Expression pattern differences between osteoarthritic chondrocytes and mesenchymal stem cells during chondrogenic differentiation

Osteoarthritis and Cartilage 18 (2010) 1596e1607

Expression pattern differences between osteoarthritic chondrocytes and mesenchymal stem cells during chondrogenic differentiation P. Bernstein y *, C. Sticht z, A. Jacobi y, C. Liebers y, S. Manthey x, M. Stiehler y y Department of Orthopaedics, University Hospital Carl Gustav Carus, Fetscherstr. 74, 01307 Dresden, Germany z Medical Research Center, University of Heidelberg, Theodor-Kutzer-Ufer, 68167 Mannheim, Germany x Department of Traumatology, University Hospital Carl Gustav Carus, Fetscherstr. 74, 01307 Dresden, Germany

a r t i c l e i n f o

s u m m a r y

Article history: Received 20 May 2010 Accepted 17 September 2010

Objective: The use of mesenchymal stem cells (MSCs) for cartilage regeneration is hampered by lack of knowledge about the underlying molecular differences between chondrogenically stimulated chondrocytes and MSCs. The aim of this study was to evaluate differences in phenotype and gene expression between primary human chondrocytes and MSCs during chondrogenic differentiation in three-dimensional (3D) pellet culture (PC). Materials and methods: Chondrocytes isolated from cartilage samples obtained during total knee alloarthroplastic procedure (N ¼ 8) and MSCs, purified from bone marrow aspirates of healthy donors (N ¼ 8), were cultivated in PC under chondrogenic conditions. Immunohistology and quantitative reverse transcribing PCR (RT-PCR) were performed for chondrogenic-specific markers (i.e., Sox9, Collagen II). Global gene expression of the so-cultivated chondrocytes and MSCs was assessed by a novel approach of microarray-based pathway analysis. Refinement of data was done by hypothesis-driven gene expression omnibus (GEO) dataset comparison. Validation was performed with separate samples in transforming growth factor (TGF)bþ or TGFb conditions by use of quantitative real-time RT-PCR. Results/conclusions: Chondrogenic commitment of both cell types was observed. Interestingly, chondrocytes demonstrated an upregulated fatty acid/cholesterol metabolism which may give hints for future optimization of culture conditions. The novel microarray-based pathway analysis applied in this study seems suitable for the evaluation of whole-genome based array datasets in case when hypotheses can be backed with already existing GEO datasets. Future experiments should further explore the different metabolic behaviour of chondrocytes and MSC. Ó 2010 Osteoarthritis Research Society International. Published by Elsevier Ltd. All rights reserved.

Keywords: Mesenchymal stromal cells Chondrocyte Pellet culture 3D culture Microarray Gene expression Human

Introduction Osteoarthritis (OA) constitutes a major health problem and biological, non-endoprosthetical osteochondral joint reconstruction remains a yet unsolved task in the interdisciplinary field of cartilage and bone tissue engineering research. Regenerative therapy of OA joints has to focus on several features of the disease1,2, e.g., mechanical imbalance and biological alterations (chondrocyte senescence or apoptosis, decreased synthesis of cartilage matrix proteins, inflammation). While some of those issues can be addressed surgically and pharmacologically, a major challenge remains the regeneration of

* Address correspondence and reprint requests to: Peter Bernstein, Department of Orthopaedics, University Hospital Carl Gustav Carus, Fetscherstraße 74, Building 29, D-01307 Dresden, Germany. Tel: 49-351-458-18174; Fax: 49-351-449-210413. E-mail address: [email protected] (P. Bernstein).

damaged cartilage by tissue engineering strategies. Although chondrocytes derived from osteoarthritic donors can be used for cartilage tissue engineering, their limited availability, quantity, and viability depict major drawbacks2. The use of alternative cell types with chondrogenic differentiation potential may offer a solution. Mesenchymal stem cells (MSCs) have shown to be able to differentiate into cartilage3,4. However, recent data gives conflicting results concerning mechanical quality of MSC-derived cartilage constructs, suggesting inferior properties of chondrogenic differentiated MSCs5e7. Interestingly, MSCs only exhibit inferior compressive properties, while tensile properties do not seem to differ from chondrocytes6,8. The amount of produced extracellular matrix varies interindividually and depends on the scaffold matrix used for tissue engineering7. Beside the evaluation of these mechanical facts, a detailed subcellular and phenotypic comparison of MSCs with chondrocytes is missing. Apart from the cellular source, a key factor for successful cartilage tissue engineering is three-dimensional (3D) cultivation.

1063-4584/$ e see front matter Ó 2010 Osteoarthritis Research Society International. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.joca.2010.09.007

P. Bernstein et al. / Osteoarthritis and Cartilage 18 (2010) 1596e1607

Chondrocytes undergo rapid dedifferentiation when cultivated in a two-dimensional (2D) environment. Sustained chondrogenic differentiation can only be achieved by the embedding of cells into biomaterials or by material-free pellet culture (PC)9,10. PC is a matrix-free 3D cultivation method which has shown to elicit a considerable chondrogenic effect on chondrocytes and MSCs9,11,12. The aim of the study was to compare temporal global gene expression patterns of PC-conditioned human primary chondrocytes vs human bone marrow-derived MSCs using microarray-based global gene expression analysis. As major drawbacks of microarray experiments are the questionable biological relevance of their results at considerable high costs we figured out a new approach to cope with those issues13. To pave the way for further experiments we wanted to exemplify the validation of our data in transforming growth factor (TGF)bþ or TGFb conditions. TGFb is a potent stimulator of collagen biosynthesis, extracellular matrix (ECM) accumulation as well as mesenchymal cell proliferation and activation14. It therefore can possibly pronounce pro-chondrogenic effects. In accordance to our previous work we chose TGF-b3 for our experiments9,15. Materials and methods All experiments using OA cartilage tissue were approved by the local institutional review board (IRB) (protocol number: EK264122004). MSCs were obtained from the Medical clinic of the university hospital for which a separate IRB vote exists. All patients’ material was obtained after informed consent. Isolation and culture of cells Primary chondrocytes were obtained from osteoarthritic knee joints during alloarthroplastic procedures from five female patients (mean age 74 years, range 61e85 years) for microarray experiments and from one male and two female patients (mean age 55, range 48e60 years) for polymerase chain reaction (PCR) validation experiments. MSCs were isolated from iliac crest bone marrow from five healthy female donors (mean age 45 years, range 42e48 years) for the microarray experiments and from one male and two female healthy donors (mean age 54 years, range 42e71 years) for PCR-validation experiments. Chondrogenic commitment of the cells used was proofed by quantitative RT-PCR analysis (Suppl. Table I). Chondrocytes were isolated by collagenase (Collagenase Type I, Worthington; Cat.No: CLS1) digestion overnight and expanded in Dulbecco’s modified eagle medium (DMEM, Low Glucose, 1x, Invitrogen, Germany) supplemented with 10% fetal calf serum (FCS); 1% Penicillin/Streptomycin. Ficoll-gradient separated MSCs obtained from iliac crest-harvested bone marrow aspirates were kindly provided by Prof. Martin Bornhäuser. FACS analysis revealed positivity for CD166, CD105, CD73, CDw90 (all 99%) and CD146 (30e60%) and negativity for CD45, CD34 and CD14 (data not shown). Passage 0 primary chondrocytes and MSCs, respectively, were used for the experiments.

