These authors contributed equally to this work.
Growth differentiation factor 5 modulation of chondrogenesis of self-assembled constructs involves gap junction-mediated intercellular communication
Article first published online: 4 NOV 2012
© 2012 The Authors Development, Growth & Differentiation © 2012 Japanese Society of Developmental Biologists
Development, Growth & Differentiation
Volume 54, Issue 9, pages 809–817, December 2012
How to Cite
Sun, Z., Zhang, Y., Yang, S., Jia, J., Ye, S., Chen, D. and Mo, F. (2012), Growth differentiation factor 5 modulation of chondrogenesis of self-assembled constructs involves gap junction-mediated intercellular communication. Development, Growth & Differentiation, 54: 809–817. doi: 10.1111/dgd.12009
- Issue published online: 10 DEC 2012
- Article first published online: 4 NOV 2012
- Manuscript Accepted: 14 SEP 2012
- Manuscript Revised: 3 SEP 2012
- Manuscript Received: 4 JUN 2012
- National Natural Science Foundation of China. Grant Number: 30800654
- Graduates' Innovation Foundation of Huazhong University of Science & Technology. Grant Number: HF-08-22-2011-530
- gap junction intercellular communication;
- growth differentiation factor-5;
- human mesenchymal stem cells;
- self-assembled cartilages
A novel scaffold-free self-assembled cartilage construct has been generated and used to repair particular chondral defects effectively. However, the mechanisms related to the construction of these self-assembled cartilages have not yet been fully elucidated. We hypothesize that gap junction intercellular communication (GJIC) plays a critical role in the development of self-assembled constructs upon GDF-5 induction. In this study, we investigated the effect of connexin 43 (C×43) mediated GJIC on GDF-5 modulation of chondrogenesis from two aspects, cell monolayer culture and 3-D self-assembly culture. We induced cells or self-assembled constructs with chondrogenic media (CM), growth differentiation factor 5 (GDF-5) or 1-heptanol for 3 weeks. At the end of that time, the results of quantitative fluorescence redistribution after photobleaching (FRAP) assay and immunofluorescence demonstrated that GDF-5 improved both GJIC and chondrogenic differentiation to a significant degree while 1-heptanol nearly offset the expected improvements in chondrogenesis. Biochemical assay and histology showed that GDF-5 can obviously enhance GAG, C×43 and type II collagen expressions. Conversely, we also showed that while 1-heptanol weakened GAG and type II collagen expression in self-assembled constructs, it had no effect on C×43 expression. Furthermore, real-time polymerase chain reaction showed that GDF-5 enhanced GAG and type II collagen transcription while 1-heptanol reduced them, but was affectless on C×43 transcription. This suggests that the generation of scaffold-free self-assembled cartilage from human mesenchymal stem cells upon GDF-5 induction may be mediated, at least in part, via the modulation of GJIC.
The appendicular skeleton vertebrate limbs are derived from an initial accumulation of mesenchymal cells that initially form prechondrogenic aggregates which, over time, ossify to form the adult bony skeleton. These pre-cartilage condensations, which come from the areas with increased cellular density, act as templates for the size, shape and position of skeletal elements (Summerbell & Wolpert 1972). It appears that this condensation stage of chondrogenesis is mediated by intercellular communication, most probably via gap junctions (Meyer et al. 1997). It is necessary to plate limb bud cells at high densities for the initiation of chondrogenesis (Ahrens et al. 1977). Our conclusions are that the contact between mesenchymal cells is necessary for the intercellular signal exchanges, which contribute to the initiation of condensation and the subsequent formation of chondroprogenitor cells.
Gap junctions are the specialized membrane channels, composed of connexons located in the cell membrane and act as the mechanism for the direct intercellular exchange of signaling molecules and metabolites (Stains & Civitelli 2005). Each connexon is composed of six molecules of a transmembrane protein known as connexin, all surrounding a centrally located pore. When appropriate signals are accessed, connexons in adjacent cells transiently join and then form a hydrophylic electrical and metabolic pathway linking the cytoplasms of neighboring cells.
