Scaffold composition affects cytoskeleton organization, cell–matrix interaction and the cellular fate of human mesenchymal stem cells upon chondrogenic differentiation

Biomaterials 52 (2015) 208e220

Contents lists available at ScienceDirect

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

Scaffold composition affects cytoskeleton organization, cellematrix interaction and the cellular fate of human mesenchymal stem cells upon chondrogenic differentiation Yuk Yin Li, Tze Hang Choy, Fu Chak Ho, Pui Barbara Chan* Tissue Engineering Laboratory, Department of Mechanical Engineering, The University of Hong Kong, Pokfulam Road, Hong Kong Special Administrative Region, China

a r t i c l e i n f o

a b s t r a c t

Article history: Received 11 September 2014 Received in revised form 30 January 2015 Accepted 6 February 2015 Available online 26 February 2015

The stem cell niche, or microenvironment, consists of soluble, matrix, cell and mechanical factors that together determine the cellular fates and/or differentiation patterns of stem cells. Collagen and glycosaminoglycans (GAGs) are important scaffolding materials that can mimic the natural matrix niche. Here, we hypothesize that imposing changes in the scaffold composition or, more specifically, incorporating GAGs into the collagen meshwork, will affect the morphology, cytoskeletal organization and integrin expression profiles, and hence the fate of human mesenchymal stem cells (MSCs) upon the induction of differentiation. Using chondrogenesis as an example, we microencapsulated MSCs in three scaffold systems that had varying matrix compositions: collagen alone (C), aminated collagen (AC) and aminated collagen with GAGs (ACG). We then induced the MSCs to differentiate toward a chondrogenic lineage, after which, we characterized the cell viability and morphology, as well as the level of cytoskeletal organization and the integrin expression profile. We also studied the fate of the MSCs by evaluating the major chondrogenic markers at both the gene and protein level. In C, MSC chondrogenesis was successfully induced and MSCs that spread in the scaffolds had a clear actin cytoskeleton; they expressed integrin a2b1, a5 and av; promoted sox9 nuclear localization transcription activation; and upregulated the expression of chondrogenic matrix markers. In AC, MSC chondrogenesis was completely inhibited but the scaffold still supported cell survival. The MSCs did not spread and they had no actin cytoskeleton; did not express integrin a2 or av; they failed to differentiate into chondrogenic lineage cells even on chemical induction; and there was little colocalization or functional interaction between integrin a5 and fibronectin. In ACG, although the MSCs did not express integrin a2, they did express integrin av and there was strong co-localization and hence functional binding between av and fibronectin. In addition, vimentin was the dominant cytoskeletal protein in these cells, and the chondrogenic marker genes were expressed but at a much lower level than in the MSCs encapsulated in C alone. This work suggests the importance of controlling the matrix composition as a strategy to manipulate cellematrix interactions (through changes in the integrin expression profile and cytoskeleton organization), and hence stem cell fates. © 2015 Elsevier Ltd. All rights reserved.

Keywords: Cellematrix interaction Cellular fates Mesenchymal stem cell Collagen Glycosaminoglycans Chondrogenesis

1. Introduction The stem cell niche refers to the microenvironment that determines the cellular fates or differentiation patterns of stem cells [1,2]. It is comprised of a biological component including

* Corresponding author. Tissue Engineering Laboratory, Medical Engineering Programme, Department of Mechanical Engineering, The University of Hong Kong, Pokfulam Road, Hong Kong Special Administrative Region, China. Tel.: þ852 28592632; fax: þ852 2885415. E-mail address: [email protected] (P.B. Chan). http://dx.doi.org/10.1016/j.biomaterials.2015.02.037 0142-9612/© 2015 Elsevier Ltd. All rights reserved.

soluble growth factors; a cellular component such as the neighboring cells and their secreted factors; a matrix component such as the extracellular matrix (ECM) and its organization; and a mechanical component such as substrate compliance and mechanical loading. Providing a combination of topographical, mechanical and biochemical input factors, the matrix component of the niche regulates cellular-fate processes including cell proliferation, differentiation, shape and migration [3e6]. Collagen and glycosaminoglycans (GAGs), are two major components of the ECM in many tissues, and are important scaffold candidates that mimic the

Y.Y. Li et al. / Biomaterials 52 (2015) 208e220

209

natural matrix microenvironment [7]. Nevertheless, GAGs are highly hydrophilic and are rapidly soluble in water, making them difficult to retain in a solid matrix for biomedical applications. However, by chemically modifying the net charge of collagen using amination, we were able to incorporate and retain GAGs in a solid meshwork of collagen via ionic interactions [8]. Such a collagenGAG matrix might therefore play an important role as a biomimetic scaffold for GAG-rich tissues such as the intervertebral disc and cartilage. However, whether a variation in matrix composition also affects stem cell fate, via changes in cellematrix interactions such as the integrin expression profile and binding, is not yet known. Two important mediators of cellematrix interactions, which are induced by variations in the matrix composition, are integrin expression and cytoskeleton organization. Firstly, cellematrix interactions mediated by integrins have been reported to influence stem cell fate [9]. The importance of integrins in chondrogenesis has been suggested by the detection of a1, a3, a5, aV, b1, b3, and b5 integrin subunits in adult articular chondrocytes [10e12]. In addition, blocking integrin-mediated ECM interactions has been shown to inhibit chondrocyte differentiation [13]. Moreover, collagen remodeling was found to be dependent on the a2b1 integrin during chondrogenesis of MSCs [14]. Secondly, cytoskeletal organization also plays an important role in regulating stem cell fate [15]. The cytoskeleton consists of microfilaments (which are comprised of actin), microtubules (comprised of tubulin) and intermediate filaments (comprised of a family of related proteins such as keratin, lamin and vimentin). It has been shown that actin transduces the chondro-inhibitory effects of integrin-mediated adhesion in MSCs seeded into an agarose gel [16]. Whereas an inhibition of actin polymerization was shown to induce a chondrogenic phenotype by enhancing the expression of the chondrogenic transcription factor Sox9, the inhibition of tubulin polymerization completely blocked Sox9 expression and resulted in an accumulation of GAGs in highdensity cultures of mouse embryo limb bud mesenchymal cells [17]. Nevertheless, the effect of the scaffold matrix composition on integrin expression, cytoskeleton organization, and hence the fate of MSCs upon chondrogenic differentiation, is still not known. In the current study, we hypothesize that imposing changes in the composition of the scaffold, by, incorporating GAGs into a collagen meshwork, will affect the morphology, cytoskeletal organization and cellematrix interactions of MSCs, and this in turn will alter their cell fate upon the induction of differentiation. We microencapsulated human MSCs (hMSCs) in scaffold systems of varying matrix compositions, namely collagen (C), aminated collagen (AC) and aminated collagen-GAG (ACG), and induced the cells to differentiate towards a chondrogenic lineage. We then characterized matrix compliance, cell viability and morphology, the organization of the cytoskeletal and integrin expression profile, as well as studying the major chondrogenic markers at the level of both the gene and protein.

The hMSCs encapsulated in each of the three types of scaffold were induced to undergo chondrogenic differentiation, as previously reported [22,23]. Specifically, the constructs were cultured in chondrogenic differentiation medium consisting of DMEM-high glucose supplemented with 10 ng/ml recombinant human transforming growth factor-b3 (TGF-b3; Merck, Darmstadt, Germany), 100 nM dexamethasone (SigmaeAldrich Co. LLC), 6 mg/ml insulin (Merck & Co. Inc., Kenilworth, NJ, USA), 100 mM 2-phospho-L-ascorbate (Fluka, St. Louis, MO, USA), 1 mM sodium pyruvate (Gibco Invitrogen Corp., Carlsbad, CA, USA), 6 mg/ml transferrin (SigmaeAldrich Co. LLC), 0.35 mM L-proline (Merck & Co. Inc.), and 1.25 mg/ml BSA (SigmaeAldrich Co. LLC). At day 7, 14, and 21, the samples were harvested for subsequent characterization.

