Physioxia stimulates extra-cellular matrix deposition and increases mechanical properties of human chondrocyte-derived tissue-engineered cartilage

Cartilage tissue has been recalcitrant to tissue engineering approaches. In this study, human chondrocytes were formed into self-assembled cartilage sheets, cultured in physiologic (5%) and atmospheric (20%) oxygen conditions and underwent biochemical, histological and biomechanical analysis at one- and two-months. The results indicated that sheets formed at physiological oxygen tension were thicker, contained greater amounts of glycosaminoglycans (GAGs) and type II collagen, and had greater compressive and tensile properties than those cultured in atmospheric oxygen. In all cases, cartilage sheets stained throughout for extracellular matrix components. Type II-IX-XI collagen heteropolymer formed in the neo-cartilage and fibrils were stabilized by trivalent pyridinoline cross-links. Collagen cross-links were not significantly affected by oxygen tension but increased with time in culture. Physiological oxygen tension and longer culture periods both served to increase extracellular matrix components. The foremost correlation was found between compressive stiffness and the GAG to collagen ratio. Summary Tissue-engineered cartilage formed from human articular chondrocytes produces thicker, stiffer, more extracellular-matrix rich cartilage tissue when grown under physiological (5%) vs. atmospheric oxygen (20%) tension.


Introduction 33
Cartilage tissue has very poor intrinsic repair capacity. Whilst osteoarthritis is a complex, multifaceted disease, 34 cartilage degradation is a core component. Autologous chondrocyte implantation and matrix assisted 35 autologous chondrocyte implantation have provided relief to patients but commonly result in fibrocartilage 36 repair [1] . Tissue engineering could potentially address this through in vitro culture methods to produce 37 functional hyaline cartilage tissue, with several examples currently in clinical trials [2] . We, and others, have 38 investigated media supplements and growth factors to improve the expansion and re-differentiation of the 39 expanded chondrocytes [3][4][5][6] . 40 It is increasingly apparent that physiological oxygen tension should be the standard culture method to grow 41 tissue engineered human articular cartilage whether it be from mesenchymal stem cells [7] , chondrocytes [8][9][10][11][12][13][14][15][16][17] or 42 chondroprogenitors [18] . The selection of cell type for engineered tissue raises some interesting issues, 43 mesenchymal stem cells commonly progress to hypertrophy [19] as do iPSCs driven down the mesenchymal 44 pathway [20] , a negative scenario for the production of hyaline cartilage. Hypertrophy has been reduced but not 45 eliminated during MSC culture [21] and subcutaneous implants by pharmacological and/or culture-dependent 46 methods [22] . This study focuses on the use of human articular chondrocytes derived from discarded total joint 47 replacement tissue as both a clinically relevant and non-hypertrophic cell source. We, and others, have 48 focused on scaffold-free self-assembly of tissue-engineered cartilage, as significant similarities to native tissue 49 structure can be achieved [17,18,[23][24][25][26][27][28][29] . Significant expansion, up to 8 population doublings, of human 50 chondrocytes whilst maintaining their differentiation capacity has been achieved through their culture on 51 devitalized synoviocyte matrix [13] . Using these methods, we produced sheets of human articular chondrocyte-52 derived cartilage and investigated the effect of low, physiological, oxygen tension and duration of culture on 53 cartilage quality in terms of biomechanics and biochemical content. It was hypothesized that physiological 54 oxygen tension and increased culture duration would improve the mechanical properties of the tissue 55 engineered cartilage through an increase in extracellular matrix content. This study adds to the current body of 56 literature by expanding the donor pool, focuses on chondrocytes isolated from total joint replacement tissue as 57 a cell source, and includes analysis of cartilage-typic collagen heteropolymer formation, collagen cross-links 58 and tensile properties of tissue engineered cartilage. 59 60

