Effects of cell–cell contact and oxygen tension on chondrogenic differentiation of stem cells

Biomaterials 64 (2015) 21e32

Contents lists available at ScienceDirect

Biomaterials journal homepage: www.elsevier.com/locate/biomaterials

Effects of cellecell contact and oxygen tension on chondrogenic differentiation of stem cells Bin Cao, Zhenhua Li, Rong Peng, Jiandong Ding* State Key Laboratory of Molecular Engineering of Polymers, Department of Macromolecular Science, Advanced Materials Laboratory, Fudan University, Shanghai 200433, China

a r t i c l e i n f o

a b s t r a c t

Article history: Received 5 March 2015 Received in revised form 8 June 2015 Accepted 11 June 2015 Available online 15 June 2015

While cell condensation has been thought to enhance chondrogenesis, no direct evidence so far confirms that cellecell contact itself increases chondrogenic differentiation of stem cells, since the change of cell ecell contact is usually coupled with those of other cell geometry cues and soluble factors in cell culture. The present study semi-quantitatively examined the effect of cellecell contact in a decoupled way. We fabricated two-dimensional micropatterns with cell-adhesive peptide arginine-glycine-aspartate (RGD) microdomains on a nonfouling poly(ethylene glycol) (PEG) hydrogel. Mesenchymal stem cells (MSCs) were well localized on the microdomains for a long time. Based on our micropattern design, single MSCs or cell clusters with given cell numbers (1, 2, 3, 6 and 15) and a similar spreading area per cell were achieved on the same substrate, thus the interference of soluble factor difference from cell autocrine and that of cell spreading area were ruled out. After 9-day chondrogenic induction, collagen II was stained to characterize the chondrogenic induction results; the mRNA expression levels of SOX9, collagen II, aggrecan, HIF-1a and collagen I were also detected. The statistics confirmed unambiguously that the extent of the chondrogenic differentiation increased with cellecell contact, and even a linear relation between differentiation extent and contact extent was established within the examined range. The cell ecell contact effect worked under both hypoxia (5% O2) and normoxia (21% O2) conditions, and the hypoxia condition promoted the chondrogenic induction of MSCs on adhesive microdomains more efficiently than the normoxia condition under the same cellecell contact extents. © 2015 Elsevier Ltd. All rights reserved.

Keywords: Chondrogenic differentiation Mesenchymal stem cell Cellecell contact Oxygen tension Micropatterning PEG hydrogel

1. Introduction Cell-based cartilage tissue engineering is a promising approach for cartilage repair since impaired cartilage is difficult to self-heal due to the nonvascularized and noninnervated nature of any cartilage. The most important pertinent seeding cells used are chondrocytes and mesenchymal stem cells (MSCs). Previous studies on MSC-based cartilage tissue engineering focused on the optimization of chondrogenic induction protocols and the combination of MSCs with different scaffolds and stimulations to improve repair outcomes [1e5]. Among the key factors that facilitate successful restoration of cartilage lesion, cell condensation is of special importance, which is also a natural progress during cartilage formation in vivo. This process increases the extent of cellecell contact, which might be beneficial for chondrogenesis as thought by many researchers [6e9].

* Corresponding author. E-mail address: [email protected] (J. Ding). http://dx.doi.org/10.1016/j.biomaterials.2015.06.018 0142-9612/© 2015 Elsevier Ltd. All rights reserved.

However, once the cellecell contact is altered by cell condensation or an increase of cell density, other cell geometry cues such as cell spreading sizes might also be altered, especially in the case of a two-dimensional (2D) culture. Meanwhile, the concentration of the soluble factors might also be changed by autocrine of the cultured cells with cell condensation. Therefore, decoupling of the cellecell contact effect from those interference factors is essential in order to conclude the cellecell contact effect solely on chondrogenic induction of stem cells. A deterministic experiment is thus strongly called for. In the present study, we investigated the effect of cellecell contact on the chondrogenic induction extent in vitro in a decoupled way, as schematically presented in Fig.1A. A series of adhesive microdomains on a persistently nonfouling background were generated for the adhesion of single cells or cell clusters with given cell numbers, where the interference of different soluble factors could be ruled out. We also expected to fix other cell geometry cues such as cell sizes on microdomains. The prerequisite is localization of cell clusters with given cell numbers on cell-adhesive microdomains on the same substrate.

22

B. Cao et al. / Biomaterials 64 (2015) 21e32

Fig. 1. A) Schematic presentation of the basic idea to examine whether cellecell contact influences the chondrogenic induction. Stem cells on the microdomains of different microisland numbers experience the same microenvironment except contact extent. B) Fabrication flow chart of surface micropatterning with cell-adhesive microdomains of different microisland numbers on a persistent nonfouling PEG background. MSCs will be seeded on the micropattern and cultured in the chondrogenic medium for 9 days under hypoxia (5% O2) and normoxia (21% O2) conditions. The effects of cellecell contact and oxygen tension on chondrogenic differentiation will be examined.

Various micropatterning and nanopatterning techniques for cell studies have been developed to explore the interaction between cells and biomaterials [10e15]. In this study, we employed an arginine-glycine-aspirate aspartate (RGD) peptide micropatterning technique on poly(ethylene glycol) (PEG) hydrogels that we previously reported [16e19]. The overall preparation process is schematically described in Fig. 1B. Briefly, the micropatterning technique includes three steps including (1) lift-off photolithography to generate a micropattern of gold on glass, (2) micropattern transferring from glass surface to the surface of a nonfouling poly(ethylene glycol) (PEG) hydrogel, and (3) formation of a selfassembly monolayer of cell-adhesive RGD peptide on the gold microislands, which can enhance cell adhesion [20,21]. The above three steps enable the successful acquisition of RGD microdomains on the persistently nonfouling PEG hydrogel. On these micropatterns, cells can be located within the microdomains for a long time. In this study, we carried out the chondrogenic induction of MSCs on the micropatterned surfaces for 9 days in a chondrogenic medium. The micropatterns were designed to contain microdomains with microisland numbers 1, 2, 3, 6, and 15. Each microisland had the same area, which probably locates only one cell. The extent of cellecell contact was thus well controlled by numbers of cells adhering on those microdomains that correspond with microisland numbers. Furthermore, the cell numbers on each microdomain will not change during induction, as adhesive areas are limited and a proliferation inhibitor aphidicolin will be introduced. Hence, the