2. Methodology

2.7. Scanning electron microscopy (SEM)

2.1. Overall research design

To examine the microstructure of the scaffolds, 1-h post-encapsulated samples were rinsed with PBS and fixed with 0.05% glutaraldehyde for 1 h. After dehydration in a graded series of ethanol, samples were critical point dried and fractured to reveal their internal structure. The samples were then gold sputter coated before being examined with a scanning electron microscope (S-4800, Hitachi, Tokyo, Japan).

Fig. 1 shows the overall design of our research plan. hMSCs were microencapsulated in scaffolds with incremental changes in the composition of the matrix as follows: unmodified collagen (C), aminated collagen (AC), and aminated collagen plus GAGs (ACG). MSCs in these scaffolds were induced to differentiate toward a chondrogenic lineage for 7, 14, or 21 days. At the desired time point, samples from each group were processed, the proportion of live and dead cells was determined, and various mechanical characterization methods were used. In addition, the ultrastructural properties of the cells were revealed and cells were subjected to various histological; immunohistochemical and immunofluorescence methods to characterize the level and localization of different proteins, and to real time RT-PCR to determine the expression of genes, using markers related to the organization of the cytoskeleton, integrin-based cell matrix interactions, extracellular matrix synthesis and readiness to differentiate into chondrocytes.

2.2. Culture of human MSCs Human MSCs were obtained from the Texas A&M Health Science Center College of Medicine Institute for Regenerative Medicine (Scott & White Hospital, Bryan, TX, USA). hMSCs were cultured in Dulbecco's modified Eagle's medium (DMEM) containing low glucose and supplemented with 10% fetal bovine serum, 100 U/ml penicillin, 100 mg/ml streptomycin and 2 mM L-glutamine. The culture medium was replaced every 3e4 days. Cells at passage 6 were used for subsequent microencapsulation. 2.3. Microencapsulation of hMSCs in unmodified collagen (C) Human MSC-collagen microspheres were prepared as described previously [18]. Briefly, hMSCs in suspension, at a final cell density of 2  106 cells/ml, were mixed with neutralized rat tail type I collagen solution (BD Biosciences, Bedford, MA, USA) at a final concentration of 2 mg/ml in an ice-bath. Droplets of 50 ml were pipetted into a 100 mm-diameter Petri dish (Sterilin Ltd., Newport, UK) covered with UVirradiated parafilm. After a 1-h incubation at 37  C, the gelated MSC-collagen (C) microspheres were ready for subsequent chondrogenic differentiation. 2.4. Microencapsulation of hMSCs in aminated collagen (AC) In order to incorporate more negatively-charged GAGs into a collagen matrix, chemical modification amination was used to conjugate carboxyl groups (such as glutamic acid and aspartic acid) with ethylenediamine (EDA), a molecule that has positive amino groups (i.e., COOeNHCH2CH2NH3þ), thus making the collagen molecule more positive [19]. In order to study whether aminated collagen might affect the differentiation potential of MSCs, aminated collagen was used as a control group and prepared as described previously [8]. Briefly, rat tail type I collagen solution (BD Biosciences) was aminated with a solution of EDA (SigmaeAldrich Co. LLC, St. Louis, MO, USA) and 1-ethyl-3(3-dimethylaminopropryl) carbodiimide (EDC) (SigmaeAldrich Co. LLC). Assuming that the molar amount of carboxylic acid (COOH) was 1.2 mmol/g in collagen [20,21], the molar amounts of EDA and ECC used, were 5000 and 300 multiples of COOH, respectively. To microencapsulate hMSCs in aminated collagen, a suspension of hMSCs was mixed with neutralized aminated collagen to a final amount of 2  106 cells per 2 mg aminated collagen in an ice-bath. Aliquots were dispensed into microtubes and gently vortexed for 10 s. After a 1-h incubation at 37  C, the reaction mixture was centrifuged at 500 g for 8 min. The supernatant was discarded, leaving behind a co-precipitate consisting of hMSCs encapsulated in aminated collagen (AC), which were ready for subsequent chondrogenic differentiation. 2.5. Microencapsulation of hMSCs in aminated collagen-GAG (ACG) The technique used to fabricate an aminated collagen-GAG scaffold was modified from a previous report [8]. In brief, aminated collagen was mixed with excess GAGs (i.e., chondroitin-6-sulfate; SigmaeAldrich Co. LLC) at 4 mg/ml in PBS by gentle vortexing. After a 1-h incubation at 37  C, the reaction mixture was centrifuged at 500 g for 8 min. The supernatant was discarded and the resulting coprecipitate consisted of hMSCs encapsulated in an aminated collagen-GAG (ACG) scaffold, which were ready for subsequent chondrogenic differentiation. 2.6. Induction of chondrogenic differentiation

2.8. Quantification of sulphated GAGs and hydroxyproline (HYP) content All co-precipitates were solubilized with 0.6 U papain (P4762, SigmaeAldrich Co. LLC) in 0.2 ml buffer solution (50 mM phosphate buffer, 5 mM EDTA, and 5 mM Lcysteine, pH 6.5). Digestion was carried out at 60  C overnight. The amount of sulphated GAGs in the digested samples was diluted and assayed using the 1,9dimethylmethylene blue-GAG binding method [24]. Absorbance at 656 nm was measured using a microplate reader (Asys UVM340, Biochrom Ltd., Cambridge, UK).

210

Y.Y. Li et al. / Biomaterials 52 (2015) 208e220

Fig. 1. Schematic diagram showing the overall research design.

The amount of GAGs in the samples was determined using a calibration curve prepared using a chondroitin sulphate standard, with a linear region between 1.25 mg/ml and 20 mg/ml. A portion of the papain-digested samples was acidified to 3 M with hydrochloric acid and was hydrolyzed by reflux at 110  C overnight. The hydrolyzed samples were then adjusted to pH 6 to 7 with sodium hydroxide and the amount of hydroxyproline (HYP) in the sample was assayed using the chloramine Tdimethylaminobenzaldehyde method [25]. The absorbance at 557 nm was measured and the amount of hydroxyproline was determined using a calibration curve with a linear region between 0.625 mg/ml and 10 mg/ml. The amount of HYP was calculated to be 13% of collagen (w/w). 2.9. Measurement of reduced elastic moduli of the constructs The reduced elastic modulus, which represents the resilience of a construct, was measured using the microplate compression method previously developed [22]. The compression system setup is illustrated in Supplementary Information 1. In brief, the system consisted of a flexible and a rigid glass plate, prepared from borosilicate glass (Sutter Instrument Co., Novato, CA, USA) and coverslip (Paul €nigshofen, Germany), respectively. The plates Marienfeld GmbH & Co. KG, Lauda-Ko were mounted onto a Piezoelectric translator, which was controlled in the X, Y and Z planes both manually and automatically, the latter via an MP285 micromanipulator controller (Sutter Instrument Co). The whole system was connected to a DMIRB inverted microscope (Leica Microsystems, Wetzlar, Germany) onto which was mounted a CoolSNAP ES digital camera (Photometrics, Tucson, AZ, USA) for video recording. The stiffness of the flexible plate was pre-calibrated before mounting. During an experiment, the construct was attached to the rigid plate first. The flexible plate was then slowly moved until the edge just touched the construct, but without deforming it. The thickness of the construct was measured with MetaMorph (Molecular Devices, LLC, Sunnyvale, CA, USA). The compression was programmed as a ‘Step Change’ experiment such that the flexible plate was