1597

CO2 with medium changes performed every third day. For PCRvalidation experiments, cells were cultivated in the presence or absence of TGF-ß3 (TGF-bþ/TGF-b). Histology For histological analysis, pellets cultured for 2 weeks were fixed with paraformaldehyde (1%), dehydrated, paraffin-embedded and cut into slices of 1 mm thickness. Immunohistology was performed using the following antibodies: Collagen I (Clone: COL-1, #C2456, Sigma, Germany), 1:200; Collagen II (Clone: 2B1.5, #MS-235-P, ThermoScientific, Germany), 1:200; Collagen X (Clone: COL-10, #C7974, Sigma, Germany), 1:500; Sox9 (polyclonal, #AB5535, Chemicon, Germany), 1:100. After detection with the appropriate biotinylated secondary antibody and AB-complex (ABC Standard Kit, Linaris, Germany), Romulin AEC Chromogen kit (Zytomed, Germany) was used for staining. All antibodies were test-run with positive controls, every section was accompanied by a negative control (without primary antibody) to rule out unspecific effects. Histology results were backed by quantitative reverse transcribing PCR (RT-PCR) data (Suppl. Table I). Isolation of RNA Total RNA was extracted after 0, 3, 7 and 14 days using a twostep extraction protocol as described by Hoemann et al.16. Briefly, one pellet sample was mixed with 350 ml of RLT-b-mercaptoethanol solution (RNeasy kit, Qiagen, Germany) and incubated for 30 min at 4 C in a thermomixer shaking vigorously. After centrifugation at 15,000  g for 10 min at 4 C the procedure was repeated adding 250 ml of RLT-b-mercapto-ethanol solution followed by

Pellet culture (PC) P0-cells were cultured in high-density PCs, as described previously9. Briefly, 1.5  106 cells were placed into a conical polypropylene tube in chondrogenic medium (see below). The cells were centrifuged at 400  g for 10 min. Cell pellets were cultured in DMEM with 1% ITS (Insulin-Transferrin-Selenium 100x, Invitrogen, Germany), 100 U/ml penicillin, 100 mg/ml streptomycin, 210 mmol ascorbic acid, 10 nmol dexamethasone (SigmaeAldrich, Germany) and 10 ng/ml TGF-ß3 (Strathmann Biotec, Germany) at 37 C, 5%

Fig. 1. Data analysis pathway, depicting the three major steps of the analysis process. (1): microarray-screening and filtering through paths A and B by comparing gene expression trends, resulting in gene list 1; (2): hypothesis-driven selection of biological relevant genes, resulting in gene list 2; (3): validation by quantitative RT-PCR.

1598

P. Bernstein et al. / Osteoarthritis and Cartilage 18 (2010) 1596e1607

Table I Gene list 1 after trend-analysis: Analysis path A: initial change from day 0 to day 3 of at least 20,5 and continuous up- or downregulation, differing from the other cell type. Analysis path B: threshold was set to zero; only those genes showing permanent departure in one distinct direction (up or down) contrasting to the curve of the other cell Group

Path A

TGF-b related

ID3 FST ID4 TBL1XR1 ADAMTS5 ACAN1 MGAT4B PLOD1 HS3ST3A1 LIMA1 CNN3 ACTN1 WDR1 ACTR2 MYH9 DIAPH3 CAPG ACTR1A ADAM10 ITGA8 GPR124 TUBA1C FLNB RHOB LPXN TNFRSF11B HEG1 CFL2 POPDC3 BEX1 PAFAH1B1 MAP1B NTRK2 BDNF SEMA3A NR4A2 RNF141 LDLR SQLE FADS3 SMPDL3A FDFT1 ANXA6 SC5DL CYP51A1 SGMS1 ACAT1 FDPS HMGCS1 SC4MOL

Wnt-pathway Glycans

Actin metabolism

Integrins/motility

Bone development Muscle development

Neuronal development

Sperm development Lipid metabolism

Path B

Group

Path A

Vesicle transport

M6PR STXBP5 SCRN1 TRAM1 SLC39A10 SH3RF1 CTSB UBE2A DDEF2 SSR4 A2M RNF144A NNT ABCB10 FOXO1 GYG1 PRC1 TPX2 BAG2 BOK EPAS1 EIF2AK4 PENK KCTD20 SLC38A6 CLIC4 PTPLAD1 FOSB FRMD6 LMCD1 LEPROT APOBEC3B GRAMD3 RCAN2 FAM62A CCND1 RAB7L1 FAM83D MRGPRF CXorf6 GCLC SNX25 FAM46A DUSP23 RPS20