contact effect on the chondrogenic extent can be examined without the disturbance of other environment factors. While it has been recognized for a long time that cellecell contact might enhance chondrogenesis, the present study will, based upon the unique micropatterning technique to control cellecell contacts, set up the first semi-quantitative relation between in vitro chondrogenic induction extents and cellecell contact extents. In addition, low oxygen tension has been reported to be beneficial for the chondrogenesis of MSCs [22e28]. One possible reason is that MSCs proliferate faster in the hypoxia condition than that in the normoxia condition [29,30]. However, since faster proliferation in the hypoxia condition leads to higher cellecell contact extents, it remains unknown whether the increased chondrogenesis is caused directly by hypoxia or indirectly by higher cellecell contact extent derived from higher cell proliferation. One strategy to answer this argument seems to be from examination of chondrogenic induction of MSCs in parallel experiments of different oxygen conditions but under the same cellecell contact (say, with the same initial cell density on tissue culture plates) with addition of a proliferation inhibitor such as aphidicolin. However, according to our trial, it is impossible to keep the same global cell density in culture under the two oxygen conditions with a fixed inhibitor concentration; on the other hand, if one increased the aphidicolin concentration in the hypoxia condition, the inhibitor concentration became another interference factor, which makes the comparison between the “direct” effect of the hypoxia condition versus the normoxia condition still not conclusive. Taking advantage of our micropatterning

B. Cao et al. / Biomaterials 64 (2015) 21e32

technique, we put forward the deterministic strategy to make comparison between the two oxygen conditions under the same “local” cell density, which can be achieved by controlling numbers of MSC cells on adhesive RGD microdomains separated by a nonfouling PEG hydrogel background. The present study observed in vitro chondrogenic induction of MSCs on a 2D surface in both hypoxia (5% O2) and normoxia (21% O2) conditions under a given aphidicolin concentration (0.5 mg/ml). Our unique material technique to maintain cell numbers on microdomains enabled the examination whether the cellecell contact regulates chondrogenic differentiation of MSCs under different oxygen conditions and whether oxygen tension has a direct effect on chondrogenic induction of stem cells under the same cellecell contact. 2. Materials and methods 2.1. Fabrication of gold microarrays on PEG hydrogel The method to fabricate micropatterns with persistent nonfouling background has been described previously [31,32]. Briefly, gold microdomains were first prepared on glass slides by photolithography with a chrome mask of a pre-designed micropattern, and then transferred onto the PEG hydrogel. The transferring step is the key to successful preparation of micropatterns. In this procedure, glass slides with gold microdomains were immersed into 1 mM N, N’-bis(acryloyl) cystamine (Sigma), a bifunctional linker solution, for at least 2 h. The linkers were bound on gold microdomains via the AueS bond. Extra linkers were washed away with ethanol 3 times (15 min for each), and the glass slides were dried by nitrogen gas. After that, poly(ethylene glycol) diacrylate (PEGDA with molar mass 700 Da, Sigma) mixed with photoinitiator 2hydroxy-4’-(2-hydroxyethoxy)-2-methylpropiophenone (Sigma) was applied onto the glass surface. Due to good performance [33], PEG and its block copolymers are widely used in biomedical applications such as post-operative adhesion [34], cell encapsulation [35] and drug release [36,37]. The PEG macromonomer was then cross-linked under UV irradiation in nitrogen atmosphere for 1 h. The gold microdomains were transferred from the glass onto the PEG hydrogel after the cross-linked network was peeled off the glass. RGD peptides were self-assembled on the gold microdomains before seeding cells. 2.2. Isolation and culturing of MSCs MSCs were isolated from 7-day old neonatal Sprague Dawley (SD) rats. Both ends of the femur were cut off, and the contents in the femur cavity were rushed out using an injection syringe with low-glucose Dulbecco's modified Eagle medium (DMEM, Gibco) thoroughly. The suspension was centrifuged at 800 rpm for 10 min. The supernatant was aspirated and the sediment was re-suspended in a growth medium. The growth medium is low-glucose DMEM supplemented with 10% fetal bovine serum (FBS, Gibco), 100 U/ml penicillin, 100 mg/ml streptomycin (Gibco), and 2 mM L-glutamine (Gibco). After 2 days, the non-adhesive cells were discarded by changing medium. Cells were passaged at about 70e80% confluence. The cells at passage 2 were used for later experiments.

23

Afterwards, the as-prepared micropatterns were disinfected in 70% ethanol for 30 min, and then thoroughly washed with phosphate buffer saline solution (PBS, Hyclone) for 6 times (30 min for each), then put into 12-well culture plates (Corning). MSCs were seeded at the density of 5  104 per well with growth medium. After incubation for 1 h, non-adhesive MSCs were absorbed off, and a fresh chondrogenesis medium was added. The cells were incubated under either hypoxia (5% O2) or normoxia (21% O2) condition for 9 days. The chondrogenic medium was high-glucose DMEM (Gibco) supplemented with 10 ng/mL transforming growth factorbeta 1 (TGF-b1, R&D), 100 nM dexamethasone (Sigma), 50 mg/mL ascorbate-2-phosphate (Sigma), ITSþ1 Liquid Media Supplement (100x) (Sigma), 5% FBS, 100 U/ml penicillin, 100 mg/ml streptomycin, and 2 mM L-glutamine. The chondrogenic medium was replaced every 3 days. In order to inhibit cell proliferation, 0.5 mg/ml of aphidicolin (Sigma) was added throughout the incubation period. In some samples, 50 mM of 18-a-glycyrrhetinic acid (AGA, Sigma) was applied to block gap junctions. 2.4. Live/dead staining of cells on micropatterns After 9-day culture, the chondrogenic medium was aspirated, and MSCs on micropatterns were rinsed with PBS gently. Then the PBS solution of ethidium homodimer-1 (1 mM) and calcein AM (2 mM) (Live/Dead Viability/Cytotoxicity Kit, Invitrogen) was added into each well and incubated at 37
C for 30 min. The dye liquor was replaced with PBS for later observation. Live cells emitted green fluorescence, while dead cells emitted red fluorescence. 2.5. Immunofluorescence staining of cells on micropatterns At day 9, cells on micropatterns were rinsed with PBS gently. Samples were then fixed in 4% paraformaldehyde for 10 min, treated with 0.1% Triton X-100 for membrane permeation for 10 min and blocked with 5% bovine serum albumin (BSA) for 20 min. For observation of vinculins, samples were incubated with the primary antibody (mouse monoclonal anti-vinculin, Sigma) at the dilution of 1:100 at 4
C overnight. The secondary antibody (Alexa Flour 488-conjugate goat anti-mouse IgG, Invitrogen) was added on the next day at a concentration of 20 mg/mL, and incubated at room temperature in dark for 1 h. In order to identify filamentous actins (F-actins) of MSCs, 1 mg/ml phalloidin-TRITC (Sigma) was added, and cells were incubated at room temperature for 30 min. Finally, 2 mg/mL 40 ,6-diamidino-2-phenylindole (DAPI, Sigma) was added to label the nuclei. For collagen II staining, after 9-day culture in the chondrogenic medium, cells on micropatterns were fixed with 4% paraformaldehyde, treated with 0.1% Triton X-100, blocked with 5% BSA. Cells were then incubated with the primary antibody (mouse monoclonal anti-collagen II IgG20, Santa Cruz Biotech) at 4
C overnight, and on the next day, further incubated with a streptavidin-biotin-Cy3 complex immumofluorescence staining kit (SABC-Cy3) (SA1072, Boster) under the manufacturer's instruction. The stained collagen II emitted red fluorescence. 2.6. Observation and analysis of the stained cells on micropatterns