GDF-5 is one of the bone morphogenetic proteins and belongs to the transforming growth factor-β (TGF-β) superfamily (Hatakeyama et al. 2004). And among the growth factors that target mesenchymal stem cells (MSCs), GDF-5 is able to promote the differentiation of MSCs into chondrocytes (Bai et al. 2004). Among humans, mutations in GDF-5 gene usually result in defects in the development of the appendicular skeleton while the mutations in mice are known to cause brachypodism (Storm et al. 1994; Thomas et al. 1996). Clearly, GDF-5 is closely involved in cartilage development and differentiation (Thomas et al. 1997). Other studies have demonstrated similar spatiotemporal mRNA expression patterns involving C×43 and GDF-5 during the development of limbs, tendons, spines and heart tissues (Coleman et al. 2003). And, as one of the gap junction protein families, C×43 can be detected during MSCs differentiating into chondrocytes (Dorshkind et al. 1993).
Self-assembly cultural techniques represent a new development in the field of tissue-engineered cartilage. This has opened new avenues for practitioners. As an example, a scaffold-free, self-assembled cartilage structure has recently been generated from GDF-5 pre-differentiated human MSCs (hMSCs) and effectively used to repair cartilage defects (Zhang et al. 2011). However, the exact mechanism via which these self-assembled cartilages form has not yet been determined. We believe that GJIC may play a significant role in GDF-5 induced self-assembled cartilages. So, to test the validity of this hypothesis, the effect of GJIC on the GDF-5-mediated chondrogenesis of monolayer cells and self-assembled constructs was investigated.
Materials and methods
Isolation and culture of MSCs
Under the auspices of an experimental protocol approved by the Ethics Committee of Wuhan Union Hospital, bone marrow specimens were obtained from the posterior iliac crests of three accident victims, ranging in age from 27 to 43 years old. The three individuals gave informed consent. To that end, hMSCs were isolated from bone marrow aspirates and cultured as described previously (Yu et al. 2004). Adherent cells were cultured for 12–14 days until they attained a confluence of greater than 80%. The cells were then digested with a solution of 0.25% trypsin and 0.02% Ethylenediaminetetraacetic acid (EdTA) (Invitrogen) and replated at a 1:2 dilution for the initial subculture. hMSCs underwent this treatment three times before they were collected for further use.
Self-assembly of MSCs
Self-assembly plates were treated as follows: Initially, 500 μL of 2% agarose solution was deposited into each well of a 24-well plate. Then the plate was gently agitated so the agarose was evenly distributed upon the well walls. Finally, the plate was inverted, allowing excess agarose to run off. At that point the remaining agarose was allowed to coagulate at room temperature before use. The hMSCs (Passage 3) were digested and resuspended in either CM (group A), CM + 100 ng/mL GDF-5 and 2.5 μmol/L 1-heptanol (Group B), CM + 100 ng/mL GDF-5, or CM + 2.5 μmol/L 1-heptanol (Group D) at a density of 2.5 × 107 cells/mL. The cells were plated to the agarose coated 24-well plates as a 400 μL drop and left to self-assemble. The media were changed once per day. The CM consisted of DMEM-high glucose medium (Invitrogen), 1% ITS + 1 (Sigma), 10−7 M dexamethasone, 1% fetal bovine serum (FBS), 40 μg/mL L-Proline, 100 μg/mL sodium pyruvate, and 50 μg/mL ascorbate 2-phosphate.
Determination of 1-heptanol's optimal concentration
To screen out the optimal concentration of 1-heptanol, cultures were exposed to 0, 1, 2.5, 5, 7.5 μmol/L 1-heptanol and sulfated GAG and cell number were quantified at t = 1 week. The dose response curve was drawn.
To verify screened concentrations, cell proliferation/viability was evaluated via a Cell Counting Kit-8 assay (CCK-8, Dojindo Molecular Technologies). In brief, hMSCs were plated in 96-well plates at a density of 2×103 cells/well. Then, after adherence to the plate, the initial defining media were aspirated away and replaced with either CM, CM + 2.5 μmol/L 1-heptanol or CM + 100 ng/mL GDF-5. At pre-determined time points, cell proliferation was assayed by the CCK-8 as the manufacturer's instructions.
Assessment of coupling between cells
Fluorescence redistribution after photobleaching assay for GJIC was performed as previously reported using an ACAS Ultima laser cytometer (Meridian Instruments) (Reits & Neefjes 2001). Cells were incubated with 5,6-carboxyfluorescein diacetate (10 μg/mL in Ca2+/Mg2+-PBS; Molecular Probes, Sigma) at 37°C for 15 min. Target cells were then randomly selected under a 20× objective microscope and photobleached to 10–30% of their original fluorescence intensity, at which point they were then examined for fluorescence in order to obtain the rate of recovery. The rate of recovery was calculated as follows: F(t) = (F(t) − F(0)) / (F(∞) − Fa) × 100%. F(t) is the rate of recovery for target cells at a fixed time point. F(t) is the relative fluorescence intensity for target cells. F(0) is the relative fluorescence intensity for target cells at the instant of bleaching. F(∞) is the relative fluorescence intensity for unbleached controls. And Fa is the relative fluorescence intensity for ground at the instant of bleaching.