Cellular energy

Mitosis SPINT2 GBL

Apoptosis Stress reaction

Channel/GTPases

Unclassified BEX1

FAM12A ATP8A2

a centrifugation step at 21,000  g for 10 min. Both supernatants were mixed and the RNeasy procedure was carried out as described by the manufacturer. Microarray analysis RNA samples from all five patients of each group were pooled to obtain sufficient material for further analysis and to mitigate for interindividual variations. Gene expression profiling was performed using N ¼ 8 HG-U133 Plus 2.0 arrays (Affymetrix, Santa Clara, CA, USA). Furthermore biotinylated antisense cRNA was prepared according to the Affymetrix standard labelling protocol. Afterwards, the hybridization was performed using a GeneChip Hybridization oven 640, subsequently dyed in a GeneChip Fluidics Station 450 and thereafter scanned with a GeneChip Scanner 3000 (all hardware from Affymetrix, High Wycombe, UK). A Custom CDF (Chip Definition File) Version 10 with Entrezbased gene definitions was used to annotate the arrays. The Raw

Path B

MTUS1 XIST SATB2 TDRD1 SYNE2 FTL PTGER2

fluorescence intensity values were normalized applying quantile normalization method, using a commercial software package SAS JMP7 Genomics, version 3.1, from SAS (SAS Institute, Cary, NC, USA). All gene expression values were calculated as ratios of the same cell type relatively to day 0. The raw and normalized data is deposited in the Gene Expression Omnibus (GEO) database (http:// www.ncbi.nlm.nih.gov/geo/; accession No. GSE19664). PCR experiments RNA was reverse transcribed into cDNA using the Omniscript RT-Kit (Qiagen, Germany) and dT18-oligonucleotides. Quantitative PCR experiments were performed on days 7 and 14 on a MX4000 Multiplex Quantitative PCR machine (Stratagene, The Netherlands) using the appropriate PCR kits (Brilliant Multiplex and SYBR green Master Mix) and selected oligonucleotides (sequences can be obtained from the author). The reaction protocol included an initial denaturation step for 10 min at 95 C which was

P. Bernstein et al. / Osteoarthritis and Cartilage 18 (2010) 1596e1607

followed by 40 cycles (30 s 95 Ce1 min 60 Ce30 s 72 C). Quantification was carried out in duplicates, gene transcription was normalized to the GAPDH-value of the same experiment and time point. PCR reactions were carried out in a Stratagene Real-Time PCR System. Data analysis Data analysis was splitted in a screening process (Microarray step ¼ gene list 1), interpretative analysis against GEO datasets (gene list 2) and a quantitative validation step via quantitative RT-PCR. The whole analysis process is outlined in Fig. 1. Gene list 1: microarray data trend-analysis Gene expression dynamics were analyzed from day 3 to day 14 as follows. All genes with an initial fold-change in expression from day 0 to day 3 of at least >20,5 were included for further analysis. Curves were classified as either constantly upregulated, constantly downregulated, or not constantly (¼unsteady) regulated. If curve classification revealed inequality between MSCs and chondrocytes (e.g., MSC ¼ up, chondrocytes ¼ unsteady) the respective gene was added to gene list 1 (Fig. 1, path A). Curves of the genes not fulfilling the initial threshold of 20,5 but showing constant dissimilarity of both curves (e.g., MSC ¼ up, chondrocytes ¼ down) were also added to gene list 1 (Fig. 1, path B). Genes were annotated to functionally distinct groups by use of gene ontology (GO) terminology. Gene list 2: trend comparison Biologically relevant background data was retrieved by a GEO search that included keywords such as “chondrocyte” or “MSC”. GEO profile datasets were required to represent a time-based observation under standardized culture conditions. It therefore resulted in a panel of GEO datasets covering processes of mesenchymal limb bud development, tissue engineering, and regeneration after injury (see Table II). Regulation curves of the genes in gene list 1 (Table I) were manually compared with the respective gene regulation of the selected GEO dataset from Table II. It was decided, if the curve trend fitted to the MSC or the chondrocyte behaviour. In cases with multiple differently regulated samples, even a both-like (MSC and

1599

chondrocyte) behaviour could be attested. Qualifying genes were added to gene list 2 (Table III). Otherwise genes were given the status “no similarity”. If the specific gene was not present in the GEO dataset, the status “not available (NA)” was given. Validation step: PCR quantification and statistics Functionally distinct genes from gene list 2 (Fig. 1) were further analyzed via quantitative RT-PCR method. All primers used were validated for specificity by gel electrophoresis of their respective products (data not shown). Data is presented as means. Statistical significance was calculated with ANOVA, using SPSSÒ 17 (SPSS Inc., Chicago, US). P-values < 0.05 were considered significant. Results Immunohistology PC-conditioned chondrocytes showed Sox9-expression comparable to MSCs with a trend towards more intranuclear retention [Fig. 2 (A) and (B)]. Collagen staining demonstrated a clear superiority of collagen type II over type I. This effect was pronounced in chondrocytes, whereas staining in MSCs appeared to be weak [Fig. 2 (C)e(F)]. Both cell types stained highly positive for collagen X [Fig. 2 (G) and (H)]. Validation was carried out by quantitative RT-PCR of the respective genes and showed a significant higher Sox9-expression in chondrocytes, whereas collagen type I, II and X did not differ significantly (Suppl. Table I). Microarray analysis Whole-genome expression data were compared between chondrocytes and MSCs at various time points (days 0, 3, 7, 14, Fig. 3, line A). Most similarity, i.e., 95%, of genetic expression between MSCs and chondrocytes existed at time point zero (start). During cultivation cell type-specific differences were observed, reducing similarity to 92% at day 14. When comparing monolayer-cultivated P0-MSCs (MSC, day 0) with different stages of chondrocyte pellet redifferentiation (days 3, 7, 14, Fig. 3, line C), decreasing similarity from 95% down to 83%

Table II Panel of GEO datasets used for comparison and selection of genes backed by the justifying hypotheses GEO dataset and reference

Biological process Tissue culture system

GDS1865 James et al.23

Endochondral differenttiation pathway

GDS3062 Larson et al.24

Transition from proliferation to differenttiation

GDS1404 Cameron Retina regeneration et al.46

GDS234 Zhao et al.47

Muscle regeneration

GDS63 Di Giovanni Neuronal regeneration et al.48

Cultivation

Murine micromass culture of E11.5 mouse embryo limb buds

2:3 DMEM/F12, 10% FBS, 0.5 mM glutamine, 25 U/ml penicillin, 25 mg/ml streptomycin, 0.25 mM ascorbic acid, 1 mM b-glycerophosphat P2-Human MSC 2D culture; a-minimal essential medium, 17% FBS, 2 mM glutamine, 5% confluence until 100 U/ml penicillin, 70% confluence 100 mg/ml streptomycin

Analysis

Justifying hypothesis for selection as a comparison tool

 Affymetrix MOE430A (w14,000 probe sets)  Three replicates, aiming at the difference between d15 and d3  d0, d3, d6, d9, d12, d15  Affymetrix HG-U133 Plus 2.0 (w31,000 probe sets)  Three replicates aiming at the difference between d2 and d7  d2, d7  Affymetrix Zebrafish Genome Array  d2, d3, d5, d14

MSC differentiation in limb buds will resemble most parts of MSC differentiation towards the chondrogenic lineage.