2.3. Seeding MSCs on micropatterns Before seeding cells, the prepared micropatterns (as described in Section 2.1) were immersed in a water solution containing 25 mM cyclic peptides c(-RGDfK-)-OEG-COCH2CH2SH (MW 836.4 Da) (R: Larginine, G: L-glycine, D: L-aspartic acid, f: D-phenylalanine, and K:
L-lysine) (Peptides International, USA) at 4 C for 4 h. The RGD was bound to the gold microdomains through SeAu bond.

After 9-day culture, MSCs were immunofluorescently stained with collagen II, and the micrographs were captured by a charge coupled device (CCD) in an inverted fluorescence microscope (AXIOVERT 200, Zeiss). The original micrograph was treated by the free software Image J (NIH), and split into red, blue and green channels. As the collagen II was stained with dye Cy3, the red channel was preserved and changed into an 8-bit greyscale image.

24

B. Cao et al. / Biomaterials 64 (2015) 21e32

Then the outline of the cells on microdomains was depicted, and the mean grey value of collagen II staining (I) and the cell area (Scell) were got by the software Image J. A background region near the microdomain with a similar area was chosen to obtain the mean grey value (I0) as well, as demonstrated in Fig. 2. The cell numbers (N) on each microdomain were counted based on the corresponding fluorescence micrograph of cells with nuclei stained by DAPI. The relative intensity of collagen II per cell on each microdomain was calculated from

mean ± standard deviation. The mean of group D1 was used for normalization. Data from qRT-PCR with independent groups (n ¼ 3) were performed by 2DDCt method. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was chosen as the housekeeping gene. Results were normalized by group D1 under the normoxia condition. One-way ANOVA was performed to evaluate the results among different groups. The difference was considered as significant if p < 0.05.

Col II ¼ ½Scell ðI  Io Þ=N

3. Results

The average was gained first on the appropriate microdomains with different microisland numbers (single cells denoted as D1, 2 cells as D2, 3 cells as D3, 5e7 cells as D6, and 11e19 cells for D15). A total of 5 independent experiments were performed (n ¼ 5). For each sample at least 500 appropriate microdomains were measured. 2.7. Detection of the related mRNA expression of cells on micropatterns Eight (23) groups of cells on microdomains were used to detect gene expression: single cells (D1) or cells in contact (D15), presence or absence of the gap junction inhibitor AGA, normoxia or hypoxia condition. The total RNAs were extracted by MagneSil Total RNA mini-Isolation System (Z3351, Promega) according to the manufacture's instruction, then reversely transcribed into cDNAs by applying PrimeScript RT reagent Kit (RR047A, Takara). The mRNA expression level was examined by quantitative real-time polymerase chain reaction (qRT-PCR, Rotor Gene Q System, Qiagen) with SYBR Green PCR Kit (Qiagen). The primers (Invitrogen) sequences are listed in Table 1 in Ref. [38]. Samples were run at 95
C for 5 min, then 40 circles of 95
C for 5 s, 65
C for 10 s and 60
C for 20 s. In order to perform PCR for D1 and D15, two new masks for fabricating micropattern were adopted. The microisland numbers in the basic repeat units are the same as the mask for other cell analysis, as seen in Fig. 1A, B and C in Ref [38], which guaranteed a similar global cell density on the three micropatterned samples. 2.8. Distribution of MSCs on microislands The numbers of cells on microdomains of different microisland numbers after 9-day chondrogenic induction were counted based on DAPI-stained fluorescence images of cells. For each type of microdomains, at least 300 microdomains were counted to obtain the population distribution of adhesive MSCs on microdomains of various microisland numbers.

3.1. Preparation of micropatterns with RGD microdomains of different microisland numbers With the micropatterning technique described in Section 2.1, gold microdomains with different microisland numbers (1, 2, 3, 6, and 15) were fabricated on a PEG hydrogel, as illustrated in Fig. 3. A bifunctional linker was used in transferring the micropattern from the glass surface to the PEG hydrogel surface. The microislands were covalently bound to the underneath PEG hydrogel, and thus the micropattern was very stable under cell culture medium. Since different microdomains were generated on the same substrate, the cellecell contact effect was studied without the influence of other physical and chemical factors. The MSC culture and induction on micropatterns were examined under both hypoxia (5% O2) and normoxia (21% O2) conditions. We will first show the results from the hypoxia condition in the following subsections. 3.2. Achievement of cell localization on microdomains The examination of the cellecell contact effect should be based upon a lasting cellecell contact during a relatively long time for cell induction (9 days in the present study). The cell localization efficacy was thus first examined. As shown in Fig. 4, cells were well localized on the adhesive microdomains separated by the nonfouling PEG hydrogel after 9-day chondrogenic induction. Also, the adherent cells on micropatterns exhibited very good viability, for most cells emitted green fluorescence. During the culture period, cells didn't spread out of the microdomains, as demonstrated in Fig. 5 with F-actins, vinculins and nuclei fluorescently stained. It confirms that the micropatterns with cell-adhesive microdomains and nonfouling PEG background well controlled the localization of cells for a long time.

2.9. Data analysis Data from immunofluorescence staining of collagen II with independent groups (n ¼ 5) were averaged, represented as

Fig. 2. Method to obtain the intensity of collagen II staining for cells on a microdomain. Yellow dotted line traces out the stained region and a nearby background with a similar area. I represents the mean grey value of collagen II staining region, I0 represents the mean grey value of the nearby background. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 3. Bright-field optical micrograph of as-prepared microdomains of 1, 2, 3, 6 and 15 microislands on the PEG hydrogel. The diameter of every microisland was 30 mm. The microislands refer to gold microsheets half-trapped into the PEG hydrogel and with the exterior surface coated by an RGD monolayer.

B. Cao et al. / Biomaterials 64 (2015) 21e32

25

Fig. 4. Fluorescence micrographs of cells with live/dead staining. MSCs experienced 9-day chondrogenic induction under 5% O2 before observations. The micrographs indicate that cells were only located within the microdomains and exhibited viability during the culture and induction. Dashed lines indicate the boundaries of cell-adhesive microdomains. (Live cells emit green fluorescence while dead cells emit red.) (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

3.3. Achievement of the cell numbers on each microdomain with respect to the number of microislands Cell adhesion on each microdomain was imaged from the fluorescence observation of cells with cytoskeleton and cell nuclei stained, as shown in Fig. 5. Especially, the DAPI staining of DNA in cell nuclei affords a reliable way to count the number of cells on each microdomain.