-Biochamber model and setup
A circular biochamber design with a 1.13 cm 2 cell seeding area (1) was used with three seeding chambers (3) stacked on top of each other giving a 2.1 ml seeding volume. Stainless steel screws (2) were used to raise the biochambers 1.2 cm allowing media access to the cell sheet through the polyester membrane (7) sandwiched between the bottom 0.2 cm plate and the seeding chambers. Biochambers were assembled and placed in Nalgene containers (6) with a ceramic filter on the lid (4). Defined chondrogenic media was added to the level of the membrane, cells seeded, then media added to 0.2 cm below the top of the topmost seeding chamber.
were punched from the membrane and incubated in cetylpyridinium chloride solution (10%, Alpha Aesar) at 88 37 °C for 20 min. Collagen cross-link analysis -Samples were dried, weighed and then acid hydrolyzed in 6M HCl, 110 °C 105 for 24h. Hydroxylysyl pyridinoline (HP) and Lysyl pyridinoline (LP) cross-linking residues were resolved and 106 quantified by C-18 reverse phase HPLC with fluorescence detection (excitation 297 nm, emission 396 nm) and 107 total collagen content was determined as described [31] . 108 Collagen heteropolymer analysis -Unused portions of the samples used for mechanical analyses were 109 used to qualitatively fingerprint cross-linked collagen types by western blots. The heteropolymeric collagen 110 network formed in the samples was depolymerized in equal volumes of 0.5 M acetic acid containing 100 g/ml 111 pepsin for 18 h at 4 °C. Equal aliquots of solubilized collagen were analyzed by SDS-PAGE and the separated 112 collagen chains visualized by Coomassie blue staining. Pepsin-extracted type II collagen from rabbit articular 113 cartilage was used as a control. The separated collagen chains were also blotted onto PVDF membranes and 114 probed with mAb 1C10 to identify 1(II) chains and with mAb 10F2, pAb 5890, mAb 2B4 to identify collagen 115 chains cross-linked to the C-telopeptide of 1(II), to the N-telopeptide of 1(XI) chains and to the 1(IX) chains 116 respectively [32] . As we have described before, this validates if a heteropolymer of type II and type XI collagen 117 had formed [33] . 118 Histological analysis: Samples were fixed in neutral buffered formalin overnight at 4°C then switched to PBS at 119 4°C until embedded. Samples were embedded by sequential dehydration in graded ethanols, xylene and 120 paraffin. Paraffin sections (8 µm) were deparaffinized and hydrated before staining with safranin-O (Sigma-121 Aldrich) for GAG with a Fast Green (AA16520-06, Alfa Aesar) counterstain. For immunohistochemistry, 122 hydrated sections were subjected to antigen retrieval by pronase (1 mg/ml in PBS containing 5 mM calcium 123 chloride; Sigma-Aldrich) incubation for 10 minutes at room temperature. Primary antibodies against type I 124 collagen (631703, MP Biomedical, 1:1000), type II collagen (DSHB II-II6B3 cell culture supernatant, 1:500) and 125 type X collagen (kind gift of Gary Gibson, Henry Ford Hospital, Detroit, MI; 1:500) were incubated with tissue 126 sections at 4°C overnight. Sections were then rinsed and stained with secondary antibody (biotinylated horse 127 anti-mouse; Gibco; BA2000; 1:2000) for 1h at room temperature before rinsing and incubation with 128 streptavidin-HRP (SNN1004, Invitrogen, 1:5000) 30 min at room temperature. Detection was then made with 129 VIP substrate (Vectashield) by incubation at room temperature for 10 minutes. Slides were rinsed and 130 counterstained with Fast Green before mounting. 131 Mechanical analysis: Samples were thawed in PBS solution equilibrated to room temperature for at least 30 132 minutes. Punches were measured 3 times with digital calipers to assess thickness. Compressive equilibrium 133 moduli were determined as previously described [34] . Briefly, after an initial tare load of 0.2 N, 4 sequential 134 strains of 5%, 10%, 15%, and 20% were applied, with a stress-equilibration period of 30 minutes between each 135 strain step. The stress measured at the end of each strain period was taken as the apparent stress at the 136 corresponding strain level. The equilibrium modulus was determined as the slope between the apparent stress 137 and the strain. 138 To test elastic tensile Young's moduli, a custom dogbone punch was made from skin biopsy punches and 139 punches taken from the 5mm punch ( Fig. 2A). Custom holders were made from overhead projector sheets 140   Biochemical assays: There was a significant increase in extracellular matrix content, both in terms of 166 GAG/DNA (Fig. 3A) and collagen/DNA in human tissue-engineered cartilage sheets when cultured in physioxia 167