Constructs harvested at t = 1, 2, 3 weeks were digested in papain (125 mg/mL) (Sigma) in 50 mmol phosphate buffer containing 2 mmol N-acetyl cysteine (Sigma) and 2 mmol EDTA (Sigma) at 65°C overnight. A Blyscan Sulfated Glycosaminoglycan Assay kit (Biocolor) was then used to quantify sulfated GAG. And a Quant-iT PicoGreen Cell Proliferation Assay kit (Invitrogen) was then used to measure total DNA content.
Histological and immunohistochemical analysis
Self-assembled constructs were fixed in 4% paraformaldehyde for 24 h, embedded in paraffin and then sectioned at 5 μm thick. Their subtle structure and organization can be observed by H&E staining. Toluidine blue (Sigma) staining was performed to detect extracellular sulfated glycosaminoglycan (GAG) deposition, whereas immunostaining was used to detect the expression of type II collagen and C×43, respectively. Each section was deparaffinized and rehydrated, followed by treatment with 10% goat serum to block nonspecific antibody binding sites, and finally incubated with primary antibodies at a dilution of 1:100 at 4°C overnight. After being rinsed with PBS, the sections were incubated with the secondary antibody, goat anti-rabbit immunoglobulin (Vectastain ABC kit), for 20 min at 37°C followed by incubations with an avidin-biotin-peroxidase complex (Amresco) for 20 min at 37°C and 3,3-diaminobenzidine (Amresco) for 3 min. Finally, the sections were counterstained with hematoxylin. Immunofluorescent staining for monolayer cells was also performed as mentioned above.
Total RNA was extracted from these constructs via standard protocols using standard commercial kits (TRIZOL Reagent, Invitrogen). Real-time polymerase chain reactions (PCRs) were performed using SYBR Green Master mix according to the protocols of the supplier (Invitrogen). The primer sequences were as follows: collagen II, sense 5′-ATG ACA ATC TGG CTC CCA AC-3′ and antisense 5′-GAA CCT GCT ATT GCC CTC TG-3′; aggrecan, sense 5′-AGA CAG TGA CCT GGC CTG AC-3′ and antisense 5′-TGG CCT CTC CAG TCT CAT TC-3′; C×43, sense 5′-GGA GAT GAG CAG TCT GCC TTT C-3′ and antisense 5′-TGA GCC AGG TAC AAG AGT GTG G-3′; GAPDH, sense 5′-GGC ACA GTC AAG GCT GAG AAT G-3′ and antisense 5′-ATG GTG GTG AAG ACG CCA GTA-3′. The SYBR Green signal was detected by a StepOne real-time PCR machine (ABI). The relative levels of transcript expression were quantified using the ΔΔCt method. All real-time PCRs were run in triplicate and gene expression was analyzed using an ABI PRISM 7900HT Sequence detection system (Applied Biosystems).
The statistical analysis was carried out via SPSS version 12.0 for Windows (SPSS). Significance of difference was determined using a one-way analysis of variance (anova). Data are presented as mean ± standard deviation (SD). Probabilities lower than 5% (P < 0.05) were considered statistically significant.
The formation of self-assembled cartilages
When cell suspensions were injected into agarose coated wells, cells initially sank into the surface of the agarose within 6 h (Fig. 1A), and then coalesced into a curved edge translucent membrane-like mass within 24 h (Fig. 1B). And within a further 4–5 days, the membrane-like mass further contracted into either a round or oval construct (Fig. 1C). Then, over the following weeks, the cambium exhibited a smooth, spherical and cartilage-like appearance (Fig. 1D).
Determination of 1-heptanol's optimal concentrations
The optimal concentration was initially screened out by biochemical analyses followed by cell viability assay. Cultures were exposed to varied dosages of 1-heptanol and sulfated GAG and cell number were quantified at t = 1 week. The dose response curve exhibited significant responses to 1-heptanol at doses of 1 μmol/L and above. At the dose of 2.5 μmol/L, there was an approximately 43.4% reduction in sulfated GAG with a statistically insignificant change in cell numbers (Fig. 2A).