programmed to compress the construct from between 0% and 30% of the construct thickness at a speed of 1 mm/s, followed by compression by another 30% of the construct thickness at a speed of 2 mm/s. This compression process was recorded via time-lapse videomicroscopy using a speed frame rate of 1 frame/s. From the time-lapse recording, the displacement of the flexible plate and time during compression (relative to the starting position of 0 mm and 0 s) was obtained by using the “Track Objects” function in MetaMorph. The data obtained were output to Excel for further processing. The slope (displacement change against time) before and after the speed change were calculated. Based on the two slopes obtained, the reduced elastic modulus was calculated according to the equation described in Li et al [22]. _ d_ 0 Þ  1ÞÞ was used. For collagen microspheres, Er ¼ ðE=ð1  n2 ÞÞ ¼ ððk=aÞððDd=D _ d_ 0 Þ  1Þ For aminated collagen and aminated collagen-GAGs, E ¼ ðkc=pa2 ÞððDd=D was used, where “a” and “c” are the radius and thickness of the constructs, 0 respectively; Dd_ and Dd_ are the slope before and after the speed change, respectively and “k” is the stiffness of the flexible plate. 2.10. Cell viability of hMSCs encapsulated in different scaffolds At 7, 14, and 21 days after the induction of differentiation, hMSCs encapsulated in C, AC or ACG scaffolds were incubated with 2 mM calcein AM and 4 mM ethidium homodimer-1 from the LIVE/DEAD Viability/Cytotoxicity kit (Molecular Probes®, Life Technologies, Grand Island, NY, USA) for 45 min in the dark, after which they were examined using a laser confocal scanning microscope (LSM710, Carl Zeiss, Germany) and stacks of optical sections were collected. 2.11. Histological and immunofluorescence staining On days 7, 14, and 21, hMSCs encapsulated in C, AC and ACG scaffolds were fixed with 4% PBS-buffered paraformaldehyde for 20 min and cut into 18 mm

Y.Y. Li et al. / Biomaterials 52 (2015) 208e220 frozen sections. Routine hematoxylin and eosin staining was then conducted to reveal the cell morphology, whereas Alcian blue staining was used to reveal the presence of GAGs. In addition, some sections were processed for immunofluorescence; unless otherwise specified, all the primary antibodies used were purchased from Abcam (Cambridge, UK). To examine cellematrix interactions, antibodies against integrins av, a5 and a2b1 were used. After permeabilization with 0.5% Tween 20 for 10 min, sections were incubated with the appropriate primary antibody at 4  C overnight. Alexa Fluor 546 goat anti-mouse IgG (Molecular Probes) was then used as secondary antibody for integrins av and a5, whereas Alexa Fluor 488 goat anti-mouse IgG was used to identify integrin a2b1. To examine the cytoskeleton, rhodamine phalloidin (Molecular Probes) was used to label F-actin whereas antibodies against a-tubulin and vimentin were used to label the microtubules and intermediate filaments, respectively. The labeling procedure used for a-tubulin was the same as that for the integrins and Alexa Fluor 488 goat anti-rabbit IgG (Molecular Probes) was used as the secondary antibody. On the other hand, the sections used for the vimentin labeling were permeabilized and then incubated with sodium citrate buffer (10 mM sodium citrate, 0.05% Tween 20, pH 6.0) at 95  C for 20 min. After an overnight incubation with the vimentin primary antibody, the sections were incubated with Alexa Fluor 488 goat anti-mouse IgG for 1 h. To evaluate the presence of cellular markers specific to the chondrogenic lineage and cellular markers of proliferative cells, antibodies against Sox9 and Ki-67 were used, respectively. After permeabilization, antigen retrieval with sodium citrate buffer and overnight incubation with primary antibodies at 4  C overnight, sections were incubated with Alexa Fluor 488 goat anti-mouse IgG (for Sox9) or Alexa Fluor 488 goat antirabbit IgG (for Ki-67), for 1 h. Finally, the sections were mounted under Fluoro-gel II mounting medium containing DAPI (Electron Microscopy Sciences, Hatfield, PA, USA). Images of the labeled sections were all acquired using multiphoton laser confocal scanning microscopy (LSM710, Carl Zeiss) and acquired using the same detector and scanning setup with 40 or 63 objective lenses. The Pearson's correlation coefficient was used to find the extent of colocalization between integrin a5 and fibronectin, by using the Colocalization module of the Imaris software (Bitplane AG, Zürich, Switzerland). For the colocalization analysis four sections were chosen, one from the upper, one from the lower, and two from the middle regions of the image stacks of the entire sample, after performing an automatic thresholding, which removes the any background noise. Each section contained 8e25 cells in the AC and ACG groups and 3e36 cells in the C group. 2.12. Gene expression of chondrogenic markers by real-time RT-PCR The expression levels of major chondrogenic markers including SOX9, aggrecan (ACAN), collagen type II (COL2A1) and collagen type I (COL1A2) were used to evaluate whether the encapsulated hMSCs differentiated into cells of the chondrogenic lineage (Table 1). At 7, 14, and 21 days, around 30e40 constructs from the AC and ACG groups were digested with 2 mg/ml pronase (SigmaeAldrich Co. LLC), at 37  C for 30e45 min and then digested with 3 mg/ml collagenase and 1 mg/ml hyaluronidase (both from SigmaeAldrich Co. LLC), at 37  C for 45e60 min. Around 10 collagen constructs were digested with collagenase and hyaluronidase only. The digested constructs were then homogenized with 1 ml TRI reagent (Molecular Research Center, Inc., Cincinnati, OH, USA) for subsequent extraction of total mRNA, according to the manufacturer's instructions. Total RNA was quantified using a NanoDrop™ 2000 spectrophotometer (NanoDrop Products, Wilmington, DE, USA) and was transcribed into cDNA using a TaqMan Reverse Transcription kit (Applied Biosystems, Foster City, CA). For SOX9, ACAN, and COL1A2, real-time PCR was performed using the TaqMan® Gene Expression Master Mix (Applied Biosystems) using standard thermal conditions, and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was used as the endogenous control. For COL2A1, real-time PCR was performed using the Power SYBR® Green PCR Master Mix (Applied Biosystems) and GADPH was used as the endogenous control. The data were analyzed using the relative quantification method using Ct values obtained from collagen microspheres cultured in hMSC normal growth medium for 7 days as the experimental calibrator. Table 1 Primers used for RT-PCR. Reagent: Power SYBR® Green PCR Master mix Gene COL2A1 GAPDH

Forward primer 50 -TCACGTACACTGCCCTGAAG-30 50 -GAGTCAACGGATTTGGTCGT-30

Reverse primer 50 -TGCAACGGATTGTGTTGTT-30 50 -TTGATTTTGGAGGGATCTCG-30

Reagent: TaqMan® Gene Expression Master Mix Gene SOX9 ACAN COL1A2 GAPDH

Assay ID according to Applied Biosystems Hs00165814_m1 Hs00202971_m1 Hs00164099_m1 NM_002046.3 (Accession Number)

211

2.13. Data analysis and statistics All quantitative data were expressed as the mean ± standard error unless otherwise specified. Two-way analysis of variance (ANOVA) with appropriate posthoc tests were used to reveal differences in expression levels of genes among the different groups. A significance level was set at 0.05, and SPSS 19.0 (SPSS Inc., Chicago, IL) was used for the analysis.