Figure 3 -Extracellular matrix deposition in tissue-engineered human cartilage sheets
Tissue-engineered human cartilage sheets had higher glycosaminoglycan content in Physioxia vs. Atm O2, time in culture had no appreciable effect (A). Sheets had more collagen when grown in Physioxia at 7 weeks (B). Trivalent Collagen collagen cross-links density increased over time, with no appreciable effect of oxygen tension (C). Note, all measures are shown on a log (base 2) scale.

GAG/DNA ( g/ g) Collagen Crosslink Density (HP+LP/Collagen)
vs. atm O 2 (Fig. 3B). This increase in GAG/DNA was not significantly affected by time in culture (Fig. 3A). Only 168 under physioxia was the accumulation of collagen/DNA greater at the 7-week time point (Fig. 3B). 169 Collagen trivalent hydroxylysyl and lysyl pyridinoline cross-links were formed in culture at both oxygen 170 tensions and increased with time in culture, with no apparent effect of oxygen tension (Fig. 3C). To determine if 171 cartilage-specific collagen type II-IX-XI heteropolymer formed in culture we used specific antibodies to 172 qualitatively fingerprint cross-linked collagen chains as we previously established [33] . Physioxia. Purified human type II collagen (lane1) and tissue engineered human cartilage is shown in the lanes 175 2-5. Lanes 2, 4 and lanes 3, 5 were normalized to wet weight of unused portion of neo-cartilage after 176 mechanical tests. Only a qualitative evaluation of collagen in these samples was possible. The major 177

Figure 4 -Collagen heteropolymer crosslink analysis
Collagen in human tissue-engineered cartilage was analyzed by SDS page and western blot. Lane 1) human type II collagen (control); 2) 7-week Atm O 2 ; 3) 7-week Physioxia; 4) 3-week Atm O 2 ; 5) 3-week Physioxia. Panels: A) Coomassie blue stain; B) Western blot type using II collagen antibody to 1(II) chain in helical region; C) Western blot using antibody to C-telopeptide of 1(II) chains of type II collagen; D) Western blot using antibody to N-telopeptide of 1(XI) chains of type IX collagen; E) Western blot of antibody that recognizes C-telopeptide of the non-collagenous domain of 1(IX) chains of type IX collagen of type IX collagen. F) Molecular interpretations of collagen heteropolymer assembly from Western blot analysis. Locations of the epitopes in the 1(II) chain or in telopeptide stubs cross-linked to the chains are also shown Coomassie blue stained pepsin-resistant chain observed was the 1(II) collagen chain which migrates similarly 178 to the chain seen for human type II collagen purified from adult articular cartilage. Two other pepsin-resistant 179 chains of varying intensities were observed (lanes 2-5), migrating above the 1(II) chains. The chains migrate 180 similarly to the 1(XI) and 2(XI) chains in type XI collagens. The 2(I) and 2(I) chain characteristic of type I 181 collagen is a minor component of the tissue engineered cartilages. 182 Western Blot using the antibody 1C10 (Fig 4B) confirmed the 1(II) and the 1(II) chains of type II 183 collagen (lanes 2-5) indicating the chondrocytes elaborated an extensive extracellular matrix containing type II 184 collagen. The antibody 10F2 reacted with the 1(II) chain as expected for a cross-linked type II collagen 185 polymer (Fig 4C), (C-telopeptide of type II collagen cross-linked to 1(II) collagen chains) indicating a cross-186 linked collagen network assembled in the all the neo-cartilages. Following an extended exposure, the antibody 187 also reacted with 1(XI) collagen chains implying this chain was cross-linked to the C-telopeptide of type II 188 collagen and that type XI collagen was copolymerized and cross-linked to C-telopeptides of type II collagen 189 Mechanical Assays: Tissue-engineered human cartilage sheet thickness increased with time in culture at both 201 physiological and atmospheric oxygen tensions (Fig. 5A). Sheets produced in physioxia were significantly 202 thicker than those produced in atm O 2 (Fig. 5A). Compressive stiffness of the cartilage sheets was greater in 203 sheets grown in physioxia at both 3-weeks and 7-weeks (Fig. 5B), time in culture was only a significant factor 204 for sheets grown in physioxia. Tensile stiffness of sheets increased with time in culture at both physiological 205 and atmospheric oxygen tensions (Fig. 5C), for this measure there was no appreciable effect of oxygen 206 tension. The highest correlation for compressive mechanical stiffness with biochemical measures was 207 achieved with the ratio of GAG/collagen (D). None of the biochemical data showed significant correlation with 208 elastic tensile modulus (Data not shown). 209