Monolayer cultures treated with 2.5 μmol/L 1-heptanol or 100 ng/mL GDF-5 showed the same level of absorbance as did the control across all time points measured. There were no significant differences among the three groups at any given time point (P > 0.05), indicating that GDF-5 did not increase cell proliferation and that 1-heptanol at a dosage of 2.5 μmol/L exhibited no cytotoxicity nor did it suppress cell proliferation (Fig. 2B).
FRAP assay and immunofluorescence of cellular monolayers
Figure 3 shows a representation of digitized images performed in all of the groups. From them we learn that the fluorescence recovery rate was markedly enhanced for those target cells treated with GDF-5 (42.45% ± 2.52%, Fig. 3 C1–C3). However, the effect of GDF-5 on GJIC was clearly inhibited by 2.5 μmol/L 1-heptanol (P < 0.01, Fig. 3 B1–B3). Its mean fluorescence recovery rate was 22.75% ± 1.86%, which proved to be statistically insignificant compared with the control, whose recovery rate was 20.18% ± 2.03% (P > 0.05, Fig. 3 A1–A3). And those only treated with 1-heptanol had the lowest recovery rate (9.82% ± 1.43%, P < 0.01, Fig. 3 D1–D3).
Accompanying the high fluorescence recovery rate, those target cells treated with GDF-5 exhibited enhanced expression of type II collagen and GAG as compared with other groups (Fig. 4C,G). Affected by 1-heptanol, the hMSCs treated with CM + 100 ng/mL GDF-5 + 2.5 μmol/L 1-heptanol had less synthesis of collagen type II and GAG, but slightly higher than those only treated with CM (Fig.4A,B,E,F). The hMSCs treated with CM + 2.5 μmol/L 1-heptanol almost showed the weakest staining of all (Fig. 4D,H).
Dry weights (DW) were measured to normalize the biochemical contents of the self-assembled constructs. Constructs C contained significantly higher GAG per DW at any given time point when compared to the other groups (P < 0.05). The GAG/DW of construct C showed an obvious increase at t = 3 weeks while that of construct A and B showed only a slight increase in that same time frame and construct D had an obvious reduction in GAG (Fig. 5).
Histological and immunohistochemical analysis
Hematoxylin and eosin staining showed the cartilage-like lacunae found in all of the constructs (Fig. 6 A1,B1,C1,D1). Positive toluidine blue staining was observed in all of the groups. Construct C had strong positive staining, whereas constructs A and B showed weak staining (Fig.6 A2,B2,C2,D2). Collagen type II immunohistochemistry showed a strong positive staining in construct C, a weak staining in construct A and B, and a nearly negative staining in construct D (Fig.6 A3,B3,C3,D3). As for the immunostaining for C×43, construct C showed a dense deposit, which may correlate with the widespread degree of chondrogenic differentiation seen in construct C (Fig. 6 C4). Whereas, the strong expression of C×43, unaffected by 2.5 μmol/L heptanol, could be still observed in construct B (Fig. 6 B4). In the construct A and D, there was only a weak positive signal on the periphery, an area less mature than the center (Fig. 6 A4 and D4).
To evaluate the gene expression patterns of self-assembled constructs from four groups, the relative levels of the mRNA transcripts for collagen II, aggrecan and C×43 were determined.
GDF-5 treatment demonstrated additional prochondrogenic effect beyond that of CM standing alone. Construct C exhibited the highest collagen II and aggrecan expression levels among those constructs (P < 0.01, Fig. 7B and C). However, the effect of GDF-5 on chondrogenic differentiation was weakened by exposure to 1-heptanol. Figure 7B and C shows that exposure to 1-heptanol accounted for a 54% reduction in aggrecan mRNA levels and a 66% reduction in collagen II mRNA levels as compared to construct C (P < 0.05). Noticeably, C×43 expression in construct B was almost identical with that of construct C (P > 0.05), indicating that C×43 expression was not affected by 1-heptanol (Fig. 7A). Exposed to 1-heptanol, construct D exhibited the lowest collagen II, aggrecan and C×43 expression, compared to other constructs (P < 0.05, Fig. 7A–C).