3. Results 3.1. Ultrastructural, physical and biochemical properties of hMSCs in the various scaffolds Fig. 2AeC are SEM images of three scaffolds to show the different ultrastructural characteristics of each, immediately after fabrication. In the C scaffold, random meshworks of nano-sized collagen fibers were found (Fig. 2A), while in the AC scaffold there was a similar fibrous meshwork with some granule-like structures (Fig. 2B). In the ACG scaffold, a lot of granule-like ground substances, which correspond to GAGs, were found intercalating the fibrous meshwork (Fig. 2C). Fig. 2D shows the GAG to HYP ratio of the different scaffolds immediately after fabrication; whereas the ACG scaffold showed a GAG:HYP ratio of almost 5, the ratio of the other two groups was well below unity, suggesting that ACG might have incorporated a lot of GAGs into the solid meshwork during fabrication. One-way ANOVA showed significant differences among the different groups (at p < 0.001) while Bonferroni's posthoc tests showed that the difference was between ACG and the other groups (at p < 0.001). As chondrogenesis continues for 7 days, cells were significantly contracted in C but not in AC and ACG. This resulted in microspheres with volumes of 0.2 mm3, 0.8 mm3 and 1.0 mm3 for the C, AC and ACG groups, respectively (Fig. 2E). Nevertheless, one-way ANOVA showed no significant difference among the groups (p ¼ 0.238). As cells are interacting with the scaffolds and responding to soluble signals in the chondrogenic differentiation medium in real time, other co-varying parameters such as elastic modulus also varies. Fig. 2F shows that at day 7, the mean ± standard error reduced elastic modulus (Er) of C, AC and ACG was 6.63 ± 1.92 kPa, 0.69 ± 0.35 kPa, and 1.58 ± 1.89 kPa, respectively. One-way ANOVA showed significant differences among groups (at p < 0.001) while Bonferroni's post-hoc test showed that the C group was significantly different from the AC group (at p < 0.001) and the ACG group (at p ¼ 0.002), but no significant difference was found between the AC and ACG groups (p ¼ 1.000). 3.2. MSCs are largely viable but have low proliferative activity in all three scaffolds during chondrogenic differentiation Fig. 3 shows the gross appearance of the hMSC-encapsulated microspheres, as well as the number of live and dead cells, and proliferative activity of MSCs in the different scaffolds at different time points during chondrogenic differentiation. In general, the C microspheres contracted from day 7 to day 14 but then maintained the same size at 14 and 21 days post-differentiation (Fig. 3A1eA3). For the AC (Fig. 3B1eB3) and ACG groups (Fig. 3C1eC3), there was less contraction and the microspheres maintained a similar size throughout the differentiation period. The live/dead staining showed that the MSCs were largely viable in all three types of scaffolds (Fig. 3D1eF3), as shown by the green fluorescence. In both the C and ACG groups, the MSCs maintained an elongated shape at the three time points tested, (see inserts of Fig. 3D1eD3 and Fig. 3F1eF3, respectively, whereas in the AC group, the cells maintained a more rounded morphology (see inserts of Fig. 3E1eE3) throughout the induction period. The proliferative activity of MSCs was low in all three groups, as demonstrated by the

212

Y.Y. Li et al. / Biomaterials 52 (2015) 208e220

Fig. 2. Ultra-structural, physical and biochemical characterization of the three different scaffolds. (AeC) SEM images showing the ultra-structural properties of the (A) collagen, (B) aminated collagen and (C) aminated collagen-GAG scaffolds (Magnification: 100KX; Scale bars: 200 nm); DeF: Box plots showing the (D) GAG:HYP ratio (n ¼ 3), (E) volume (n ¼ 4) and (F) reduced elastic modulus (n ¼ 3e4) of the three scaffolds.

absence of Ki-67 expression in the nuclei throughout the induction period (Fig. 3G1eI3). 3.3. Cytoskeletal organization of the hMSCs in the scaffolds during chondrogenic differentiation Fig. 4 shows H&E staining and organization of the cytoskeleton of MSCs in the different scaffolds. The H&E staining shows the typical morphology of the cartilage-like tissues in which the MSCs resided at Day 7, 14 and 21. In the C group, the cells were located in lacunae-like structures in the collagen matrix (Fig. 4A1eA3), whereas clusters of MSCs were found in both the AC (Fig. 4B1eB3) and ACG groups (Fig. 4C1eC3) during chondrogenic differentiation. F-actin was strongly expressed in the C group (Fig. 4D1eD3), and moderately expressed in the ACG group (Fig. 4F1eF3) but not expressed at all in the AC group (Fig. 4E1eE3). Intense expression of a-tubulin was also observed in the C group during the course of differentiation (Fig. 4G1eG3), but only some cells expressed atubulin in the AC (Fig. 4H1) and ACG (Fig. 4I1) groups at 7 days postdifferentiation and little expression was observed at 14 and 21 days post-differentiation (Fig. 4H2eH3, I2eI3). Cells encapsulated in all three types of scaffold expressed vimentin throughout chondrogenic differentiation (Fig. 4J1eJ3, K1eK3, L1eL3), with the highest expression being observed in the ACG group (Fig. 4L1eL3). 3.4. Differential expression of integrins in the three scaffolds during chondrogenic differentiation Fig. 5 shows the labeling patterns of integrin av, a5 and a2b1 in hMSCs in the three scaffolds during chondrogenic differentiation. In the C group, the expression of integrin av increased between day 7 and day 21 of chondrogenic differentiation (Fig. 5A1eA3), whereas it was moderately expressed throughout differentiation in the ACG group (Fig. 5C1eC3), and it was not expressed much at all at any of the differentiation time points in the AC group (Fig. 5B1eB3). On the other hand, integrin a5 was expressed constantly throughout the period of differentiation in group C

(Fig. 5D1eD3), group AC (Fig. 5E1eE3) and group ACG (Fig. 5F1eF3). For integrin a2b1, the initial expression in the C group was very high but its expression decreased over time (Fig. 5G1eG3). No expression of integrin a2b1 was found in either the AC (Fig. 5H1eH3) or the ACG (Fig. 5I1eI3) groups. 3.5. Differential co-localization between integrin a5 and fibronectin in hMSCs in the three scaffolds during chondrogenic differentiation Fig. 6 shows the localization pattern of integrin a5 in hMSCs, and its well-known extracellular matrix protein ligand fibronectin, as well as their pattern of co-localization in the various scaffolds during chondrogenic differentiation. In group C, apart from day 0 (Fig. 6A1eA4), hMSCs expressed integrin a5 and deposited fibronectin extracellularly during chondrogenic differentiation (Fig. 6D1eD4, G1eG4 and J1eJ4). Nevertheless, there was very little co-localization of this ligandereceptor pair throughout all time points, as shown by the Pearson coefficient of <0.1 (Fig. 6M), which indicates no or only a very weak correlation, and suggests a lack of functional binding or adhesion (Fig. 6A3eA4, D3eD4, G3eG4 and J3eJ4). In the AC group, hMSCs expressed both integrin a5 and fibronectin within a few hours after encapsulation on day 0 (Fig. 6B1eB4). However, again the Pearson coefficient was only slightly higher than 0.1 (Fig. 6M), which suggests very weak correlation or colocalization (Fig. 6B3eB4 and M). As chondrogenic differentiation proceeded, hMSCs in the AC scaffolds increasingly expressed both integrin a5 and fibronectin (Fig. 6E1eE4, H1eH4 and K1eK4) but again the level of co-localization was negligible with close-to-zero Pearson coefficients (Fig. 6M). In the ACG group, the hMSCs also expressed both integrin a5 and fibronectin immediately after encapsulation on day 0 (Fig. 6C1eC4) although the level of colocalization was negligible at this stage (Fig. 6M). However, the level of expression of both integrin a5 and fibronectin increased over time (Fig. 6F1eF4, I1eI4 and L1eL4). Notably, the hMSCs in the ACG scaffolds exhibited an increasing level of colocalization between integrin a5 and fibronectin, with Pearson coefficients of >0.1 and >0.4 on day 7 (Fig. 6F3eF4 & M) and 14

Y.Y. Li et al. / Biomaterials 52 (2015) 208e220

213

Fig. 3. Morphology, viability and proliferation of hMSCs in the different scaffolds during chondrogenic differentiation. hMSC were microencapsulated in (A,D,G) C, (B,E,H) AC and (C,F,I) ACG and then induced to differentiate toward a chondrogenic lineage for 7 (1), 14 (2) or 21 (3) days. (A1eC3) Phase contrast images show the morphology of the hMSCmicroencapsulated microspheres. (D1eF3) The number of viable hMSCs are shown by the fluorescence images whereby live cells are green and dead cells are red. (G1eI3) Immunofluorescence images of Ki-67 (green) and DAPI (blue) showing a negligible level of proliferation. The panel inserts show magnified views of representative cells. Scale bars: 0.5 mm for AeC; 70 mm for DeF; 20 mm for GeI; and 4 mm for the inserts. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