Figure 5 -Mechanical testing of tissue-engineered human cartilage sheets
Tissue-engineered human cartilage sheets grew significantly thicker over time in both Physioxia and Atm O 2 , with thicker sheets overall in Physioxia (A). Sheets were stiffer in compression when grown at physiological oxygen tension vs. Atm O 2 both at 3 weeks and 7 weeks (B). When tested for tensile properties, sheets gained tensile stiffness over time but oxygen tension had no appreciable effect (C). Equilibrium modulus correlated with GAG/Collagen ratio (D).

D)
Histological analyses: Tissue-engineered human cartilage sheets were stained for glycosaminoglycan content 210 Tissue-engineered human cartilage sheets from a single donor are shown (see supplemental data for other donors), they were stained for glycosaminoglycan (safranin-O, column A), type I collagen (column B), type II collagen (column C) and type X collagen (column D). Rows 1 and 2 show data from the 3-week time point and rows 3 and 4 show data from week-7. Atmospheric O2 sheets are shown in rows 1 and 3, physioxic sheets are shown in rows 2 and 4. The scale bar indicates a 1 mm distance. 0.42 vs. 0.52 ± 0.19 mm; mean ± S.D.). Glycosaminoglycan staining was more intense in sheets grown under 214 physioxic conditions (Fig. 6A2) vs. atmospheric oxygen tension (Fig. 6A1). Type I collagen and type II collagen 215 staining was similar under both oxygen tensions at 3-weeks (Fig. 6B1, 6B2, 6C1, 6C2). Type X collagen 216 staining was slightly increased under atmospheric oxygen tension (Fig. 6D1) vs. physioxic (Fig. 6D2) 217 conditions at week 3. At the 7-week time point, safranin-O staining in sheets grown in atmospheric oxygen 218 tension (Fig. 6A3) had increased to a similar level as sheets grown in physioxic conditions (Fig. 6A4). Type I 219 collagen staining under atmospheric oxygen tension at week 7 (Fig. 6B3) has decreased intensity vs the 3-220 week time point (Fig. 6B1). Similarly, the sheets grown under physioxic conditions at week 7 have reduced or 221 minimal staining for type I collagen (Fig. 6B4) vs. (Fig. 6B2). Type II collagen was relatively intense under both 222 oxygen tensions (Fig. 6C3 and 6C4). Type X collagen staining was more intense on the upper surface of the 223 sheet grown under atmospheric oxygen tension at week 7 (Fig. 6D3). In sheets grown under physioxic 224 conditions, type X collagen staining was predominantly intracellular vs. extracellular (Fig. 6D4). remarkable. Similarly, a consistent increase in total collagen deposition was also found through culture under 232 physiological oxygen conditions. Unfortunately, the biochemical assay for collagen does not discriminate 233 between the different types of collagen. This shortcoming is apparent when looking at the histological data, 234 where temporal and regional variation in collagen type and intensity are evident. Initial expression and 235 replacement of type I collagen has been documented developmentally in vivo [36,37] and, as we have also 236 shown in human bone marrow derived stem cell neo-cartilage, engineered in vitro [33] . This could indicate that 237 expression and replacement is a normal progression in tissue-engineered cartilage development and that 238 replacement of type I with type II collagen is aided by physioxia. There was no apparent effect of oxygen 239 tension on trivalent collagen cross-linking, but significant increases in total hydroxylysyl and lysyl pyridinoline 240 cross-link formation were observed with longer culture duration. This indicates that under both physioxic and 241 normoxic conditions a fibrillar network of collagen with mature cross-links