Zebrafish retina injury

Fresh tissue

Cardiotoxin e mediated muscle injury (mouse model)

Fresh tissue



Spinal cord injury (rat model)

Fresh tissue

    

An increasing confluence of proliferating MSCs will resemble the pellet situation in a timely delayed manner.

In vivo regenerative processes will resemble MSC differentiation in PC, depending on the similarity of tissues. Affymetrix Murine Genome Pellet MSC would behave U74A version 1 more similar to muscle Two replicates regeneration than retina or d0, 12 h, d1, d2, d4, d10 neuronal regeneration. Affymetrix rat U34A array Two replicates 30 min, 4 h, 24 h, 7d

1600

P. Bernstein et al. / Osteoarthritis and Cartilage 18 (2010) 1596e1607

Table III Gene list 2: results of the trend comparison of GEO datasets with chondrocyte or MSC PC. C ¼ GEO gene behaves like chondrocytic PC, M ¼ GEO gene behaves like MSC PC, CþM ¼ GEO gene is similar to chondrocytes and MSC GEO-comparison with chondrocyte PC (C) and MSC-pellet culture (M) Functional group/gene

Actin-metabolism

Glycan-metabolism

Integrins TGFb related

Muscle development

Neuronal development

Lipids/steroids/ cholesterol

Vesicle transport

Channels Cellular energy

Not further classified

GDS1865 “limb bud” ARP2 actin-related protein 2 homolog (ACTR2) Capping protein (actin filament, CAPG) Myosin, heavy chain 9, non-muscle (MYH9) LIM domain and actin binding 1 (LIMA1) Actinin, alpha 1 (ACTN1) Calponin 3, acidic (CNN3) Aggrecan (ACAN) ADAM metallopeptidase with thrombospondin motif (ADAMTS5) Mannosyl (alpha-1,3-)-glycoprotein beta-1,4-N-acetylglucosaminyltransferase, isozyme B (MGAT4B) ADAM metallopeptidase domain 10 (ADAM10) Inhibitor of DNA binding 3, dominant negative (ID3) Inhibitor of DNA binding 4, dominant negative (ID4) Follistatin (FST) HEG homolog 1 (zebrafish, HEG1) POPDC3 Cofilin 2 (muscle, CFL2) Brain-derived neurotrophic factor (BDNF) Brain expressed, X-linked 1 (BEX1) Platelet-activating factor acetylhydrolase, isoform Ib, alpha subunit 45 kDa (PAFAH1B1) Nuclear receptor sub-family 4, group A, member 2 (NR4A2) Acetyl-Coenzyme A acetyltransferase 1 (ACAT1) Cytochrome P450 oxidase, family 51, sub-family A , polypeptide 1 (CYP51A1) Farnesyl-diphosphate farnesyltransferase (FDFT1) Farnesyl-diphosphate synthase (FDPS) Low density lipoprotein receptor (LDLR) Squalene epoxidase (SQLE) Sterol-C4-methyl oxidase-like (SC4MOL) Fatty acid desaturase 3 (FADS3) HMGCS1 Cathepsin B (CTSB) Solute carrier family 39, type 10 (SLC39A10) Syntaxin binding protein 5 (tomosyn, STXBP5) Translocation associated membrane protein 1 (TRAM1) CLIC4 NNT ATP-binding cassette, sub-family B (ABCB10) Glycogenin 1 (GYG1) Ferritin, light polypeptide (FTL) Cyclin D1 (CCND1) FBJ murine osteosarcoma viral oncogene homolog B (FOSB) Leptin receptor overlapping transcript (LEPROT) FERM domain containing 6 (FRMD6) MAS-related GPR, member F (MRGPRF) RAB7, member RAS oncogene family-like 1 (RAB7L1) Spectrin repeat containing, nuclear envelope 2 (SYNE2) Tudor domain containing 1 (TDRD1) Family with sequence similarity 83, member D (FAM83D) Regulator of calcineurin 2 (RCAN2) Dual specificity phosphatase 23 (DUSP23) Family with sequence similarity 62 (FAM62A) GRAM domain containing 3 (GRAMD3)

occurred. When comparing dedifferentiated monolayer-cultivated P0-chondrocytes (chondrocytes, day 0) with different stages of chondrogenic MSC-pellet cultivation (days 3, 7, 14, Fig. 3, line B), the gain of difference between both cell types was smaller (compare lines C vs B), ranging from 95% down to 90% similarity. Microarray data trend-analysis was done as shown in Fig. 1. Data reduction using path A resulted in 95 genes (gene list 1, see Table I), of which most belonged to lipid/cholesterol pathway, actin and integrin pathway as well as to vesicle transport systems. Applying path B resulted in 12 genes, belonging to integrins, sperm and lipid metabolism. The gene brain expressed, X-linked 1 (BEX1) is represented in both analysis pathways as it shows characteristics of path A and B.