The population distribution of cells on microdomains is shown in Fig. 6A. Taking advantage of our micropatterning technique and the appropriate selection of microisland area, a good correspondence was achieved between the most probable cell number on each microdomain and the microisland number: 1 for D1, 2 for D2, 3 for D3, 6 for D6 and 13e15 for D15. The average cell numbers perfectly matched their corresponding microisland numbers, as indicated in Fig. 6B.

Fig. 5. Fluorescence micrographs of cells with F-actins, vinculins and nuclei stained. MSCs experienced 9-day chondrogenic induction under 5% O2 before observations. Cell numbers could be counted through the images after staining of cell nuclei by DAPI. The dashed lines indicate the positions of the cell-adhesive microdomains.

26

B. Cao et al. / Biomaterials 64 (2015) 21e32

Fig. 6. Number of adherent cells on RGD microdomains on the PEG hydrogel. MSCs were cultured under 5% O2. A) Distribution of cell number on microdomains with indicated microisland numbers. B) Average cell number on microdomains as a function of microisland number.

3.4. Collagen II staining and the expression of collagen II per cell on each microdomain in hypoxia condition (5% O2) After 9-day chondrogenic induction, collagen II (a marker for chondrocytes) was immunofluorescently stained with the method described in Section 2.5. Some representative images are shown in Fig. 7A. With the method of quantitative fluorescence microscopy introduced in Section 2.6 along with the scheme presented in Fig. 2, the relative intensity of collagen II per cell on each microdomain was obtained with the statistical results shown in Fig. 7B. The intensity of collagen II per cell increased with the microisland number, which indicates that the extent of chondrogenic induction is positively related to the cellecell contact. 3.5. Chondrogenic induction of MSCs under normoxia condition (21% O2) The cellecell contact effects on the chondrogenic induction of MSCs also exist under normoxia condition (21% O2). The data are summarized in Fig. 8. After 9-day chondrogenic induction, cells did not spread out of the microdomains and exhibited good viability (Fig. 8A and D). The population and average cell number on each microdomain well matched with corresponding microisland numbers (Fig. 8B and C). The images with collagen II stained in Fig. 8E and the statistic results in Fig. 8F indicated that the intensity of collagen II per cell increased with the microisland number. Therefore, the cellecell contact effect on cell differentiation is a universal phenomenon. 3.6. Comparison of the chondrogenic induction extents between hypoxia and normoxia conditions While the trends of chondrogenic induction as a function of cellecell contact are similar, the chondrogenic extents under hypoxia and normoxia conditions are not necessarily identical to each other. After integrating the statistic results of collagen II staining under both hypoxia and normoxia conditions in Fig. 9A, it is clear that the intensity of collagen II staining per cell is higher in the hypoxia condition than that in the normoxia condition. In order to confirm the conclusion, we further detected the expression of characteristic genes, using the qRT-PCR method described in Section 2.7. Cells on D1 and D15 both in the hypoxia and normoxia conditions were quantified. According to Fig. 9BeF, the gene expressions of SOX9, collagen II, aggrecan and HIF-1a were higher under the hypoxia condition than that under normoxia

Fig. 7. Results of quantitative fluorescence microscopy for the chondrogenic induction of MSCs on the micropatterned surfaces under 5% O2. A) The upper row shows representative micrographs of cells with collagen II staining after 9-day chondrogenic induction. The lower row displays the corresponding images with nucleus staining. The dashed lines indicate the positions of the cell-adhesive microislands. B) Relative intensity of collagen II per cell on microdomains with microisland numbers 1, 2, 3, 6 and 15. The data were normalized by the mean of microdomain A with respect to isolated cells. Comparison results between two nearest groups are marked in the figure with “*” (p < 0.05) and “***” (p < 0.001). The global test among all of the groups analyzed from One-way ANOVA read p ¼ 1.3  109 << 0.001.

B. Cao et al. / Biomaterials 64 (2015) 21e32

27

Fig. 8. Adhesion and chondrogenic induction of MSCs on micropatterns for 9 days under normoxia condition (21% O2). A) Fluorescence micrographs of cells on microdomains with different microisland numbers. (Red: F-actin, green: vinculin, blue: nucleus.) B) Distribution of cell number on microdomains with indicated microisland numbers. C) Average cell numbers on microdomains versus microisland numbers. The dashed line indicates data y ¼ x. D) Live/dead staining results. (Green fluorescence indicates live, and red fluorescence indicates dead.) E) Representative micrographs of collagen II staining of cells (upper) and the corresponding micrographs of nucleus staining (lower). Dashed lines indicate the boundaries of adhesive microdomains. F) Relative intensity of collagen II per cell on microdomains with microisland numbers 1, 2, 3, 6 and 15. Comparison results between two nearest groups are marked with “*” (p < 0.05) and “***” (p < 0.001). The global comparison between all of the five groups analyzed from One-way ANOVA resulted in p ¼ 6.9  1012 << 0.001. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

condition, indicating an enhanced chondrogenic induction. Furthermore, under the same oxygen tension, the expressions of these four genes were higher on D15 when compared with D1. Collagen I expressions were lower under D15 and the hypoxia condition. The low expressions of collagen I and the high expressions of collagen II are consistent with each other, as both of which characterize the chondrogenic induction.

3.7. Linear relation of extent of cellecell contact to that of chondrogenic induction As the extent of chondrogenic differentiation increases with that of cellecell contact, we further tried to find possible semiquantitative relationship between them. Both coordination number and microisland number were chosen as parameters to semi-

28

B. Cao et al. / Biomaterials 64 (2015) 21e32

Fig. 9. Comparison of chondrogenic induction of MSCs on micropatterned surfaces between hypoxia and normoxia conditions. A) The intensity of collagen II staining per cell on microdomains with microisland numbers 1, 2, 3, 6 and 15. BeF) qRT-PCR results on indicated microdomains (1 and 15) for relative expression levels of genes of B) SOX9, C) collagen II, D) aggrecan, E) collagen I, and F) HIF-1a. All the data are normalized by the means of microdomain with microisland number 1 under the normoxia condition. Significant differences are marked with “*” (p < 0.05) and “***” (p < 0.001).

quantify the extent of cellecell contact. As shown in Fig. 10A, a coordination number denotes the average cell number that one cell can contact with on each microdomain, and a microisland number signifies the number of cells on each microdomain. Then we put the coordination number or microisland number as the x-coordinate, and the corresponding intensity of collagen II staining per cell as the y-coordinate to progress linear fitting. Fig. 10B and C demonstrated that the coordination number has a

better linear fitting outcome (R2 ¼ 0.92 for hypoxia and R2 ¼ 0.99 for normoxia) than the microisland number (R2 ¼ 0.87 for the hypoxia condition and R2 ¼ 0.92 for the normoxia condition). So, the coordination number is more proper as a parameter to quantify the extent of cellecell contact. The data with respect to coordination number 0 or microisland number 1 are not involved in linear fitting, due to absence of cellecell contact for single cells in such a case.