had formed in neo-cartilage. Western 242 blot analyses using established antibodies to specific collagen peptides involved in covalent cross-link 243 formation [32,35] indicated that a cross-linked hetropolymer of type II-IX-XI collagen had formed in the tissue 244 engineered cartilage. This cross-linked collagen heteropolymer is typical of cartilage and is essential in the 245 proper assembly of the cartilage collagen fibril [38] . Our findings that a similar nascent heteropolymeric template 246 is formed in human neo-cartilage with increased cross-linking with time in culture point to a progressive 247 formation of type II collagen based fibril network typical of cartilage. 248 249 250 Subjective assessment of the sheets (physical handling) indicate that longer culture durations gave stronger 251 sheets in all cases, although this was not fully supported by the equilibrium moduli which only showed a benefit 252 in sheets grown under physioxia. This is potentially due to untestable sheets formed under atmospheric 253 oxygen culture (11 of 16 sheets). However, the results did show an increase in both total collagen content and 254 collagen cross-link content with time. These increases correlated with the increased tensile properties of the 255 tissues.  also found that hypoxia increased collagen crosslinks in tissue engineered bovine 256 cartilage constructs but that they also found weak correlations to compressive mechanical properties [39] . The There remain significant challenges in producing autologous, tissue-engineered cartilage. Current clinical trials 262 cover a wide range of approaches, many using allogenic cells (for review see [2] ). The advantages of a well 263 characterized allogenic cell bank are clear given the range of GAG and collagen contents due to donor 264 variability. In chondrocyte progenitor experiments, this has been further focused in on showing clonal 265 variability [40] . Interestingly, while there was a wide range of extracellular matrix component concentrations 266 detected, the distribution in mechanical properties was actually relatively narrow. Evans et al. (2005) found that 267 cartilage compressive stiffness correlated with GAG density in native tissues [41] . Roeder et al. found that tensile 268 modulus increased with increasing concentration of type I collagen engineered constructs in a linear 269 manner [42] . Our data showed the greatest correlation of biochemical measures with compressive moduli when 270 the GAG/Collagen ratio was used but, probably due to the 4 conditions and 6 donors, this correlation was 271 relatively weak. When looking at biochemical content correlations with tensile properties, nothing gave a 272 significant correlation; this analysis was hampered by the number of conditions analyzed and lack of sheet 273 formation for two of the 4 donors at atmospheric oxygen tensions. Similarly, Williamson et al. (2003) found no 274 significant correlation between bovine fetal cartilage and tensile tests [43] . Overall, the data indicate that 275 physiological oxygen tension is beneficial to chondrogenesis of human tissue-engineered cartilage sheets 276 formed through scaffold-free culture of human articular chondrocytes. 277 278

Conclusions 279
Tissue-engineered human cartilage sheets, formed through scaffold free self-assembly of articular 280 chondrocytes, have significantly more extracellular matrix with correlative increases in compressive stiffness. 281 Temporal increases in collagen crosslinks and type were more evident in sheets grown under physiological 282 oxygen tension and these correlated with increased tensile stiffness.