GDS1404 “retina”

GDS234 “muscle”

C C C e e e C C C

GDS3062 “MSC confluence” CþM e CþM CþM e e C C M

e e e C e e e e e

e e e M M M C e e

e C C C e e e e e e

C M M e C C CþM M M C

e e e e e e e e e e

M M e M e e e M C C

e M M M M M M e C e C e e e C M e e C C C C C C C C C e e e e e

M e C M M e M M C C C CþM CþM CþM C C e e C M M C e e C e e C M CþM CþM CþM

C e e e e e e e e e e e e e C e e e e e e e e e e e e e e e e e

e e e e M M M e e e e e e e e C C C M e M e e e e e e e e e e e

Most genes of gene list 1 demonstrate a final gene expression increase in chondrocytes during pellet cultivation (Fig. 4). GEO trend comparison GEO database search resulted in five datasets which are described more precisely in Table II. One hundred and seven genes from gene list 1 were manually compared with the trends of the genes or their homologues in the mentioned datasets. It could be demonstrated that in the group “differentiating limb buds” (GDS1865) 45 genes showed similar regulation to either MSCs or chondrocytes or both; in the group “confluently growing MSCs” (GDS3062) there were 62 genes showing similar patterns. In the

P. Bernstein et al. / Osteoarthritis and Cartilage 18 (2010) 1596e1607

1601

Fig. 2. Immunohistological staining of chondrocyte (A, C, E, G) and MSC pellets (B, D, F, H) reveals sufficient Sox9-expression (A, B) and only limited intranuclear retention in MSC (B). Collagen I is produced on a low level in both cell types (C, D). Collagen II is highly upregulated in chondrocytes (E) as opposed to MSCs (F). Both cell types show an upregulated expression of collagen X (G, H). Sections are representative for larger areas of the specimen. Scale bars ¼ 20 mm.

groups “retina regeneration” (GDS1404) and “muscle regeneration” (GDS234) only nine and 30 genes, respectively, showed a similar regulation. The group “neuronal regeneration” (GDS63) resulted in 15 comparable genes. Due to this low sample size, further analysis within this group was not carried out.

The result of the similarity of chondrocytes and MSCs with GEO datasets in temporal gene expression levels is summarized in gene list 2 (Table III). A minimum of one gene out of each functional group, occurring in more than one GEO datasets and matching to chondrocyte or

1602

P. Bernstein et al. / Osteoarthritis and Cartilage 18 (2010) 1596e1607

Fig. 3. Multivariate correlation matrix of genetic expression similarity between chondrocytes and MSCs compared with the monolayer startpoint (day 0) during pellet cultivation until day 14. Line A: Similarity between chondrocytes and MSCs; line B: similarity between dedifferentiated day-0 chondrocytes and pellet-cultured MSCs; line C: similarity between day-0 MSCs and pellet-cultured chondrocytes.

MSC behaviour in PC, was arbitrarily selected for further PCR experiments (if possible, at least one gene out of each GO-group with key relevance to the respective pathways). PCR quantitative analysis Regardless of TGFb supplementation, aggrecan (ACAN1) expression levels were approximately 20e30-fold increased in chondrocytes compared to MSCs in the first and second weeks of cultivation (Fig. 5). Of all the other genes, only inhibitor of DNA binding (ID4) was expressed significantly (30-fold) by chondrocytes and MSCs after 2 weeks of cultivation with TGFb. Low-value-difference patterns were visible for chloride intracellular channel 4 (CLIC4), being expressed 1,1e1,5-fold (week 1eweek 2, TGFbþ) or 2,1e2,6-fold (TGFb) higher in chondrocytes, reaching significance only in the TGFb group. A similar trend could be observed for the 3-hydroxy-3-methylglutaryl-Coenzyme A synthase 1 (HMGCS1), which was 1,1e1,6-fold (TGFbþ) or 2,1e2,8-fold (TGFb) higher in chondrocytes, although only close to significance. Significant upregulation could be seen in the chondrocytic ADAM metallopeptidase (ADAMTS5) only under TGFb conditions (2,3e2,8-fold), in the first-week chondrocytic nicotinamide nucleotide transhydrogenase (NNT, 2,1-fold) and in the second-week chondrocytic cytochrome P450 oxidase (CYP51A1, 3,7-fold). The chondrocytic gene popeye domain containing 3 (POPDC3) showed a significant reduction of expression in the second week under TGFb-stimulation to 30% of the MSC-value. The other genes were low and not significantly regulated (Fig. 5). Of the significantly regulated genes, chondrocytic aggrecan (ACAN1) and POPDC3 showed the least sensitivity to TGFb-regulation, whereas a reduction of chondrocytic CYP51A1 (0,77-fold), CLIC4 (0,56-fold), NNT (0,54-fold), ADAMTS5 (0,4-fold) and ID4 (0,09-fold) could be observed (Fig. 5). Discussion In the field of cartilage tissue engineering there is a major need for a sufficient yield of cells. Since healthy cartilage offers only 5%

cells/volume, an even lower amount of chondrocytes is usually harvested from diseased joints17. MSC has shown to possess suitable chondrogenic potency18. However, MSC-derived chondrocytes show inferior mechanical properties and produce less extracellular matrix proteins compared with primary chondrocytes19. We therefore investigated whether we could depict gene expression differences between chondrocytes and MSCs during chondrogenic redifferentiation by PC. Commitment to chondrogenic differentiation We have started our experiments by confirming chondrogenic commitment of PC-conditioned chondrocytes and PC-conditioned MSC immunohistologically and by quantitative RT-PCR analysis of Sox9, collagen type I, II and X. Chondrocytes and MSC, showed a ready progression towards the chondrogenic lineage. Sox9-expression was higher in chondrocytes accompanied by an immunohistological notion of more intranuclear retention. Although there were no significant differences in collagen I and II mRNA expression, our data suggests that MSC’s may be less efficient at translating, processing or incorporating collagen into the extracellular matrix. Chondrogenic arrest is missed and both cell types proceed towards hypertrophy. This, being a still unsolved major drawback of in vitro culture systems, has been described before in more detail19,20. Analysis strategy Classical ways of microarray analysis use a number of replicates for each study point and try to filter biological information in a more or less computed fashion21. The involved procedures rely on statistical methods that are still flawed by difficulties (e.g., multiple testing) and uncertainties about the biological relevance of the results13,22. Alternatively, we used a serial approach, including microarray-screening, GEO-comparison and PCR validation of

P. Bernstein et al. / Osteoarthritis and Cartilage 18 (2010) 1596e1607

1603

Fig. 4. Trend-distribution matrix showing the final difference in gene expression between day 14 and day 3 (dynamics) of selected genes from gene list 1 from chondrocytes and MSCs, cultivated as pellets. A dynamic value > 0 represents an increasing expression.

candidate genes (Fig. 1). As the microarray data is retrieved through pooling of samples, statistical relevance cannot be calculated. Hence, this denotes a potential limitation of the study that was dealt within the further experimental design. First, biological variation was already incorporated through the pooling of samples. Second, we underweighted microarray data and did not attempt to implement statistical safety in the beginning. We used the two microarray “clouds” of chondrocyte and MSC behaviour to filter out the differences. Omitting doubtful statistical procedures had the advantage of gaining a high sensitivity in the screening process. Low specificity is raised in a step-wise manner as described in the methods section. We analyzed the microarray data against an arbitrarily selected, biologically relevant background of regenerative processes, assessed through carefully conducted microarray experiments. In contrast to a non-hypothesis-driven analysis, often performed in whole-genome expression analysis experiments, we generated hypotheses for every GEO dataset (Table. II). Notably, this approach facilitated to narrow the scope of candidate genes and allowed for the interpretation from a system-biological point of view. However, there are clear limitations: hypothesis-driven microarray analysis only works when statistically safe background data is available.