B. Cao et al. / Biomaterials 64 (2015) 21e32

29

Fig. 10. Finding of a linear relationship between chondrogenic extent and coordination number. A) Schematic illustration of the microdomains with corresponding coordination numbers and microisland numbers. B and C) Resulting collagen II intensity per cell as a function of B) coordination number and C) microisland number for microdomains. The linear fitting was performed for data points except that for single cells with coordination number 0 or microisland number 1. The squared correlation efficiencies are marked in the figures, and a better correlation was seen in B).

3.8. qRT-PCR measurements after blocking gap junctions Further experiments were carried out to preliminarily explore the mechanism of the cellecell contact effect on the chondrogenic induction of MSCs. Cells in contact might establish communications, and gap junction is one of the common communication ways. So, AGA, which can block the gap junction formed between cells, was added during the chondrogenic induction. The microdomains with coordination numbers 0 (C0) and 4 (C4) under the hypoxia and normoxia conditions were focused on here, and parallel groups without addition of AGA were taken as controls. On the basis of the qRT-PCR results showed in Fig. 11, for the group of C0 (single cells), the addition of AGA did not significantly affect the gene expression levels (expression of SOX9, collagen II, aggrecan, HIF-1a and collagen I) both in the hypoxia and normoxia conditions. In contrast, for C4 with cellecell contact, the expression of SOX9, collagen II, aggrecan and HIF-1a were down-regulated, while collagen I was up-regulated with the addition of AGA. Therefore, gap junction plays a role in the underlying cellecell contact effect. 4. Discussion 4.1. Cell-cell contacts enhance the chondrogenic differentiation of stem cells In recent decades, MSCs have attracted much attention as they are relatively easy to obtain, expand and induce into various cell types such as osteocytes, adipocytes, myocytes and chondrocytes [39e41]. MSCs are a very promising cell source for cartilage repair [42e45]. For MSC-based cartilage tissue engineering, how to promote the chondrogenic induction of MSCs is an important topic,

which affects the outcome of cartilage repair significantly. However, fully-repaired cartilage has not yet been achieved, and meanwhile, the mechanism in the natural process of cartilage formation has not been extensively understood. Hence, besides developing new techniques of cartilage repair, the fundamental research of chondrogenic induction of MSCs is highly required. It has been known that in vivo cell condensation represents an important step for the beginning of cartilage formation during limb skeletogenesis. In the process of cell condensation, the space among cells reduces, along with the increase of contact extent. Thus it is thought that cellecell contact plays a critical role in chondrogenesis [6,9,46,47]. Researchers in in vitro chondrogenic culture often adopt pellet culture, which enhances the cellecell contact extent to a large extent [48,49]. Nevertheless, it is not easy to directly study the contact effect on chondrogenic induction. Herein with the micropatterning technique that we developed previously, microdomains with different numbers (1, 2, 3, 6 and 15) of microislands were designed and fabricated, as shown in Fig. 3. On these microdomains, cells were cultured under the same microenvironment, and the main difference was the cellecell contact extent. MSCs exhibited good viability and did not spread out of the microdomains even after 9-day chondrogenic induction, as presented in Figs. 4 and 5. Cell numbers on each microdomain well matched the corresponding microisland number (Fig. 6). While the contact extent increased with the microisland number, the average spreading area per cell was kept the same, and thus our technique offers a proper way to study the effect of cellecell contact effect without the interference of other factors. Statistics of immunofluorescence staining (Fig. 7) indicated that the intensity of collagen II per cell increased with cellecell contact extent. The cellecell contact effect on chondrogenesis applied to both hypoxia and normoxia conditions (Fig. 8).

30

B. Cao et al. / Biomaterials 64 (2015) 21e32

Fig. 11. Examination of the role of gap junction in cellecell contact by addition of the inhibitor 18-a-glycyrrhetinic acid (AGA). Expressions of the indicated characteristic genes were quantified by qRT-PCR of cells on microdomains with coordination number 0 or 4 after MSCs underwent 9-day chondrogenic induction with or without the addition of AGA. All the data are normalized by those of the microdomains with coordination number 0 (single cells) under the normoxia condition. “þA” represents the presence of AGA. Significant differences are marked in the figure with “*” (p < 0.05), “**” (p < 0.01) and “***” (p < 0.001).

It is known that during the process of chondrogenesis, SOX9, collagen II, aggrecan and HIF-1a are up-regulated and collagen I is down-regulated. These changes were observed in the corresponding mRNA expressions on microdomains with higher contact extent (Fig. 9BeF). Combining collagen II staining and mRNA detection together makes it very conclusive that the increase of cellecell contact enhances the extent of chondrogenic differentiation. Furthermore, we found a quite good linear relationship between cellecell contact and chondrogenic induction according to the results presented in Fig. 10. Such a linear relation seems worthy as a highlight of the present study, considering that semiquantitative relations are rare in biological sciences. Our 2D micropattern guarantees the same microenvironment of MSCs except cellecell contact, which is the advantage of the unique material technique to reveal the basic sciences of cell-material and cellecell interactions. Our finding implies that when cells begin to contact with each other some connections might be established to promote differentiation. According to previous work [50e52], gap junction is very likely to play a role here. Therefore, AGA was added to block gap junction, which was found to down-regulate the expression of SOX9, collagen II, aggrecan and HIF-1a on microdomains with contacted cells (Fig. 11). While gap junction was confirmed to be a determination factor in the chondrogenic induction of MSCs, our work also implies that it is not the only factor underlying the cellecell contact, for even after the addition of AGA the normalized extent of chondrogenic induction was still higher for contacted cells than for single cells. Besides gap junction, some other factors such as N-cadherin [6,8,53,54] might join in the cellecell contact effect. We also think that beyond the classic biological cues, the physical contact might also afford an enhanced stiffness of a “matrix” around a cell, since its neighboring cells must be stiffer than the cell culture medium. The stiffness interpretation might also