214

Y.Y. Li et al. / Biomaterials 52 (2015) 208e220

Fig. 4. Morphology of the hMSCs and expression of cytoskeletal components in the various scaffolds during chondrogenic differentiation. (Fig. 4A1eC3) H&E staining revealed the morphology of the hMSCs in the different scaffolds.; (Fig. 4D1eI3) Fluorescent labeling of (Fig. 4D1eF3) F-actin (red);, (Fig. 4G1-I3) a-tubulin (green) and (Fig. 4J1eL3) vimentin (green) in hMSCs in the different scaffolds at (1) 7, (2) 14 and (3) 21 days post-differentiation. In all the fluorescence images, the nuclei (blue) were stained with DAPI. The panel inserts show representative individual cells. (Scale bars: 50 mm for A1eC3; 5 mm for D1eF3, and 4 mm for G1eL3). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

(Fig. 6I3eI4 & M), respectively, suggesting an increasing level of correlation during these stages of chondrogenic differentiation. These data indicate strong interactions, functional binding and/or stable integrin a5-based adhesion formation between the hMSCs and the ACG scaffolds. On the other hand, at day 21 even with the strong expression of both integrin a5 and fibronectin there was little colocalization at this stage of chondrogenic differentiation. Fig. 6M is a bar chart showing the colocalization index of integrin a5 and fibronectin (as represented by the Pearson correlation coefficient) in the various scaffolds and at the different time points. Twoway ANOVA showed that both time and group factors significantly affected the colocalization index (at p < 0.001). In addition, Bonferroni's posthoc tests showed that the colocalization index of the ACG group was significantly different from both the C and AC groups (at p < 0.001) while the colocalization index at day 14 of

chondrogenic differentiation was significantly different from all the other time points (at p  0.002). 3.6. The cell fates of hMSCs in the various scaffolds during chondrogenic differentiation Fig. 7 shows the expression of Sox9, one of the major chondrogenic markers, in hMSCs in the different scaffolds during chondrogenic differentiation. Nuclear localization of Sox9 was found in the C group (Fig. 7A1eA3), suggesting activation of this major transcription factor for chondrogenesis, but no such localization was observed in either the AC (Fig. 7B1eB3) or ACG (Fig. 7C1eC3) groups. Fig. 7G shows the level of SOX9 gene expression among the different groups during chondrogenic differentiation. Two-way ANOVA showed that SOX9 expression was

Y.Y. Li et al. / Biomaterials 52 (2015) 208e220

215

Fig. 5. Immunofluorescence labeling of integrins av, a5 and a2b1 in scaffold-encapsulated hMSCs during chondrogenic differentiation. Expression of (A1eC3) integrin av (red), (D1eF3) integrin a5 (red), and (G1eI3) integrin a2b1 (green) expression in hMSCs encapsulated in (A1e3, D1e3, G1e3) C, (B1e3, E1e3, H1e3) AC or (C1e3, F1e3, I1e3) ACG at day 7 (1), 14 (2) and 21 (3) post-chondrogenic differentiation. The nuclei (blue) were stained with DAPI. (Scale bars: 4 mm for A1eF3 and 5 mm for G1eI3). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

significantly different in the various scaffold groups (at p < 0.001) but that within each group, expression of the gene was not affected by the sampling time (p ¼ 0.475). A comparison was made between SOX9 expression in the various scaffold types, with that in a

reference group, where cells were encapsulated in collagen (like the C group) but then cultured in normal medium (NM). Thus, SOX9 gene expression increased by ~3 fold in the C group but only by ~1.5 fold in the ACG group, and it was undetectable in the AC group. In

216

Y.Y. Li et al. / Biomaterials 52 (2015) 208e220

Fig. 6. Immunofluorescence labeling of integrin a5 and fibronectin in hMSCs encapsulated in the different scaffolds during chondrogenic differentiation. Expression of (A1eL1) integrin a5 (red) and (A2eL2) fibronectin (green), as well as (A3eL3) the two channels when merged. (A4eL4) Higher magnification views of a few cells from panels A3eL3. In A3eL3 and A4eL4, regions of integrin a5 and fibronectin co-localization are shown in yellow. hMSCs were encapsulated in scaffolds made of (A1e4, D1e4, G1e4 and J1e4) collagen, (B1e4, E1e4, H1e4 and K1e4) aminated collagen or(C1e4, F1e4, I1e4 and L1e4) aminated collagen-GAG, and sampled at (A1eC4) day 0 of differentiation as well as (D1eF4) day 7, (G1eI4) day 14 and (J1eL4) day 21 post-differentiation. Scale bars are 4 mm for A4 to L4 and 20 mm for all the other images. (M) Bar chart to show the colocalization index of fibronectin and integrin a5 in the three scaffold groups at the different time point during chondrogenic differentiation. The colocalization index is the Pearson correlation coefficient, which was determined from the semi-quantitative image analysis data acquired from the immunofluorescence labeling. Measurements were based on the analysis of 8e25 cells per section for the AC and ACG groups, and 3e36 cells per section for the C group, in 4 sections representing different planes of the image stack of the specimens. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Y.Y. Li et al. / Biomaterials 52 (2015) 208e220

217

Fig. 7. Expression of chondrogenic markers in hMSCs encapsulated in different scaffolds at different time points during chondrogenic differentiation. Immunofluorescence labeling of Sox9 (A1eC3) and Alcian blue staining of GAGs (D1eF3) in hMSCs in the C (A1eA3, D1eD3), AC (B1eB3, E1eE3) and ACG (C1eC3, F1eF3) groups at day 7 (1), 14 (2) and 21 (3) post-chondrogenic differentiation. Scale bars are 4 mm for A1eC3 and 50 mm for D1eF3). (GeJ) Bar charts showing the expression of (G) SOX9, (H) COL2A1, (I) ACAN, and (J) COL1A1 The data are presented as mean ± 1SE of n ¼ 3 to 6 experiments).

addition, Alcian blue staining indicated that GAGs were most prominent in the C group (Fig. 7D1eD3), far less obvious in the ACG group (Fig. 7F1eF3) and absent in the AC group (Fig. 7E1eE3). Fig. 7H shows the COL2A1 gene expression among different groups during chondrogenic differentiation. Two-way ANOVA showed that

COL2A1 expression was significantly different both in the three scaffold groups (p ¼ 0.005) and at the three time points (p ¼ 0.026). Comparing the level of COL2A1 expression in the different scaffolds with the control group, it was shown to increase by nearly 150 fold in the C group and by ~13 fold in the ACG at day 21 post-