Result of hypothesis testing For example, in contrast to the initial hypothesis, chondrocyte PC resembled limb bud outgrowth23 more likely than MSC did. Regulatory similarities were more likely to be seen in chondrocytic PC and were more closely related to the chondrogenic pathway. Another hypothesis e MSC PCs would be closer related to 2D-MSCs reaching confluence e was also falsified: 2D-expansion of MSCs resembled changes occurring in chondrocytic PCs more likely than in MSC PCs. This accounts especially for differences that were validated for the cholesterol pathway and for chondrogenic marker genes. One might be tempted to accuse the high-serum conditions used in the reference experiment “confluently growing MSCs”24 to explain the unexpected similarity between cell layer-cultivated MSC and chondrocytes cultured in 3D PC. To gain more insight into the validity of the observed trends, validation experiments were performed using quantitative RT-PCR. Triggers of chondrogenic effects In contrast to the observed phenotypic and functional differences, chondrocytes expressed significantly more aggrecan

1604

P. Bernstein et al. / Osteoarthritis and Cartilage 18 (2010) 1596e1607

Fig. 5. Results of quantitative RT-PCR validation of gene expression differences between chondrocytes and MSC in the first and second week of PC in TGFb and TGFbþ culture conditions (n ¼ 6 patients, see material part). P-values of significant differences are indicated above the bars. Error bars show 95% confidence interval.

compared with MSCs. This has not been shown before quantitatively. Supposedly to support aggrecan production, the aggrecanase ADAMTS5 was downregulated in chondrocytes. This occurred only under the influence of TGF-b as has been shown before25e27 and correlates well with other already described pro-chondrogenic effects of TGF-b, e.g., its chondro-protective effect in osteoarthritic joints28, its role in preventing terminal endochondral differentiation29, and its beneficial use in cartilage tissue engineering30. TGF-b-driven effects denote an example of the sensitivity of cellbased constructs towards culture conditions.

Although not included in RT-PCR-validation, chondrocytic upregulation of the glucosaminyltrans-ferase MGAT4B demonstrated the increased glycano-anabolic activity in chondrocytes when compared with MSCs. Fatty acid metabolism and cholesterol biosynthesis Most strikingly, chondrocytes showed an upregulated fatty acid metabolism (CYP51A1, FADS3, SMPDL3A) and cholesterol-/sterol biosynthesis-pathway (FDFT1, FDPS, HMGCS1, LDLR, SC4MOL,

P. Bernstein et al. / Osteoarthritis and Cartilage 18 (2010) 1596e1607

SC5DL, SQLE) in the microarray-screening. Chondrocytic upregulation was validated for HMGCS1 and the cytochrome P450 oxidase CYP51A1, thus verifying an increased cholesterol-biosynthetic activity in chondrocytes. This phenomenon was discovered earlier as a specific feature of growth plate chondrocytes31. The importance of the cholesterolbiosynthetic pathway for chondrocyte and limb bud development has been demonstrated by linking it to hedgehog signalling32e34, nuclear receptor roralpha35 and to liver X and retinoid receptors36. MSCs, however, do not upregulate their cholesterol biosynthesis, which could be interpreted as a hint that their original cellular signalling context is located farther away from the joint line e more towards bone37. MSC in their relation to bone Osteoprotegerin (TNFRSF11B) upregulation underlines the specific feature of MSCs to readily undergo osseous differentiation38,39. Compared to chondrocytes, their suitability for a regenerative cartilage therapy seems therefore to be limited e at least under the conditions used in our experimental setting. Other differences between chondrocytes and MSC ID-Proteins function as Corepressors by binding of Helix-LoopHelix motifs of tissue-specific transcription factors. Higher expression of the inhibitors of DNA binding ID3 and ID4 in MSCs e as was detected in our experiments e can be indicative for a sustained undifferentiated state40. Downregulation of the eukaryotic translation initiation factor EIF2AK4 marks an increase of protein biosynthesis in MSCs41. Proenkephalin (PENK) and endothelial PAS domain protein 1 (EPAS1) are hypoxia-responsive genes that support angiogenesis42. This seems to be an answer of oxygen-deprived MSC (due to PC condition) that cannot be seen in chondrocytes. Chondrocytes expressed more of the intracellular chloride channel CLIC4, a conserved gene among vertebrates. CLIC4 was shown to be a mediator of TGF-b signalling via its translocation to the nucleus. Although its function is not yet clear, it seems to have some relevance in endothelial tubule formation, anti-apoptotic action and fibroblast-to-myofibroblast-transdifferentiation in cancer cells43,44. Despite its primary description as an intracellular channel, involvement in cellular volume regulation could be a function of the untranslocated protein45. Conclusion Phenotype and gene expression of chondrocytes and MSCs differ during chondrogenic PC. Our data suggests, that MSC in simple PC do not seem to reach the same stage of chondrogenic differentiation as chondrocytes do, regardless of TGFb-supplementation and despite harvesting from younger and healthier donors. Most strikingly, MSCs show a reduced expression of chondrogenic matrix proteins, e.g., collagen II and aggrecan. On the other side, MSCs are more rapidly prepared for an osseous transformation. A specific metabolic aspect of chondrocytes, that has not been described in comparison with MSCs before, is their upregulated fatty acid/cholesterol biosynthesi-pathway. This pathway may be a potential target for optimization of chondrogenic MSC culture conditions. We interpreted our data in the background of already existing datasets, thus bringing it closer to biological relevance. This may be an advantageous choice in cases when hypotheses can be backed by already existing expression data. However, this might not be the