partially account for that three dimensional (3D) culture of cells surrounded by a hydrogel is beneficial for chondrogenesis more than 2D culture [55]. 4.2. Hypoxia favors chondrogenesis Previous reports demonstrate that low oxygen tension promotes the chondrogenesis of MSCs [22,23,26e30,56e58]. In most of these studies, the proliferation rate of MSCs under different oxygen tension are much different, and thus besides oxygen, cellecell contact effect might play an important role. A question remains open, whether or not the enhanced chondrogenesis is raised by the higher proliferation rates (and thus higher cellecell contact extent) or directly by the oxygen tension (even under the same cellecell contact). In the present study, the cellecell contact could be controlled by our material micropatterning technique, and thus the oxygen tension effect and the cellecell contact effect on in vitro chondrogenic induction of MSCs could be decoupled with each other. The collagen II staining data from hypoxia and normoxia conditions were compared on a series of contact extents with single cells or with 2, 3, 6 and 15 cells (Fig. 9A). It is very evident that the chondrogenic extents were greater under the hypoxia condition than those under the normoxia condition. Furthermore, mRNA detection was performed on microdomains of D1 and D15 in both the hypoxia and normoxia conditions. For MSCs induced in the hypoxia condition, the expressions of SOX9, collagen II, aggrecan and HIF-1a were higher, and that of collagen I was lower compared to those in the normoxia condition (Fig. 9BeF). It is known that the chondrogenic induction promotes the up-regulation of SOX9, collagen II, aggrecan, but down-regulation of collagen I, and low oxygen tension promotes the up-regulation of HIF-1a during chondrogenesis. So, the present study is consistent with previous reports that hypoxia promotes the chondrogenic induction of MSCs

B. Cao et al. / Biomaterials 64 (2015) 21e32

and meanwhile sheds a new insight that the oxygen tension has a direct impact on chondrogenic induction. As the oxygen tension in cartilage is about 0.5%e5% [59], a hypoxia condition in vivo facilitates the chondrogenesis of MSCs. To sum up, both cellecell contact and low oxygen tension promote chondrogenic differentiation of MSCs. Combining these two factors together can achieve a better outcome than each single factor. To obtain fully repaired cartilage, more cytokines (transforming growth factor-b, bone morphogenetic protein etc.) and physical factors (compression, magnetic field etc.) could be combined to optimize the chondrogenesis. Some signal pathways (Rac1 [53], Wnt [60], RhoA/ROCK [61]) possibly influence chondrogenesis as well. The detailed mechanism of the impact of cellecell contact and oxygen tension on chondrogenic induction of MSCs needs to be investigated in the future. Another extension might be relevant 3D studies. Cell aggregation occurs during both in vivo and in vitro chondrogenesis as well as embryonic condensation in three dimensions. Some differences between 2D and 3D results have been indicated [55,62] and 3D environment is more beneficial for chondrogenesis than 2D environment. One of the reasons might come from more cellecell contacts in 3D. The cellecell interplay in a 3D cell aggregate might also be mediated by adhesion of two neighboring cells with their common extracellular matrix (ECM) [63]. It can be imaged that the cell-ECM contact might gradually dominate over the cellecell contact in the case of 3D condensation of a large amount of cells. The present studies of a small amount of cells on 2D patterned surfaces reflect, to a certain extent, the very early stage of chondrogenesis, and it is interesting that the direct cellecell contact enhanced the expression of collagen II, a key ECM protein for chondrogenesis and might accelerate the formation of the cell-ECM contact. From a translational perspective, both cellecell and cellmaterial or cell-ECM interactions should be taken into consideration in the process of chondrogenesis.

[3]

[4]

[5]

[6]

[7] [8]

[9]

[10]

[11] [12]

[13] [14] [15]

[16]

[17]

[18]

5. Conclusions Based on the design of appropriate micropatterns and in vitro 2D examination of chondrogenic induction, this study illustrated that cellecell contact significantly enhances the chondrogenic differentiation of MSCs, and found that the differentiation extent is linearly related to that of cellecell contacts, which could be semiquantified by corresponding coordination number. Gap junction was proved to be one of the deterministic factors underlying the cellecell contact effect, but not the only one. The regulation applies to both the hypoxia and normoxia conditions. Hypoxia directly promotes the induction process even under the same cellecell contact extent. More investigations are called for to explore the other hidden factors underlying the cellecell contact effect and the oxygen tension effect on lineage commitments of various stem cells.

[19] [20]

[21]

[22]

[23]

[24]

[25]

[26]

Acknowledgments The authors are grateful for the financial supports from NSF of China (grant No. 51273046), Chinese Ministry of Science and Technology (973 program No. 2011CB606203), and Science and Technology Developing Foundation of Shanghai (grant No. 14JC1490400). References [1] B.C. Heng, T. Cao, E.H. Lee, Directing stem cell differentiation into the chondrogenic lineage in vitro, Stem Cells 22 (2004) 1152e1167. [2] C.Y.C. Huang, K.L. Hagar, L.E. Frost, Y.B. Sun, H.S. Cheung, Effects of cyclic

[27]

[28]

[29]

[30]