Currently, the most common approach to cartilage tissue engineering involves a structurally, mechanically and biocompatible sound scaffold seeded with either chondrocytes or MSCs. However, scaffold-based tissue engineering cartilages are subject to a number of significant limitations including poor adhesion, immune rejection, material-induced inflammatory responses, toxicity arising from material degradation and the unfavorable mechanical properties of the scaffolds (Furukawa et al. 2008; Mayer-Wagner et al. 2010). Therefore, the development of a more sophisticated tissue-engineered alternative has received a great deal of recent attention. Specifically, scaffold-free tissue engineering is currently undergoing intense investigation. And recently, a scaffold-free self-assembled construct of hMSCs has been generated upon GDF-5 induction by our research group (Zhang et al. 2011). We believe the self-assembly approach seems significant because when cells are seeded at high densities into agarose-coated substrates, they are inhibited by the agarose from attaching and flattening. As a result, the attachment substrate for these cells are other cells, which seemingly reinforces the intercellular communication necessary for the maintenance of the chondrocytic phenotype and recapitulates the developmental process of articular cartilage (Hu & Athanasiou 2006; Elder & Athanasiou 2008; Ofek et al. 2008). In this case, cartilage-like constructs self-formed via a process we term “self-assembly”. However, the mechanisms involved in the technique have not yet been elucidated. Hence, in this study, we focused on the functional role played by C×43 in the development of self-assembled constructs upon GDF-5 induction.
1-heptanol, the long chain alcohol, was reported to block GJIC effectively (Weingart & Bukauskas 1998). This effect was not limited to C×43 gap junctions, but to all gap junctions. V.L. Keevil found that 1-heptanol could effectively block GJIC between adjacent rat heart muscle cells (Keevil et al. 2000). 1-heptanol's effects are rapidly reversible at low dosages, while high dosages seemingly cause toxic inhibition, with effects on the transcription of housekeeping genes as well as on specific aspects of C×43. Our study confirmed that the biological activity of 2.5 μmol/L 1-heptanol was not related to its toxicity.
We then investigated the effects of C×43 mediated GJIC on GDF-5 modulation of chondrogenesis from two aspects, cell monolayer culture and 3-D self-assembly culture. Examining the monolayer cultures after a period of 3 weeks, we observed that the GDF-5 treated cultures showed strengthened expressions of type II collagen and aggrecan. Furthermore, we also found that the strengthened expressions would be weakened by the presence of 2.5 μmol/L 1-heptanol while FARP assays showed that 100 ng/mL GDF-5 obviously strengthened the rate of fluorescence recovery, which was as well weakened by the presence of 2.5 μmol/L 1-heptanol. From these results we inferred that GJIC may very well be involved in GDF-5 modulation of chondrogenesis of monolayer cells.
Regarding the 3-D self-assembly cultures, we learned that construct C had both a five-fold increase in the level of C×43 transcription as well as a wider expression of C×43 protein when compared to controls after a 3-week culture period. The concordance between C×43 mRNA levels and C×43 protein expression following GDF-5 induction suggests that GDF-5 regulates C×43 expression at a transcriptional level. However, the question of whether GDF-5 can also act at a post-transcriptional or translational level is still unknown. We have chosen to examine this issue as our next area of study. In addition, our experiments showed an 11-fold increase in collagen II transcription and a seven-fold increase in aggrecan transcription as compared with controls. As might be expected, 2.5 μmol/L 1-heptanol inhibited expected levels of increase in collagen II and aggrecan transcription responding to GDF-5. Based on histological staining and biochemical analysis, our data indicated that GDF-5 induced hMSCs to form cartilage tissue during the self-assembly process by enhancing C×43 mediated GJIC. This result would be consistent with previous studies, which demonstrated that 100 ng/mL of GDF-5 was highly effective at inducing the differentiation of fibroblasts, periosteum-derived cells, and stem cells into chondrocytes while at the same time promoting the secretion of a chondrocyte-specific collagen and proteoglycan matrix (Gruber et al. 2001; Bai et al. 2004; Feng et al. 2008; Yin et al. 2010).
Low dosage 1-heptanol, by acting as a gap junction signal blocker, inhibited the expected increase in the chondrogenic differentiation of monolayer cells or self-assembled constructs in response to GDF-5, while the FRAP assays also showed a reduction in the expected enhancement in fluorescence recovery in response to GDF-5. These negative phenomena seem to imply that gap junction coupling plays an important role in GDF-5 modulation of chondrogenesis. We believe that gap junctions dictate the type of signals, second messenger signals and metabolites that propagate within cellular communications. Thus, these cells seem to form a functional entity, with the advantage that every cultured cell has maximal opportunities to receive the various types of differentiation signals. But once this gap junction was blocked by 1-heptanol, this functional entity was blocked, meaning that cells received fewer differentiation signals. We found it interesting that there was no reduction either in C×43 protein expression or in the levels of C×43 messaging even when gap junction uncoupling had occurred. And from that we deduced that 2.5 μmol/L 1-heptanol exerted no influence on either C×43 transcription or translation. As regards the underlying mechanism, we plan to research it as part of our next series of experiments.