218

Y.Y. Li et al. / Biomaterials 52 (2015) 208e220

differentiation, but it was undetectable in the AC group. Fig. 7I shows the expression of ACAN among the different scaffold groups during chondrogenic differentiation. Two-way ANOVA showed that ACAN expression was significantly different in the various scaffold groups (p ¼ 0.018) but that within each group, ACAN did not change over time (p ¼ 0.657). Comparing ACAN expression in the three scaffold types, with the control collagen group in normal medium, indicated that gene expression increased by ~15 fold in the C group and by ~3 fold in the ACG group, but was undetectable in the AC group, during chondrogenic differentiation. Fig. 7J shows the expression of COL1A1 among the different groups during chondrogenic differentiation. Two-way ANOVA showed that COL1A1 expression was significantly different in the various scaffold groups (at p < 0.0001) but that it didn't change over time between day 7 and day 21 (p ¼ 0.402). A comparison of COL1A1 gene expression in the different scaffolds with that of the control group, showed that COL1A1 increased in all three groups, i.e., by ~8 fold in the C group, and by ~3 fold in both the ACG and AC groups during chondrogenic differentiation. 4. Discussion The current study demonstrates that the composition of the matrix used as a scaffold for MSCs may affect the fate of the cells upon chondrogenic differentiation, perhaps via changes in integrin expression and/or organization of the cytoskeleton. In the early stages of differentiation, chondrogenically differentiating MSCs in the C group express only integrin a2b1, which is a major collagen receptor [26] and is involved in collagen remodeling [14,27]. This is reasonable because the initial template matrix in the C group is type I collagen. As differentiation proceeds, changes in the ECM, such as deposition of proteoglycan and type II collagen, and hence alterations in the cellematrix interactions occur. Specifically, we showed that there is an increase in integrin av and a decrease in integrin a2b1 in conjunction with a relatively constant level of integrin a5 expression. This pattern of integrin expression in chondrogenically differentiating MSCs encapsulated in collagen microspheres is similar to that reported in chondrocytes themselves, which express high levels of integrin a5, a2b1 and av for ECM components such as fibronectin, collagen and vitronectin [28,29], as well as far lower levels of integrin a2 [29]. On the other hand, the initial interaction of MSCs with a collagen matrix via integrin a2 is essential in that it leads to a rapid contraction during early differentiation, and thus simulates the mesenchymal condensation and pre-cartilaginous stages during chondrogenesis [30]. This contraction is important in further increasing the density of cells, which has been shown to facilitate chondrogenic differentiation [23,31]. The importance of an enhanced cellematrix interaction in stimulating contraction was shown by the association between integrin av expression and increased traction force generated by MSCs. The intimate and developing cellematrix interaction that occurs during chondrogenic differentiation is also reflected by the stable actin cytoskeleton, which suggests that MSCs might be able to adhere, spread and generate traction force on the matrix. We showed that there was stable expression of integrin a5 but no co-localization with its major ligand, fibronectin, in hMSCs in the C group. This may be related to the high levels of expression of other integrins, including av and a2, which may alone be sufficient to stimulate chondrogenesis. All in all, collagen alone has been shown to be a conducive scaffold for stimulating chondrogenesis in hMSCs, which corroborates our previous studies [22,23]. The AC group was used as a control for the ACG group because we wanted to find out if amination of collagen alone has an effect on MSC chondrogenesis. Indeed, amination of collagen was shown to completely inhibits the chondrogenic ability of the collagen

scaffold. MSCs neither adhere to nor spread on aminated collagen, as shown by the rounded morphology and the absence of actin cytoskeleton throughout the chondrogenic differentiation period. The rounded morphology may be due to poor adhesion of the MSCs in the AC scaffold. In addition, the absence of the collagen receptor, integrin a2b1, in the AC group suggests that the structure of collagen in and around the integrin a2b1 binding site may have been altered by the amination process, which might explain why MSC adhesion was inhibited. Moreover, cations such as Mg2þ are known to be necessary for the attachment of chondrocytes to matrices such as fibronectin and collagen [32]. Aminated collagen, however, itself has an increased positive charge and this may repel the cations and hence prevent cell adhesion to such matrices. Moreover, the hMSCs did not express integrin av, which is known to bind to both fibronectin and vitronectin [33]. This suggests that cell adhesion to aminated collagen might have been inhibited due to a lack an integrin av-based adhesion mechanism. Anoikis and subsequent cell death may result if anchorage-dependent cells such as MSCs do not recognize and adhere to the substrate. Fibronectin was not added extrinsically in the scaffold system, but hMSCs were able to express both fibronectin and integrin a5 within a few hours after microencapsulation. A previous study suggests that increased expression of integrin a5 may enhance the chondrogenic differentiation of MSCs [34] but in the current study, the stable expression of integrin a5 in aminated collagen did not appear to be sufficient to support chondrogenic differentiation of MSCs. This might be due to the fact that there is only a low level of co-localization (and hence functional binding and the creation of attachments) between integrin a5 and its ligand, fibronectin. On the other hand, the early expression of fibronectin in the AC group might be associated with a mechanism for rapid attachment to ensure the survival of adherent cells such as hMSCs. From the rounded morphology of the fibronectin expressing hMSCs, we speculate that cells might efficiently synthesize and secrete fibronectin as a pericellular matrix. Integrin a5b1 has been shown to support survival of cells [35]. Therefore, the ability of MSCs to survive at all in the AC scaffold, where they do not bind via either integrins a2 or av, may be due to the rapid production of fibronectin and expression of integrin a5, at a time when survival is a priority over differentiation. Previous studies have shown that cells with a rounded morphology have a dispersed actin cytoskeleton and few adhesions [15], and thus spreading is restricted [36]. In addition, when the actin cytoskeleton is disrupted then chondrogenesis of embryonic stem cells [37] and embryo limb bud mesenchymal cells [17,38,39], as well as dedifferentiation of chondrocytes [40] is stimulated. The current study on the other hand is somewhat contradictory to the previous reports [15,17,36e39], in that, despite maintaining a rounded morphology, with minimal spreading, having a diffuse actin cytoskeleton and an absence of integrin a2 and av-based adhesion formation, the MSCs failed to differentiate towards a chondrogenic lineage. This discrepancy suggests that having a rounded shape with little spreading and minimal actin cytoskeleton might not be enough to encourage chondrogenic differentiation. Other factors, such as the presence of an optimal matrix, functional binding with a suitable integrin, and the appropriate mechanical stiffness might be more important for the chondro-conductivity of scaffolds. In summary, the AC scaffold was shown to support survival but prevent chondrogenesis of hMSCs. The ACG scaffold was developed to incorporate more GAGs into the solid meshwork of collagen [8], therefore simulating the ECM of GAG-rich tissues such as articular cartilage and intervertebral disc. The amount of GAGs incorporated into collagen is controlled by chemically modifying the charged functional groups of collagen. In the current study, excess GAGs were used to exhaust bind the aminated collagen groups, such that the MSCs were not directly

Y.Y. Li et al. / Biomaterials 52 (2015) 208e220

exposed to the positive charges of the AC group. The main cytoskeletal protein in MSCs encapsulated in the ACG scaffold, was vimentin rather than actin. Vimentin is an intermediate filament protein, which has the function of maintaining cell shape and integrity [41,42]. Previous studies have demonstrated that an intact vimentin network is crucial for the maintenance of a chondrocytic phenotype [43], and it acts as a positive regulator of chondrogenesis in human adult multipotent progenitor cells [44]. However, vimentin-null mice are healthy without any skeletal malformations [45], suggesting that vimentin alone is insufficient to support chondrogenesis. In this study, we showed that the interaction of MSCs with the cell matrix in the ACG scaffold is significantly affected. No integrin a2b1 collagen receptor was expressed at all, similar to that in the AC group, but integrin av was positively expressed in the ACG group throughout the differentiation period, suggesting that the expression of av is crucial for chondrogenic differentiation. Integrin av is expressed in both chondrocytes [29] and MSCs [46]. It appears to be responsible for chondrocyte adhesion to cartilage [47] and it regulates proteoglycan synthesis in chondrocytes seeded on agarose gel [48]. Friedl et al. [49] demonstrated a significant correlation between integrin av and Sox9 gene expression upon mechanical stimulation of MSCs. It has also previously been suggested that integrin av is important in mediating the survival signal [50]. In the ACG scaffold, therefore, av may play a role in cell adhesion and to support chondrogenesis. In addition, there was a high level of co-localization between fibronectin and integrin a5 in the ACG group. This suggests that, as previously suggested [51], specific and functional binding or the formation of stable adhesions in relation to integrin a5 may be dominant in enabling subsequent signaling for chondrogenesis, particularly when integrin av signaling is not optimal. In contrast to the AC group, gene expression of chondrogenic markers including Sox9, Col2 and aggrecan are all positive in the ACG group, suggesting that an ACG scaffold can support chondrogenesis of MSCs, even though differentiation is not as efficient as that in the C scaffold. This work suggests that varying the matrix composition imposes changes in the cellematrix interactions via the expression of different integrins. These might work in a complex and complementary manner, and might be associated with events including cytoskeletal organization and substrate stiffness, in facilitating chondrogenesis. The differential roles played by the various integrins during chondrogenesis warrant further investigation. Our results, however, also demonstrate that changing the chemical composition of the collagen-based matrix may result in the MSCs having totally different fates of upon the induction of chondrogenesis. In addition, varying factors such as mechanical compliance and the ultrastructural characteristics of the resulting matrix may also contribute. It is generally accepted that substrate stiffness alone might affect the fate of MSCs in 2D [52,53] and 3D [54,55] models. In microencapsulation models such as the current study, the MSCs actively interact with the collagen-based materials by expressing integrin a2 as they are entrapped in the solidifying meshwork or hydrogel, and this results in real-time contraction. However, no such cellecollagen interaction occurred in the aminated collagen or aminated collagen-GAG scaffolds. This resulted in different extents of contraction and hence different mechanical compliances. The stiffer matrix (i.e., >6 kPa) in the C scaffold appeared to better support MSC chondrogenesis, whereas the more compliant matrices (i.e., <1 kPa) in the AC and ACG scaffolds did not. Similar to all encapsulation materials previously tested, it was not possible in the current study to decouple the chemical factor (matrix composition) from the mechanical factor (elastic modulus) and/or other factors such as the ultrastructural properties. This is because the cells interact with the scaffolds in real-time whereas the associating factors co-vary in real-time. We