31

compressive loading on chondrogenesis of rabbit bone-marrow derived mesenchymal stem cells, Stem Cells 22 (2004) 313e323. I.L. Kim, S. Khetan, B.M. Baker, C.S. Chen, J.A. Burdick, Fibrous hyaluronic acid hydrogels that direct MSC chondrogenesis through mechanical and adhesive cues, Biomaterials 34 (2013) 5571e5580. P.G. Duan, Z. Pan, L. Cao, Y. He, H.R. Wang, Z.H. Qu, et al., The effects of pore size in bilayered poly(lactide-co-glycolide) scaffolds on restoring osteochondral defects in rabbits, J. Biomed. Mater. Res. A 102 (2014) 180e192. J. Lam, S. Lu, E.J. Lee, J.E. Trachtenberg, V.V. Meretoja, R.L. Dahlin, et al., Osteochondral defect repair using bilayered hydrogels encapsulating both chondrogenically and osteogenically pre-differentiated mesenchymal stem cells in a rabbit model, Osteoarthr. Cartil. 22 (2014) 1291e1300. S. Tavella, P. Raffo, C. Tacchetti, R. Cancedda, P. Castagnola, N-CAM and Ncadherin expression during in vitro chondrogenesis, Exp. Cell. Res. 215 (1994) 354e362. B.K. Hall, T. Miyake, Divide, accumulate, differentiate: cell condensation in skeletal development revisited, Int. J. Dev. Biol. 39 (1995) 881e893. J. Chimal-Monroy, L.D. de Leon, Expression of N-cadherin, N-CAM, fibronectin and tenascin is stimulated by TGF-beta 1, beta 2, beta 3 and beta 5 during the formation of precartilage condensations, Int. J. Dev. Biol. 43 (1999) 59e67. W.S. Toh, Z. Yang, H. Liu, B.C. Heng, E.H. Lee, T. Cao, Effects of culture conditions and bone morphogenetic protein 2 on extent of chondrogenesis from human embryonic stem cells, Stem Cells 25 (2007) 950e960. H. Takahashi, K. Emoto, M. Dubey, D.G. Castner, D.W. Grainger, Imaging surface immobilization chemistry: correlation with cell patterning on nonadhesive hydrogel thin films, Adv. Funct. Mater. 18 (2008) 2079e2088. D.F. Williams, On the nature of biomaterials, Biomaterials 30 (2009) 5897e5909. E.K. Yim, E.M. Darling, K. Kulangara, F. Guilak, K.W. Leong, Nanotopographyinduced changes in focal adhesions, cytoskeletal organization, and mechanical properties of human mesenchymal stem cells, Biomaterials 31 (2010) 1299e1306. Z. Pan, C. Yan, R. Peng, Y.C. Zhao, Y. He, J.D. Ding, Control of cell nucleus shapes via micropillar patterns, Biomaterials 33 (2012) 1730e1735. X. Wang, C. Yan, K. Ye, Y. He, Z.H. Li, J.D. Ding, Effect of RGD nanospacing on differentiation of stem cells, Biomaterials 34 (2013) 2865e2874. B. Cao, R. Peng, Z.H. Li, J.D. Ding, Effects of spreading areas and aspect ratios of single cells on dedifferentiation of chondrocytes, Biomaterials 35 (2014) 6871e6881. J.G. Sun, S.V. Graeter, L. Yu, S.F. Duan, J.P. Spatz, J.D. Ding, Technique of surface modification of a cell-adhesion-resistant hydrogel by a cell-adhesionavailable inorganic microarray, Biomacromolecules 9 (2008) 2569e2572. R. Peng, X. Yao, J.D. Ding, Effect of cell anisotropy on differentiation of stem cells on micropatterned surfaces through the controlled single cell adhesion, Biomaterials 32 (2011) 8048e8057. C. Yan, J.G. Sun, J.D. Ding, Critical areas of cell adhesion on micropatterned surfaces, Biomaterials 32 (2011) 3931e3938. X. Yao, R. Peng, J.D. Ding, Cell-material interactions revealed via material techniques of surface patterning, Adv. Mater. 25 (2013) 5257e5286. R. Chen, S.J. Curran, J.M. Curran, J.A. Hunt, The use of poly(L-lactide) and RGD modified microspheres as cell carriers in a flow intermittency bioreactor for tissue engineering cartilage, Biomaterials 27 (2006) 4453e4460. Y.X. Lai, C. Xie, Z. Zhang, W.Y. Lu, J.D. Ding, Design and synthesis of a potent peptide containing both specific and non-specific cell-adhesion motifs, Biomaterials 31 (2010) 4809e4817. H. Betre, S.R. Ong, F. Guilak, A. Chilkoti, B. Fermor, L.A. Setton, Chondrocytic differentiation of human adipose-derived adult stem cells in elastin-like polypeptide, Biomaterials 27 (2006) 91e99. M. Zscharnack, C. Poesel, J. Galle, A. Bader, Low oxygen expansion improves subsequent chondrogenesis of ovine bone-marrow-derived mesenchymal stem cells in collagen type i hydrogel, Cells Tissues Organs 190 (2009) 81e93. B.D. Markway, G.K. Tan, G. Brooke, J.E. Hudson, J.J. Cooper-White, M.R. Doran, Enhanced chondrogenic differentiation of human bone marrow-derived mesenchymal stem cells in low oxygen environment micropellet cultures, Cell. Transpl. 19 (2010) 29e42. E.G. Meyer, C.T. Buckley, S.D. Thorpe, D.J. Kelly, Low oxygen tension is a more potent promoter of chondrogenic differentiation than dynamic compression, J. Biomech. 43 (2010) 2516e2523. E. Duval, C. Bauge, R. Andriamanalijaona, H. Benateau, S. Leclercq, S. Dutoit, et al., Molecular mechanism of hypoxia-induced chondrogenesis and its application in in vivo cartilage tissue engineering, Biomaterials 33 (2012) 6042e6051. E.J. Sheehy, C.T. Buckley, D.J. Kelly, Oxygen tension regulates the osteogenic, chondrogenic and endochondral phenotype of bone marrow derived mesenchymal stem cells, Biochem. Biophys. Res. Co. 417 (2012) 305e310. V.V. Meretoja, R.L. Dahlin, S. Wright, F.K. Kasper, A.G. Mikos, The effect of hypoxia on the chondrogenic differentiation of co-cultured articular chondrocytes and mesenchymal stem cells in scaffolds, Biomaterials 34 (2013) 4266e4273.  pez-Holgado, F.M. Sa nchez-Guijo, E. Villaro n, V. Barbado, S. Carrancio, N. Lo S. Tabera, et al., Optimization of mesenchymal stem cell expansion procedures by cell separation and culture conditions modification, Exp. Hematol. 36 (2008) 1014e1021. A. Krinner, M. Zscharnack, A. Bader, D. Drasdo, J. Galle, Impact of oxygen environment on mesenchymal stem cell expansion and chondrogenic