1605

case for other situations when such data does not exist and statistical safety is important to the whole-genome dataset. Future experiments should explore the different metabolic behaviour of chondrocytes and MSC and their biological relevance. Author contributions Peter Bernstein: conception, design, analysis and data interpretation, drafting, final approval, responsibility for integrity. Carsten Sticht: analysis and data interpretation, statistical expertise, critical revision and final approval. Angela Jacobi: conception and design, analysis and data interpretation, critical revision. Cornelia Liebers: technical support, provision of study material, critical revision. Suzanne Manthey: technical support, critical revision. Maik Stiehler: conception and design, data interpretation, multiple critical revisions, final approval. Role of funding source This study was funded by the Saxonian Ministry for Science and Arts (SMWK), Germany and the German Academic Exchange Service/German Federal Ministry of Education and Research (grant # D/09/04774, MS). Conflict of interest All authors hereby disclose financial and personal relationships that would have biased this work. Acknowledgements Special thanks belong to Stefan Fickert, MD, who initiated parts of the project. We are very grateful to Prof. Martin Bornhäuser, MD and Katrin Müller for providing primary MSC, Juliane Rauh, PhD and Henriette Friedrich for technical assistance and to Denis Corbeil, PhD for providing laboratory infrastructure. Supplementary Material Supplementary data associated with this article can be found in the online version at doi:10.1016/j.joca.2010.09.007. References 1. Goldring MB, Goldring SR. Osteoarthritis. J Cell Phys 2007;213:626e34. 2. Nesic D, Whiteside R, Brittberg M, Wendt D, Martin I, MainilVarlet P. Cartilage tissue engineering for degenerative joint disease. Adv Drug Deliv Rev 2006;58:300e22. 3. Johnstone B, Hering TM, Caplan AI, Goldberg VM, Yoo JU. In vitro chondrogenesis of bone marrow-derived mesenchymal progenitor cells. Exp Cell Res 1998;238:265e72. 4. Pittenger MF, Mackay AM, Beck SC, Jaiswal RK, Douglas R, Mosca JD, et al. Multilineage potential of adult human mesenchymal stem cells. Science 1999;284:143e7. 5. Karlsson C, Brantsing C, Svensson T, Brisby H, Asp J, Tallheden T, et al. Differentiation of human mesenchymal stem cells and articular chondrocytes: analysis of chondrogenic potential and expression pattern of differentiation-related transcription factors. J Orth Res 2007;25:152e63. 6. Mauck RL, Yuan X, Tuan RS. Chondrogenic differentiation and functional maturation of bovine mesenchymal stem cells in long-term agarose culture. Osteoarthritis Cartilage 2006;14: 179e89.

1606

P. Bernstein et al. / Osteoarthritis and Cartilage 18 (2010) 1596e1607

7. Erickson IE, Huang AH, Chung C, Li RT, Burdick JA, Mauck RL. Differential maturation and structure-function relationships in mesenchymal stem cell- and chondrocyte-seeded hydrogels. Tissue Eng Part A 2009;15:1041e52. 8. Huang AH, Yeger-McKeever M, Stein A, Mauck RL. Tensile properties of engineered cartilage formed from chondrocyteand MSC-laden hydrogels. Osteoarthritis Cartilage 2008;16: 1074e82. 9. Bernstein P, Dong M, Corbeil D, Gelinsky M, Günther K-P, Fickert S. Pellet culture elicits superior chondrogenic redifferentiation than alginate-based systems. Biotechnol Prog 2008;25:1146e52. 10. Bernstein P, Dong M, Graupner S, Corbeil D, Gelinsky M, Günther KP, et al. Sox9 expression of alginate-encapsulated chondrocytes is stimulated by low cell density. J Biomed Mater Res A 2009;91A:910e8. 11. Tallheden T, Karlsson C, Brunner A, Van Der Lee J, Hagg R, Tommasini R, et al. Gene expression during redifferentiation of human articular chondrocytes. Osteoarthritis Cartilage 2004;12:525e35. 12. Ichinose S, Tagami M, Muneta T, Sekiya I. Morphological examination during in vitro cartilage formation by human mesenchymal stem cells. Cell Tissue Res 2005;322:217e26. 13. Sims AH. Bioinformatics and breast cancer: what can highthroughput genomic approaches actually tell us? J Clin Pathol 2009;62:879e85. 14. Ramirez F, Tanaka S, Bou-Gharios G. Transcriptional regulation of the human alpha2(I) collagen gene (COL1A2), an informative model system to study fibrotic diseases. Matrix Biol 2006;25:365e72. 15. Tare RS, Howard D, Pound JC, Roach HI, Oreffo RO. Tissue engineering strategies for cartilage generationemicromass and three dimensional cultures using human chondrocytes and a continuous cell line. Biochem Biophys Res Commun 2005;333:609e21. 16. Hoemann CD, Sun J, Chrzanowski V, Buschmann MD. A multivalent assay to detect glycosaminoglycan, protein, collagen, RNA, and DNA content in milligram samples of cartilage or hydrogel-based repair cartilage. Anal Biochem 2002;300:1e10. 17. Mollenhauer JA. Perspectives on articular cartilage biology and osteoarthritis. Injury 2008;39(Suppl 1):S5e12. 18. Yoo JU, Barthel TS, Nishimura K, Solchaga L, Caplan AI, Goldberg VM, et al. The chondrogenic potential of human bone-marrow-derived mesenchymal progenitor cells. J Bone Joint Surg Am 1998;80:1745e57. 19. Pelttari K, Steck E, Richter W. The use of mesenchymal stem cells for chondrogenesis. Injury 2008;39(Suppl 1):S58e65. 20. Mwale F, Stachura D, Roughley P, Antoniou J. Limitations of using aggrecan and type X collagen as markers of chondrogenesis in mesenchymal stem cell differentiation. J Orthop Res 2006;24:1791e8. 21. Murphy D. Gene expression studies using microarrays: principles, problems, and prospects. Adv Physiol Educ 2002;26:256e70. 22. Pounds SB. Estimation and control of multiple testing error rates for microarray studies. Brief Bioinform 2006;7:25e36. 23. James CG, Appleton CT, Ulici V, Underhill TM, Beier F. Microarray analyses of gene expression during chondrocyte differentiation identifies novel regulators of hypertrophy. Mol Biol Cell 2005;16:5316e33. 24. Larson BL, Ylostalo J, Prockop DJ. Human multipotent stromal cells undergo sharp transition from division to development in culture. Stem Cells 2008;26:193e201. 25. Gendron C, Kashiwagi M, Lim NH, Enghild JJ, Thogersen IB, Hughes C, et al. Proteolytic activities of human ADAMTS-5:

26. 27.

