunofluorescence staining also showed a universal presence of collagen I in two TCNF bioinks (Figures 6D,E) and the COL bioink (Figure 6F). The COL bioink showed more densely packed collagen I compared to the two TCNF/ALG bioinks. However, collagen II was present in the two TCNF/ALG bioinks (Figures 6D,E) and not in the COL bioink (Figure 6F). The collagen II matrix was most concentrated at the outer edges of the constructs. There was no detectable collagen X immunofluorescence in any constructs (Figures 6G–I). The SEM images of the 7030 and 8020 bioinks after 6 weeks of in vitro chondrogenic culture are depicted in Figure 7. Newly synthesized ECM was observed on the surface of the bioprinted constructs for both 7030 and 8020 bioinks. The TCNF material was visible under the ECM.

The mRNA expression of fibrocartilage-related (ACAN, COL1A2, COL2A1, SOX9, and VCAN) and bone-related (COL10A1, MMP13, ALPL, and RUNX2) genes are shown in Figure 8. For all these genes, no significant differences were observed between the 7030 and 8020 bioinks. The fibrous markers COL1A2 and VCAN in the COL bioink were higher than in 7030 bioinks with near statistical significance (i.e., p = 0.086 and p = 0.077, respectively). The cartilaginous marker COL2A1 in the COL bioink was significantly lower compared to the 7030 bioink. No statistically significant difference was observed among the three bioinks in gene expression of ACAN and SOX9. No significant difference was noted in ALPL, and RUNX2 among the three groups for these bone formation-related genes. On the other hand, there was an upregulation of MMP13 in the COL bioink compared to 7030 and 8020 with near statistical significance (i.e., p = 0.066 and p = 0.076, respectively). The expression of COL10A1 in the COL bioink for each donor increased but with no detectable significant difference.

After the strain-controlled unconfined compression test, the peak modulus (instantaneous modulus) was calculated using the force change between the peak and equilibrium forces at each relaxation period. The calculated values were normalized by the cross-sectional area of the 3D bio-printed tissues. The stepwise stress-relaxation as a function of time of one donor (0–10%, 10–20%, 20–30%, and 30–40% strain for each force jump) and the peak modulus as a function of the cumulative strain is shown in Figures 9A,B. No significant interaction was observed between material type and strain in the two-way ANOVA. Therefore, only the main effects of each factor, i.e., strain and bioink material, were tested. Strain was the only significant main effect for the compressive modulus, with the significant differences between each strain level, except for 30% vs. 40%.