32

B. Cao et al. / Biomaterials 64 (2015) 21e32

differentiation, Cell. Prolif 42 (2009) 471e484. [31] R. Peng, X. Yao, B. Cao, J. Tang, J.D. Ding, The effect of culture conditions on the adipogenic and osteogenic inductions of mesenchymal stem cells on micropatterned surfaces, Biomaterials 33 (2012) 6008e6019. [32] X. Yao, Y.W. Hu, B. Cao, R. Peng, J.D. Ding, Effects of surface molecular chirality on adhesion and differentiation of stem cells, Biomaterials 34 (2013) 9001e9009. [33] L. Chen, T.Y. Ci, T. Li, L. Yu, J.D. Ding, Effects of molecular weight distribution of amphiphilic block copolymers on their solubility, micellization, and temperature-induced sol gel transition in water, Macromolecules 47 (2014) 5895e5903. [34] Z. Zhang, J. Ni, L. Chen, L. Yu, J.W. Xu, J.D. Ding, Biodegradable and thermoreversible PCLA-PEG-PCLA hydrogel as a barrier for prevention of postoperative adhesion, Biomaterials 32 (2011) 4725e4736. [35] J.A. Burdick, K.S. Anseth, Photoencapsulation of osteoblasts in injectable RGDmodified PEG hydrogels for bone tissue engineering, Biomaterials 23 (2002) 4315e4323. [36] S.J. Lee, Y. Bae, K. Kataoka, D. Kim, D.S. Lee, S.C. Kim, In vitro release and in vivo anti-tumor efficacy of doxorubicin from biodegradable temperature-sensitive star-shaped plga-peg block copolymer hydrogel, Polym. J. 40 (2007) 171e176. [37] T.Y. Ci, T. Li, G.T. Chang, L. Yu, J.D. Ding, Simply mixing with poly(ethylene glycol) enhances the fraction of the active chemical form of antitumor drugs of camptothecin family, J. Control Release 169 (2013) 329e335. [38] B. Cao, Z.H. Li, R. Peng, J.D. Ding, Data in support of effects of cell-cell contact and oxygen tension on chondrogenic differentiation of stem cells, Data Brief (2015) (submitted for publication). [39] M.F. Pittenger, A.M. Mackay, S.C. Beck, R.K. Jaiswal, R. Douglas, J.D. Mosca, et al., Multilineage potential of adult human mesenchymal stem cells, Science 284 (1999) 143e147. [40] A. Uccelli, L. Moretta, V. Pistoia, Mesenchymal stem cells in health and disease, Nat. Rev. Immunol. 8 (2008) 726e736. [41] X. Wang, S. Li, C. Yan, P. Liu, J.D. Ding, Fabrication of RGD micro/nanopattern and corresponding study of stem cell differentiation, Nano Lett. 15 (2015) 1457e1467. [42] S. Wakitani, T. Goto, S.J. Pineda, R.G. Young, J.M. Mansour, A.I. Caplan, et al., Mesenchymal cell-based repair of large, full-thickness defects of articularcartilage, J. Bone Jt. Surg. Am. 76A (1994) 579e592. [43] R.R. Varshney, R.J. Zhou, J.H. Hao, S.S. Yeo, W.H. Chooi, J.B. Fan, et al., Chondrogenesis of synovium-derived mesenchymal stem cells in gene-transferred co-culture system, Biomaterials 31 (2010) 6876e6891. [44] Q.Q. Han, L.T. Fan, B.C. Heng, Z.G. Ge, Apoptosis and metabolism of mesenchymal stem cells during chondrogenic differentiation in vitro, Int. J. Tissue Reg. 4 (2013) 61e64. [45] C.R. Rowland, D.P. Lennon, A.I. Caplan, F. Guilak, The effects of crosslinking of scaffolds engineered from cartilage ECM on the chondrogenic differentiation of MSCs, Biomaterials 34 (2013) 5802e5812. [46] K. Kavalkovich, R. Boynton, J. Mary Murphy, F. Barry, Chondrogenic differentiation of human mesenchymal stem cells within an alginate layer culture system, In Vitro Cell Dev Biol Animal 38 (2002) 457e466. [47] K. Zhang, L.L. Wang, Q.Q. Han, B.C. Heng, Z. Yang, Z.G. Ge, Relationship between cell function and initial cell seeding density of primary porcine

chondrocytes in vitro, Biomed. Eng App Bas C 25 (2013) 1340001. [48] B. Johnstone, T.M. Hering, A.I. Caplan, V.M. Goldberg, J.U. Yoo, In vitro chondrogenesis of bone marrow-derived mesenchymal progenitor cells, Exp. Cell. Res. 238 (1998) 265e272. [49] S. Shirasawa, I. Sekiya, Y. Sakaguchi, K. Yagishita, S. Ichinose, T. Muneta, In vitro chondrogenesis of human synovium-derived mesenchymal stem cells: optimal condition and comparison with bone marrow-derived cells, J. Cell. Biochem. 97 (2006) 84e97. [50] N.M. Kumar, N.B. Gilula, The gap junction communication channel, Cell 84 (1996) 381e388. [51] J. Tang, R. Peng, J.D. Ding, The regulation of stem cell differentiation by cellcell contact on micropatterned material surfaces, Biomaterials 31 (2010) 2470e2476. [52] N. Faucheux, J.M. Zahm, N. Bonnet, G. Legeay, M.D. Nagel, Gap junction communication between cells aggregated on a cellulosecoated polystyrene: influence of connexin 43 phosphorylation, Biomaterials 25 (2004) 2501e2506. [53] A. Woods, G.Y. Wang, H. Dupuis, Z.H. Shao, F. Beier, Rac1 signaling stimulates N-cadherin expression, mesenchymal condensation, and chondrogenesis, J. Biol. Chem. 282 (2007) 23500e23508. [54] L.M. Bian, M. Guvendiren, R.L. Mauck, J.A. Burdick, Hydrogels that mimic developmentally relevant matrix and N-cadherin interactions enhance MSC chondrogenesis, Proc. Natl. Acad. Sci. U. S. A. 110 (2013) 10117e10122. [55] L.P. Cao, B. Cao, C. Lu, G.W. Wang, L. Yu, J.D. Ding, An injectable hydrogel formed by in situ cross-linking of glycol chitosan and multi-benzaldehyde functionalized PEG analogues for cartilage tissue engineering, J. Mater. Chem. B 3 (2015) 1268e1280. [56] A.B. Adesida, A. Mulet-Sierra, N.M. Jomha, Hypoxia mediated isolation and expansion enhances the chondrogenic capacity of bone marrow mesenchymal stromal cells, Stem Cell Res. Ther. 3 (2012) 9. [57] J.R. Choi, B. Pingguan-Murphy, W.A.B. Wan Abas, M.A. Noor Azmi, S.Z. Omar, K.H. Chua, et al., Impact of low oxygen tension on stemness, proliferation and differentiation potential of human adipose-derived stem cells, Biochem. Biophys. Res. Co. 448 (2014) 218e224. [58] J. Shang, H. Liu, J. Li, Y. Zhou, Roles of hypoxia during the chondrogenic differentiation of mesenchymal stem cells, Curr. Stem Cell Res. Ther. 9 (2014) 141e147. [59] K. Lund-Olesen, Oxygen tension in synovial fluids, Arthritis Rheumatism 13 (1970) 769e776. [60] S.H. Hsu, G.S. Huang, Substrate-dependent Wnt signaling in MSC differentiation within biomaterial-derived 3D spheroids, Biomaterials 34 (2013) 4725e4738. [61] A. Woods, G.Y. Wang, F. Beier, RhoA/ROCK signaling regulates Sox9 expression and actin organization during chondrogenesis, J. Biol. Chem. 280 (2005) 11626e11634. [62] Y. Dou, N. Li, Y. Zheng, Z.G. Ge, Effects of fluctuant magnesium concentration on phenotype of the primary chondrocytes, J. Biomed. Mater. Res. A 102 (2014) 4455e4463. [63] A. Gigout, M. Jolicoeur, M. Nelea, N. Raynal, R. Farndale, M.D. Buschmann, Chondrocyte aggregation in suspension culture is GFOGER-GPP- and beta1 integrin-dependent, J. Biol. Chem. 283 (2008) 31522e31530.