219

speculate that the biochemical composition might play a more important role than the other factors, including the mechanical compliance and ultrastructural properties. This is because the AC and ACG scaffolds have a very similar elastic modulus but very different morphology, cytoskeletal organization, integrin expression and fate determination upon chondrogenic differentiation. Nevertheless, the interplay among the associated factors including the surface chemistry and charge density, matrix composition, mechanical compliance, ultrastructural properties, cellematrix interaction, cytoskeleton pattern and the fates of the MSCs during chondrogenic differentiation is extremely complex, and there is likely to be significant cross-talk between the various factors, all of which warrant further investigation. For example, it has been reported that an interaction between matrix stiffness and adhesive ligand presentation regulates MSC differentiation [56]; that matrix stiffness controls cell fate by modulating integrin binding and reorganization of adhesion ligands [52]; and that integrinmediated adhesion inhibits chondrogenic differentiation of MSCs via the actin cytoskeleton [16]. 5. Conclusions This work demonstrates the impact that the microenvironment has on stem cell fate. By varying the composition of the scaffolding matrix, we showed that the fate of MSCs is significantly affected upon chondrogenic differentiation, perhaps via changes in integrin expression and cytoskeleton organization. Unmodified collagen scaffolds successfully support MSC chondrogenesis, such that the cells were able to spread, integrin a2 was expressed and the actin cytoskeleton was prominent. As the MSCs differentiated into chondrogenic lineage cells with increasing levels of expression of specific chondrogenic markers, the expression of integrin a2 was reduced while that of integrin av increased. Chemical modification of collagen by amination prevented MSC chondrogenesis although the cells remained viable. In addition, the MSCs did not spread on the aminated collagen scaffold and they had no actin cytoskeleton. Furthermore, the cells did not express either integrin a2 or av, but they did exhibit a low level of functional interaction between integrin a5 and fibronectin, and ultimately they failed to differentiate into cells of a chondrogenic lineage upon chemical induction. Electrostatic modification of the aminated collagen with GAGs partially enabled MSC chondrogenesis. Although the MSCs did not express integrin a2, they were able to express integrin av, and they showed strong co-localization of integrin a5 with fibronectin. In addition, the cells exhibited a well-spread morphology, they had a vimentin-based (i.e., rather than actin-based) cytoskeleton, and displayed an increase in expression of chondrogenic markers, albeit at a much lower level that of the unmodified collagen. This work suggests the importance of controlling matrix composition as a strategy to manipulate stem cell fates. Acknowledgments This work was supported by a grant from The University of Hong Kong Small Project Funding (201109176132) and a Hong Kong Research Grants Council General Research Fund award (760111). The authors thank the Texas A&M Health Science Center College of Medicine Institute for Regenerative Medicine at Scott & White for providing the hMSCs, and Dr. H.J. Diao for his assistance with the real-time PCR analysis. Appendix A. Supplementary data Supplementary data related to this article can be found online at http://dx.doi.org/10.1016/j.biomaterials.2015.02.037.

220

Y.Y. Li et al. / Biomaterials 52 (2015) 208e220

References [1] Watt FM, Hogan BL. Out of Eden: stem cells and their niches. Science 2000;287:1427e30. [2] Moore KA, Lemischka IR. Stem cells and their niches. Science 2006;311: 1880e5. [3] Bi Y, Stuelten CH, Kilts T, Wadhwa S, Iozzo RV, Robey PG, et al. Extracellular matrix proteoglycans control the fate of bone marrow stromal cells. J Biol Chem 2005;280:30481e9. €nherr E. Proteoglycans of the extracellular matrix and growth [4] Kresse H, Scho control. J Cell Physiol 2001;189:266e74. [5] Guilak F, Cohen DM, Estes BT, Gimble JM, Liedtke W, Chen CS. Control of stem cell fate by physical interactions with the extracellular matrix. Cell Stem Cell 2009;5:17e26. [6] Flanagan LA, Rebaza LM, Derzic S, Schwartz PH, Monuki ES. Regulation of human neural precursor cells by laminin and integrins. J Neurosci Res 2006;83:845e56. [7] Renth AM, Detamore MS. Leveraging “raw materials” as building blocks and bioactive signals in regenerative medicine. Tissue Eng Part B Rev 2012;18: 341e62. [8] Choy AT, Leong KW, Chan BP. Chemical modification of collagen improves glycosaminoglycan retention of their co-precipitates. Acta Biomater 2013;9: 4661e72. [9] Docheva D, Popov C, Mutschler W, Schieker M. Human mesenchymal stem cells in contact with their environment: surface characteristics and the integrin system. J Cell Mol Med 2007;11:21e38. [10] Loeser RF. Integrin-mediated attachment of articular chondrocytes to extracellular matrix proteins. Arthritis Rheum 1993;36:1103e10. [11] Salter DM, Hughes DE, Simpson R, Gardner DL. Integrin expression by human articular chondrocytes. Br J Rheumatol 1992;31:231e4. [12] Woods VL, Schreck PJ, Gesink DS, Pacheco HO, Amiel D, Akeson WH, et al. Integrin expression by human articular chondrocytes. Arthritis Rheum 1994;37:537e44. [13] Yasuda T, Shimizu K, Nakagawa Y, Ishikawa H, Nishihara H, Nakamura T. Possible involvement of RGD (Arg-Gly-Asp)-containing extracellular matrix proteins in rat growth plate chondrocyte differentiation in culture. J Bone Miner Res 1996;11:1430e7. [14] Chang CF, Lee MW, Kuo PY, Wang YJ, Tu YH, Hung SC. Three-dimensional collagen fiber remodeling by mesenchymal stem cells requires the integrinmatrix interaction. J Biomed Mater Res A 2007;80:466e74. [15] Mathieu PS, Loboa EG. Cytoskeletal and focal adhesion influences on mesenchymal stem cell shape, mechanical properties, and differentiation down osteogenic, adipogenic, and chondrogenic pathways. Tissue Eng Part B Rev 2012;18:436e44. [16] Connelly JT, Garcia AJ, Levenston ME. Interactions between integrin ligand density and cytoskeletal integrity regulate BMSC chondrogenesis. J Cell Physiol 2008;217:145e54. [17] Woods A, Wang G, Beier F. RhoA/ROCK signaling regulates Sox9 expression and actin organization during chondrogenesis. J Biol Chem 2005;280: 11626e34. [18] Chan BP, Hui TY, Yeung CW, Li J, Mo I, Chan GC. Self-assembled collagenhuman mesenchymal stem cell microspheres for regenerative medicine. Biomaterials 2007;28:4652e66. [19] Tiller JC, Bonner G, Pan LC, Klibanov AM. Improving biomaterial properties of collagen films by chemical modification. Biotechnol Bioeng 2001;73:246e52. [20] Eastoe JE. The amino acid composition of mammalian collagen and gelatin. Biochem J 1955;61:589e600. [21] Olde Damink LH, Dijkstra PJ, van Luyn MJ, van Wachem PB, Nieuwenhuis P, Feijen J. Cross-linking of dermal sheep collagen using a water-soluble carbodiimide. Biomaterials 1996;17:765e73. [22] Li CH, Chik TK, Ngan AH, Chan SC, Shum DK, Chan BP. Correlation between compositional and mechanical properties of human mesenchymal stem cellcollagen microspheres during chondrogenic differentiation. Tissue Eng Part A 2011;17:777e88. [23] Hui TY, Cheung KM, Cheung WL, Chan D, Chan BP. In vitro chondrogenic differentiation of human mesenchymal stem cells in collagen microspheres: influence of cell seeding density and collagen concentration. Biomaterials 2008;29:3201e12. re V, Caruelle JP, Martelly I, Papy[24] Barbosa I, Garcia S, Barbier-Chassefie García D. Improved and simple microassay for sulfated glycosaminoglycans quantification in biological extracts and its use in skin and muscle tissue studies. Glycobiology 2003;13(9):647e53. [25] Woessner JF. The determination of hydroxyproline in tissue and protein samples containing small proportions of this imino acid. Arch Biochem Biophys 1961;93:440e7. [26] Leitinger B, Hohenester E. Mammalian collagen receptors. Matrix Biol 2007;26:146e55. [27] Phillips JA, Bonassar LJ. Matrix metalloproteinase activity synergizes with a2b1 integrins to enhance collagen remodeling. Exp Cell Res 2005;310:79e87. [28] Loeser RF. Chondrocyte integrin expression and function. Biorheology 2000;37:109e16.