28.

29. 30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

41.

comparative studies with ADAMTS-4. J Biol Chem 2007;282:18294e306. Huang K, Wu LD. Aggrecanase and aggrecan degradation in osteoarthritis: a review. J Int Med Res 2008;36:1149e60. Yamanishi Y, Boyle DL, Clark M, Maki RA, Tortorella MD, Arner EC, et al. Expression and regulation of aggrecanase in arthritis: the role of TGF-beta. J Immunol 2002;168:1405e12. Pujol JP, Chadjichristos C, Legendre F, Bauge C, Beauchef G, Andriamanalijaona R, et al. Interleukin-1 and transforming growth factor-beta 1 as crucial factors in osteoarthritic cartilage metabolism. Connect Tissue Res 2008;49:293e7. Li TF, O’Keefe RJ, Chen D. TGF-beta signaling in chondrocytes. Front Biosci 2005;10:681e8. van der Kraan PM, Buma P, van Kuppevelt T, van den Berg WB. Interaction of chondrocytes, extracellular matrix and growth factors: relevance for articular cartilage tissue engineering. Osteoarthritis Cartilage 2002;10:631e7. Le Lous M, Corvol MT, Maroteaux P. Lipid composition of two types of chondrocytes in primary culture. Calcif Tissue Int 1981;33:403e7. Aguilar A, Wu S, De Luca F. P450 oxidoreductase expressed in rat chondrocytes modulates chondrogenesis via cholesteroland Indian Hedgehog-dependent mechanisms. Endocrinology 2009;150:2732e9. Schmidt K, Hughes C, Chudek JA, Goodyear SR, Aspden RM, Talbot R, et al. Cholesterol metabolism: the main pathway acting downstream of cytochrome P450 oxidoreductase in skeletal development of the limb. Mol Cell Biol 2009;29: 2716e29. Wu S, De Luca F. Role of cholesterol in the regulation of growth plate chondrogenesis and longitudinal bone growth. J Biol Chem 2004;279:4642e7. Woods A, James CG, Wang G, Dupuis H, Beier F. Control of chondrocyte gene expression by actin dynamics: a novel role of cholesterol/roralpha signaling in endochondral bone growth. J Cell Mol Med 2009;28:28. Gentili C, Tutolo G, Pianezzi A, Cancedda R, Descalzi Cancedda F. Cholesterol secretion and homeostasis in chondrocytes: a liver X receptor and retinoid X receptor heterodimer mediates apolipoprotein A1 expression. Matrix Biol 2005;24:35e44. Mak KK, Chen MH, Day TF, Chuang PT, Yang Y. Wnt/betacatenin signaling interacts differentially with Ihh signaling in controlling endochondral bone and synovial joint formation. Development 2006;133:3695e707. Mauney JR, Kirker-Head C, Abrahamson L, Gronowicz G, Volloch V, Kaplan DL. Matrix-mediated retention of in vitro osteogenic differentiation potential and in vivo bone-forming capacity by human adult bone marrow-derived mesenchymal stem cells during ex vivo expansion. J Biomed Mater Res 2006;79A:464e75. Niemeyer P, Kasten P, Simank HG, Fellenberg J, Seckinger A, Kreuz PC, et al. Transplantation of mesenchymal stromal cells on mineralized collagen leads to ectopic matrix synthesis in vivo independently from prior in vitro differentiation. Cytotherapy 2006;8:354e66. Bonora-Centelles A, Jover R, Mirabet V, Lahoz A, Carbonell F, et al. Sequential hepatogenic transdifferentiation of adipose tissue-derived stem cells: relevance of different extracellular signaling molecules, transcription factors involved and expression of new key marker genes. Cell Transplant 2009;18 (12):1319e40. Lu W, Laszlo CF, Miao Z, Chen H, Wu S. The role of nitric-oxide synthase in the regulation of UVB light-induced phosphorylation of the alpha subunit of eukaryotic initiation factor 2. J Biol Chem 2009;284:24281e8.

P. Bernstein et al. / Osteoarthritis and Cartilage 18 (2010) 1596e1607

42. Takeda N, Maemura K, Imai Y, Harada T, Kawanami D, Nojiri T, et al. Endothelial PAS domain protein 1 gene promotes angiogenesis through the transactivation of both vascular endothelial growth factor and its receptor, Flt-1. Circ Res 2004;95:146e53. 43. Singh H, Cousin MA, Ashley RH. Functional reconstitution of mammalian “chloride intracellular channels” CLIC1, CLIC4 and CLIC5 reveals differential regulation by cytoskeletal actin. Febs J 2007;274:6306e16. 44. Suh KS, Malik M, Shukla A, Yuspa SH. CLIC4, skin homeostasis and cutaneous cancer: surprising connections. Mol Carcinog 2007;46:599e604.

1607

45. Jentsch TJ. CLC chloride channels and transporters: from genes to protein structure, pathology and physiology. Crit Rev Biochem Mol Biol 2008;43:3e36. 46. Cameron DA, Gentile KL, Middleton FA, Yurco P. Gene expression profiles of intact and regenerating zebrafish retina. Mol Vis 2005;11:775e91. 47. Zhao P, Iezzi S, Carver E, Dressman D, Gridley T, Sartorelli V, et al. Slug is a novel downstream target of MyoD. Temporal profiling in muscle regeneration. J Biol Chem 2002;277:30091e101. 48. Di Giovanni S, Knoblach SM, Brandoli C, Aden SA, Hoffman EP, Faden AI. Gene profiling in spinal cord injury shows role of cell cycle in neuronal death. Ann Neurol 2003;53:454e68.