In summary, this study has demonstrated for the first time that GJIC was, at least partially, involved in the generation of scaffold-free tissue-engineered cartilage. However, more specific studies are needed to further elucidate other mechanisms involved in the development of self-assembled engineered cartilage. Looking towards the future, we hope to manipulate the seeding cells with the C×43 gene in such a way as to allow us to harvest the self-assembled constructs with improved cartilage characteristics.
This work was supported by a grant from National Natural Science Foundation of China (No. 30800654) and a grant from Graduates' Innovation Foundation of Huazhong University of Science & Technology (No. HF-08-22-2011-530).
- 1977. Stage-related capacity for limb chondrogenesis in cell culture. Dev. Biol. 60, 69–82. , &
- 2004. Cartilage-derived morphogenetic protein-1 promotes the differentiation of mesenchymal stem cells into chondrocytes. Biochem. Biophys. Res. Commun. 325, 453–460. , , , , , &
- 2003. Correlation of GDF5 and connexin 43 mRNA expression during embryonic development. Anat. Rec. A. Discov. Mol. Cell. Evol. Biol. 275, 1117–1121. , , ,
- 1993. Connexin-43-type gap junctions mediate communication between bone marrow stromal cells. Blood 82, 38–45. , ,
- 2008. Effects of confinement on the mechanical properties of self-assembled articular cartilage constructs in the direction orthogonal to the confinement surface. J. Orthop. Res. 26, 238–246. &
- 2008. Adenovirus-mediated expression of growth and differentiation factor-5 promotes chondrogenesis of adipose stem cells. Growth Factors 26, 132–142. , , , &
- 2008. Scaffold-free cartilage by rotational culture for tissue engineering. J. Biotechnol. 133, 134–145. , , .
- 2001. Effects of cartilage-derived morphogenetic proteins and osteogenic protein-1 on osteochondrogenic differentiation of periosteum-derived cells. Endocrinology 142, 2087–2094. , , , , , &
- 2004. Distinct functions of BMP4 and GDF5 in the regulation of chondrogenesis. J. Cell. Biochem. 91, 1204–1217. , &
- 2006. A self-assembling process in articular cartilage tissue engineering. Tissue Eng. 12, 969–979. &
- 2000. The effect of heptanol on the electrical and contractile function of the isolated, perfused rabbit heart. Pflugers Arch. 440, 275–282. , , , &
- 2010. Membrane-based cultures generate scaffold-free neocartilage in vitro: influence of growth factors. Tissue Eng. Part A 16, 513–521. , , , , , , &
- 1997. Developmental regulation and asymmetric expression of the gene encoding Cx43 gap junctions in the mouse limb bud. Dev. Genet. 21, 290–300. , , , , , &
- 2008. Matrix development in self-assembly of articular cartilage. PLoS ONE 3, e2795. , , , , &
- 2001. From fixed to FRAP: measuring protein mobility and activity in living cells. Nat. Cell Biol. 3, E145–E147. &
- 2005. Gap junctions regulate extracellular signal-regulated kinase signaling to affect gene transcription. Mol. Biol. Cell 16, 64–72. &
- 1972. Cell density and cell division in the early morphogenesis of the chick wing. Nat. New. Biol. 239, 24–26. &
- 1996. A human chondrodysplasia due to a mutation in a TGF-beta superfamily member. Nat. Genet. 12, 315–317. , , , , &
- 1997. Disruption of human limb morphogenesis by a dominant negative mutation in CDMP1. Nat. Genet. 17, 58–64. , , , , , , &
- 1998. Long-chain n-alkanols and arachidonic acid interfere with the Vm-sensitive gating mechanism of gap junction channels. Pflugers Arch. 435, 310–319. &
- 2004. Mid-trimester fetal blood-derived adherent cells share characteristics similar to mesenchymal stem cells but full-term umbilical cord blood does not. Br. J. Haematol. 124, 666–675. , , ,
- 2010. Chondrogenic transdifferentiation of human dermal fibroblasts stimulated with cartilage-derived morphogenetic protein 1. Tissue Eng. Part A 16, 1633–1643. , , , , , , &
- 2011. Human mesenchymal stem cells induced by growth differentiation factor 5: an improved self-assembly tissue engineering method for cartilage repair. Tissue. Eng. Part C. Methods. 17, 1189–1199. , , , , , &