[29] Woods Jr VL, Schreck PJ, Gesink DS, Pacheco HO, Amiel D, Akeson WH, et al. Integrin expression by human articular chondrocytes. Arthritis Rheum 1994;37:537e44. [30] Singh P, Schwarzbauer JE. Fibronectin and stem cell differentiation e lessons from chondrogenesis. J Cell Sci 2012;125:3703e12. [31] Bobick BE, Chen FH, Le AM, Tuan RS. Regulation of the chondrogenic phenotype in culture. Birth Defects Res C Embryo Today 2009;87:351e71. [32] Loeser RF. Modulation of integrin-mediated attachment of chondrocytes to extracellular matrix proteins by cations, retinoic acid, and transforming growth factor b. Exp Cell Res 1994;211:17e23. [33] Boettiger D, Lynch L, Blystone S, Huber F. Distinct ligand-binding modes for integrin avb3-mediated adhesion to fibronectin versus vitronectin. J Biol Chem 2001;276:31684e90. [34] Connelly JT, García AJ, Levenston ME. Inhibition of in vitro chondrogenesis in RGD-modified three-dimensional alginate gels. Biomaterials 2007;28: 1071e83. [35] Zhang Z, Vuori K, Reed JC, Ruoslahti E. The a5b1 integrin supports survival of cells on fibronectin and up-regulates Bcl-2 expression. Proc Natl Acad Sci U S A 1995;92(13):6161e5. [36] Gao L, McBeath R, Chen CS. Stem cell shape regulates a chondrogenic versus myogenic fate through Rac1 and N-cadherin. Stem Cells 2010;28:564e72. [37] Zhang Z, Messana J, Hwang NS, Elisseeff JH. Reorganization of actin filaments enhances chondrogenic differentiation of cells derived from murine embryonic stem cells. Biochem Biophys Res Commun 2006;348:421e7. [38] Zanetti NC, Dress VM, Solursh M. Comparison between ectoderm-conditioned medium and fibronectin in their effects on chondrogenesis by limb bud mesenchymal cells. Dev Biol 1990;139:383e95. [39] Zanetti NC, Solursh M. Induction of chondrogenesis in limb mesenchymal cultures by disruption of the actin cytoskeleton. J Cell Biol 1984;99:115e23. [40] Loty S, Forest N, Boulekbache H, Sautier JM. Cytochalasin D induces changes in cell shape and promotes in vitro chondrogenesis: a morphological study. Biol Cell 1995;83:149e61. [41] Goldman RD, Khuon S, Chou YH, Opal P, Steinert PM. The function of intermediate filaments in cell shape and cytoskeletal integrity. J Cell Biol 1996;134: 971e83. [42] Eckes B, Dogic D, Colucci-Guyon E, Wang N, Maniotis A, Ingber D, et al. Impaired mechanical stability, migration and contractile capacity in vimentindeficient fibroblasts. J Cell Sci 1998;111:1897e907. [43] Blain EJ, Gilbert SJ, Hayes AJ, Duance VC. Disassembly of the vimentin cytoskeleton disrupts articular cartilage chondrocyte homeostasis. Matrix Biol 2006;25:398e408. [44] Bobick BE, Tuan RS, Chen FH. The intermediate filament vimentin regulates chondrogenesis of adult human bone marrow-derived multipotent progenitor cells. J Cell Biochem 2010;109:265e76. [45] Colucci-Guyon E, Portier MM, Dunia I, Paulin D, Pournin S, Babinet C. Mice lacking vimentin develop and reproduce without an obvious phenotype. Cell 1994;79:679e94. [46] Gronthos S, Simmons PJ, Graves SE, Robey PG. Integrin-mediated interactions between human bone marrow stromal precursor cells and the extracellular matrix. Bone 2001;28:174e81. [47] Kurtis MS, Schmidt TA, Bugbee WD, Loeser RF, Sah RL. Integrin-mediated adhesion of human articular chondrocytes to cartilage. Arthritis Rheum 2003;48:110e8. [48] Chai DH, Arner EC, Griggs DW, Grodzinsky AJ. av and b1 integrins regulate dynamic compression-induced proteoglycan synthesis in 3D gel culture by distinct complementary pathways. Osteoarthr Cartil 2010;18:249e56. [49] Friedl G, Schmidt H, Rehak I, Kostner G, Schauenstein K, Windhager R. Undifferentiated human mesenchymal stem cells (hMSCs) are highly sensitive to mechanical strain: transcriptionally controlled early osteo-chondrogenic response in vitro. Osteoarthr Cartil 2007;15:1293e300. [50] Townsend PA, Villanova I, Uhlmann E, Peyman A, Knolle J, Baron R, et al. An antisense oligonucleotide targeting the alphaV integrin gene inhibits adhesion and induces apoptosis in breast cancer cells. Eur J Cancer 2000;36(3): 397e409. [51] Augello A, De Bari C. The regulation of differentiation in mesenchymal stem cells. Hum Gene Ther 2010;21:1226e38. [52] Engler AJ, Sen S, Sweeney HL, Discher DE. Matrix elasticity directs stem cell lineage specification. Cell 2006;126:677e89. [53] Park JS, Chu JS, Tsou AD, Diop R, Tang Z, Wang A, et al. The effect of matrix stiffness on the differentiation of mesenchymal stem cells in response to TGFb. Biomaterials 2011;32:3921e30. [54] Huebsch N, Arany PR, Mao AS, Shvartsman D, Ali OA, Bencherif SA, et al. Harnessing traction-mediated manipulation of the cell/matrix interface to control stem-cell fate. Nat Mater 2010;9:518e26. [55] Murphy CM, Matsiko A, Haugh MG, Gleeson JP, O'Brien FJ. Mesenchymal stem cell fate is regulated by the composition and mechanical properties of collageneglycosaminoglycan scaffolds. J Mech Behav Biomed Mater 2012;11: 53e62. [56] Rowlands AS, George PA, Cooper-White JJ. Directing osteogenic and myogenic differentiation of MSCs: interplay of stiffness and adhesive ligand presentation. Am J Physiol Cell Physiol 2008;295:C1037e44.